text
stringlengths 0
1.19M
|
|---|
m−1 () and the contact pressure had varied from 460 to 55 MPa. On the polished coatings in ambient air, in the Ringer’s solution and the synthetic serum, the final wear rate is low. It was hardly measurable in the solution and the serum with values equal or lower than about 1×10−9 |
mm3 |
N−1 |
m−1. This can be shown in that present the pin profiles before and after the friction test carried out in the Ringer’s solution and the final image of the pin tip, respectively. In these cases, it is obvious that the contact pressures had remained high during all the test. The SEM and EDX images of the disc tracks provided some evidence of the transfer of titanium on the diamond surface. However, this feature was seen with great difficulty and only in the coating tested in ambient air.With the Co–28Cr–6Mo alloy, the friction coefficient in air is about 0.1 at the beginning of the test. After about 500 m it reaches a nearly constant value of about 0.2 (). With the Ringer’s solution and the synthetic serum the friction coefficient is in the 0.03–0.08 range. The final wear rate is higher in ambient air than in the Ringer’s solution and the synthetic serum where the values are in the (1–1.5)×10−9 |
mm3 |
N−1 |
m−1 range (With the 316L steel, the friction coefficients is about 0.1 in ambient air and it is also lower with the Ringer’s solution and the synthetic serum (). The final wear rates are reported in . It can be noted that the wear rate are nearly equal in all environmental conditions. It seems that the influence of corrosion in the Ringer’s solution and the synthetic serum is negligible in conditions of low wear rate by abrasion on the smooth diamond surface. Only some pits and corroded crevices were evident on the edge of the wear scars produced on the 316L pin while most of the surface was only polished (This result can be extended to the other alloys. Considering all these different metal-on-diamond friction tests, it seems that a tribo-corrosion phenomena does not noticeably influence the wear rates of the metallic counterfaces (). This result should be confirmed in alternating pin-on-disc and fretting tests to conclude that a synergistic interaction of corrosion with mechanical effects can be neglected in the control of the amount of wear particles and ions released in a body fluid. Results on wear debris and their size could not be obtained in our testing conditions. If crevice corrosion would be a non-negligible mode for material damaging in these mechanical conditions, it is known however that Co–Cr–Mo alloys are more resistant than 316L steel. Moreover, Ti–6Al–4V alloy is insensible to this phenomenon.In a third embodiment, orthopedic implants could utilize a UHMWPE on diamond coating contact. It is well established that unidirectional sliding of UHMWPE does not well simulate the multi-directional sliding results for wear In ambient air, the friction coefficient of UHMWPE sliding against the polished coating increases to high values reaching about 0.6 as shown in . In the Ringer’s solution and the synthetic serum, the friction coefficients are lower than 0.1. Still lower values probably would have been obtained if the load had not been changed because the friction coefficient increased at each change and then decreased to lower values, for example about 0.03 at the end of the test in the Ringer’s solution.Considering the final wear rates of the UHMWPE, it is obvious that it decreases as a function of the environment, from ambient air to the Ringer’s solution (). A lower wear rate is still obtained in synthetic serum, with a moderate value of 2.5×10−8 |
mm3 |
N−1 |
m−1, in accordance with the well-known influence of serum proteins that act as “lubricant” in UHMWPE-containing biomechanical contacts Adherent as-deposited smooth diamond coatings on Ti–6Al–4V alloy can be considered for biomechanical use. They permit low friction coefficient and low wear, from 10−10 to a few 10−10 |
mm3 |
N−1 |
m−1 in Ringer’s solution and synthetic serum, respectively, when both counterface parts are coated. A self-polishing mechanism allows the roughness to decrease during the wear test (from 15 to 30 nm) for the as-deposited coating to a few nanometers, 3–5 nm. In accordance, the wear rate is very low, even with relatively high contact pressures.Polished nano-smooth diamond coatings deposited on Ti-alloy can also be considered for biomechanical use with a metallic alloy counterface. Except in ambient air with the Co–28Cr–6Mo alloy, the friction coefficients are always lower than 0.1. The lowest wear rates are obtained with Ti–6Al–4V and Co–28Cr–6Mo alloys, with values equal to or lower than about 10−9 |
mm3 |
N−1 |
m−1. No noticeable tribo-corrosion phenomena have been evidenced either in the saline solution or in the synthetic serum. This could be due to the very small abrasion of the alloys on the nano-smooth diamond surfaces. However, more thorough studies should be carried out in the future to gain more insight on that subject, especially in alternating and multi-directional sliding and in fretting conditions.With the UHMWPE on diamond contact, the beneficial influence of proteins is observed with a lower wear rate in synthetic serum than in Ringer’s solution. This “lubricant” effect permits the wear rate of UHMWPE to decrease into the 10−8 |
mm3 |
N−1 |
m−1 range.This preliminary study allows smooth diamond coatings on Ti–6Al–4V alloy to be considered as a material for various biomechanical applications. Further study, especially with good simulators are however necessary to conclude about the interest of this material.On bend-stretch forming of aluminum extruded tubes — I: experimentsTubular aluminum frame parts for automotive applications are best produced by extrusion. The tubes are then cold formed to the required shape by prestretching, pressurizing and bending them over rigid dies. Tension prevents buckling of the compressed side and significantly reduces the springback on unloading. An unwanted byproduct of the process is distortion of the cross section. It has been found that modest levels of pressure can reduce this distortion. The selection of the level of tension and pressure for optimum forming is presently empirical. The study discussed herein seeks to develop a scientific basis for optimizing forming processes such that buckling is avoided and distortion and springback are minimized. Part I describes a custom bend-stretch-pressure forming facility developed for the study. The facility is operated by one pneumatic and two servohydraulic closed-loop systems. This allows computer control of the process, and affords selectable loading histories. The planar forming process was modeled by approximating the tube as a nonlinear elastic–plastic beam which can undergo large rotations. The model was shown capable of reproducing accurately the loading history experienced by different sections along the length of the part during forming. Representative results from forming experiments involving rectangular aluminum are presented. The results are used to discuss the effect of friction, tension and pressure on the cross-sectional distortion, springback and net elongation of the part. Part II presents a model for establishing the cross-sectional distortion induced during forming. The model is used in conjunction with experimental results to establish ways of optimizing the process.The majority of current car frames are made by stamping steel sheets into the required shapes and spot welding them together (unibody concept). By contrast, a significant number of modern automotive frame components are cold formed, extruded aluminum tubes. Such tubes are assembled to form a space frame; a revolutionary design concept which has been used in several low production volume all aluminum cars such as the Audi A8. It is composed of aluminum extruded tubes joined with cast aluminum connecting nodes. The joints are arc-welded and the skins are attached to the frame by rivets. Before the car could be built, appropriate alloys for the frame, for the nodes and for the skins had to be developed. Innovations were also necessary in advanced extrusion processes, in new cold forming machines and processes, in robotic arc-welding techniques, in robotic assembling processes, etc. Typically, the tubes are stretch formed. In this process, they are prestretched, internally pressurized and bent to the required shapes over dies.Automated assembly requires unusually tight dimensional tolerances on the formed tubes. Typical cross-sectional dimensions are 50–150 mm and wall thickness ranges between 1 and 4 mm. The cross-sectional distortions resulting from forming must be kept within ±0.5 mm and shape deviations along the length of the members must be equally small. For such aluminum designs to become a serious alternative to current unibody steel designs, they must be amenable to automated manufacturing processes for high production speed and lower cost. For these reasons, the space frame may not be easily adaptable to mass produced cars. However, cold formed aluminum extruded tubes remain a competitive option for use in other hybrid frame design concepts where tubular components are incorporated into the more traditional unibody design (e.g. Honda Insight).The extent to which hollow closed sections can be bent is limited by various forms of buckling which, for the materials and geometries of interest here, occur in the plastic range. Thus, the onset and evolution of prevalent buckling modes are a strong function of the section geometry and the stress–strain behavior of the alloy. One way of delaying the onset of such bending instabilities is application of tension during forming. Tension reduces, and can even eliminate, compressive stresses due to bending, and thus buckling is avoided. A practical way of accomplishing this is to bend the part over a curved mandrel. The axial tension is reacted by contact with the mandrel (see review by Welo An unwanted byproduct of bending of thin-walled sections is distortion of their cross section The current state of the art remains empirical, requiring extensive trial and error testing, and several die modifications before an acceptable forming process (i.e., die shape, prestretch load, level of internal pressure) is established for a new part. It is important to note that the cross-sectional shapes of many of the car components are rather complex and their design is often decided by esthetic concerns rather than structural performance. Any new car design will have a large number of different tubes for which forming dies must be made. The empiricism of the present way of designing the manufacturing process is inefficient.The objective of this study was to develop a scientific basis for designing a forming process for each part which ensures accuracy in shape with minimum distortion of its cross section, is free of wrinkles and with minimum or predictable springback. These goals are pursued through a combination of experiment and analysis. In Part I, we first describe a custom bend-stretch-pressure forming facility developed for the study. Two series of experiments involving rectangular aluminum tubes were performed using the forming facility to determine how the applied loads interact to determine the final shape of the formed tubes. In Part II of the series, a relatively simple two-dimensional model of the forming process is presented which is able to accurately capture the cross-sectional distortion induced by the forming process. The model is validated by comparing the predicted shapes with corresponding experimental results. The model, in conjunction with experiments, is then used to investigate alternative loading histories. Finally, the insight gained is used to formulate a methodology for selecting the optimum loading history for a particular formed part.A stretch-forming machine of the type used in industry to cold form tubes of various cross sections is shown schematically in . The part is gripped in open-sided jaws mounted on two large hydraulic actuators. The grips also seal the ends. The tube is first pressurized to the required level and is subsequently stretched to a plastic stress state. The pressure and axial load are then kept constant while the horizontal table moves forward and the die engages the part forcing it to conform to the required shape. As the die moves forward, the actuators rotate so as to remain approximately tangential to the two ends of the tube. The pressure and the axial load are subsequently released, the die is retracted, and the part is removed from the machine. Typical forming cycles currently range from 30 to 60 s.The forming machine described is prohibitively large for laboratory testing. In addition, we wanted to add feedback and computer control to our facility. For these reasons, we opted to develop a smaller custom facility for this experimental program which had the desired flexibility and features.A photograph of the bend-stretch-pressure testing facility developed is shown in . The major components of the facility are identified in the scaled drawings in . It consists of a rigid reaction frame (length 108 in or 2743 mm) made of structural tubing (8×4×0.5 in or 203×102×13 mm). The frame is mounted horizontally on Π-frame legs equipped with wheels and leveling mounts and is at a height of 50 in (1270 mm) off the floor. While the actual machine () operates in a horizontal plane, this one operates in a vertical plane. This change removed the bending loads due to gravity and eliminated the need for linear support slides for the moving die. Bending is limited to just one plane and twisting of the specimen is precluded. (An alternate laboratory stretch forming facility is described in Refs. Tension is applied to the test specimen via two hydraulic actuators with the following characteristics: bore 3.25 in (83 mm), rod diameter 1.375 in (35 mm) and rod length 12.5 in (318 mm). They are connected to the frame via trunnion mounts which slide into specially designed bearings mounted to the sides of the fame. The bearings allow the actuators to rotate freely about pivots so that they can follow the motion of the ends of the part and remain approximately tangential to it during the forming process (cf. ). The applied tension is monitored by in-line load cells (20 kips or 90 kN capacity) placed between each actuator and pin connection. The extension of each actuator is monitored by an LVDT displacement transducer.Several circular forming dies have been fabricated out of 3 in (76 mm) thick solid steel plate. They were milled to precision and ground finished using NC milling machines. They have radii in the range of 50–20 in (1270–508 mm). The dies are connected to two vertical parallel actuators 14 in (356 mm) apart with the following characteristics: 3.25 in (83 mm) bore, 1.75 in (44 mm) rod and 26 in (660 mm) extension. The two actuators are mechanically and hydraulically connected so they displace equally (see ). The double actuator arrangement was adopted because it provides additional torsional and bending rigidity to the system required due to the length of rods when fully extended. This allowed us to eliminate the slides used in the actual system (see ). Low profile, high stiffness load cells (20 kips or 90 kN capacity) are attached between each actuator and the connecting plate. The position of the die is monitored by a magnetostrictive displacement transducer. The load and displacement capacities of the facility are listed in The results presented here come from rectangular cross-sectional tubes with dimensions of 50×30×1.8 mm custom made by Alcoa for this research program (Al-C210-T4). Steel end-plugs are bonded to the ends of the tube and pin-connected to the actuator rods. High-strength Hysol (EA9430) epoxy was used over an area of 8 in2 (5160 mm2) so that the strength of the bonded joint exceeded that of the tubes. The pin-to-pin length of the specimen can vary, but 1000 mm is typical. By using different end-plugs, the specimen can be bent either about its wide or narrow side. Switching to tubes of other shapes mainly involves the construction of new pairs of compatible end-plugs.The test specimen can be pressurized internally by compressed air or, alternatively, it can be evacuated through ports in the tube end-plugs (see ). The pressure is set by a pneumatic servovalve with a capacity of −1–10 atm (bar) and monitored by a pressure transducer.The most significant improvement in the testing facility over commercial units is the addition of the instrumentation to each of the three modes of loading and inclusion of feedback control. These in turn allow the execution of a forming test by a computer and afford some freedoms in the choice of the loading path followed. The stretching (H) and forming (V) modes of loading are connected to a general purpose hydraulic power unit (3000 psi or 207 bar). These two loading modes are controlled via closed-loop servohydraulic control systems shown schematically in . They can be operated either under load or displacement control. In the experiments that follow, the H-mode was operated under load control while the V-mode was operated under displacement control. The P-mode has its own feedback loop with pressure as the feedback signal.The distortion of the cross section of the tube is monitored by a custom-built biaxial transducer shown in , usually mounted at the mid-span of the specimen. The transducer monitors simultaneously the maximum deflection of the top face of the rectangular tube (see ) and the maximum change in the width of the tube (bulging see ). The moving parts of the transducer contact the specimen via rollers.The transducer rests on the circular die with wheels so that by interrupting a test it can be rolled along the die to record cross-sectional deflections along the length of the specimen. The transducer position along the die is monitored as follows. A thin tape with regularly spaced black and silver markings is bonded to one edge of the curved surface of the die. An emitter/receiver photodiode pair mounted on the traveling transducer produces an electrical pulse whenever it encounters a silver band (see ). Position along the die is then related to the count of pulses from the center of the specimen. Typically, a signal is recorded every 0.2 in (5.1 mm) on the curved surface of the die. (The importance of the final shape of formed parts and an alternate method of measuring the shape is discussed in Ref. The signals of one-load cell and one-displacement transducer from each of the V- and H-mode are conditioned via a four-unit MTS 458.20 control console. The console also enables control of the two modes via the feedback signals of choice. The remaining transducer signals from the V- and H-mode are conditioned separately as shown in the block diagram in . The pressure transducer and the associated conditioner and controller are an integral part of the pneumatic servovalve. The signals from the two-displacement transducers in the biaxial distortion device are conditioned as shown in The command signals for the three closed loading loops are generated by a data acquisition/control system operated by a PC (200 MHz Pentium Pro). Two data acquisition cards, each with 2 D/A output channels, are used to generate the command signals (12 bit resolution) for the V-, H-, and P-modes. The ±10 V signals are generated at a maximum update rate of 1 MS/s. The control program uses the LabVIEW environment and generates the command signals which allow a choice of the tension-pressure-vertical displacement path that will be followed in a test.Concurrently, the data acquisition cards record the signals from the 11 transducers present in the test facility. Each of the cards is equipped with 8 differential input D/A channels which operate with 12-bit resolution at a maximum sampling rate of 1.25 MS/s. When all 11 transducers are monitored, the sampling rate for an individual channel is approximately 100 kS/s. The signals are processed through LabVIEW so as to have real-time display of all of the variables. In its standard operation mode, the system continuously monitors and records the transducer signals. When an axial scan of the cross-sectional deflections is performed, recording of measurements is triggered by the voltage pulses generated by the photodiode. For redundancy and safety, several of the key variables are also monitored via digital volt-meters and analog plotters.Five test specimens were cut from each of the tubes provided by Alcoa. The following geometric measurements were made on each using a micrometer. Three measurements of the tube height (h) and width (b) were made at the two ends and at the middle of the specimen. The thickness (t) was measured at 12 locations (three measurements of each of the four walls, identified by subscripts i=1,4) on each end of the specimen. The mean values and the variations of each of the variables are listed in . The wall thickness was usually nearly constant for three of the four tube walls. The fourth wall (one of the two short sides) was usually about 5% smaller.The excess material was used to obtain the mechanical properties of each length of tube used in the study. Uniaxial tension tests were performed on strips cut from the tubes. The mean value and the variation of Young's modulus (E) and the yield stress (σ0) of the tubes used in this series of tests are given in . Tests performed using strips from different side walls of the tube did not show variation in the material properties. Tensile tests were also performed on specimens cut transverse to the tube axis. The difference between the axial and transverse yield stress was insignificantly small. The tubes came from the same batch thus differences in mechanical properties from tube to tube were relatively small.The surfaces of the end-plugs and the tube were prepared for bonding by sanding with a medium-grit sandpaper and cleaning with 1–1–1 trichloroethane. The adhesive was cured at 180° F (82°C) for 1 h.To minimize the effect of friction, the surface of the die was covered with an 8 mil Teflon tape while simultaneously the contacting surface of each test specimen was coated in light grease. The effect of friction will be discussed later in light of the results.A typical forming test was comprised of the steps listed below. Tests in which we deviated from this procedure will be described and discussed in the section on experimental results.Two series of forming experiments were performed using the custom bend-stretch-pressure testing facility and procedure described in the previous section. The goal of these experiments was to determine how the final curvature, cross-sectional shape and overall extension were affected by the forming process, and in particular, the magnitude of the applied loads. Tubes were formed either about the tall or the short side of the cross section. For the experiments discussed here, the die radius was 20 in (508 mm). The pressure, tension and the loading history were varied sufficiently to establish trends. All specimens were approximately 36.5 in (927 mm) long (2L) which includes the 1.5 in (38 mm) end sections bonded to the solid inserts. In all experiments reported here, the die was moved at a constant velocity of ḋ=1.3 in/min (33 mm/min). This relatively slow forming rate was adhered to in order to enable complete acquisition of the signals from all variables discussed earlier. The experiment was usually terminated when approximately 82% of the initial length of the tube was in contact with the die. In other words, just before the plugged ends came into contact with the die to prevent failure of the bonds. through a sequence of photographs of the tube as it progressively conforms to the shape of the die. This tube was bent with the long sides of its cross section vertical. A tension T1=4000 lbf (18 kN — i.e. 62% of the yield tension, T0) was applied first and was subsequently held constant during forming. No internal pressure (or vacuum) was applied in this case. shows a plot of the vertical force seen by the die (Fν) normalized by 2T1 versus the die displacement (d). The relationship is nearly linear. (Note that Fν/2T1 represents the sine of the current angle the axis of the tension actuators make with the horizontal direction.) Marked on the force–displacement response are the locations corresponding to the numbered photographs in . The initial straight configuration is not visible from the observation angle used. Bending is initially localized at the mid-span where the tube first comes into contact with the mandrel. After this point has conformed to the mandrel's curvature, the active zone moves away from the mid-span. As the die moves vertically, the deforming tube emerges from behind the horizontal support beam of the device. In configuration , the die displacement is already 5.28 in (134 mm). By configuration , the full length of the tube becomes visible while the ends of the tension actuators enter the viewing frame in configuration . It is interesting to observe that the sections of the tube not in contact with the die remain relatively straight and that the tension actuators rotate so that they remain approximately tangential to the nearly straight ends. This is responsible for the nearly linear relationship between Fν and d seen in . The test was terminated at configuration when the die displacement was 21.6 in (550 mm) and 80% of the original length of the tube was in contact with the die.The performance envelope of our testing machine was established using a numerical model of the forming process in which the tube is represented as a nonlinear elastic–plastic beam as described in (i.e. beam cross-sectional distortions are neglected). The calculated die force–displacement response is included in and is seen to be in good agreement with the measured response. Included in the figure is the calculated contact length (sA/L in ) as a function of the die displacement. In general, the contact length is slightly smaller than measured values primarily due to the higher rigidity of the assumed “beam” cross section and due to the initial delay in coming into contact with the die exhibited by the model (see ). A sequence of calculated deformed beam configurations corresponding to the same die displacements as those of the sequence of photographs in . The progression of events is very similar to that in the photographic sequence.We will now take a more detailed look at a test which closely follows present practice. The tube cross section is oriented with its long sides parallel to the axis of bending. The tube was pre-stretched to a tension T1=0.95T0 and then pressurized internally to 2.3 bar (P). The tension and pressure were then held constant during forming. shows a plot of the vertical force seen by the die (Fν/2T1) versus the die displacement. The relationship is again nearly linear.The cross section of the deforming tube distorts as it comes into contact with the die. The distortion is partly due to the bending and can be mitigated by a relatively small value of internal pressure. The extent of distortion is defined by the variables Δ (or ) and Γ defined in the same figure. The distortion variables will be expressed as ratios of the tube mid-surface height (h0) and width (b0), respectively (see ). Because the tube cross section has an aspect ratio of approximately 1.62 the values of the ratios will be distinctly different in the two tube orientations used.The evolution of distortion at the mid-span of the tube was monitored by the biaxial distortion transducer discussed in . The measured values of Δ and Γ are plotted against the displacement of the die in . Initially, the top surface of the tube bulges slightly outward due to the internal pressure. As the tube is bent further, the top surface begins to sag inward and reaches its final value (Δ/h0≈−5.7%) when the die displacement is approximately 8 in (203 mm). The small discontinuities on the two trajectories correspond to the locations when the test was interrupted and an axial scan of the cross section of the part of the tube in conduct with the die was performed. The side walls distort in a similar manner but much less (Γ/b0≃1%). Indeed, this distortion is coupled to the distortion of the top surface in order to maintain compatibility of rotations at the corners of the cross section. During the initial phase when the top surface is bulging, the tube walls are bent inward causing the device to measure the position of the tube corners. After the die has displaced 2 in (51 mm), the side walls begin to deform outward and engage the device. Local distortion ceases once this location has conformed to the shape of the die.In this test, the forming was stopped periodically and an axial scan of the portion of the tube in contact with the die was performed using the biaxial distortion transducer. Results from four scans taken at the vertical die displacements indicated are shown in (sDis a measure of distance along the circular surface of the die). The distortion is symmetric about the mid-span and increases slightly away from the center. Once again, distortion ceases at a particular cross section once it has locally conformed to the shape of the die (i.e. bending ceases). The relative magnitude of the two measures of distortion remains similar to that seen at mid-span.Measurements of the distortion made after the specimen was removed from the test facility are presented in (s is a measure of distance along the inner surface of the deformed tube). Here Δ* represents the sum of the distortion of the top and bottom surfaces of the tube, measured individually, relative to the corners (see ). (Note that the contribution to Δ* from the bottom surface of the tube is rather small.) This removes the contribution from the change in h0 due to the Poisson effect which is included in the variable Δ measured by the biaxial transducer during the experiment. As a result, this more accurate representation of distortion is less than that seen in . A similar symmetric distribution is observed with the minimum distortion occurring at the mid-span and increasing as the distance from the center increases.As we will see, this pattern of the distortion at mid-span being smaller than in the rest of the tube is persistent in all results and for both tube orientations. One of the reasons for this is the fact that the central part initially bends with a shear point force which starts at zero and grows to certain value when the section has conformed to the radius of the die (see ). A second reason is that the deforming sector of the tube is supported on both sides by undeformed sections which add to its stiffness. By contrast, for sections away from the mid-span, the local shear is at the fully developed value which tends to grow slightly as the contact length increases. In addition, now the deforming sector sees a deformed sector on one side and an undeformed sector on the other which makes it a bit more compliant.The final radius of curvature (ρf) of the tube was measured along its length and is compared to the radius of the die (ρ) in . The tube sprang back approximately 4.9%. The springback is relatively uniform along the length with a slight increase close to the ends. This is partly due to the influence of the solid insert plug at the ends of the test specimen.We now consider a test in which the cross section is oriented so that the long sides are vertical. The test specimen was also prestretched to a tension T1=0.95T0 but no internal pressure was applied because, with the tube in this orientation, internal pressure only exacerbates the bending induced distortions. The vertical force seen by the die is plotted against its displacement in . The relationship is again nearly linear. The distortion variables Δ and Γ measured at the mid-span are plotted in against the die displacement. The distortions increase monotonically to values of Δ/h0≈−4.76% and Γ/b0≈6.62% achieved at a die displacement of approximately 8 in (203 mm). Subsequently, the distortions do not change. In practice, both values are too high so that this part would be rendered a reject.Again axial scans of the distortion, in the portion of the tube in contact with the die, were taken during periodic pauses of the motion of the die. Four such scans of the two distortion variables are shown in . As in the previous case, the distortion profiles have depressions around the mid-span and grow slightly towards the ends. We observe once more that, at any given axial position, the distortion reaches certain values at a particular die position and subsequently remains unchanged. Clearly, for this orientation of the cross section, the distortion of the longer side walls is more pronounced. shows how the distortion variable Γ varied along the length of the formed tube after it was removed from the machine. The distribution along the length is similar to that in (including the small asymmetry about the mid-span). The maximum value is approximately 9.8%.The final local radius of curvature of the tube is compared to the radius of the die in the same figure. The tube sprang back approximately 4.34%. Once again, the springback is relatively uniform along the length with a slight increase close to the ends.Friction can play a role in the forming process, especially if the tension is changed once the tube is in contact with the die. Frictional effects were minimized by covering the surface of the die with a thin Teflon tape and simultaneously lubricating the surface of the specimen which contacts the die. The adequacy of this scheme was evaluated in special tests in which the axial strains at different positions along the length of the tube were compared. Results from such a test are shown in . The tube is formed with the long sides vertical. Pairs of strain gages were bonded on each side wall of the tube (see inset in ) at the mid-span and 7 in (178 mm, S/L=0.39) away from it at a height of 0.5 in (12.7 mm) above the surface of the die. The tube was prestretched to T1=0.77T0 and evacuated to a pressure of −0.48 bar. shows a plot of the applied axial force versus the average strain (εs) at each of the two locations. At the end of forming, the strains at the two locations differ by 3%. Once the tube came into contact with the die the tension was increased to T5=0.98T0 while the vacuum was held at the same level. (Numeric variable subscripts refer to the generic loading history shown schematically in . In this case, no postpressure was applied thus points coincide.) The strains at the two locations increased by essentially the same amount as indicated in by the nearly parallel trajectories traced during this phase. It is thus concluded that the lubrication adopted is effective.The strain histories at the two locations as a function of the die displacement are compared in . We observe that the strain at mid-span goes through a maximum (d=3.35 in or 85 mm) and eases down to its steady-state value of 1.03% at a die displacement of 5 in (127 mm). Subsequently, εS remains unchanged until the tension is increased at the end of forming. The beam model of the forming process yields that the mid-span conforms to the shape of the die at d=3.68 in which compares well with the time the maximum distortion is recorded. By contrast, in we saw that in that experiment the distortion variables reached their full values at a die displacement of approximately 8 in (203 mm). Although most of the distortion is induced during the time the section in question is bent, a small additional increase takes place subsequently as the neighboring sections bend in the process reducing their support of the cross section at mid-span.The evolution of εS at S/L=0.39 is significantly different. This section bends gradually due to beam action up to a die displacement of approximately 12 in (305 mm). At this time, the strain experiences a sharp increase as the tube locally conforms to the curvature of the die. The steady-state value is reached when d=15 in (380 mm). As mentioned above, the steady-state values at the two locations differ by approximately 3%.The final distortion (Γ) distribution along the length of the tube is similar to that in but with a somewhat reduced amplitude due to the internal pressure. The tube was found to have sprang back approximately 5.05%. Once again, the springback was relatively uniform along the length.Tension plays two important roles in bend-stretch forming of tubes. First, it is used to eliminate (or reduce) compression on the concave side of the part and thus prevent wrinkling or local buckling for tubes bent with their long sides horizontal and in for tubes bent with the long sides vertical. The first set shows the final radius of curvature along the length of four tubes formed at tensions of T1/T0=0.54, 0.65, 0.95 and 1.08. As the tension increases, the springback is reduced from approximately 7.3% to approximately 4.5%. The effect of tension on springback for the other tube orientation () is similar. In both sets of experiments, the shape is quite uniform along the length although a tendency to become more uniform at higher tensions is seen in for three tubes formed with the long sides horizontal. An internal pressure of 2.28 bar was applied while the tension had values of T1/T0=0.65, 0.95 and 1.08. The plot compares the distortion profiles (Δ*) along the length (measured after unloading) for the three cases. As was the case for the tubes discussed earlier, the distortion profiles exhibit depressions around the mid-span which grow with tension. Away from the mid-span, the distortion clearly grows with tension (by nearly a factor of two between the lowest and highest values of T1).The effect of the bend-stretch process on the cross section becomes more pronounced if we compare the final distortion profiles of two tubes formed at the same tension with and without internal pressure. shows such a comparison of Δ*/h0 profiles for two tubes formed at T1/T0=0.95; one had no internal pressure, while for the second, a constant pressure of P2=2.28 bar was applied. In the presence of pressure, the average distortion was reduced from −15.2 to −3.1%. (The dramatic improvement becomes even more evident in the comparison of photographs of the two cross section in of Part II.) The springback was mildly increased by the pressure from 3.9 to 4.9% as evidenced in Because of this detrimental effect of tension on cross-sectional distortion, until recently, stretch forming was not favored for cold forming tubular members. At the same time, the method is attractive because of the high speed with which parts can be manufactured. It was found that modest values of internal pressure (usually compressed air at less than 3 bar) can help reduce distortion of the cross section In the course of this study, it was observed that internal pressure is not suitable for forming tubes of all cross sections. In fact, when forming tubes with tall sections, internal pressure can aggravate the distortion. In such cases, vacuum was found to be beneficial. This is illustrated in for tubes bent with the tall sides vertical at a tension of T1/T0=0.62. The first one was bent at atmospheric pressure while the second had a vacuum of 0.97 bar applied to it. The effect of the vacuum on the final distortion variable Γ is quite dramatic as it is reduced from 11.4 to 1.05% (; see also pictures of cross sections in of Part II). In addition, vacuum decreased the springback from 6.4 to 5.7% (). Additional results for loading histories which include vacuum will be discussed in Part II.In this Part I of this two-part series of papers, a model bend-stretch-pressure testing facility has been described. The facility is operated by one pneumatic and two servohydraulic systems. It differs from comparable commercial machines in that each of the three modes of loading is operated under feedback control. This allows computer control of the process, and affords flexibility in the loading history through which the part is taken. An additional feature of the testing facility is a biaxial transducer which monitors cross-sectional changes of the part during forming. The planar forming process was modeled by approximating the tube as a nonlinear elastic–plastic beam which can undergo large rotations. The model was shown capable of reproducing accurately the loading history experienced by different sections along the length of the part during the forming process.Representative results from forming experiments involving rectangular aluminum tubes representative of those used in automotive applications were presented. The results were used to discuss the effect of friction, tension and pressure on the cross-sectional distortion, springback and net elongation of the part. The repeatability of the experimental results was in general found to be good. The major conclusions from the experiments are listed in the introduction of Part II of this study that follows. There, a relatively simple, two-dimensional model is presented which is capable of predicting the cross-sectional distortion resulting from bend-stretch-pressure forming. Results from the model, supported with additional experimental results, will be used to illustrate strengths and weaknesses of customary and alternate forming histories. The conclusions drawn from the study as a whole appear at the end of Part II.Numerical simulations of the stretch forming process as performed in our model facility were conducted through a nonlinear beam analysis (see also Ref. in such a way that its line of action always passes through point representing the pivot of the horizontal actuator. During the forming process, a portion of the tube, section , comes into smooth contact with the die. The remaining suspended section is analyzed through the following small strain, large deflection beam equations Here (x,y) are the coordinates of points on the mid-surface of the beam, θ is the angle between the mid-surface of the beam and the x-axis, κ is the curvature and e the strain of the mid-surface while S and s are the undeformed and deformed lengths of the mid-surface. H and V are forces defined in and M is the moment acting on the beam cross section. The axial force T and shear force Q can be related to H and V as follows:The material stress–strain relationship is based on a multilinear representation of the response measured in uniaxial tensile tests on strips cut from the tubes tested. T and M are given byThe following incremental relationships relate The following boundary conditions complete the formulation:The problem is solved incrementally in three phases. Initially, the tension is incremented until the desired value is reached. Subsequently, the tension is held constant. In the second phase, the curvature at the mid-span is incremented until it reaches the curvature of the die. In the third phase, the length of the tube in contact with the die (sA) is prescribed incrementally. In each phase, the resultant two-point boundary value problem is discretized through a finite difference scheme and solved through Newton's method using the IMSL package BVPFD routine (see Refs. The performance of the model will be illustrated through an example of a tube formed with the long sides horizontal at a tension of T/T0=1.08 and an internal pressure of 2.28 bar (the beam model does not account for the effect of pressure). In this case, an additional feature was added to the experimental setup which enabled us to monitor the length of tube in contact with the die. A special resistance paper, which has regularly spaced, continuous, conductive lines of silver deposited on one side, is bonded to the die. The tube is mounted to the forming machine in the usual manner, but the resistance paper and the tube are connected forming two constant current circuits as shown in . The circuits are calibrated prior to the experiment. As the tube deforms and comes into contact with the die, the contact lengths sA1 and sA2 are determined from the changes in voltage in each of the two circuits (technique first used in Ref. The contact lengths recorded are plotted as a function of the die displacement in . Contact length is seen to be symmetric about the mid-span, indicating that the beam does not slide once in place. The contact lengths exhibit an initial nonlinearity during which the neighborhood of the mid-span conforms to the circular shape of the die. Subsequently, contact length increases nearly linearly with the die displacement. are predictions of the contact length(s) from the nonlinear beam model. The calculated contact length remains zero until the mid-span conforms to the curvature of the die at a displacement of 2.75 in (70 mm). Subsequently, the contact length grows essentially in the same manner as recorded in the experiment. The stiffer initial response exhibited by the model is partly due to the mathematically sharp point contact and point reaction inherent to beam theory, but also due to the fact that the presence of the internal pressure was neglected in the beam model. (The long walls exhibited an initial mild bulging due the pressure which made the initial contact problem three dimensional.)Another view of the evolution of contact between the tube and the die is shown in a set of 10 calculated beam/die deformed configurations in . The main features of the configurations are similar to those of the experiment discussed in shows a comparison between the measured and calculated die force-displacement response. The predictions are in very good agreement with the measured results. A variable thought to play a role in cross-sectional distortion is the local shear force in the neighborhood of the liftoff point of the beam from the die. In the beam model, this is idealized as a point reaction force (QA in shows how its value evolves with the die displacement. Initially, as the mid-span gradually bends and comes into contact with the die, QA grows linearly. At a die displacement of 2.75 in (70 mm), the mid-span has conformed to the die curvature and the point of liftoff starts to move outwards. QA is seen to remain essentially unchanged until full contact is achieved. (Note that for lower values of tension, QA grows to some degree with contact length.)A byproduct of this forming process is a net elongation of the part. This is illustrated in which shows a plot of the evolution of axial strain (e≡ strain of mid-surface of beam) with the die displacement at a location of S/L=0.38. The beam develops an initial strain of 0.9% due to the pretension. The point monitored remains in the suspended part of the beam until d=12.2 in (310 mm). The suspended section bends gradually inducing the slow growth in the axial strain seen in the figure. As it comes closer to be in contact with the die, the local curvature grows faster and the axial strain experiences a sudden surge. Once contact is achieved, the elongation ceases to grow. Again, because of the assumptions inherent to beam theory, contact is a mathematically sharp event resulting in the sharp discontinuity in slope in the value of the strain seen in the figure.An alternative view of this coupling between the axial strain and the bending deformation is shown in where the strain variable e at the mid-span (S/L=0) and at S/L=0.38 is plotted against the curvature κ at the corresponding locations. Interestingly, the relationship between e and κ at the two points is essentially identical. The axial strain grows essentially linearly with κ and stops growing once bending deformation ceases.Controlling the microstructure and properties of wire arc additive manufactured Ti–6Al–4V with trace boron additionsThis study demonstrates that trace boron addition to Ti–6Al–4V coupons produced by additive layer manufacturing is an effective way to eliminate the deleterious anisotropic microstructures often encountered with this manufacturing technique. Trace boron additions (up to 0.13 wt.%) to this alloy eliminate grain boundary-α and colony-α, and instead produce a homogeneous α-microstructure consisting of fine equiaxed α-grains in both as-deposited and heat treated coupons. Prior-β grains remain columnar with boron addition but become narrower due to the wider solidification range and growth restricting effect of the boron solute. Compared to unmodified Ti–6Al–4V alloy, Ti–6Al–4V modified with trace boron additions showed up to 40% improvement in plasticity with no loss in strength under uniaxial compression at room temperature. Boron additions were found to inhibit twinning transmission that causes sudden large load drops during deformation of the unmodified Ti–6Al–4V alloy in the heat treated condition.It is estimated that machining operations account for up to 50% of the total product cost of titanium components used in aerospace applications A consequence of repeated melting and deposition is the epitaxial growth of Ti–6Al–4V to form coarse primary columnar β-Ti grains. These are frequently reported in titanium structures fabricated using both powder and wire feedstock with heat sources including laser The steep thermal gradients and high cooling rates experienced during the solid-state β → α transformation (occurring at approximately 1000 °C for Ti–6Al–4V) promote the formation of very fine widmanstätten or acicular-α and/or α′-martensite. Al-Bermani et al. A beneficial consequence of the fine acicular-α in Ti–6Al–4V ALM parts is high strength. The tensile strength is directly related to the width of the α-lamella as well as the α-colony size because these features determine the effective slip length Given the prevalence for large prior-β grains and the limited prospects for preventing grain boundary-α during cooling of Ti–6Al–4V ALM structures, a clear need exists to explore other methods of controlling the microstructure. Despite the volume of recent work in the area, a literature search reveals few attempts to control the evolution of ALM microstructures, particularly grain boundary-α, through trace chemical additions. In a compositionally graded Ti–Mo alloy produced by SLM, Collins and co-workers The wire arc ALM technique which utilises Gas Tungsten Arc Welding (GTAW) technology was investigated here as it is a promising technology for high volume deposition (up to several kg/h) and parts produced by this method are reported to be an order of magnitude less expensive than parts produced by laser-based powder ALM processes Samples were prepared by wire arc ALM under conditions given in . The deposition machine consists of a 350 A GTAW power source (EWM Tetrix350), a water cooled TIG welding torch mounted on a 3-axis CNC stage, a high purity argon trailing shield and automatic wire feed nozzle to ensure consistent deposition. A wrought Ti–6Al–4V base plate was used as a substrate for the deposits. Each deposit was created by moving the torch in a linear direction and feeding wire into the molten pool, which solidified to make a layer. A subsequent layer was then deposited over the first by increasing the height of the torch. Each sample is comprised of 10 deposited layers. Photographs of the deposition machine and the samples being produced are shown in Three compositions were prepared to investigate the effects of trace boron on the microstructure and properties, including pure Ti–6Al–4V (no boron added), Ti–6Al–4V with a low trace level of boron (target ≈ 0.05 wt.% B) and Ti–6Al–4V with a high trace level of boron (target ≈ 0.15 wt.% B). Boron was alloyed in situ with the Ti–6Al–4V wire by coating the build surface with specially prepared boron paint prior to each deposition. The paint was prepared by mixing amorphous boron particles (Strem Chemicals, average size 1 μm, 92–95 wt.%) with a polymer lacquer. The overall composition of the mixture was determined by Spectro ICP-AES and was found to contain approximately 15 wt.% Boron. During deposition the boron paint dissolved and alloyed with the liquid titanium. The actual chemical composition determined after deposition is given in (composition determined using Leco TC-436 for O & N; Leco CS-444 for C; Spectro ICP-OES for Al & V; and DCP6 for B).Following deposition the samples were stress relieved at 480 °C for 2 h. This is a common stress relief for titanium alloys and no microstructural changes in Ti–6Al–4V occur during this process. After stress relief, each sample was sectioned and one half of the sample was heat treated at 1050 °C for 30 min followed by a rapid furnace cool (furnace door was opened and cool air circulated throughout). The additional heat treatment was performed to remove the α′ martensite or very fine acicular-α which is expected to be present in the as-deposited state. Samples were then machined from the as-deposited and heat treated coupons for compression testing and microstructural evaluation. At least three compression samples were prepared from each group (cylinders ø |
= 4 mm, 8 mm long) and these were compression tested with a crosshead speed of 1 mm/min using an Instron testing machine equipped with a video extensometer. Samples were compression tested in the vertical build direction.The microstructure of the samples was examined using optical and electron microscopy techniques. The electron backscatter diffraction (EBSD) analyses were acquired using a Jeol 6610 SEM equipped with an HKL Channel 5 system. The EBSD data were processed by HKL Channel 5 Tango and Mambo. In boron containing samples, the nanohardness of the boride phase was analysed using a Hysitron Triboindenter equipped with Berkovich tip (radius 100 nm).The macro and microstructure of the samples is shown in over several length scales. It is observed that the addition of boron to the Ti–6Al–4V alloy has a dramatic influence on the microstructure. The external photographs show that the trace boron refines the width of the grains and produces a more elongated structure than boron free Ti–6Al–4V. The polished cross sections show that the deposited structures are comprised of large prior β-grains containing a network of fine α-laths and inter-granular β. Boron containing samples also feature fine boride particles in the microstructures.Synchrotron XRD spectra for the as-deposited Ti–6Al–4V and Ti–6Al–4V–0.13B alloys are shown in . For the Ti–6Al–4V alloy, characteristic peaks associated with the predominant hexagonal α-Ti phase were detected as well as a minor β (2 0 0) peak from the relatively small amount of retained β present in the microstructure. Other β phase peaks, (1 1 0) and (2 1 1), are not easily distinguishable as they overlap with the α (1 0 1) and (1 0 3) peaks, respectively. In the case of the Ti–6Al–4V–0.13B alloy, the α-Ti peaks are detected in conjunction with a number of minor peaks which can be indexed to the orthorhombic TiB phase. The α phase peaks for the Ti–6Al–4V–0.13B alloy also exhibit broadening in comparison to the boron-free Ti–6Al–4V alloy which is indicative of a reduction in the average α crystallite size due to the refining effects of the borides (discussed further in Section The effect of boron on individual microstructural features and mechanical properties is discussed below.Boron dramatically influences the morphology and size of the prior-β grains in samples produced by the wire arc ALM process. In unmodified Ti–6Al–4V, the grain structure is coarse and contains prior-β grains several millimetres in width and height. These are easily visible at all length scales due to the different morphologies of the α-phase within the individual prior-β grains. In some instances the prior-β grains appear equiaxed in the cross sections. However, closer examination revealed that the cross sections contain only a few primary grains that grow around one another in three dimensions, which when sectioned in 2D, give the appearance of individual equiaxed grains. These large grains epitaxially nucleate and grow through several deposited layers in near vertical directions as evident in Boron additions significantly refine the width of surface columnar grains (). Using the linear intercept method for estimating grain size, the average columnar width reduced from 1823 μm in the boron free Ti–6Al–4V to 339 μm in Ti–6Al–4V–0.13B, while 0.05 wt.% of boron addition reduced the columnar width by more than half to 819 μm. At the microstructural level it is almost impossible to distinguish prior-β grain boundaries because no clear tell-tale ‘signs’ such as grain boundary-α or other α-morphologies that ordinarily reveal these boundaries in Ti–6Al–4V are visible. Only some of the high angle grain boundaries are visible under low magnification and ideal lighting conditions, nevertheless, many of the low-angle fine columnar grains that are visible in remain difficult to observe on the 2D polished sections. shows optical and SEM images of the highly dendritic microstructure found in boron modified alloys which are visible due to the presence of fine TiB needles in the interdendritic regions. This figure also demonstrates the difficulty in resolving prior-β grain boundaries (a prior-β grain boundary is indicated by the arrows in A). The side view of the boron modified Ti–6Al–4V deposits in reveals that in the final layer the columnar grains also adopt some directionality in their growth, which initially grow vertically but then tend to grow horizontally in the direction of torch travel.The observation that boron refines the width of the prior-β columnar grain size and does not promote equiaxed grain formation is expected when considering how boron refines the β-grain size in titanium castings. Over recent decades it has been comprehensively established that effective grain refinement of metallic castings during solidification requires a population of potent heterogeneous nucleant particles in conjunction with powerful segregating solute to provide constitutional supercooling Although boron is an extremely powerful grain refining solute in the titanium system, it is only effective if a population of potent nuclei exist in the melt. In the samples produced by ALM, no heterogeneous nucleants are introduced and consequently, boron is unable to refine the prior-β grain size in the absence of these potent nucleants. In titanium castings, under some thermal conditions it has been suggested that chilled titanium crystals can separate from the mould wall during turbulent casting processes and survive in the cooling liquid to act as native nucleant particles ), these form during a eutectic reaction after primary solidification and do not provide any direct heterogeneous nucleation substrates for β-Ti.The presence of boron solute during solidification increases the freezing range of the alloy and can account for some of the directionality of the surface columnar grains shown in . According to the binary Ti-B phase diagram . In both Ti–6Al–4V with and without boron, ALM inherently produces high cooling rates during solidification. Given that the cooling rate is related to the secondary dendrite arm spacing (SDAS), the cooling rate experienced here during solidification can be estimated to be at least 150 Ks−1 based on comparison with cast titanium-boron alloys that produced a SDAS of 25–30 μm at this cooling rate A illustrates that this high cooling rate rapidly promotes growth and solidification of favourably oriented grains in a very narrow freezing window for Ti–6Al–4V. Furthermore, the steep thermal gradient and lack of potent nucleant particles prevent nucleation ahead of the growing grains. However, when boron solute is present (B) it extends the solidification period due to the larger freezing range, and the lateral rejection of boron-rich solute from the base and sides of columnar dendrites restricts lateral growth and provides an opportunity for neighbouring dendrites to grow. A very similar effect was observed when increasing the concentration of boron-rich solute in cast cobalt alloys in the absence of potent nucleants where boron also has a high Q value in the cobalt system As commonly reported in as-deposited ALM Ti–6Al–4V components, the microstructure produced here using the wire arc ALM process consists of very fine acicular-α where individual laths are of the order of about a micrometre thick and more than a hundred micrometres long (A). This structure forms due to the rapid cooling experienced during the solid state β → α transformation, which occurs at around 1000 °C for this alloy. During the additional heat treatment the cooling rate during the β → α transformation was slower, and as a result, the acicular-α coarsened to about 4 μm thick and doubled in length (B). The presence of colony-α (packets of similarly orientated α-laths) can also be found after heat treatment and these colonies are several millimetres in size. Grain boundary-α also delineates the prior-β grains in both the as-deposited and heat treated conditions. This continuous layer of α-phase is only of the order of a micrometre thick in the as-deposited state but increases in thickness tenfold after heat treatment. In both cases, the network of grain boundary-α extends throughout the macrostructure and it is not uncommon to observe single continuous segments several millimetres in length (e.g. Trace boron additions have a remarkable influence on this microstructure. TiB needles were found throughout the microstructures of the boron containing alloys and increased in volume fraction as the boron concentration increased. TiB needles were typically clustered in the interdendritic regions (refer to ), which is to be expected given that the interdendritic regions are the last to solidify and contain the highest boron solute concentration, from which the TiB forms via the eutectic reaction. Unlike castings where TiB has been reported predominately at prior-β grain boundaries The presence of TiB significantly altered all variants of the α-phase, including grain boundary-α. The average size of the α-lath length and width is given in (measured using the linear intercept technique). In the as-deposited state, although the width of the α-laths did not significantly change with boron addition, the lath lengths were dramatically reduced by more than fourfold when boron was added. The same trend also occurred after heat treatment and individual α-laths become more equiaxed in appearance (i.e. the aspect ratio reduces). In cast Ti–6Al–4V–TiB composites (14 vol.% B), Hill and co-workers Trace boron addition was also extremely effective in eliminating colony-α and grain boundary-α structures as only a handful of small, isolated colonies and segments remain after boron modification. The lack of grain boundary-α was noted in Section as its absence makes identifying prior-β grain boundaries difficult. At the high trace level boron concentration (0.13 wt.%), no grain boundary-α is discernible in the microstructure but at the lower concentration (0.05 wt.%) a few small segments in certain regions can occasionally be resolved (e.g. A). SEM examination of these segments revealed that they are semi-continuous, consisting of multiple individual α-grains rather than a single continuous α-segment (refer to Over recent years a number of researchers have investigated the mechanism by which boron refines the α-grain structure in titanium alloys. It is now well accepted that TiB is a potent nucleant for heterogeneous nucleation of the α-Ti phase. Genç, Hill, Banerjee and co-workers In summary, the potency of TiB as a heterogenous nucleant for α-titanium combined with the facts that boron has very little solid solubility in titanium (and tends to form TiB) and the fast cooling rates create a fine dendritic structure dispersed with TiB, means that trace additions are very effective in eliminating the coarse α-textures that otherwise form in Ti–6Al–4V during ALM.The compressive mechanical properties of the ALM components are presented in . It is clear that both trace boron additions and heat treatment influence the mechanical properties. It is often reported that as-deposited Ti–6Al–4V has very high strength on account of the fine acicular-α and martensite present in the microstructure, and that heat treatment can reduce this strength by coarsening these features Trace boron additions to Ti–6Al–4V did not significantly influence the compressive strength in the as-deposited state, although the boron additions did slightly improve plasticity. After heat treatment, the compressive ductility was significantly higher for the boron modified alloys with the Ti–6Al–4V–0.13B alloy exhibiting the highest strain at failure of any condition, 40% higher than the unmodified heat treated Ti–6Al–4V. It is also noteworthy that this dramatic improvement did not come at the expense of strength, as the maximum compressive strength is similar for the heat treated and as-deposited boron modified alloys. Although a small reduction in proof strength occurred after heat treating the boron modified alloys (approximately 10%), the reduction was less severe than for the unmodified Ti–6Al–4V with the average proof stress remaining at around 850 MPa in both boron modified heat treated alloys (compared to a 15% reduction in the boron-free Ti–6Al–4V). Furthermore, the boron containing alloys exhibit a high rate of strain hardening and reach maximum strengths comparable to those achieved in the as-deposited boron-free Ti–6Al–4V.An enlarged portion of the plastic flow regions of the compressive stress–strain plots is presented in . Most strikingly, after the heat treatment the boron free Ti–6Al–4V material displays extensive irregular serrated flow up to approximately 25% strain with sudden large drops in the load of up to 40 MPa. Audible ‘clicks’ were also heard emanating from the test coupons during testing. The serrations and audible phenomena are typically features associated with twinning in systems where the stress required to grow twins is much smaller than for their nucleation Longitudinal sections of the compression samples were observed for evidence of twinning using SEM and EBSD which are presented in , respectively. Many large twins are present in the boron free Ti–6Al–4V sample (A–D) reveals are {10–12} 〈10-1-1〉 type tensile twins with characteristic 85° misorientation angles. This type of twinning was also observed during compression of heavily rolled Ti–6Al–4V with strong transverse basal texture and is common in metals with a hexagonal crystal structure A and B) also reveal the strong (0 0 0 1) basal texture of the α-phase which facilitates large-scale twinning transmission across multiple α-phase laths. The boron containing Ti–6Al–4V alloys also exhibit features consistent with twins (B and C), although they are less prolific and of a much finer scale than those observed for the boron free alloy. The EBSD analysis shows that these much smaller twins are also {10–12} 〈10-1-1〉 type tensile twins with a misorientation angle of 85°. The size of the twins and their occurrence decrease with increasing levels of boron. The inverse pole figure maps (E and F) also illustrate the effectiveness of the boron additions in restricting the formation of the large-scale α-phase colonies, with each α-lath exhibiting a distinct orientation and no obvious texture apparent.The microstructural investigations have shown that trace boron additions to the ALM Ti–6Al–4V do not prevent twinning. However, they do dramatically inhibit the growth of the twins and consequently prevent large sudden load drops during deformation. The large colony-α observed in the heat treated boron free alloy facilitates large-scale twinning transmission across regions of similarly oriented α-laths which manifest as significant load drops in the stress–strain plot. In one of the few other reported observations of this behaviour in the Ti–6Al–4V alloy, the individual load reductions occurring in conjunction with twinning in highly textured rolled samples were in the order of 3–5 MPa A number of researchers have investigated the properties of titanium composites reinforced with TiB and generally it is reported that the secondary TiB phase strengthens the composite at the expense of some ductility On this basis, the estimated volume fraction of the TiB phase which forms is around 0.25% for the Ti–6Al–4V–0.05B alloy and 0.90% for the Ti–6Al–4V–0.13B alloy. shows the measured nanohardness of the TiB and the Ti–6Al–4V matrix. Although the TiB is very hard compared to the titanium matrix, the volume fraction of TiB reinforcement is not large enough to significantly impact on the bulk material properties such as strength on the basis of their direct contribution according to the rule of mixtures. In the case of titanium-TiB composites, refinement of the α-phase microstructure is also believed to contribute to strengthening While many researchers have reported increases in strength and decreases in ductility with boron additions, in the present work it is found that trace additions produced a slight decrease in strength in the as-deposited state. It is likely that the very fine acicular-α and/or α′ martensite present in as-deposited Ti–6Al–4V is already close to the optimal size to impart maximum strengthening to the alloy and that the addition of TiB at these low levels contributes little to further strengthening. However, heat treating the Ti–6Al–4V caused the α-lath width to coarsen more than fourfold (). This increases the available slip length leading to enhanced plasticity and decreased strength. In the heat treated condition, trace boron additions to Ti–6Al–4V were found to improve the strength by approximately 10%, despite the α-lath width in the boron containing alloys being larger. The mechanism for this is not yet clear and requires further investigation but it is likely that the refined equiaxed α-structure in the heat treated boron modified alloys is at least partly responsible in conjunction with possible solid solution strengthening on account of the limited solubility of boron in the hcp lattice. The addition of boron also substantially improved the plasticity of these alloys by up to 40%. It is noteworthy that the highest strain achieved in any condition corresponds to the most equiaxed α-grain microstructure (refer to F3). Fine equiaxed grain structures like this are synonymous with high levels of plasticity.Although it has been demonstrated in the literature that high levels of boron can reduce the ductility of Ti based alloys, some researchers have shown that lower levels of boron can actually improve ductility. In cast alloys modified with boron, Bilous et al. In summary, trace boron additions to Ti–6Al–4V during ALM are found to have a significant influence on the microstructural evolution during the liquid-to-β transformation as well as the solid β-to-α transformation. schematically summarises these transformations and enables comparison between the microstructural development during ALM of Ti–6Al–4V and Ti–6Al–4V modified with trace boron addition. The typically fast cooling rate encountered during ALM produces a small dendrite arm spacing in both alloys. In the boron modified alloy, a finely spaced network of TiB particles forms between the dendrite arms via the eutectic reaction. After solidification, as the temperature continues to decrease below the β-transus temperature, the α-phase begins to nucleate. In Ti–6Al–4V the α-phase nucleates at β-grain boundaries and large textured-α structures develop (colony-α). However, in the boron modified alloys the α-phase preferentially nucleates on the many finely dispersed TiB particles, and as a result, many fine equiaxed-α grains are produced. The consequential refinement of the α-phase with boron addition improves the compressive ductility and strength while also suppressing large scale twinning deformation in textured α-colonies.Trace boron additions up to 0.13 wt.% were made to the Ti–6Al–4V alloy in an attempt to control the microstructure and properties of the components produced. The boron additions were incorporated into arc-wire ALM process by applying boron paint prior to depositing the Ti–6Al–4V alloy so that the boron dissolved into the molten weld pool. A summary of the key findings are:Trace boron additions have a significant effect on the columnar-β grain morphology. In the absence of potent heterogeneous nucleants, boron additions produce thin elongated columnar structures due to lateral solute rejection during epitaxial nucleation and growth from prior substrate layers. Unlike unmodified Ti–6Al–4V which has a narrow freezing range and practically no growth restriction from Al and V solutes, the presence of trace levels of boron solute produces sufficient constitutional supercooling to restrict lateral columnar growth and allow neighbouring columnar grains the opportunity to nucleate and grow.Fine TiB needles are dispersed throughout the microstructures in trace boron modified alloys. The fast cooling rates encountered during ALM produce a small secondary dendrite arm spacing which determines the spacing of the TiB particles.Trace boron additions are very effective in eliminating grain boundary-α and colony-α, in addition to refining the α-lath width and length. The α-grains become more equiaxed with boron addition, especially after heat treatment. The TiB needles are effective in nucleating the α-phase and produce more isotropic α-microstructures.In the as-deposited state (no heat treatment), the addition of trace boron does not significantly influence strength but improves ductility. After heat treatment, boron additions offer greater strength (10%) and improved ductility (up to 40%) compared to the unmodified Ti–6Al–4V alloy.Ti–6Al–4V produced by ALM is susceptible to twinning deformation. {10–12} 〈10-1-1〉 type tensile twins were found in both boron modified and non-modified Ti–6Al–4V. Twinning transmission across textured colony-α structures in Ti–6Al–4V results in significant stress drops of up to 40 MPa during plastic deformation. The refinement of the α-phase and elimination of textured α-colonies with trace boron prevents twinning transmission and significantly decreased the scale of {10–12} 〈10-1-1〉 twinning.Trace boron additions can limit and/or inhibit the effects of twinning for strongly transverse basal textured Ti–6Al–4V which can otherwise be susceptible to sudden large load drops during plastic deformation.Geometrical and material optimisation of deformed steel fibres: Spirally deformed fibresThis paper presents an optimisation procedure of geometrical and material properties of steel fibres using the validated three-dimensional nonlinear Finite Element (FE) pullout model previously proposed by the authors. The FE model is employed as a virtual laboratory unit to investigate the pullout performance of steel fibres with various shapes such as hooked-end and spirally deformed steel fibres. The preliminary FE pullout analyses imply that the optimisation of the fibre with spiral configuration (adjustment of geometrical and material properties of the fibre) could result in an efficient high-performing steel fibre; hence such a fibre is selected for the optimisation procedure. In order to re-validate the numerical pullout model and further adjust its nonlinear parameters for enhancing the simulations prediction, a set of experimental pullout tests are conducted on four spirally deformed steel fibres with different geometrical properties. Extensive parametric studies are performed on the material and geometrical properties of the fibre to examine their effects on the fibre pullout performance. Moreover, an empirical equation is proposed for the estimation of fibre efficiency and optimum design of spirally deformed steel fibres in which geometrical and material properties are taken into account.fibre efficiency obtained from finite element simulationfibre efficiency calculated from the proposed empirical equationcharacteristic cylinder strength of concrete, MPatarget mean cylinder strength of concrete for concrete mix design, MPaouter diameter of spirally deformed fibre, mmThe idea of using discrete fibres to reinforce brittle construction materials has been in practice for many centuries. For instance, ancient Egyptians used straw to improve the post-cracking behaviour of sun-dried mud bricks for huts Findings of the investigations indicate the significant effect of the fibres on the brittle response of concrete material. After micro-cracks initiate, the contribution of the fibres starts as crack growth arrestors, which leads to an increase in the residual strength of concrete Fibre Reinforced Concrete (FRC) provides advantages over the use of conventional reinforced concrete for non-structural applications such as cost saving, distributing localised stresses, and reducing crack widths and surface permeability. In other words, fibres are a viable alternative to secondary reinforcement such as shrinkage and temperature reinforcement. FRC can be used in highway/airport runway overlays and industrial floors, thin-sheet materials, nuclear reactor shielding, pile caps, and tunnel wall linings Most types of steel fibres on the market at the volume contents generally used in practice do not have favourable structural performance and those which relatively have are not efficient and cost-effective and might demand restrictive requirements on the concrete matrix properties In this paper, a high-performing deformed steel fibre is engineered and optimised specifically for normal strength concrete to advance the adoption of such discontinuous fibres in the structural design of Reinforced Concrete (RC) members. To this end, the Finite Element (FE) pullout model previously proposed by Hajsadeghi et al. To further adjust the non-linear parameters of the FE model (re-validation), a set of experimental pullout tests are conducted on four different spirally deformed steel fibres. Using the numerical model, extensive parametric studies are conducted on the geometrical and material properties of the fibre to examine their effects on the pullout performance. Over the numerical pullout load database of various spirally deformed fibres, an empirical equation is proposed to estimate the fibre efficiency, i.e. ratio of peak pullout load to wire’s load carrying capacity, in which fibre geometrical and material properties are the input parameters. According to the required peak pullout load, the configuration and material strength of the fibre can be adjusted using the proposed equation, so that the fibre efficiency is optimum, i.e. fibre design.The Finite Element (FE) fibre pullout model proposed by Hajsadeghi et al. In the FE model, the 8-noded solid element (SOLID 185) with three degrees of freedom at each node (translations in the nodal x, y, and z directions) is used to simulate the steel fibres material. The element has plasticity, stress stiffening, large deflection, and large strain capabilities. Cementitious matrices are modelled using the 8-noded solid brick element (SOLID65) with cracking, crushing and plastic deformation capabilities The surface-to-surface contact is used to simulate the contacting fibre and matrix surfaces (the fibre-matrix interface) which are respectively specified as contact and target surfaces. The contact and target elements have the same geometric characteristics overlaying the contacting surfaces which are capable of simulating the deformable contact interface. The physicochemical bond of the fibre-matrix interface is defined by the Coulomb friction model where two contacting surfaces carry shear stresses up to a certain magnitude across their interface before they start sliding relative to each other As the first phase for developing an optimum and versatile structural steel fibre, various new fibres with different hooks arrangements as mechanical anchorages are generated (around 20 fibres) such as NF-1 and NF-2 (see ). The hooked-end fibre FE model validated by Hajsadeghi et al. The new fibres, e.g. NF-1 and NF-2, have different configurations (mechanical anchorages with different levels of deformity and effectiveness) compared with the original hooked-end fibre, therefore they all possess different peak pullout loads (all the other properties and parameters including concrete strength, fibre diameter and length, and interface characteristics are identical in all the models).To prevent fibres fracture as well as fully utilise the steel material, the optimisation of the fibres material is performed. In other words, the optimum (minimum) material strength of each fibre is separately determined to preclude the fibre fracture during the pullout process up to the complete fibre withdrawal. Depending on the mechanical anchorage of each fibre, a few consecutive simulations are performed to optimise the fibre’s material strength. For instance, the optimised yield and ultimate stresses of the steel material (fy and fu) for NF-1 and NF-2 are found to be 1600 MPa and 2000 MPa, and 1000 MPa and 1250 MPa, respectively.In order to compare the effectiveness of the fibres’ shape irrespective of the deformity level and material strength, the normalised pullout load (with respect to the steel wire’s load carrying capacity) versus slip curves of the new fibres are compared. The considered criteria for the comparison are the normalised peak pullout load, i.e. fibre efficiency (Ef = Pmax/(Af⋅fu)) and the pullout load decay (Ld = (Pmax – P25)/Pmax), i.e. the difference of the maximum load and the load at slippage of 25 mm over the maximum load.Since the main objective of this research paper is to design a versatile structural steel fibre suitable for different loading conditions including serviceability, ultimate and extreme loadings, the parameter Ld (reflecting the residual pullout strength up to 25 mm slippage) plays a major role in the fibres’ comparison. The consideration of such a large slippage (25 mm) would ensure the effective contribution of the fibre to the capacity of SFRC members even in extreme loading conditions where excessive deformations and wide cracks are expected.As examples, the normalised pullout load versus slip of two fibres, NF-1 and NF-2, outperforming the other new fibres with different hooked geometries are shown in . The normalised peak pullout load (Ef) and the pullout load decay (Ld), as well as their differences relative to those of the hooked-end fibre, are provided in , NF-1 and NF-2 possess desirable fibre efficiencies as well as slip-hardening response, whose peak pullout loads respectively sustain up to the slippages around 3.4 mm and 7 mm, contrary to the hooked-end fibre experiencing pullout load decay after the slippage of 1.8 mm. The pullout results of the investigated fibres with various hooked geometries show that the post-peak pullout response could be drastically improved compared with conventional hooked-end fibres (e.g. NF-1). However, such fibres suffer from significant pullout load decay up to the slippage of 25 mm as the conventional hooked-end fibres do.On the other hand, findings of previous research revealed that steel fibres with spiral configuration could show favourable post-peak pullout responses up to the complete fibre withdrawal; however, they cannot effectively contribute to the load-bearing capacity of structural FRC members mainly attributed to the fibres’ poor efficiency (the diameter and length of the fibre, materials properties, and contact parameters are the same as those of the hooked-end fibre model), is examined via FE simulation. The optimised yield and ultimate stresses of the fibre material are 600 MPa and 750 MPa respectively.) indicate that the fibre (NF-3) has a relatively low efficiency; however, it possesses superior post-peak response compared with its counterparts.The von Mises stress distribution of the new fibres experiencing pullout process at various slippages is shown in It should be noted that in real scenarios, the cracks might initiate at any point along the fibre and the fibre generally slips on the shorter embedded side. This could highly affect the pullout performance of fully deformed fibres such as crimped and twisted. The study of the stress distributions along the spirally deformed fibre reveals that during the pullout process the regions nearby the exit point experience considerably higher levels of stress compared with other regions, as seen in . It implies that the pullout response of such fibres is less influenced by the variation of embedment length (lem) compared with the other fully deformed steel fibres and therefore as the fibre is pulled out from the matrix, the fibre resistance substantially sustains.Besides, the FE results reveal that the spirally deformed fibres experience a complex stress state during the pullout process where the fibres undergo twist at the fibres’ exit point from the matrix as well as bending (due to the fibres straightening) and tension.Overall, the spiral configuration is selected as a potential geometry (mechanical anchorage) for further investigations to design a versatile structural steel fibre.In order to re-validate the numerical pullout model and further adjust its non-linear parameters to enhance the simulations prediction, a set of experimental pullout tests are conducted on four different spirally deformed steel fibres. Steel wires with two different diameters (dw = 0.55 mm and 0.90 mm) are deformed to the desired shapes using basic tools including a drill, steel nails, and pliers. The ultimate strength (fu) of the wires with diameters of 0.55 mm and 0.90 mm is 1000 MPa and 1600 MPa, respectively. The specifications of the considered fibres are summarised in in which N = number of turns, OD = outer diameter of the fibre, df = wire diameter (fibre diameter), and lf = length of the fibre (see As seen in the table, the fibres are labelled such that their geometrical properties can be identified by the label. For instance, the label “SD-4-2.0-0.90-50” indicates that the fibre is spirally deformed, with 4 turns, outer diameter of 2.0 mm, wire diameter of 0.90 mm, and it is 50 mm in length.To cast the specimens for the fibre pullout test in a double-sided manner, a wooden mould is fabricated (as seen in ) allowing eight identical specimens to be cast at the same time. The double-sided fibre pullout specimen consists of two separate concrete blocks each with dimensions of 65 mm × 65 mm × 100 mm with a fibre embedded. The fibre is centrally and perpendicularly aligned to the concrete blocks’ interfaces, i.e. the crack surface, and the embedment lengths of the fibre in both concrete blocks are equal It should be mentioned that one side of the specimen is cast and once the block is hardened, a thin plastic separator is placed on the top of the block and then the other part is cast. During the concrete casting, steel tubes are placed into the concrete blocks to provide a proper arrangement for the test specimen to be gripped to the pullout testing clamps (see ). Considering the specimens’ dimensions, a clamp which can be mounted to the electro-hydraulic servo Universal Testing Machines was prepared to grip both sides of the specimens (The concrete strength is fck = 35.4 ± 2.9 MPa (fm = 35 MPa) where 35.4 is the average strength and 2.9 is two times the standard deviation of strengths of eight standard cylinder concrete specimens. The concrete mix proportions are provided in The mould and the pullout testing setup are shown in . The specimens are statically loaded in a displacement-controlled manner with the loading rate of 0.6 mm/min. For each fibre, eight specimens are tested where the average result is considered as the pullout response of the fibre. The pullout response of eight “SD-5-2.0-0.90-50” fibres and the corresponding average pullout response are shown in . In the figure, mean and standard deviation values (µ and σ) of pullout loads at three different slippages (slippage corresponding to the peak pullout load, 6 mm, and 18 mm) are also included.The average pullout response of the spirally deformed steel fibres (As can be seen in the figure, spirally deformed steel fibres show slip-hardening pullout response in normal concrete which indicates their superior post-peak pullout behaviour compared with the commercially available steel fibres. Moreover, the geometrical properties of the spiral configuration (N, OD, df, and lf) greatly affect the pullout behaviour of the fibres.The experimental results revealed that the fibres are pulled out from one side. In general, the pullout resistance of the fibres (perpendicularly aligned to the crack plane) is provided by the physicochemical bond of the fibre-matrix interface and the mechanical anchorage. For a particular type of fibre and cementitious matrix, the bond and the contribution of the mechanical anchorage (for fully deformed fibres) are proportional to the embedment length. Regarding the end-deformed fibres (as long as the crack is within the two end anchorages), only the interfacial bond is proportional to the embedment length (not the contribution of the mechanical anchorage) as straight fibres. All in all, the fibres generally slip from one side (shorter embedded side or the side with less resistance due to any sort of imperfection) due to the equilibrium.The strength and composition of the matrix can affect the pullout behaviour of steel fibres A few consecutive pullout simulations are performed for all the fibres (see ) to calibrate the interface parameters through minimisation of the absolute difference between the numerical peak pullout loads as well as the loads at slippages of 6 mm and 18 mm and those loads obtained from the experiments. The absolute difference, on average for all the four fibres, is around 5% of the peak pullout load from the physical testing. The contact parameters of the FE model calibrated for the spirally deformed steel fibres (see The numerical pullout load-slip curves of the fibres, as well as their experimental counterparts, are shown in . As seen, the FE model can properly predict the pullout response of spirally deformed steel fibres; hence, it can be used as a virtual laboratory unit to optimise the material and geometrical properties of such fibres.The tensile testing on 16 steel wire specimens with diameters of 0.55 mm with fu = 1000 MPa and 0.9 mm with fu = 1600 MPa (for each diameter, three straight specimens and five spirally deformed specimens with different levels of deformity) revealed that the ultimate strength of steel decreases by 4%, on average, due to the residual stresses induced by plastic (permanent) deformations. To consider this material strength degradation in the simulations, the strength of the steel wires (fy and fu) is reduced by 4% and then assigned as the fibres material properties in the model.Furthermore, detailed studies of the stress distribution of the fibres during the pullout process reveals that spirally deformed fibres with higher diameters are prone to fracture due to the higher transverse shear stresses resulted from the twist.A parametric study is performed on the geometrical properties of the spirally deformed steel fibres, i.e. N, OD, df, and lf, to examine their effects on the pullout performance of the fibres. Furthermore, to consider the effect of fibre material properties on the pullout characteristics, two different steel materials with yield and ultimate strengths of 800 MPa and 1000 MPa, and 1200 MPa and 1500 MPa are incorporated into the study. The specifications of the FE models are summarised in . A parametric code in APDL (ANSYS Parametric Design Language) is developed to reduce the modelling time. Amongst all the possible combinations (in total 1680 fibres), 684 fibres are selected such that the effect of the geometrical parameters and material properties variation can be properly investigated.For instance, the pullout load versus slip curves of typical spirally deformed steel fibres with various numbers of turns (N = 2 to 12) are shown in where the other geometrical parameters (OD = 2.0 mm, df = 0.55 mm, and lf = 50 mm) are kept constant (fy and fu of wire material are 1200 MPa and 1500 MPa). As seen, as N increases, the peak pullout load increases; however, for N greater than 10, fibres fracture which results in a significant decrease in the dissipated pullout energy.The dissipated pullout energy (Et) up to the slippage of 25 mm for selected fibres is shown in . The fibres are chosen such that the effects of the variation of N and OD can be observed clearly. In the figure, the dotted and striped bars respectively indicate that the corresponding fibres are completely withdrawn or fractured. The effects of variations of N on the pullout energy are shown in the figure (with bars in black) in which the OD, df, and lf are kept constant. Moreover, the effects of variations of OD on the pullout energy are shown with bars in red and blue. From the figure, it is evident that with the increase in N or OD, the pullout energy increases. However, with the increase in the parameters (N and/or OD) the possibility of fibre fracture increases resulting in a drastic reduction of the pullout energy.The pullout results of all the investigated spirally deformed steel fibres including pullout load, dissipated pullout energy, and fibre efficiency are summarised in . As seen, each fibre is examined with two different steel materials. The bold values demonstrate that the corresponding fibres fracture during the pullout process.The fibre efficiency (Ef) versus geometry index (Ige) of all the investigated fibres (in where the geometry index is defined as Eq.As seen in the figure, with the increase in the geometry index, the fibre efficiency enhances, however, the probability of fibre fracture also increases. It can be roughly estimated that fibres with Ige ≥ 0.8 fracture during the pullout process.A metaheuristic inspired by the process of natural selection known as the Genetic algorithm (GA) to estimate the fibre efficiency, in which material and geometrical properties are the input parameters.Ef,pre.=0.065lf-0.08fu0.0187N0.694OD0.465df-0.617To assess the accuracy of the proposed equation, the mean absolute error (MAE), is calculated as Eq. where Ef,FE and Ef,pre. are the fibre efficiencies from the FE simulations and the empirical equation (Eq.The error (MAE) over the available database (see In this section, a spirally deformed steel fibre is designed using the proposed empirical equation (Eq.). The length of the fibre is fixed to 40 mm to provide enough embedment length and consequently ensure its effectiveness in different loading conditions (serviceability, ultimate and extreme). The design procedure of spirally deformed steel fibres is summarised as follows:Fibre geometry index (Ige), wire diameter (df), fibre length (lf), and ultimate tensile strength (fu) of steel are selected., and assuming the value of either OD or N, the other parameter is calculated.Fibre efficiency (Ef) is calculated by Eq.The initial fibre geometry index (Ige) is considered to be 0.8 (see ). Considering the intended application of the fibre and also mixing issues, the fibre diameter (df) is selected to be 0.40 mm which results in the fibre aspect ratio (lf/df) of 100. The ultimate tensile strength of the steel wire is 1500 MPa. In , the design steps of the fibre are provided.The calculated value for the outer diameter in step is rounded to 1.45 mm. Therefore the peak pullout load of the fibre, i.e. SD-9-1.45-0.40-40, is:Pmax=Ef,pre.×Affu=0.54×0.25×π×0.42×1500=102NIn order to check the predicted peak pullout load, an FE pullout simulation is performed for fibre SD-9-1.45-0.40-40 whose load-slip curve is shown in . The peak pullout load from the simulation is 108 N, which indicates the acceptable accuracy of the empirical equation (Eq.). Moreover, the outer diameter of the fibre (OD) is slightly increased (from 1.45 mm to 1.55 mm) to investigate the effect of a slight increase in the fibre geometry index (from 0.815 to 0.872) on the fibre pullout performance. As seen in , even with the small increase in OD, the fibre fractures, therefore, SD-9-1.45-0.40-40 is considered as an optimum fibre In this study, a comprehensive programme including numerical simulations and experimental tests was performed to engineer a versatile structural steel fibre.Some major findings may be summarised as follows:Contrary to the conventional fully deformed steel fibres (e.g. hooked-end), spirally deformed steel fibres possess desirable post-peak pullout response (slip-hardening). As the FE results revealed, during the pullout process, the entire embedded length of the fibres with hooked configurations contributes to the pullout resistance through repetitive bending and straightening. It implies that as such fibres are pulled out (and the embedded length decreases), their pullout resistance proportionally decreases (slip-softening). However, the pullout resistance of the spirally deformed fibres is mainly provided by a smooth and continuous fibre straightening at the fibre exit point from the matrix, causes an effective and relatively uniform resistance until the complete fibre withdrawal.The pullout performance of the spirally deformed fibres is positively correlated with the inclination angle of the fibres with respect to the loading direction. It is evident that the angle increases as the number of turns and/or fibre outer diameter increase and it decreases as the fibre length and/or diameter increase.Following steps were taken to design optimum spirally deformed steel fibres.A parametric study on the material and geometrical properties of the fibres (N, OD, df, lf, fu) to examine their effects on the pullout performance.Proposing an empirical design equation for such fibres based on the fibres’ material and geometrical properties.Finally, a spirally deformed steel fibre (SD-9–1.45–0.40–40) was designed for structural purposes.Effect of weld details on the ductility of steel column baseplate connectionsFactor to account for strain hardening (taken as 1.2)Measured strain at the initiation of neckingMaximum drift ratio sustained prior to crack initiationMaximum drift ratio sustained prior brittle failureMeasured ultimate strength of steel, MPaExpected moment calculated as Mpr=CprRyFyZx, kN mFactor to account for material overstrength, taken as 1.1 for Grade 50 (345 MPa) steelPlastic section modulus for major axis bending, mm3Column base connections are critical components in earthquake resistant structural systems, since they must maintain their ability to transfer axial forces, shear forces and moments to the foundation while sustaining large inelastic deformations. Severe damage to column base plate components has been observed in previous earthquakes, such as the 1994 Northridge Motivated by the scarcity of data comparing the relative fracture resistance of CJP and PJP weld details and the popularity of these details in high seismic regions, this paper presents experimental investigations on six two-thirds scale column base specimens tested as a part of a Network for Earthquake Engineering Simulation and Research (NEESR) project. In addition to their practical relevance to seismic design, these experiments provide data to validate a new type of micromechanics-based model, previously developed by the authors Few studies have specifically investigated failure of the welds connecting columns and base plates. However, weld fractures have been observed in tests designed to obtain general response characteristics of column base plate connections. Most previous studies emphasized the catastrophic strength loss due to connection failure, and do not always report detailed data on the precise instant and location of weld fracture initiation and subsequent growth prior to unstable propagation. and later discussion). Weld toughness was observed to have a strong influence on ductility. The specimen fabricated with the toughness-rated weld metal sustained about twice the drift before fracture (approximately 5% interstory drift ratio) as compared to the specimen with the non-toughness rated weld metal, which fractured at a drift ratio of approximately 2.5%.). All the specimens failed through weld fracture at drift levels ranging from 2% to 13%. Similar to the tests by Astaneh and Bergsma Of the previous research, five test specimens had weld details that are representative of those currently used in seismic design practice (two PJP from , the test specimen consists of a W8×67 (W200×100, A992 Gr. 50–345 MPa) cantilever column and a 457×457×57.2 mm (A572 Gr. 50–345 MPa) base plate that is supported on a rigid steel foundation plate and loaded transversely by a hydraulic actuator. Because of constraints on the allowable forces in the testing lab, the specimen dimensions were chosen to be roughly two-thirds that of a more realistically sized first floor column. This specimen scale produced the largest forces that the testing facility could safely withstand. The specimen was loaded in the direction of the column major axis bending with load applied 1.75 m above the base to approximately represent the point of inflection in the bottom story of a fixed-base moment frame (roughly 2/3rd of the story height). For clarity, the restraint system provided to prevent lateral (out of plane) deformations is not shown in The test variables considered and selected test results are summarized in . The primary test variable is the type of weld detail, CJP or PJP, as shown in . As discussed previously, the details were selected based on a survey of current design practice for SMRF systems. In the CJP detail ( summarizes measured material properties from round coupon tests, designed per the ASTM standard for tension testing of metallic materials , loads were applied according to two alternative displacement histories that are based on ATC-SAC protocol a) was applied to five specimens (Test #1, 3, 4, 5, and 6). The “near-fault” loading protocol (b) was applied to Test #2 and features a large unidirectional pulse, followed by the general cyclic loading history similar to that applied in the other tests. The general cyclic history was appended to the near-fault history in the event that fracture does not occur during the near-fault protocol cycles. The SAC loading protocols are expressed in terms of story drift ratios and are converted to displacements by multiplying them with the column height (69 in.). The results are presented in terms of story drift ratios, expressed in percent drifts, for convenient interpretation.Representative plots of the column base moment versus drift ratio are shown in a, for general cyclic loading history (Test #4), and a). The initiated crack was observed to grow steadily as the loading progressed. The initiation and growth of the ductile crack did not produce any discernible changes in the load deformation curve (refer ). Stable crack growth continued for a considerable number of cycles (approximately 10–15) until the growing ductile crack finally triggered a sudden failure at drift levels in the range of 5%–9%.b), however, for PJP tests the crack propagates partly or wholly through the web. shows a photograph of a PJP specimen (Test #5) after brittle crack propagation.For Test #2, subjected to the near-fault loading history ( summarizes the test data corresponding to ductile crack initiation, θinitiation, and complete fracture propagation, denoted as θfailure. The test results are also presented graphically in a, b, where the square and round markers are superimposed on the displacement loading history to indicate the occurrence of fracture initiation and complete fracture, respectively. The data for fracture initiation in refer to the largest drift sustained prior to initiation, while the data for complete fracture include both the maximum drift and drift cycle sustained prior to fracture and the drift at fracture. For example, for Test #1, fracture initiation occurs during the first cycle to 3% drift and complete fracture occurs during the second cycle to 5% drift at a drift ratio of 3.1%. Another index for fracture endurance is the energy norm, which is total energy dissipated by the specimen prior to fracture, normalized by the product of the yield moment and yield drift (Myθy). Total dissipated energy is calculated through a simple trapezoidal numerical integration technique. The last column of the table includes a breakdown of the energy dissipated during the near fault loading to the total energy dissipated for Test #2. is a comparison between the predicted ultimate strengths of the base connection (associated with column hinging, per FEMA 350 for an illustration), to reflect moment demands at design level earthquakes. The expected strength is determined according to the provisions of FEMA 350 (see footnote) for determining moment demands for the design of beam–column connections. The FEMA 350 procedures adjust for the expected steel yield strength (through the Ry factor) and strain hardening (through the Cpr factor) in determining the required strength. For Grade 50 (345 MPa) steel, the expected strength is equal to 1.3 (Cpr×Ry=1.2×1.1) times the nominal plastic moment of the columns., the main observations from the tests are as follows: All the specimens sustained drift levels of 5%–8% before exhibiting any type of deterioration in the load–deformation response. In all six tests, the onset of deterioration was sudden and corresponds with a complete connection failure. This stable behavior up until 5%–8% drift exceeds the expected earthquake drift demands of approximately 2% for design level earthquake loads and 4%–5% for maximum considered earthquakes (per Krawinkler et al. Overall, specimens with PJP welds performed better than those with CJP welds, with the PJP details sustaining drifts as large as 8%–9% before failure, as compared to 5%–6% drifts for the tests with the CJP detail. This is attributed to the concentration of stresses and strains in the fusion line and HAZ of the column, created by the weld access hole in the CJP specimens (b). The concentration of strains is exacerbated by the fact that material in the HAZ is generally less ductile than either the parent base metal or the weld metal a shows the crack growth just prior to the brittle fracture (Test #4), and The type of loading protocol does not appear to have a significant effect on either the mode of failure or the ductility of the specimens. Test #2, which featured the near-fault loading protocol appended by the general cyclic loading protocol, exhibited fracture ductility (with respect to complete fracture) similar to that of the other CJP specimens. This is despite the early initiation of fracture in Test #2 during the near-fault portion of the loading protocol.Generally, the energy dissipated by the specimens is proportional to the drifts observed at the complete fracture, mainly because there is little degradation in the specimen properties until the point of complete fracture and since the specimen configuration and details are similar. For test #2, approximately 50% of the energy was dissipated during the near fault loading.For these test specimens, the FEMA 350 provisions ). However, the calculated demands do exceed the measured moments at 2% drift ratios (see column of M2%/Mpr ratios in ) by about 20%. Thus, the FEMA 350 predictions appear reasonable and conservative for estimating moment demands in base connections during design level events.This paper presents findings from six two-thirds scale tests of column base connection specimens subjected to earthquake-type cyclic loading. The main objective of this paper is to examine the effect of weld details on the response of column base connections loaded in major axis bending. The data complement previous studies (Fahmy et al. High cycle fatigue life prediction of laser additive manufactured stainless steel: A machine learning approachVariations in the high cycle fatigue response of laser powder bed fusion materials can be caused by the choice of processing and post-processing strategies. The numerous influencing factors arising from the process demand an effective and unified approach to fatigue property assessment. This work examines the use of a neuro-fuzzy-based machine learning method for predicting the high cycle fatigue life of laser powder bed fusion stainless steel 316L. A dataset, consisting of fatigue life data for samples subjected to varying processing conditions (laser power, scan speed and layer thickness), post-processing treatments (annealing and hot isostatic pressing) and cyclic stresses, was constructed for simulating a complex nonlinear input-output environment. The associated fracture mechanisms, including the modes of crack initiation and deformation, were characterised. Two models, by employing the processing/post-processing parameters and the static tensile properties respectively as the inputs, were developed from the training data. Despite the diverse fatigue and fracture properties, the models demonstrated good prediction accuracy when checked against the test data, and the computationally-derived fuzzy rules agree well with understanding of the fracture mechanisms. Direct application of the model to literature results, however, yielded a range of prediction accuracies because of the variability in the reported data. Retraining the model by incorporating the literature results into the dataset led to improved modelling performance.Recent progress in additive manufacturing (AM) has encouraged the use of the technology beyond rapid prototyping to direct manufacture of functional parts Data-driven approach utilising the design of experiment technique had also been studied, where the process and fatigue life relationship of L-PBF stainless steel 316L was examined for a two-factor system Machine learning techniques are effective alternatives for solving engineering problems as they are capable of recognising patterns in complex data In particular, the adaptive neuro-fuzzy inference system (ANFIS) could be a technique that is suited for fatigue modelling. By integrating fuzzy logic into the neural network In this work, the ANFIS was examined for predicting the high cycle fatigue life of L-PBF stainless steel 316L samples, under the effects of varying processing/post-processing conditions and cyclic stresses. The dataset consists of 139 experimental fatigue data, part of which was used for training the model. Two models, using the processing/post-processing parameters and the tensile properties respectively as the inputs, were constructed. The model performance was evaluated by applying the models to the test data, as well as by cross validation with literature results.The dataset consists of 139 S-N data obtained from the authors’ prior experimental works . This resulted in parts with different defect characteristics, ranging from near fully dense parts with porosity fractions less than 0.1% to parts with porosity fractions close to 10%. Some of the S0 samples were subjected to post-processing annealing or hot isostatic pressing (HIP) treatments The fatigue failure modes of the samples are also listed in . Depending on the processing and post-processing conditions, three major types of failure behaviour were observed, i.e. (1) microstructure-driven crack initiation, (2) defect-driven crack initiation and (3) ratcheting-dominated deformation. Samples exhibiting the different failure mode constitute about 55%, 21% and 24% of the dataset respectively. This section gives a brief description of the fracture behaviours for selecting the input variables of the model; refer to the Refs. Microstructure-driven crack initiation concerns the near fully dense samples and is characterised by inter- and trans-granular fracturing. Competitive grain growth as a result of non-equilibrium solidification during L-PBF processing resulted in the formation of differently-oriented grain clusters a and b for Sample 4. Closely-packed ultrafine cellular sub-grains are clearly visible at the crack origin, indicating intergranular de-bonding. Within this processing region, cracking is very sensitive to microstructural heterogeneities, where the transition from intergranular- to transgranular-dominated fracture occurred at higher energy inputs, as shown in d for Sample 8. The longer time for diffusion at slower cooling rate could have led to preferential clustering of second phase particles at the sub-grain boundaries rather than the grain boundaries, leading to the transgranular fracture With a reduction of the laser energy input, large irregular lack of fusion defects were formed because of poor layer-layer and track-track overlapping e. With further reduction of the energy input, the defects increased both in size and number. Interaction among the closely-spaced defects led to enhanced stress fields, resulting in simultaneous crack initiation from multiple defects, as shown in f. Samples associated with such failure modes are referred to as the porous samples in this work. Note that both microstructure-driven (e) crack initiations were observed for Sample 4. As the size of the defects produced at this processing condition could be approaching the critical size that trigger the transition from microstructure-driven to defect-driven crack initiation, either failure mode could be possible depending on the microstructure and defect arrangements, e.g. size, orientation and location. This is an exemplification of the probabilistic nature of fatigue for L-PBF stainless steel 316L.Ratcheting-dominated failure applies to the post-processed samples. The annealing or hot isostatic pressing procedure degraded the strength of L-PBF stainless steel 316L via recrystallization and grain growth shows that samples exhibiting the different failure modes are associated with different trends in terms of the tensile strength and ductility properties. The near fully dense samples demonstrate the optimum tensile strength, in agreement with literature results The above results indicate that the high cycle fatigue properties of L-PBF stainless steel 316L are related to the processing/post-processing strategies and the tensile properties. Therefore, two models, employing the different variables as the inputs, were implemented, i.e.‘Process-based’ model: the processing and post-processing parameters, i.e. P, v, t and T served as the inputs, where T was assigned with levels 1, 2, 3 and 4 for representing the as-built, low-temperature annealing, high-temperature annealing and HIP conditions respectively.‘Property-based’ model: the ultimate tensile strength σb and elongation to failure δ served as the inputs.In addition, as the fatigue tests were conducted at a constant load ratio, the maximum applied cyclic stress σmax was used for characterising the loading condition.A typical feedforward neural network consists of three layers: an input layer, a set of hidden layers, and an output layer show the architectures of the ‘process-based’ and ‘property-based’ ANFIS models respectively. As the ‘property-based’ model is associated with a simpler structure, it is used for explaining the ANFIS framework here.The neural network structure of an ANFIS is formulated based on a set of ‘if-then’ rules, such as ‘if x is A, then y is B’ Rule 1: If σb is A1, δ is B1 and σmax is C1, thenRule 2: If σb is A2, δ is B2 and σmax is C2, thenwhere fi is the linear consequent function of the ith fuzzy rule and pi, qi, ri and si are the consequent parameters of the rule.Layer 1: Nodes in the first layer specify the degrees to which the given inputs belong to each of the linguistic labels in terms of the membership functions μ:Membership functions can adopt the forms of the triangular, trapezoidal, Gaussian or bell-shaped functions where αj and βj are parameters of the Gaussian distribution.Layer 2: Nodes in this layer perform the T-norm operation, where membership values corresponding to each fuzzy rule are multiplied:The output ωi is the firing strength, also known as the ‘degree of fulfilment’, of a particular rule. As the ANFIS model contains four fuzzy rules, i goes up to a value of four.Layer 3: Output of a node in this layer is equal to the product of the normalized firing strength, i.e. the ratio of the ith rule’s firing strength to the summation of the firing strengths of all the rules, and the consequent function fi of the rule:Layer 4: This layer constitutes the summation of all incoming signals, from which the fatigue life N is predicted:Note that the logarithmic values of N were used for modelling to avoid extreme differences in fatigue lives, which are within the range of 103 – 106 cycles for the dataset. The square nodes, i.e. Layer 1 and Layer 3, are adaptive nodes where the parameters are optimized by the adaptive neural network. For supervised learning, models learn by direct comparison of the predicted outputs with the actual values in order to minimise the error function, in this case, the root mean squared (RMS) error. Gradient descent-based method can be used for tuning the parameters of the ANFIS For large multivariate datasets, clustering is often performed to identify natural groupings in the data so as to allow a concise representation of the system behaviour. For ANFIS, clustering is used as a rule extraction algorithm where the cluster centres provide the basic premises of the rule base Subtractive clustering was performed on the training datasets using the Clustering tool in Matlab. The algorithm carried out the following steps for extracting the cluster centres As the ROI is directly related to the number of fuzzy rules, suitable ROI value can be determined by evaluating the performance of the corresponding ANFIS model. Prediction error generally decreases with increase in model complexity as the larger number of model parameter allows more explicit representation of the system behaviour shows the training and validation errors as a function of the number of fuzzy rules, obtained by varying the ROI. For both the ‘process-based’ and ‘property-based’ models, the training errors decrease with the number of fuzzy rules, as expected. The validation errors, however, increase following the initial reductions. This is suggestive of overfitting as the model is unable to generalise to the unseen data despite the increase in model complexity. Therefore, to prevent overfitting, the numbers of fuzzy rules were selected for the conditions with the smallest validation errors before they start diverging. This led to nine rules for the ‘process-based’ model (ROI = 0.65) and four rules for the ‘property-based’ model (ROI = 0.5), as highlighted in This section examines the validity of the computationally-derived fuzzy inference system vis-à-vis the fracture behaviours of the samples. The accuracy of life prediction is demonstrated for the training and test datasets, while the generalisation capability of the models is evaluated by applying them to literature data.For the ‘process-based’ model, the input variables P, v, t, T and σmax were partitioned into a set of 5-1-3-3-5 Gaussian-type membership functions, which are schematically shown in . By carefully examining the membership functions constituting the antecedents of the fuzzy rules, five out of the nine rules were found to characterise the effects of P and v at t ≈ 20 μm and T ≈ 1, two rules for the effects of t, with P and v at the standard processing conditions and T ≈ 1, and two rules for the effects of T, with P, v, and t at the standard processing conditions. This is in agreement with the distribution of the input data shown in , indicating that the rules were derived according to the different sample classes. shows the fatigue life contour plot, taken from a prior work . It can be seen that the locations of the ‘cluster centres’ correspond to some of the major fatigue life regions. In fact, they coincide with the processing conditions of Samples 1–5, which are samples that exemplify the different fracture modes of the as-built parts, as specified in . Based on this observation, linguistic labels such as Very Low, Low, Optimum, High and Extreme can be assigned to the membership functions for P (a). Specifically, the Optimum level (P ≈ 195 W) refers to processing conditions that produce the optimum fatigue properties; the Low (P ≈ 137 W) and Very Low (P ≈ 98 W) levels pertain to crack initiation from single and multiple defects respectively, while the High (P ≈ 254 W) and Extreme (P ≈ 293 W) levels are associated with the transition to transgranular-dominated crack initiation and the generation of critical over-heating-induced defects respectively. Note that the use of membership functions, in contrast to the Boolean logic, for describing the input variables is appropriate, as the change from one fracture mode to another is not triggered by abrupt changes in the input settings, but by a gradual transition.Similar analysis applies to the membership functions for layer thickness (c), where the primary fatigue cracks originate from a single defect at t ≈ 60 μm (Low level), and from multiple defects at t ≈ 80 μm (Very Low level). They contributed to the two rules that characterise the effects of t. For the post-processing variable T, membership functions were identified for the as-built, high-temperature annealing and hot isostatic pressing conditions (d), where the increase in post-processing intensity correlates directly with the extent of fatigue strength degradation due to cyclic plastic deformation, as explained in . This is responsible for the two rules corresponding to the effects of T. Membership functions for σmax (e) are dependent on the actual stresses applied for the fatigue tests and did not contribute to any additional fuzzy rules.A simpler ANFIS structure with 4-4-4 membership sets was obtained for the ‘property-based’ model, as illustrated in shows the ‘cluster centres’ in relation to the tensile properties of the dataset. The σb-δ pairs forming the premises of Rules 1 and 2 correspond to the tensile properties of the near fully dense samples, while those for Rules 3 and 4 correspond to the post-processed and porous samples respectively. Note that rules were generated for two different cyclic stresses for the near fully dense samples but not for the other sample types. This could be attributed to the larger percentage of input data belonging to this group (55% of the dataset). As the clustering algorithm defines cluster centre by calculating the potential of a data point in terms of its distance to all other data points, high density region with many neighbouring data, such as the near fully dense samples, are more likely to be chosen as the cluster centre In summary, the above results indicate that regardless of the input variables, subtractive clustering effectively captured the underlying fracture mechanisms of the dataset. The assignment of fuzzy rules based on the cluster centres ensured mapping of the input-output relationships in accordance to the fracture behaviours of the samples. lists the RMS errors obtained from applying the models. The overall RMS errors range from about 11% to 16% across the datasets. In a similar study by Vassilopoulos et al. The RMS errors for the individual sample types expand a bigger range, from about 9% to 21%. Worth noting is that the porous samples are associated with the greatest errors which consistently surpass the overall errors, while errors for the near fully dense samples tend to be the least and are always less than the overall errors. The good prediction accuracy for the near fully dense condition stems from the large amount of data belonging to this sample group. The more representative dataset ensured more effective training, as evidenced in the larger number of cluster centres/rules corresponding to the input space of these samples. Between the porous and post-processed conditions, which have similar data distributions, the worse modelling performance of the former could be caused by a few reasons. For the ‘process-based’ approach, fabrication of the porous samples incurred the processing variables P, v and t, while post-processing is a function of T only. The synergistic effects of the input variables for the porous samples could have resulted in greater variance, thereby the larger prediction errors. For the ‘property-based’ model, note that the tensile properties of some of the porous samples, i.e. Samples 11 and 13 (crack initiation from single defect), are very close to those of the near fully dense samples (e.g. Samples 2 and 6). Despite the different fatigue properties between the two types of samples, the similar input values will lead to similar model predictions. compares the error distributions for the samples. It can be seen that the histogram for the porous samples shows a heavy right tail, which is indicative of overestimations, while the opposite trend applies for the near fully dense samples. This is reasonable because as the model sought to minimise the sum of squared errors according to the least squares method, it arrived at solutions that approximate the average fatigue lives of the two types of samples. The resulting biased predictions, however, are not desirable, especially for the porous samples as they can lead to unconservative estimates of the fatigue properties. compares the predicted and experimental fatigue lives. The predicted results are generally within a factor of two of the experimental data, even for the porous samples which are associated with the greatest errors. presents the results for selected samples on S-N plots. With the exception of Sample 6 (d), the predicted S-N curves fit well the experimental data across the different failure modes. It is worth noting that the predicted result at each loading condition is generally within the experimental scatter band, indicating that the models are able to account for variations in the dataset due to fatigue scatter. For Samples 6 and 13, the biased predictions from applying the ‘property-based’ model are evident, with the overestimation for Sample 13 amounting to a factor of about 1.7 on life. Nonetheless, the results are acceptable considering the extent of variation in fatigue data due to scatter.In relation to the underlying working mechanism of the ANFIS, the good prediction capability, despite the diverse mix of fatigue properties and fracture behaviours of the dataset, can be attributed to the decomposition of the modelling task by the fuzzy rules The generalisation capability of the ANFIS for high cycle fatigue life prediction was demonstrated using literature data . Only the ‘property-based’ model was examined as complete input information for the ‘process-based’ model was not always available. Data pre-processing was done to adjust for the effects of fatigue loading: the Walker equation, based on a Walker parameter of 0.527 Results from applying the ‘property-based’ model to the literature data are shown in . A range of prediction accuracies was obtained depending on the dataset. Specifically, the model produced reasonable results for the heat treated samples of Leuders et al., but not for the other samples by the same authors, especially at the lower and higher fatigue life regions. Results for Spierings et al. and Mower and Long were significantly overestimated, by as much as over an order of magnitude in the extreme cases. Mismatched material properties reported by the studies were likely to have contributed to the varied modelling performance. For example, the tensile strengths of 760 MPa and 717 MPa, reported by Spierings et al. and Mower and Long respectively, are higher than those for samples with equivalent ductility values used in this work. As the higher tensile strength implies better fatigue resistance, the fatigue lives were overestimated. For the HIP samples used by Leuders et al., as the tensile properties are comparable with this work, the poor predictions are suggestive of dissimilar fatigue properties. Noteworthy also is the considerably worse predictions for the inclined samples than the horizontal samples of Mower and Long despite the similar tensile properties. This is a result of actual difference in the experimentally determined S-N data, which, as noted by the authors, could have been caused by surface defects that promoted premature failure of the inclined samples.The mismatched material properties exemplify the machine-to-machine and the associated processing variabilities of L-PBF parts A systematic approach to input parameter selection can be achieved using statistical methods. For example, by conducting design of experiments with 1000 design points, Bessa et al. Expanding the database by incorporating the literature data serves to provide the needed information for learning, which is the basis of ‘big data’ analytics , and including the load ratio as an additional input parameter, a 6-rule ANFIS model was obtained. This led to marked improvements in the prediction accuracy across the different sources of data, as shown in . The overall RMS errors are larger than those for the original model, at 17.41% for the training data and 20.25% for the testing data, because of the increased data complexity. Nonetheless, the errors are acceptable considering the fatigue scatter, which is evaluated at a factor of two on life.The better modelling performance can be appreciated by examining the fuzzy rules. shows the locations of the ‘cluster centres’ on the σb-δ plot. In comparison with the original ‘cluster centres’ in , three new ‘cluster centres’ were extracted. The locations of the cluster centres correspond to the tensile properties of the literature data, indicating that the clustering algorithm correctly captured the oddities in the data and assigned new rules for prediction. Task decomposition by the rule-based approach reduced the dimensionality of the problem, such that the use of the tensile properties as the inputs is still valid. Worth noting also is the small amount of data needed for training the model (e.g. 13 data points only for Leuders et al.). The strong versatility and adaptability of the ANFIS suggest the possibility of extending the model to highly complex systems, e.g. one that involves more processing variables, machine models and material types, provided sufficient training data are available.This study examined the adaptive neuro-fuzzy-based machine leaning technique for modelling the high cycle fatigue life of L-PBF stainless steel 316L. The following conclusions could be drawn from the results obtained:The ANFIS method successfully predicted the fatigue life of L-PBF stainless steel 316L samples showing a wide range of material properties and fracture behaviours, arising from the use of different processing/post-processing conditions, and subjected to different cyclic stress levels. The models effectively captured the characteristic failure modes of the dataset, which formed the rule base for fatigue life predictions.Both the processing/post-processing parameters and tensile properties can be used as the inputs for constructing the ANFIS, as demonstrated by the ‘process-based’ and ‘property-based approaches. For real life engineering practices, these approaches can be adopted concurrently, for quality assurance at the manufacturing stage and property assessment stage respectively.The transparency offered by the linguistic rules allowed better appreciation of the model, aided the selection of model parameters and simplified the model design and validation process. This is an advantage of the ANFIS over the non-fuzzy-based learning approaches. Besides, the use of fuzzy boundaries could have allowed better tolerance for imprecise data, such that the models could effectively account for the presence of scatter in the S-N data.Results from applying the model to literature data indicate that the applicability is restricted to the experimental space over which it is trained. The selection of representative input variables and construction of a large database for training are important for improving the generalisation capability of the model.Molecular dynamics simulations of brittle fracture in fcc crystalline materials in the presence of defectsMolecular dynamics (MD) and molecular statics (MS) simulations of crack propagation in the presence of defects in brittle crystalline materials under mode I loading are carried out on the (0 0 1)[1 0 0] crack system using the embedded atom method (EAM) interatomic potential. Substitutional impurity point defects are introduced into a 3D thin-strip slab of 160,000 atoms at various distances from the crack tip. The critical load required for the initiation of crack propagation is obtained, along with the atomic level stress distribution near the crack tip. The results indicate that the critical load is dependent on the defect species, geometry, and position. When located directly at the crack tip, the defects reduce the peak internal stress, increasing the critical load relative to the defect-free system. As the defects are moved away from the crack tip the critical load goes through a minimum and approaches the value of the pure material asymptotically. In addition, the critical loads calculated for the defect-free systems exceed the Griffith value, given by 2γs, for a pure brittle material. This difference is investigated by analyzing the lattice trapping phenomenon using a series of constrained energy minimizations (drag method) and the nudged elastic band (NEB) method in a defect free crack system.The main failure mechanism of brittle materials occurs through the creation and propagation of cracks. Encompassing an atomic scale crack tip and extending over macroscopic distances, crack propagation is inherently multi-scale in its nature. By gaining a deeper understanding of this process and the factors that affect it, the strength of materials can be more accurately predicted and methods to prevent material failure can be identified.The presence of material defects such as impurities, vacancies, dislocations and microcracks, is known to alter the propagation of a crack. Depending on their type and location, the propagation can be aided or impeded. In order to fully analyze the various imperfections and their impact on the local stress and strain fields at the crack tip, a method that can account for the atomic level inhomogeneities of the material must be used.Traditionally a continuum mechanics based approach has been implemented in the study of crack propagation. From this treatment, one can obtain predictions for the critical failure stress and strain of a homogenous linear material along with the stress, strain, and displacement field distributions near the crack tip. Ultimately, though, it is inadequate for atomic level details since it models matter as being infinitely divisible and continuous. Linear elasticity theory fails to fully describe the fine details in the vicinity of the crack tip due to the nonlinear behavior in the high stress/strain process zone.Modeling brittle fracture atomistically yields phenomena such as lattice trapping and the velocity gap, which arise due to the discrete nature of materials Ideally crack propagation would be studied by solving the time-dependent Schrodinger equation, as the modeling of a system at the atomic level is fundamentally quantum mechanical. This computational task is, however, impractical for any moderate size system and approximations must be implemented. Treating the system classically significantly reduces the amount of computational work that is required. The accuracy of the results, of course, depends on the reliability of the interatomic potential that is employed.With the multi-core evolution of processing units (CPUs & GPUs) and fast efficient parallel simulation codes, appropriate size systems can now be easily modeled from a classical atomistic standpoint. The goal of parallel computation is to achieve increased performance by breaking a given job into smaller pieces and distributing them to separate processing cores for concurrent completion. A large majority of the molecular dynamics algorithm is parallelizable, such as the force calculations along with the position and velocity updates. The simulation code utilized in this study is the Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS) which uses a spatial-decomposition algorithm and was developed by Plimpton at Sandia National Labs Over the years there have been numerous classical computational studies This present study focuses primarily on the effects of substitutional defects on crack propagation. We follow the methods of Gumbsch et al. In addition to the study of defects, the lattice trapping barrier will be investigated using a drag method and with the nudged elastic band (NEB) method. Lattice trapping is a term coined to describe the phenomenon where a crack tip becomes stuck or trapped between atoms. A periodic potential, which is comprised of a series of energy barriers, exists for the extension of the crack tip due to the atomistic nature of the crystalline material. The crack encounters an energy barrier and while there is enough energy available for crack extension stored in the strained material ahead of the crack tip, the bonds directly in front of the crack tip are not strained enough to break This study is organized as follows: In Section , the technical details such as basic analytical fracture concepts and the computational procedures are described. In Section , the results of the crack propagation simulations with and without the defects are summarized and finally in Section The internal stress and strain induced in an externally loaded material are significantly magnified at the tip of cracks. Using continuum mechanics this is clearly demonstrated as the stress field a distance r from the crack tip is predicted to have a leading term with a (1/r) dependence From an energetic point of view, a crack will propagate in a material if the energy available for crack extension is larger than the energetic cost of the extension. For a perfectly brittle material under fixed displacement loading, the energy available for crack growth is equal to the decrease in the system’s elastic energy caused by the extension of the crack. This quantity is formally defined as the energy release rate, G, which is the energy released per unit thickness and per unit length of crack tip advance. The cost or resistance to crack growth in a brittle material is the energy necessary to create the new free surfaces that are formed when the crack advances. The crack driving term acts to decrease the energy of the system by relieving the strain in the portion of the material that the crack has passed through while the crack resistance term acts to increase the total system energy since new free surfaces are formed.By analyzing a specific crack system energetically, the functional form of the energy release rate can be determined. depicts the thin strip crack geometry that will be utilized in the simulations. The system is strained in the z-direction by displacing the upper and lower surface layers from their equilibrium position and holding them rigidly fixed. This fixed-grips (constant strain) condition prevents the external load from performing any additional work on the system during crack propagation. When the crack extends in the strained linear cubic material, the elastic energy dissipated or released per unit area of advance is equal to,where ɛzz is the engineering strain defined as the ratio of change in height to unstrained height zo and E* is the effective Young’s modulus in the direction of the applied strain The energetic cost required per unit area of crack surface creation is given by the Griffith condition,where γs is the surface energy of the crack planes. The identification of the crack resistance term with twice the surface energy, originally proposed by Griffith, is known to underestimate the actual crack resistance term of a material and does not account for lattice trapping The (0 0 1)[1 0 0] crack system is studied using molecular dynamics and molecular statics simulations. In several atomistic studies The thin strip geometry allows for the study of crack propagation under a constant energy release rate, G. To ensure the functional form of the energy release rate for the thin strip geometry is applicable, a system length (x-direction) to height (z-direction) ratio of at least 4 (L |
⩾ 4 h) is used and the crack tip of the initial starter crack is located a distance of L/4 from the system’s edge. A base 160,000 atom lattice comprised of 400 × 8 × 100 atomic layers in the x, y, and z directions respectively is utilized, as shown in . The simulation box is periodic in the y-direction [0 1 0]. The starter crack region is comprised of 10 atomic layers above and below the crack plane as depicted in . The interaction between these two groups of atoms is turned off, severing the atomic bonds and allowing for the formation of a seed crack in the material. The atoms in the last layer perpendicular to the x-direction (right edge) are constrained to move only in the y and z directions. This prevents the edge from bowing inwards due to the material’s nonzero Poisson’s ratio.Two Stokes damping regions are present in the system to absorb the acoustic/shock waves emitted from the crack tip. These regions consist of the first and last 10 atomic layers perpendicular to the x-direction in which a drag force proportional to the atom’s velocity, Fd |
= |
−bv where b |
= 0.1 (eV ps/Å2), acts to drain energy out of the system.The crack tip is not only a concentrator of stress and strain, but also is a region of high temperature. Temperature fluctuations can cause a crack to be trapped or can knock a crack out of a trapped state by instantaneously increasing or decreasing the energy release rate in the vicinity of the crack tip by producing material voids, kinks, dislocations, and blunting the crack tip Three impurity defect configurations are investigated in nickel in which the native atomic species at a lattice point is replaced by a differing species, as shown in . In configurations A and B, four defect lines comprised of 4 impurity atoms each are placed throughout the thickness of the material. In configuration A, the defect lines are grouped together along the upper and lower crack planes. The pairs of defect lines in configuration B are separated by 4 atomic layers in the z-direction and lie off the upper and lower crack planes. In configuration C, two defect lines are again grouped together on the upper and lower crack planes along with two pairs, both above and below, each separated by 3 atomic layers in the z-direction. The critical load required to initiate crack propagation is first found for the defect free material. This is then recalculated for the three defect configurations with varying impurity defect species (Cu, Au, Ag, Pd and/or Pt) and distances from the initial crack tip. The stress field distribution and strain energy of the atoms along the crack plane are analyzed to interpret and explain the resulting changes in the critical loads.The material is first uniformly strained in the z-direction by scaling the atomic coordinates and holding the atoms in the top-most and bottom-most layers rigidly fixed in all directions. This initial strain value corresponds to the energy release rate or load equal to the Griffith value, Go |
= 2γ100, of the material. After the system is uniformly strained to a load corresponding to Go via scaling, the system is relaxed through energy minimization with a force tolerance convergence criterion of 1 × 10−9 (eV/Å). This leaves the crack in a lattice trapped state. The load is then increased by 0.005Go by displacing the top and bottom atomic layers and again holding them rigidly fixed. Eq. can be used to find the associated strain at a given load, G, and the total layer displacement, δ is calculated using the strain values, δ |
= |
zo(ɛf |
− |
ɛi). Load values of Go and 1.005Go are used for G in the calculations of ɛi and ɛf, respectively, for the first displacement step.A 150 ps NVT molecular dynamics simulation is then performed via the time integration of the Nosé–Hoover equations of motion within a velocity Verlet framework with a damping parameter of 0.1 ps, a temperature of 1 K, and time-step of 1 fs. The system trajectory and various physical quantities are then analyzed using a combination of visualization software (AtomEye In order to analyze the stress field distribution near the crack tip, a local measure of atomic stress is necessary. Employing the Virial theorem which relates the average kinetic and potential energy of the system to an internal atomic stress, an instantaneous atomic virial stress tensor can be defined as σijα=1Vα-mαṙiαṙjα+12∑β≠α∂Epot∂rαβriαβrjαβrαβwhere riαβ is the ith component of the separation vector rαβ, ṙiα is the ith component of the velocity vector ṙα, mα and Vα are the mass and volume of atom α, and the summation is over the interacting neighbors of atom α. The virial stress has both a kinetic and potential contribution, as evident from Eq. The lattice trapping barrier in pure Ni is investigated using two methods. With the first, which was utilized by Bernstein and Hess In order to confirm the validity of the energy path obtained from the aforementioned drag method, the nudged elastic band (NEB) The critical loads, Gc, for pure (no defects) Ni, Cu, and Au were calculated for the crack system ( with the surface energies given for the EAM implementation depicts the variation of the independent elastic constants as the three pure element crack systems are strained in the z-direction.It should be noted that, in general, there is no unique way to determine the two engineering elastic constants (Young’s modulus and Poisson’s ratio) from the three or more independent constants of a crystal material due to the overdetermined nature of the problem, however many well-known averaging methods exist (Voigt, Reuss, Hill, etc.). While quantities derived using the results of a particular averaging method may vary (i.e. critical loads), their overall trend remains consistent and independent of the particular method employed.The pure material critical load results are presented in , where Go |
= |
2γ100. As expected, for each material the critical load exceeds the value predicted by the Griffith energetic approach.In two previous atomistic studies, the critical energy release rate for Ni has been calculated and reported as 1.2Go, by Karimi et al. In order to investigate the discrepancy between Go and Gc in pure Ni, the dependency of the lattice trapping barrier on the externally applied load was analyzed using the drag method. Several energy pathways for crack extension are shown in , where the total potential energy (zeroed to the initial minimized energy) is plotted against the reaction coordinate (measure of the system’s position along the transition). The trapping barrier for the extension of the crack by one atomic layer can be identified as the path between two adjacent local minima. The general features of the energy pathways fell into three distinct regimes depending on the externally applied load, as indicated in . For G |
< |
Go, the forward energy barrier is larger than the reverse barrier. At G |
≈ |
Go, the energies at the minima on either side of the barrier are approximately equal. Finally for the case of G |
> |
Go, the forward energy barrier is smaller than the reverse one. The forward lattice trapping barrier is shown to decrease with increasing load, G. This trend will continue until the critical value Gc is reached where the barrier disappears and crack propagation initiates in the material (if thermal/kinetic excitations do not cause initiation beforehand).For the case of G |
= |
Go, the forward and reverse lattice trapping barriers were found to be 32.17 meV and 30.90 meV, respectively using the drag minimization method. This calculation was also performed using NEB (), and resulted in forward and reverse energy barriers of 31.82 meV and 30.82 meV, respectively. The close agreement of the lattice trapping barrier values calculated using these two different approaches confirms that the drag method samples very closely to the minimum energy path found using NEB. The difference in the barrier with direction is believed to be attributable to slight asymmetries between the initial and final system configurations.The critical load for the Ni crack system was calculated to be 0.32Go above the Griffith load. This corresponds to an excess energy release rate of 63.1 meV/Å2. Then, for the system geometry of , extension of the crack by a single atomic layer releases ∼1.56 eV of strain energy above the amount required to create the crack surfaces. Therefore, when crack propagation finally initiates at a load of 1.32Go, approximately 50 times the energy required for the system to overcome the G |
= |
Go lattice trapping barrier mentioned above (∼31–32 meV) is available, granted some of this excess energy is dissipated by the Nosé–Hoover thermostat and Stokes damping regions. Increasing the external load by an amount over Go that corresponds to the trapping barrier is thus not enough to initiate propagation in the material; i.e. the lattice trapping barrier alone is insufficient to account for the discrepancy between Go and Gc for a given material. The transformation of the trapping barrier with increasing load is expected to play a significant role in accurately predicting the critical load value, as are correction terms to the Griffith crack resistance such as those due to plastic deformation and the strain dependence of the elastic constants and surface energy The presence of defects can disrupt the existing homogeneity and underlying symmetry of a material and thus directly alter the internal stress and strain fields. When defects are in the vicinity of an atomically sharp crack tip, they can affect the peak stress and alter the trapping barrier due to their direct effect on its two energetic contributions. Each of these in turn has an impact on the critical load required for the initiation of crack propagation. Substitutional defects consisting of Cu, Pd, Pt, Ag, and Au were placed in Ni in the three configurations depicted in . Ni and Au defects were also considered in a Cu system. The critical loads were recalculated with the various impurities present and are reported in for Cu. It is observed that the critical load is increased for each defect configuration and impurity species as compared to the result for the pure material. This result is consistent with a 2-dimensional study by Rafii-Tabar et al. In order to gain deeper insight into the critical load increase due to the presence of defects, the stress field along the crack planes in the vicinity of the crack tip is examined for a subcritical load. A material strain corresponding to a load of G |
≈ |
Go is applied to the material via the uniform scaling method. A molecular statics geometric optimization is then performed to obtain the relaxed atomic configuration and equilibrium stress field distribution. In using a subcritical load, it is assured that the minimization will not extend the crack (), and the static stress field for a stationary crack tip can be analyzed. The atomic virial stress in the z-direction for the minimized configurations is calculated and plotted as a function of position along the crack planes in the x-direction. In the results for defect configurations A, B, and C, are given for the five different impurity species in Ni. The vertical lines indicate the position of the crack tip. The result for pure Ni is included in each of the graphs for ease of comparison. For the case of pure Ni, the σzz stress steadily increases coming from the intact material, reaches a maximum at the crack tip and then falls off to zero., the peak stress at the crack tip is observed to be reduced with the introduction of defects. Due to this reduction, a greater applied load will be required in order for the stress at the crack tip to reach the material failure value, as compared to the pure case. This is demonstrated by a simple quantitative argument. The internal stress at the crack tip can be calculated by multiplying the applied stress by some magnification factor, λp for a pure material and λd for a material with defects at the crack tip. From , it is apparent that λp |
> |
λd |
> 1. Assuming that the static seed crack begins to travel in the material when the internal stress at the crack tip reaches β (in arbitrary stress units), an applied stress of Sp |
= (β/λp) and Sd |
= (β/λd) will initiate propagation in the pure material and the material with defects, respectively. This criterion is one in which a static crack becomes linearly unstable. For the given failure stress β we have,where λp/λd |
> 1, and, therefore, Sd> |
Sp, i.e. the critical applied stress for the material with defects present is larger than the pure material. It follows then, that the corresponding critical load of the material with defects will also be larger than that of the pure material. Thus the reduction of the peak stress at the crack tip due to the presence of the defects in configurations A, B, and C acts to increase the applied critical load, as reported in . In addition, the amount of stress reduction that occurs at the crack tip for a given defect type and configuration determines which defect species will cause the greatest increase in critical load over that of the pure material. For example, in the middle column of (configuration B), the difference in the peak stresses at the crack tip between the pure Ni and defect cases ranks in increasing order as Cu, Pt, Pd, Ag, Au, which also corresponds to the order of increasing critical load. Thus, as the reduction amount increases the critical load increases as well.To complement the stress analysis that was used in the explanation of the critical load increase, the atomic strain energy along the crack plane is examined. This method was used by Karimi et al. , the atomic strain energy distribution is plotted along the crack plane for Cu defect configuration A in Ni for a load G |
= |
Go and also for pure Ni. As with the atomic stress, one observes a decrease in the peak strain energy at the crack tip when comparing the pure and defect cases. Similarly then, it will require a greater applied load for the strain energy concentration at the crack tip to reach the failure point, and thus a larger critical load is calculated. This atomic strain analysis method was only used to investigate a single defect configuration and species. However, it is expected that similar treatments for each configuration and defect species will yield the same results, as was the case for the stress analysis. Mainly that the peak atomic strain energy at the crack tip is reduced, leading to an increase in the critical load.In this section, we study fracture initiation in the presence of impurity defects away from the crack tip. In all of the previous critical load calculations, the front portion of the various defect configurations was located directly next to the seed crack atoms, in the same atomic plane of the initial crack tip (). In order to investigate the dependence of the critical load on the defect location, the defect configurations are moved 4, 8, and 16 atomic layers from the initial crack tip position, and the critical load is recalculated. To decrease the required calculations for defects in Ni, only the two species with the largest and smallest critical load reported in are considered for each defect configuration. For defects present in copper, Au and Ni are considered for defect configuration A only. In this context the critical load that is calculated corresponds to the load necessary for the crack to travel through the material until the crack tip reaches the first atomic plane containing the material defects. From here the crack arrests and requires the appropriate critical load reported in either , to re-initiate and travel further. The results are presented graphically in . The vertical axis is the calculated critical load in the presence of the particular defect, Gcd, divided by the critical load of the pure material, Gc. The horizontal axis is the number of atomic layers that the defects are located from the initial crack tip.As defects with a larger atomic radius than the native material are moved away from the crack tip, the critical load is first observed to decrease, then dip below the critical value of the pure material, reach a minimum, and finally increase and approach the critical load value of the native material. For the single case of the defects having a smaller atomic radii than the atoms of the native material (Ni defects in Cu), the critical load does not go through a minimum. Instead it decreases smoothly with an increase in the crack tip to defect distance and approaches the critical load of the pure material. In general, as the defects are moved further away from the crack tip, their impact on the local stress field will become less and less. In this context the initial decrease of the critical load along with the asymptotic behavior of each graph can be understood. The amount of the peak stress reduction, which occurs at the crack tip when defects are present, becomes less as the defects are located further from the crack tip. This in turn causes the critical load to decrease. In the limit that the defects are effectively an infinite distance from the crack tip, they no longer have any effect. Therefore the measured critical load, with defects at this large distance, becomes equal to the critical load of the pure material, Gcd |
= |
Gc.The decline in the critical load below that of the pure material (and the emergence of the minima), in the cases where the defect species has a larger atomic radius than the host material, is an indication that the peak stress at the crack tip is no longer reduced as compared to the pure material but is amplified above it. In the reverse of the earlier argument, a larger peak stress at the crack tip results in the defective material reaching the failure value for a smaller applied load than in the pure material. This amplification of the peak stress is confirmed in , where the atomic stress distribution for defect configuration A with copper impurity atoms in Ni located 4 atomic layers from the crack tip is depicted. Here it is observed that the peak stress at the crack tip for the selected defect case exceeds that of the pure material. This, therefore, accounts for the decline of the critical loads in Molecular dynamics and molecular statics simulations were utilized to study brittle fracture in a thin-strip 160,000 atom (0 0 1)[1 0 0] crack system. The atomic interactions were modeled using the embedded atom method interatomic potential. Periodic boundary conditions were employed to prevent the emission and nucleation of dislocations allowing for the study of idealized brittle FCC crystalline materials. Critical loads for the initiation of crack propagation were calculated in pure materials and with various substitutional impurity defect configurations present. By analyzing the change in the atomic stress in the vicinity of the crack tip due to the presence of the defects, one can gain insight into the fracturing process along with the resulting change that occurs in the critical load. Additionally the lattice trapping barriers for three different applied loads regimes were investigated in pure Ni.The critical loads for the pure Ni, Cu, and Au systems were found to exceed the Griffith load. The additional load that is required to initiate crack propagation above Go is typically attributed to the lattice trapping phenomenon. This was investigated using the drag and NEB methods, which were used to calculate the energy path for crack extension in pure Ni. As the external loading on the material increases, the forward lattice trapping barrier correspondingly decreases. At the Griffith load, the forward lattice trapping barrier was calculated to have a value of 31.82 meV. This static energy barrier was found to be insufficient to completely account for the additional load of 0.32Go that is required to initiate propagation.When defects were placed directly at the crack tip, the peak atomic stress was reduced, resulting in an increased critical load as compared to the pure material case. The defects were also found to decrease the strain energy at the crack tip, effectively strengthening the material. As the defects with a smaller atomic radius were displaced from the initial crack tip location, the critical load was observed to decrease monotonically and asymptotically approach the value of the pure material. For defects with a larger atomic radius than the atoms of the native material, the critical load decreased below that of the pure material and went through a minimum before approaching the value of the pure material. This behavior was attributed to an enhancement in the peak stress at the crack tip, which reduces the critical load and thus weakens the material to fracture.For the systems studied, substitutional defects with a smaller atomic radius can be used to strengthen the material, while those with a larger atomic radius can toughen or weaken the material to fracture depending on their location with respect to the crack tip.Behavior of composite segment for shield tunnelComposite segment has been developed for obtaining high capacity in lining of shield tunnel subject to high hydraulic pressure and earth pressure in deep underground, which is made of steel plates connected to an infill of high fluidity concrete with shear connectors. However, a rational design method for composite segments can not be established, because the behavior of composite segments is not clarified. The purpose of this paper is to study the behavior of the most complicated composite segment with six steel plates using finite element method and the four-point bending tests. The effects of the shear studs and the thicknesses of steel plates were verified and estimated quantitatively. Meanwhile, the failure modes of composite segments were investigated. Comparisons show that the finite element model is suitable for composite segments and predicts the behavior of composite segments accurately.The subsurface under-roads in major cities are already crowded with underground facilities such as railroads, and tunnels for electricity, gas, communications, sewage and drainage, and conduits. For example, the subsurface of a national highway in Tokyo has about 33 km of conduits per kilometer of road as shown in Previously, the use of underground space was moving from shallow underground to deep underground. However, at depths of down to 10 m below the surface, there is congestion in recent years. Therefore, new underground construction works for road tunnels of large cross-section, regulating ponds (underground rivers) have been constructed at progressively greater depths.The frequency of using underground space in urban area tends to increase, because the technologies for tunnel excavation are advanced significantly. The use of underground space is not just to excavate deeply, but also to enlarge the cross-section of tunnel and to replace the circular shape with rectangular, multi-circle or other shape of cross-section in recent years. For these reasons, hydraulic pressure and earth pressure on tunnel are high and the occurred resultant forces are very large. Therefore, segments for shield tunnel must satisfy the required performances under severe conditions.For concrete segment, it must thicken the thickness of concrete segment. The weight increases with increasing thickness of concrete segment, and it spends a large amount of labor on producing and handling in factory, transporting from the factory to site, removing from the construction yard to tunnel, and assembling at face. Meanwhile, it easily causes to damage the corner edges of concrete segment due to the segment weight and lack of tensile strength. On the other hand, the costs of producing, assembling temporarily, dividing, transporting, and re-assembling increase with increasing outside diameter of the shield machine. In addition, the treating cost of surplus soil will increase, because the amount of the excavated soil increases in proportion to the square of the excavated outside diameter. For steel segment, it must use the thick steel plates to ensure the necessary load carrying capacity and rigidity subject to high hydraulic pressure and earth pressure, jack thrust and backfill grouting pressure. However, steel segment has economic and welding disadvantages. The adopted ductile cast iron segment has economic disadvantage as well as steel segment (). Therefore, a new composite segment combined a box-shaped thin steel frame and high fluidity concrete is developed as shown in . This type of composite segment has the following advantages: (a) reducing the producing periods of composite segment by using steel form as formwork in cast; (b) obtaining high dimensional accuracy because of minimizing the deflection of steel form in welding process; (c) manufacturing an arbitrary section and an arbitrary shape because of excellent weldability; (d) not be damaged easily like concrete segment in assembling stage because of all sides covered with the steel plates, and compositing a uniform tunnel section; (e) easily assembling the composite segments because of having high stiffness, the few openings of the joints, and high dimensional accuracy; (f) decreasing the construction cost with decreasing the outside diameter of the shield machine and excavated soil, because the reduction of segment thickness can be achieved; (g) resisting large flexural moment because of having large carrying capacity; (h) resisting the particular loads for neighboring construction and sharply curved construction by increasing thickness of the skin plates; (i) a rational structure for seismic design because of having superior ductility; (j) the segment width can be enlarged for having stronger main girders of steel.However, a rational design method for composite segment can not be established, because the behavior of composite segments is still not clarified. To further optimize a rational design method of composite segment, the purpose of this paper is to study the behavior of the most complicated composite segment with six steel at the limit ultimate state. The authors used the tests and numerical analysis to discuss the restraint effect, load carrying capacity, stress distributions of steel plates, and failure modes. Recently, ductile and reinforced concrete (DRC) segment and Steel Segment with pre-filled concrete (SSPC) segment were developed (). The proposed analytical method can be used in the design codes for these type composite segments, if the behavior of the most complicated composite segment with six steel plates can be clarified.Ten steel–concrete composite model segments were designed, constructed, and air-cured. The details of the composite segment specimens are shown in . The compressive strength of the concrete at 28 d concrete was determined by testing standard 100 × 50 mm concrete cylinders according to Japanese concrete specification. Tensile and yield strengths of the structural steel were obtained by tensional testing according to Japanese structural steel Specification. Mechanical properties of structural steel (SS400) and concrete are listed in The simply supported composite segment specimens were loaded symmetrically at two points within the span using a distribution beam shown in . In these test arrangements, pure bending of the composite segment specimens can be obtained between the two loading points without the presence of shear and axial forces. The Biaxial Structure Test Machine of 5000 kN capability was used to apply monotonic load. The load applied was measured and recorded using a load cell. Initial loading control was based on readings of the load applied through the load cell, until 80% of the predicted load carrying capacity. Thereafter, loading was shifted to displacement control based on increments of midspan deflection, which was measured using displacement transducers. Strain gauges were installed on the selected locations in an attempt to measure the stress profiles. The arrangements of the strain gauges and displacement transducers in the composite segment specimens are shown in gives the values of the ultimate load, the midspan deflections at failure, the midspan yield deflections, and failure modes for each of the composite segment specimens. The midspan yield deflections were taken as the average deflections (from the displacement transducers located at the midspan measuring the deflections) corresponding to yield loads that are equal to about 80% of the ultimate loads. shows that the deflections of the composite segments with the welded shear studs are equal to about 60–80% of the ultimate deflections after yielding. shows that the composite segments with no welded shear studs (Cases 1–7) failed by local buckling of the top skin plate. shows that the composite segment of Case 8 with the welded shear studs failed by local buckling of the top skin plate between two adjacent rows of shear studs and concrete crushing, because shear studs were welded to the top skin plate, pulled out of the concrete, causing the top skin plate to separate from the concrete. shows concrete crushing failure of Case 8 after the buckled top skin plate has been removed from the surface to reveal the concrete. shows that the composite segments of Cases 9 and 10, failed by concrete crushing. These two composite segments failed in a different mode because the necessary pull-out strength of shear stud was ensured.An investigation was performed to evaluate the sensitivity of the overall response of composite segments (represented by ultimate load) to likely variations in the thicknesses of the steel plates. Composite segment specimens of Cases 1–7 described in were used for the sensitivity study. The dimension of the compared standard case is L |
× |
W |
× |
H |
= 900 × 200 × 100 mm; all the thicknesses of skin plates, main girders, and joint plates are 4.5 mm; the shear studs are not welded. It can be observed from that the thicknesses of skin plates contribute obviously the ultimate carrying capacity of composite segment. Meanwhile, the effect of the joint plates on load carrying capacity is similar to the rib shear connectors or shear studs.Load and midspan deflection relationships can be described by the load–midspan deflection curves shown in for all tested specimens. It can be observed from that the load–midspan deflection curves are linear up to a load corresponding to the yield point. After yielding, composite segments represent very large deflection or excellent ductility, and maintain load carrying capacities to achieve the ultimate deflections. shows the measured strain distribution across the steel plates along the midspan section of Case 10 as a typical case. It was noticed that the strain values in the skin plates increase and decrease repeatedly, and the strain distribution shows markedly wavy change with increasing load. The strain response was caused by the shear studs welded in the skin plates. Some values of the strains in the skin plates are larger than the values of the strains in the edges of the main girders. The restraint effect of the main girders on the deflection of the skin plates is greater. Based on this investigation, it can be observed that the shear studs and the main girders can affect not only the overall response (e.g., shape of the load–deflection curve), but also local results (e.g., strain distributions along the skin plates).The present study used the finite element program MSC. MARC version2005 () to quantify the mechanical behavior of composite segments under flexural loads. A three-dimensional (3D) finite element model has been developed to account for geometric and material nonlinear behavior of composite segment as shown in . The eight-node solid elements with reduced integration were used to model concrete, steel plates (including the skin plates, the joint plates, and the main girders). Discrete shear studs were modeled by using truss elements, which assume a perfect bond between the shear studs and the surrounding concrete. The cross-sectional area of the truss element was modified to make it equivalent in both strength and stiffness to the actual shear studs in composite segment. Springs were adopted to trace the behavior of the bond–slip between the steel plates and the concrete. The bond-friction model was used to model the interface between the steel plates and concrete surface, which was distinguished according to contact interface and non-contact interface (The constitutive relationship for mortar as an infill of material is subject to the constitutive relationship of the concrete, because mortar has a similar property to the concrete. Material properties of mortar were defined by experimental results of flexural and compressive tests. A nonlinear stress–strain relationship shown in was used for concrete sections in both compression and tension. The elastic–plastic material behavior of compressive concrete with strain softening was modeled by the modified equation based on the proposed criterion by where σc is compressive stress in concrete (N/mm2); ɛc is strain in concrete; fc′ is uniaxial compressive strength of concrete (N/mm2); εc′ is strain corresponding to fc′; and α is defined by the following equation (The coefficient of variability β increases with increasing the compressive strength of the concrete. Therefore, when fc′=42.5N/mm2 then β |
= 32.4 and when fc′=88.0N/mm2 then β |
= 71.4, the other values of β are calculated by using the linear interpolation method.The stress–strain relationship for the concrete in tension assumes that the tensile stress increases linearly with increasing tensile strain up to concrete cracking. After concrete cracking, the tensile stress decreases linearly to zero as the concrete softens. The smeared crack model was used to model the cracking behavior of the concrete. The cracks occur when the stress reaches the tensile strength of concrete ft′, and the full shear retention is assumed. The Buyukozturk yield criterion was used in nonlinear analysis to identify the yielding condition of concrete (It is difficult to measure directly the tensile strength of concrete, therefore, which can be calculated using the following equation based on flexural tests of the concrete specimens.where ft′ is the tensile strength of concrete (N/mm2).The reduction in shear modulus due to concrete cracking was defined as a function of direct strain across the crack in the shear retention model. The shear modulus of cracked concrete is defined as G |
= |
ϕGe, where Ge is elastic shear modulus of uncracked concrete; ϕ is reduction factor, which is given by the following equation (where ɛcd is direct strain across the crack. The shear retention model states that the shear stiffness of open cracks reduces linearly to zero as the crack opening increases.Steel plates and shear studs were modeled in nonlinear analysis as an elastic–plastic material with strain hardening. A trilinear stress–strain relationship was used for steel and shear studs in both compression and tension, as shown in . In nonlinear analysis, the von Mise yield criterion was used to identify the yielding condition of steel plates and shear studs.All composite segment specimens were analyzed by using 3D finite element method as shown in . The analyzed results of some specimens were compared with the experimental results in this section. Meanwhile, the similar results of others specimens were obtained by this study.The calculation of the flexural moment–curvature relationship is based on the following assumptions:Linear strain distribution through the full depth of the section.The stress–strain relationship for steel is assumed to be elastic–plastic.The stress–strain relationship for concrete is assumed to follow Eq. (1).The flexural moment–curvature relationship was obtained by subdividing the cross-section into a large number of horizontal layers. It can be observed that the concrete strain in compressive edge reaches the ultimate strain based on the measured values of strains installed on the concrete surfaces of composite segment specimens with the welded shear studs. Therefore, concrete strain in the compressive edge was incremented using the measured strain until reaching the ultimate strain which was assumed as 0.0044 as recommended by the compressive tests of this study. For each step or increment of strain, the depth to the neutral axes was determined by the strain distribution in main girders, when the sum of all the forces acting on the section becomes zero (i.e., equilibrium of forces is satisfied); then the moment of all the forces (acting on the section layers) about the neutral axis is calculated and the curvature is determined by dividing the concrete strain by the depth of the neutral axis. The entire curve was plotted by repeating the above procedure until the ultimate strain in the concrete was reached. The last point in the moment–curvature curve was the flexural capacity of the section. The flexural moment–curvature curves obtained by using nonlinear analysis are compared with the experimental results as shown in that the predicted result of flexural moment–curvature by using nonlinear analysis is closer to the experimental results. Meanwhile, the flexural moment–curvature curves of composite segments are linear up to a flexural moment corresponding to the yield point. Therefore, composite segment represents unvaried flexural rigidity before yielding. show the load–midspan deflection curves of the experiment and nonlinear analysis. It can be observed from that the load–midspan deflection response predicted by using nonlinear analysis is similar to the corresponding experimental plots. The calculated flexural rigidity of the composite section after yield load becomes similar to the experimental flexural rigidity gradually. However, the calculated flexural rigidity before yield load is larger than the experimental flexural rigidity. It is considered that local buckling or yielding due to residual stress reduces the rigidity of the composite segment section as well as cracking.The behavior of composite segment was studied by using the experimental tests and finite element method in this paper. Meanwhile, the analyzed results by using nonlinear analysis were compared with the experimental results. The following conclusions can be concluded from this paper:Composite segment with the shear studs having sufficient pull-out strength failed by the concrete crushing, and composite segment with no shear studs failed by local buckling of the top skin plate. In addition, composite segment with weak shear studs represents the failure modes of concrete crushing and local buckling of the top skin plate.Sensitivity study has demonstrated that the skin plates make a significant contribution to the ultimate carrying capacity of composite segment. The effect of the joint plates on load carrying capacity is similar to the rib shear connectors or shear studs.The strain distribution in the skin plates shows markedly wavy change with increasing load. This strain response was caused by the shear studs welded in the skin plates. Therefore, the effects of the shear studs must be considered in the design codes for composite segment.Composite segment represents very large deflection or excellent ductility after yielding, which is a rational structure for seismic design.It can be observed that the predicted results using the proposed finite element model were close to the experimental results. Therefore, the proposed finite element model is suitable for studying the behavior of composite segment. Annals of the CIRP Vol. 56/1/2007 -73- doi:10.1016/j.cirp.2007.05.020 Fundamental Wear Mechanisms when Machining Austempered Ductile Iron (ADI) F. Klocke 1 (1), C. Klöpper 2 , D. Lung 1 , C. Essig 1 1 WZL, Laboratory for Machine Tools and Production Engineering, Aachen University, Germany 2 Alfred H. Schütte GmbH & Co. KG, Köln, Germany Abstract Austempered Ductile Iron (ADI) is characterised by improved mechanical properties but low machinability compared to conventional ductile iron materials and steels of similar strengths. The mechanical properties of ADI are achieved by a very fine austenitic-ferritic microstructure. However this unusual microstructure significantly affects mechanical and thermal machining properties. A keen understanding for the interactions of microstructure, chip formation, machining properties, cutting material and wear mechanisms is essential for the optimisation of the cutting process. This paper describes material and machining investigations as well as cutting simulations to reveal the wear mechanisms being responsible for the low machinability of ADI. Keywords: Machinability, Microstructure, Finite Element Method (FEM) 1 INTRODUCTION In many industries increasing demands concerning cost- and weight-efficiency call for new construction materials. Most efforts to achieve these requirements either apply new materials with similar strengths but lower densities or increase the strength of the traditional material by alloying or by heat treatment. The alternative chosen depends on parameters such as the mechanical and thermal loads or on boundary conditions such as manufacturing costs, recyclability, public acceptance and machinability. Cast iron materials offer a high freedom in shape and relatively low costs in production. For parts with high mechanical and thermal loads and special damping capacity requirements (e.g. engine blocks) compacted graphite iron (CGI) has proven suitable. Providing material properties comparable to high-quality forging steels, the fields of application of austempered ductile iron (ADI) extend beyond those of conventional ductile iron materials. Due to a special heat treatment, material properties uncommon to cast irons such as high strength and toughness as well as ductility and good wear resistance can be achieved. High performance transmission parts and chassis components are typical applications of ADI. As a high accuracy is required, the machining process takes place in the heat-treated condition. In order to meet standards for serial production, demands on the tool life are high. In case forged parts or conventional cast iron parts are being substituted, the costs of the new parts must furthermore not exceed the former costs [1]. In order to perform cutting tool and cutting parameter optimisation, detailed knowledge concerning the cutting properties is therefore essential [2]. In this paper, the cutting properties of ADI, especially its material-specific properties, are examined. 2 MATERIAL PROPERTIES Investigations on the material properties of ADI relevant to machinability focus mainly on three fundamental aspects: First of all, the mechanical properties determining the dominance either of abrasive or of adhesive wear mechanisms are examined. Furthermore the extreme conditions during chip formation, the high-strain material behaviour and strain rates are of interest. Finally, the influence of the ADI-specific microstructure is examined; especially the formation and detection of martensite and its effect on machinability are analysed in this part [3]. 2.1 Mechanical properties The mechanical properties of ADI cast iron grades can be modified within a wide range by changing the transformation temperature. Hardness and high wear resistance are achieved by applying lower temperatures (280 - 350°C), whereas higher temperatures (350 - 390°C) lead to a combination of high strength and ductility which is exceptional for cast irons. The minimum basic mechanical properties are standardised in Europe and the US [4, 5]. In this paper, one grade with high ductility and strength (ADI-900) as well as one with high hardness and wear resistance (ADI-1200) are examined. ADI can be characterised by twice the nominal strength of conventional cast irons with similar toughness. Additionally, ADI is highly ductile despite the inner notch effect common to globular cast irons. With respect to this specific set of material properties, ADI can be compared to quenched and tempered steels. The ratio of density to elastic limit as well as the damping behaviour of ADI is superior to that of steel and of most light-weight metals. The insertion of inherent compressive stress by strain- hardening additionally increases the resistance to fatigue and wear significantly. These effects can be attributed to the transformation of the austenite to martensite in the rim zone. 2.2 Machinability Despite similar mechanical properties, a significantly shorter tool life based on the same cutting parameters is attained for ADI in relation to comparable steels and conventional cast irons. Extreme crater wear located very close to the cutting edge is the characteristic tool wear phenomenon. This effect destabilises the cutting edge and leads to a fracture of the crater lip. In order to understand this wear phenomenon, a more precise understanding of the machinability behaviour of ADI is necessary. -74- Compared to other ductile irons, the potential to form segmented chips is even more pronounced when cutting ADI. From the process perspective, this specific chip formation leads to low mean cutting forces with a strong dynamic proportion, but the peak values of these forces clearly exceed those of quenched and tempered steels. Additionally, the size of the contact zone between the chip and the rake is small and located close to the cutting edge. Moreover the austenitic-ferritic microstructure has a tendency to adhere with tungsten-based carbides, which are commonly utilized for cutting cast iron materials. Besides adhesive also abrasive wear effects are observed when machining ADI. This abrasive wear is caused by the hardness of the austenitic-ferritic microstructure as well as by segregation-effected hardness fluctuations. Beside the hardness of its structural constituents, the abrasion resistance and hardness of ADI is determined by its filigree microstructure. For ADI grades with higher strengths, the microstructure includes additional carbides, which cause stronger abrasive wear [3]. 2.3 Microstructure The unusual combination of adhesive and abrasive wear mechanisms is caused by the special austenitic-ferritic microstructure of ADI. The microstructure of ductile ADI- grades (comp. Figure 1) consists of a fine structure of acicular ferrite, which is free of bainitic ε -carbides and reacted stable austenite. 50 μm 10 μm globular graphite stabilised austenite acicular ferrite austenitic-ferritic microstructure initial austenite grain size residual austenite Figure 1: Microstructure of a ductile ADI grade (ADI-900). Reducing the isotherm transmission temperatures leads to a finer austenitic-ferritic structure. Furthermore, the diffusion of graphite is inhibited and ε -carbides are separated at the boundaries of the ferrite lamellas. This particular microstructure is called bainitic-ferrite. The availability of graphite and the presence of silicon prevent the segregation of iron carbides and thus potentiate reacted stable austenite. The shape and amount of graphite inclusions clearly influence the plasticity of the material. This can be attributed to the interruption of the microstructure by the graphite. Behaviour at high strain and strain rates Because of its direct influence on the thermal and mechanical loads on the cutting edge during chip formation, the high strain material behaviour (with respect to both, nominal strain values and strain rates) is crucial. The plastic behaviour under high strain rates is a basic material property for the explanation of wear phenomena. The behaviour at low strain rates is evaluated by compression tests carried out on a conventional press. The results are displayed in Figure 2. With the exception of ADI-900, all materials show a similar increase of compression force depending on the deformation degree. The failure of the ADI-900 specimen resembles that of ferrite ductile iron at a deformation degree of about 40%, yet the compression force is more than twice as high. In contrast to quenched and tempered steels, the plastic deformation phase of all cast irons is interrupted as soon as the first cracks appear. 0% 10% 20% 30% 40% 50% 60% Deformation degree Co mpres s ion forc e / N 0 100 200 300 400 500 600 EN-GJS-400-15 EN-GJS-700-2 42CrMo4+QT ADI 900 : crack initiation : failure Co mpres s ion forc e / N Figure 2: Force-displacement diagram of compression tests. Flow stress curves are generated by applying so-called Split-Hopkinson-Bar tests for quasi-static conditions and strain rates between 10 3 s -1 and 10 4 s -1 . The quasi-static tests lead to results comparable to the outcome of the preliminary tests on the conventional press. Compared to ADI-1200, strain hardening of ADI-900 is more pronounced. At higher strain rates this effect degenerates. This can be attributed to the thermal softening due to plastic elongation. In spite of the high ductility of ADI-900 at quasi-static conditions, the deformability at higher strain rates decreases similarly to other cast iron materials with globular graphite. Micro and nano hardness When evaluating machinability, micro hardness represents a stronger determining factor than global hardness, which includes the hardness of graphite inclusions. In comparison to cast irons with globular graphite, an approximately linear dependence between the mean micro hardness and the nominal tensile strength was identified. Furthermore, an increase of the variance of the micro hardness values at higher strengths could be identified. As individual constituents, austenite and ferrite possess relatively low hardnesses. The strength of the ADI microstructure is caused by its finely-striped structure. The compliance of the austenite and ferrite crystallographic constituents with respect to those of conventional microstructures is evaluated by nano hardness measurements. The austenite in ADI is crystallographically modified by its high-graphite content and its hardness is significantly increased to approx. 70%. Due to the presence of silicon, the hardness in the ferrite phase increases by approx. 25%. Within ADI grades with higher strengths the hardness diverges in places to higher values. This effect can be attributed to inherent iron carbides. The austenite is mechanically and thermally stable due to carbon diffusion during heat treatment, but it can transform into martensite during the plastic deformation phase. With respect to chip formation, this effect can lead to a very high mechanical specific load on the cutting edge. The strain hardening effects described in crush and flow curves can also be detected by micro hardness measurements. Compared to ADI-1200 and other cast iron materials with globular graphite, ADI-900 shows the strongest increase in micro hardness by quasi-static deformation. One explanation is provided by measuring the residual austenite: In ductile ADI grades more austenite can be detected in the microstructure due to an increased graphite content as a result of higher tempering -75- temperatures. This austenite transforms almost completely into martensite during deformation independently of the strain rate. As a result, the micro hardness increases due to a higher level of martensite. 3 DERIVATION OF THE CONSTITUTIVE MATERIAL LAW In order to find an adequate constitutive material law, it is necessary to consider certain basic machining conditions. Thereby the extreme conditions in the chip formation zone such as strain, strain rates and temperature are relevant. According to the fundamental approach of Brodmann [6], several special material laws are developed for this application. These laws are based on a strain-related characterisation and an influence of the strain rates in the form of viscose dampening. It was modified by a term which describes the strain dependency and allows a better translation of the appropriated flow stress curves into the constitutive material law. The suggested material law says )( 1 TKK n < � H � K H � V with m T TT eT 0 )( E < (1) and [MPa] flow stress V 1 K [MPa] material parameter [MPa] material parameter K true plastic strain H H [s -1 ] true strain rate n strain hardening exponent K [Ns/m²] damping constant )(T < temperature function T [K] temperature 0 T [K] default temperature m T [K] melting temperature E [MPa/K] material constant of temperature function. 3.1 Comparison of ADI with materials of similar strength In view of a quantitative comparison, the acquired material parameters are diagrammed depending on their micro hardness in Figure 3. Beside the comparison to perlitic and ferrite ductile iron, the two ADI grades ADI-900 and ADI-1200 are also compared to a quenched and tempered steel 42CrMo4 in different heat-treated conditions. With respect to the hardness behaviour of the parameter K an approximately linear increase can be observed with a slight deviation for the ADI grades. The damping constant progresses exponentially depending on the micro hardness. According to Figure 3, the austenitic-ferrite microstructure of ADI has a decreased damping capacity compared to 42CrMo4. Yet the damping properties of ADI exceed those of steel if they are related to the macro hardness. This is caused by the high damping capacity of graphite and its decreasing influence on macro hardness. Both conventional cast irons and steel show an exponential progression. For ADI a distinct unsteadiness is visible, which can be attributed to the strain hardening coefficient of ADI-900. This value is quadrupled compared to quenched and tempered steels. Even for ADI-1200 this coefficient is twice as high as that of 42CrMo4. 4 SIMULATION The constitutive material law provides input for the approximation of flow stresses. In order to gain a better understanding of the altering strain conditions on the cutting edge due to segmented chip formation, the chip formation process is FEM simulated and a two- dimensional model approach for solving non-linear flow and deformation problems is applied. The programme utilises implicit numerical algorithms and enables the integration of damage criteria or of special flow conditions. The results, i.e. the characteristic chip shape, the dynamic cutting forces and temperatures, are then verified on an experimental basis. 4.1 Segmented chip formation Two criteria are necessary in order to simulate the segmented chip formation process: Firstly a failure factor for the simulation of heterogeneous microstructures and secondly a criterion that initiates cracks on the material surface. In front of the main cutting edge, high hydrostatic pressure and compressive stress in all principle axes are effective. Due to high shear stress and the heterogeneous microstructure composed of globular graphite and a metallic matrix, the chip formation process becomes increasingly unstable and leads to accelerated gliding. In order to simulate this phenomenon with a homogeneous matrix, the constitutive material law has to be supplemented by a failure factor. The material deformation capacity is exhausted as soon as the tensile maximum principle stresses become effective on interface of the workpiece. As a consequence, cracks appear in the shear zone. This effect can be simulated by a factor which sets the fraction of tensile strength in proportion to the highest shear tension. The damage criterion for element erase is determined by this state variable. 4.2 Verification The simulation of the chip formation process shows a distinct analogy to experimental chip root investigations and to all common theories on segmented chip formation. The periodical deflection parallel to the tool face and alienation of the primary shear zone close to the cutting edge is demonstrated in the simulation and can be verified by the split shear zone theory of van Luttervelt [7]. As a result, the wear on the tool face close to the cutting edge Quenched and tempered steel 42CrMo4+QT (several ageing grades) Cast irons with globular graphite GJS-700 ADI-900 ADI-1200 St rain hard ening exp o n en t n 0 0,05 0,10 0,15 0,20 200 300 400 500 600 Vickers hardness HV 300 mN Paramet e r K / G P a 0 0,5 1,0 1,5 2,0 2,5 3,0 200 300 400 500 600 D a mping const a nt K 0 0,05 0,10 0,15 0,20 200 300 400 500 600 Vickers hardness HV 300 mN Vickers hardness HV 300 mN St rain hard ening exp o n en t n St rain hard ening exp o n en t n Paramet e r K / G P a Paramet e r K / G P a D a mping const a nt K D a mping const a nt K D a mping const a nt K Figure 3: Material parameters according to equation 1 depending on micro hardness. -76- increases. This is caused by the adhesion of relatively hot and reactive material. The plasticity of the material and the absence of continuous side gliding lead to the accumulation and compression of material on the tool face. During the subsequent gliding phase, this bond is torn away. The maximum values of the simulated cutting force correlate well to the average of peak values measured with a slim line force sensor mounted between the tool holder and the indexable insert. This assembly shows compared to other force measurement devices a relatively high resonance frequency of about 5 kHz. This enables the system to resolute the chip segmentation frequency by applying a high sample rate. The frequency analysis shows a dominant frequency of 2800 Hz at a cutting speed of v C = 70 m/min and a feed of f = 0,25 mm. This is equal to a period of approx. 0,35 ms during which the chip segment is formed. Therefore the simulated period of T ≈ 0,4 ms corresponds closely to the period gained by simulation (compare Figure 4). Simulation Measurement 0 0,4 0,8 1,2 S p ecifi c cutti ng forc e F c /b / N/mm 0 100 200 300 400 500 600 700 800 0 0,4 0,8 1,2 Cutting time t c / ms Cutting time t c / ms S p ecifi c cutti ng forc e F c /b / N/mm Figure 4 : Simulated and measured cutting forces. The simulation indicates that constant high temperatures are effective only close to the cutting edge. On the chip surface however, altering thermal loads are dominant. Temperature measurements in this area of the rake showed a relatively low temperature as the dynamic thermal loads cannot be resolved. With respect to the zone of constant temperature close to the cutting edge, the measurement shows a good correlation to the results of the simulation. Another effect related to segmented chip formation is adhesive wear. This can be attributed to a strong bonding between the tool and the material during the compression phase followed by the erosion of this bond during the gliding phase, i.e. single grains and small particles of the tungsten carbide are torn out of the tungsten-cobalt composite. The adequateness of the FEM-simulation as a means to analyse the local and temporal resolution of mechanical and thermal loads on the cutting edge during segmented chip formation could thereby be verified. 5 CONSEQUENCES ON TOOL IMPROVEMENTS Cutting tools are exposed to a combination of both high abrasive and adhesive wear as well as to specific loads on the cutting edge when machining ADI. Furthermore, cutting tests showed that an important role for the improvement of tool wear behaviour can be attributed to the design of the cutting edge. An optimal compromise between sufficient stability and a high sharpness of the cutting edge has to be found. On the one hand, breakages should be prevented and on the other hand, the influence of segmented chip formation on the adhesive wear phenomena should be reduced. An optimal cutting tool material combines high abrasive wear resistance as well as very good edge strength. By empirical research the importance of a good coating adherence to withstand high dynamic cutting forces could be identified and underlined. Additionally the coating must prevent the adhesive wear of the tungsten-carbide substrate. Further optimisation steps can be deduced from these demands. Tools for machining steel and conventional cast irons are not suited to the special wear phenomena that become effective when machining ADI. If further improvements are to be achieved, the complete system of tool geometry, tool material substrate and coating must be optimised. 6 SUMMARY Austempered ductile iron (ADI) offers a high potential for future applications in series production. During the ADI machining process, cutting tools are exposed to a combination of both high abrasive and adhesive wear as well as to high specific loads of the cutting edge. The aptitude of ADI to segmented chip formation is even more pronounced compared to other ductile irons with globular graphite; therefore a high dynamic proportion within the cutting forces and a deep crater wear close to the cutting edge can be detected. In compression tests at high strain rates, a disproportional decrease of the high ductility of ADI was identified. The exceptional position of ADI among other ductile irons with globular graphite was elaborated during the search and definition of coefficients for a constitutive material law. Furthermore, a simulation model was developed in order to support the determination of mechanical and thermal loads during segmented chip formation process. 7 REFERENCES [1] Abele, R., et al., 2002, Wear Mechanism when Machining Compacted Graphite Iron. Annals of the CIRP, 51/1:53-56 [2] Hoppe, S., Experimental and numerical analysis of chip formation in metal cutting, 2003, Aachen University, Dissertation [3] Klöpper, C., Untersuchungen zur Zerspanbarkeit von austenitisch-ferritischem Gusseisen mit Kugelgraphit (ADI), 2006, Aachen University, Dissertation [4] DIN EN 1564, 2003, Giessereiwesen: Gusseisen mit Kugelgraphit, Beuth Verlag, Berlin [5] ASTM A 897M-03, 2003, Standard Specification for Austempered Ductile Iron Castings, ASTM, Conshohocken, PA [6] Brodmann, M., 2001, LFW Mitteilung – Schädigungsmodell für schlagartige Beanspruch-ung metallischer Werkstoffe, Aachen University, Dissertation [7] van Luttvelt, C. A., Pekelharing, A. J., 1977, The Split-Shearzone – Mechanism of Chip Seg- mentation, Annals of the CIRP, 25/1:33-38 ure measurements in this area of the rake showed a relatively low temperature as the dynamic thermal loads cannot be resolved. With respect to the zone of constant temperature close to the cutting edge, the measurement shows a good correlation to the results of the simulation. Another effect related to segmented chip formation is adhesive wear. This can be attributed to a strong bonding between the tool and the material during the compression phase followed by the erosion of this bond during the gliding phase, i.e. single grains and small particles of the tungsten carbide are torn out of the tungsten-cobalt composite. The adequateness of the FEM-simulation as a means to analyse the local and temporal resolution of mechanical and thermal loads on the cutting edge during segmented chip formation could thereby be verified. 5 CONSEQUENCES ON TOOL IMPROVEMENTS Cutting tools are exposed to a combination of both high abrasive and adhesive wear as well as to specific loads on the cutting edge when machining ADI. Furthermore, cutting tests showed that an important role for the improvement of tool wear behaviour can be attributed to the design of the cutting edge. An optimal compromise between sufficient stability and a high sharpness of the cutting edge has to be found. On the one hand, breakages should be prevented and on the other hand, the influence of segmented chip formation on the adhesive wear phenomena should be reduced. An optimal cutting tool material combines high abrasive wear resistFundamental Wear Mechanisms when Machining Austempered Ductile Iron (ADI)Austempered Ductile Iron (ADI) is characterised by improved mechanical properties but low machinability compared to conventional ductile iron materials and steels of similar strengths. The mechanical properties of ADI are achieved by a very fine austenitic-ferritic microstructure. However this unusual microstructure significantly affects mechanical and thermal machining properties. A keen understanding for the interactions of microstructure, chip formation, machining properties, cutting material and wear mechanisms is essential for the optimisation of the cutting process. This paper describes material and machining investigations as well as cutting simulations to reveal the wear mechanisms being responsible for the low machinability of ADI.Hybrid thermal, mechanical and chemical surface post-treatments for improved fatigue behavior of laser powder bed fusion AlSi10Mg notched samplesComplex geometries can be produced by laser powder bed fusion (LPBF) techniques in a layer-by-layer manner. These parts exhibit inhomogeneous microstructure and poor surface quality in their as-built state. Performing post-treatments to modify these imperfections can play a substantial role in enhancing the performance of LPBF parts. However, the effects of post treatments on local geometrical irregularities are not still well documented. In this study, four different post-treatments including heat treatment, mechanical and chemical surface treatments as well as their combination were considered. Their effect was studied on microstructure, surface, and mechanical properties of LPBF V-notched AlSi10Mg samples. The as-built samples were subjected to two different shot peening processes (using different Almen intensity, shot diameter, and shot hardness), chemical polishing and electro-chemical polishing, in individual and combined configurations. Comprehensive microstructural characterization was carried out and the surface state of the samples was studied in detail in terms of surface morphology and roughness. In addition, mechanical properties including microhardness and residual stresses were measured and finally the fatigue behaviors of the samples were analyzed and compared at a constant stress level. All post treatments led to improved fatigue life. The combination of the aforementioned post-treatments led to a remarkable fatigue life improvement up to 414 times higher compared to the as-built state.Laser powder bed fusion (LPBF) is one of the most widely used additive manufacturing (AM) technologies, used to fabricate mechanical components with intricate geometries. LPBF metallic materials, due to the complex physical phenomena of melting, solidification and rapid cooling occurring during the fabrication process are characterized with different types of surface and bulk defects Each AM technology can be controlled by a set of specific parameters known as process parameters the alteration of which directly affects the properties of the fabricated material in terms of internal and surface quality The defects of as-built AM material and in particular the presence of irregular and inhomogeneous surface features act as stress concentration sites causing early crack initiation On the other hand, surface post-treatments based on no material removal such as shot peening (SP), induce homogeneous and regular surface morphological features besides other beneficial effects such as surface layer hardening and inducing compressive residual stresses Most studies till now have been dealing with smooth surfaces and standard geometries for AM fabricated samples. However, more intricate geometries, that are indeed the point of strength of AM, should be treated differently as they can respond to the post-treatments in a distinct way different from the smooth surfaces. Notched AM parts are known to represent a more inhomogeneous surface roughness depending on the orientation of the surfaces with respect to the build direction. Past studies have shown that the notch area especially on its downward face contains higher extent of surface imperfections that can have adverse effects on the part's fatigue behavior Following our previous study, which surveyed the effects of different post-treatments of HT, SP and their combination on fatigue strength of notched LPBF samples Gas atomized spherical powder of AlSi10Mg (SLM solutions Group AG, DE) with particle size of 20–63 μm and mean diameter of 46.65 μm was used for LPBF samples manufacturing. shows the nominal chemical composition of the feedstock material. The samples were fabricated via SLM 500 HL systems (SLM Solution Group AG, DE) with Yttrium fiber lasers. LPBF process parameters included spot diameter of 78 μm, laser power of 350 W, and scan speed of 1150 mm /s, layer thickness of 50 μm and hatch distance of 170 μm. In this study, the LPBF process parameters are referenced to a previous work of the authors . The baseplate was pre-heated to 150 °C and the chamber was flooded with argon gas, keeping the oxygen content below 0.2% during the fabrication process. These parameters were optimized in a previous study performed by some of the authors b illustrates the shape and size of the notched fatigue samples.Three different post-treatments were performed . The first one, HT, was aimed at modifying the microstructural and residual stresses, whereas the other two were surface treatments mainly dealing with surface imperfections.T6 thermal treatment as a common treatment for Al alloys was performed on half of the samples according to the time and temperature intervals recommended by Aboulkhair et al. Two different SP treatments were performed using different types of impacting media and also different Almen intensities. represents the details of considered SP processes with steel and ceramic shots applied on both as-built and heat treated samples. Almen intensities were determined according to the SAE J443 standard In addition, two different chemical treatments including CP and ECP were carried out on both as-built and heat treated series. As there is no available data in the literature about the optimized conditions of CP and ECP processes on LPBF AlSi10Mg, different process parameters were considered for both chemical treatments. 12 different conditions were designed for each CP and ECP on as-built state and then the optimized parameters were selected for both treatments in terms of highest roughness reduction.In CP process, two parameters of temperature and time were considered to control the process. The CP experiments were performed using bath of solution of 900 mL H2O + 200 g Nitrate (NO-3) + 375 g NaOH in three different temperatures of 25 °C (room temperature), 50 °C and 75 °C and 4 different polishing times of 60 s, 120 s, 180 s and 240 s. In ECP treatments, two parameters of voltage and exposure time were considered for controlling the process. ECPs were carried out using bath of 400 mL solution with 94% Acetic acid (CH3COOH) + 6% Perchloric (HClO4) acid at three different voltages of 5 V, 10 V and 15 V and 4 different times of 60 s, 120 s, 180 s and 240 s. After the chemical treatments, the samples were immediately washed with water and then ultrasonically cleaned in a solution of 50% acetone and 50% desalinated water. Selected CP and ECP parameters were applied to as-built sample leading to two series of AB+CP and AB+ECP. Then, between the CP and ECP, the treatment that had the highest effect on fatigue behavior improvement of as-built state was chosen to apply on the other sets of heat treated and shot peened samples (mentioned above). reveals the schematic illustration of the applied surface post-treatments in this study including SP, CP and ECP considering details of each process.Monotonic tensile tests were carried out on dog-bone as-built and heat treated sets using three samples per each. The samples were tested following ISO 6892-1 on an MTS Alliance RT/100 machine at a strain rate of 0.7 mm/min up to 2% strain with constant rate of 2 mm/min. To measure the material elongation during the experiment, an extensometer was attached to the samples.Firstly, samples were cut in longitudinal and transversal sections with respect to the build direction and then were impregnated in hot mounting resin. The polishing steps were carried out with final step of polishing using silica suspensions. The polished cross-sections were chemically etched for 20 s in Keller's reagent (95% pure H2O, 1% HF, 1.5% HCl, 2.5% HNO3). The microstructural characterization was carried out using a Nikon Eclipse LV150NL optical microscope (OM, Nikon Corporation, Japan) and a high resolution Zeiss Sigma 500 VP field-emission scanning electron microscope (FE-SEM, Carl Zeiss Microscopy GmbH, Germany) equipped with energy dispersive spectrometry (EDS) and electron backscattered diffraction (EBSD). AZtecHKL software was used to process the EDS and EBSD data.Image based analyses were employed for porosity measurements on the notch area of the samples. Three different surfaces including top (xy-plane), middle (yz-plane) and bottom (xy-plane) surfaces of notch area were considered for SEM observations. Three back scattered electron SEM (BSE-SEM) images were taken from random areas of each polished surfaces. Three additional OM images were used for sub-surface porosity analysis of all sets of samples. ImageJ software Surface roughness measurements were performed via Mahr Perthometer (PCMESS 7024357) equipped with MFW 250 probe with a tip diameter of 5 μm. EN ISO 4287 standard Microhardness tests were performed on finely polished cross-sections of samples using a Leica WMHT30A micro Vickers hardness tester. Loads of 25 gf and dwell time of 15 s were used for each indentation. Measurements were performed on transversal section (xy-plane) and 5 different paths were considered for each sample starting at a depth of 40 μm from the sub-surface towards the core material up to a depth of 740 μm with an interval distance of 50 μm.X-ray diffraction (XRD) was employed to obtain the distributions of residual stresses along a perpendicular path with respect to the build direction starting from top surface. AST X-Stress 3000 portable X-ray diffractometer with CrKα radiation, λK alpha 1 = 2.2898 Å, irradiated area of 4 mm diameter, and sin2(ψ) method, was used. Diffraction angle (2θ) of 139° corresponding to {311}-reflex was scanned with a total of 7 Chi tilts between 45° and − 45° along three rotations of 0°, 45°and 90° with a constant step size of 0.028°. Measurements were applied gradually by removing a very thin layer of material about 20 μm up to the depth of 200 μm; the step size was increased to 40 μm for depths>200 μm; material removal was performed through electro-chemical polishing with a solution of acetic acid (94%) and perchloric acid (6%) at a voltage of 40 V for 45 s. Precision Mitutoyo micrometer (IDCH0530/05060) was used to determine the depth of material removal in each removing step. XRD is the most widely used method for residual stress measurements and also electro-polishing is mostly used for in-depth material removal as this method does not have any considerable effects on inducing other residual stresses if the process parameters are selected correctly Fatigue behavior of LPBF V-notched AlSi10Mg were analyzed via rotating bending fatigue tests at a stress ratio of R = −1 at room temperature and air relative humidity of about 50%. Italsigma (IT) equipment with rotational speed around 2500 rpm was used. In order to compare the influence of the applied post-treatments on fatigue life of the notched samples, fatigue tests were performed at fixed amplitude stress of 110 MPa with a run-out limit set to 6 × 106 cycles for all sets. Three samples were tested at mentioned stress level and the average of obtained fatigue lives of each set is reported. In addition, fractography analysis was carried out on the failed samples STEREO discovery V12 Zeiss.Quasi-static mechanical properties of the AB and AB+HT samples were obtained using tensile test on three samples Different analyses using OM, FESEM, EBSD and XRD were performed for microstructural characterization of the LPBF AlSi10Mg. Firstly, the AB and AB+HT samples were investigated in detail and then the plastically deformed surface layers of the shot peened samples were analyzed. presents the microstructural observations for AB and AB+HT samples. a, reveals the comparison of the general microstructure of the AB and AB+HT samples in transversal (xy-plane) and longitudinal (yz-plane) cross sections obtained by OM. In longitudinal section of AB, the melt pool morphologies and hatching lines are clear and grains elongated along the build direction can be observed. Whereas micrographs from the transversal section of AB, exhibit the melt pool tracks and the inhomogeneous microstructure orientated following the 67° rotation strategy implemented between the subsequent layers. On the AB+HT sample, on the other hand, melt pool morphologies and hatching traces are mostly faded leading to a notable homogeneity in both cross-sections. Also, spherical and irregular pores can be seen in both AB and AB+HT samples.b illustrates the FESEM observations of AB and AB+HT samples on their longitudinal cross-sections (yz-plane). In AB sample, fibrous Si networks surrounded by α-Al matrix can be observed; while in the AB+HT sample, remarkable homogeneity was obtained by transformation of the Si networks to Si spherical particles that were homogeneously dispersed in the Al matrix, as also previously reported by Wei et al. b; these maps clearly indicate the segregation of Si particles spread in boundaries of α-Al matrix in AB+HT sample.c represents the output of EBSD analyses in transversal cross-sections (xy-plane) of AB and AB+HT samples. The band contrast images highlight a remarkable inhomogeneity in grains morphology for the AB sample in comparison with the AB+HT one. As the solidification of grains in LPBF AlSi10Mg is facilitated along the build direction (z-direction) due to epitaxial growth depicts the XRD patterns of all sets of samples and b illustrates the primary and secondary peaks of the XRD patterns. Compared to the as-built and heat treated states, the intensities of the primary and secondary peaks are enhanced in the shot peened samples and the peaks are shifted towards different diffraction angles. The presence of Mg2Si peak confirms the FESEM observations. Mg2Si peak intensities are not so high due to the low content of Mg (0.35 wt%) in the feedstock alloy. In addition texture coefficient analysis was carried out using Harris' equation, which calculates the degree of preferred orientation through ratios between the measured peak intensities and database intensities for the same material c. Texture coefficient of 1 is related to material's database and the texture coefficient of 3.5 is corresponding to the complete preferred orientation in the plane. Peaks of Al (111), Al (200), Si (111), Si (311) and Mg2Si (220) were considered. AB+HT + SP2 sample exhibited the highest texture coefficient with 3.24 in Al (200) followed by AB+HT + SP1, AB+SP2, AB, AB+SP1 and AB+HT samples with 2.53, 1.85, 1.42, 1.1 and 0.64, respectively; this observation indicates considerable shifts from the preferred orientation in the columnar grown considering texture coefficient of 1. In addition, texture coefficients for Mg2Si (220) were only obtained in the heat treated samples (AB+HT, AB+HT + SP1 and AB+HT + SP2).The results of different analyses of crystallographic orientation, grain boundary, recrystallization, Kernel average misorientation (KAM), geometrically necessary dislocation (GND) and strain contouring performed by processing the EBSD results for the surface layer of the shot peened samples are shown in . Transversal cross-section (xy-plane) along the building direction was considered for all shot peened sets. IPF-Z maps reveal the random orientation of the grains in the surface layer compared to the dominant (001) orientation in the core material.Grain boundaries with orientation of 2° < and 5° < were considered for processing. The grain boundary graphs, a higher density grain indicating some extent of grain refinement could be observed near to the top surface of SP series. The observations confirmed the presence of gradient microstructures on SP series starting with ultrafine-grained (UFG) structure (< 0.5 μm) on the top surface layers gradually increasing to reach the original grain size at subsurface. SP1 treatment with higher intensity was found to be more effective regarding in-depth grain refinement on both AB and AB+HT series upto a depth of about 25 and 20 μm, respectively. SP2, on the other hand, showed thinner grain-refined layers with depths of about 12 and 7 μm on AB and AB+HT samples, respectively. It should be noted that both SP treatments were found to be more effective on AB samples rather than the AB+HT ones, due to the lower ductility and conformability of the material in AB condition.Recrystallization graphs depict the distributions of recrystallized, substructured and deformed grains. In AB+SP2, AB+HT + SP1 and AB+HT + SP2 samples, deformed grains are mostly seen on the top surface layer upto the depth of about 15 μm; these samples have relatively high amounts of substructured grains and less scattered recrystallized grains. However, in the AB+SP1, due to the higher kinetic energy of the Applied SP treatment combined with the lower ductility of the base material in AB configuration, a very dense layer of highly deformed grains is generated on the top surface layer, which is gradually decreased through the ending part of the scanned area. In this sample, due to the presence of large number of deformed grains, the density of recrystallized and substructured grains are reduced considerably compared to the other SP series.KAM values can be used as an index of high plastic deformation Strain contouring maps can reflect the plastic strain variations of material after SP. The obtained strain contouring maps indicate that the strain increased from the bulk material to the shot peened surface. The strain values of the localized deformation for AB+SP1, AB+SP2, AB+HT + SP1 and AB+HT + SP2 samples are determined as 9.4, 6.4, 5.7 and 5.1 respectively. High grain refinement and surface layer hardening after SP lead to the increase of the plastic strain; as expected, the treatment with the highest intensity and kinetic energy (SP1) was the most efficient in inducing plastic strain. The variations of strain values are in agreement with the variations in the grain boundaries and misorientation distributions. In addition, it can be observed that the maximum values of strain were in the sub-surface and not exactly on the top treated surface, which is typical for the surface plastically deformed materials . Based on the EBSD results obtained within the scanned area of 32 × 10 μm2, the average grain areas of 2.15, 0.56, 0.86, 5.5, 3.69 and 5.03 μm2 were determined for AB, AB+SP1, AB+SP2, AB+HT, AB+HT + SP1 and AB+HT + SP2 samples, respectively (b). These results confirm grain size increase after heat treatment and the effect of SP in grain refinement. Similar trends were obtained for grain aspect ratio, which were decreased after SP (Formation of columnar grains along the build direction in FCC metals clearly confirm the presence of (001) fiber textures in the both AB and AB+HT samples indicating that the applied heat treatment did not considerably affect the grain orientation, although it increase microstructural homogeneity. However, it can be seen that after SP, the grains are oriented in different random directions with different dominated texture orientations. For instance in AB+SP1 sample, (111) orientation is the dominated with strength of about 1.The porosity measurements were carried out on all sets based on image analysis of BSE-SEM and OM micrographs in the notch area. The analysis was mainly focused on the internal and sub-surface porosities around the notch as presented in , respectively. Similar trends were observed in the top and bottom areas with respect to the notch in all samples although the porosity in top surface was found to be slightly lower. The lowest porosity was detected in the mid area of the notch. OM observations near to the treated surface indicated that the SP treatments had no remarkable influence on the reduction of sub-surface porosities due to the low kinetic energy of the SP process parameters used in this study. However, as reported by Lesyk et al. Quantitative image analysis results for porosity measurements illustrated in indicate a porosity range of 0.40–0.48% for the inner parts of the samples in the notch root zone while in the sub-surface the porosities are slightly higher in the range of 0.51–0.56%.Vickers microhardness tests were performed from the depth of 40 up to 740 μm in the transversal section of all sets as shown in . The AB+HT series exhibit a remarkably reduced microhardness; AB and AB+HT samples showed hardness values of about 105 and 65 Hv, respectively. Similar trend for hardness reduction after thermal treatment was also reported in the study carried out by Zakay Residual stresses distributions were obtained by in-depth XRD analyses along a path perpendicular to the build direction as presented in a. In the AB sample, tensile residual stresses with a maximum of 31 MPa were determined. However, these tensile stresses were mostly relaxed after applying HT. In the shot peened samples, on-surface compressive stresses of −83 and −65 MPa, were continuously raised to reach a maximum value of −157 MPa at the depth of 180 μm and −170 MPa at the depth of 200 μm in AB+SP1 and AB+SP2 samples, respectively. The total depths affected by compressive stresses were measured as 530 μm and 480 μm for AB+SP1 and AB+SP2 samples, respectively. The total depth affected by compressive residual stresses was higher in the AB+SP1 sample in comparison with AB+SP2, caused by the higher Almen intensity of SP1 treatment.On the other hand, in the heat treated condition although more compressive residual stresses are induced after relaxation of the tensile stresses, the results revealed lower surface residual stresses after applying SP on heat treated samples. This can be attributed to the lower hardness and higher ductility of these sets as previously reported by Bagherifard et al. Generally, LPBF materials have poor surface quality due to the formation of irregular surface features. In LPBF parts with complex geometries such as notches, which have downward and upward faces, more surface irregularities are observed especially in the downward surfaces, caused by the particular thermal history and the lower cooling rate depict the surface morphologies of an AB sample in notch area and a smooth area, located 1 cm above the notch. In the smooth area, poor surface quality can be seen clearly. In the notch area, however, three distinct regions of upward face, notch root and downwards face can be identified with specific surface morphologies; the qualitative comparison shows the lowest surface quality for the downward face with a high density of surface irregularities followed by the upward face and then the notch root. c reveals different types of common surface defects including spatter, balling, unmelted and partially melted powders on downward face, which are specified based on the characteristics of surface imperfections in LPBF materials, categorized in the literature To modify the surface morphology and reduce the surface roughness, initially CP and ECP treatments were performed on the smooth area of AB samples, using 12 different parameter settings. The treated samples were screened in terms of qualitative analysis of surface morphology and quantitative effects on surface roughness reduction, and the best combination were applied on notched samples for further analysis. represent the SEM observations of CP and ECP treated samples using different combinations of temperature or voltage and exposure times. It can be observed that by increasing time and temperature of CP, more regular surface morphologies were obtained with almost no signs of surface imperfections that were observed in the as-built condition. However, some oxidized areas and indications of pitting corrosion can be observed on the samples' surface, especially at higher exposure time. Similarly, in the ECP series, by increasing the time and voltage of the process, the efficiency of the treatment in inducing surface regularity became more notable.The obtained surface roughness parameters and mass loss are illustrated in c and d for ECP samples, respectively; roughness values in terms of Rz, Rzmax and Rt are presented in ), for all measurements. Surface roughness was continuously reduced by simultaneous increasing of considered process parameters both for CP and ECP. The best combinations induced a roughness reduction of 4.49 μm in terms of Ra to 3.19 μm after CP with parameters of 75 °C and 240 s; the same initial Ra was reduced to 2.96 μm after ECP with a parameter combination of 15 V and 240 s. Therefore, these two chemical treatments were selected as optimum conditions of CP and ECP, respectively to be used for further analysis. These treatments correspond to mass loss of 8.65 and 1.87% for the CP and ECP series, respectively. It was reported by Scherillo The selected CP and ECP treatments were performed on AB notched samples to assess the contribution of these treatments to their fatigue behavior. depicts the SEM micrographs taken from the notch area in AB, AB+CP and AB+ECP samples. The observations confirm that after CP, almost all the surface imperfections existed in the as-built state were eliminated; however, due to the high rate of corrosion, few new surface irregularities were formed on the surface of the material corresponding to the laser tracks. However, in the AB+ECP sample, many the surface defects were removed with no deep corrosion pits observed on the surface. In comparison with the effects of CP and ECP on the smooth area (a), the SEM observations at the notch area indicate that the efficiencies of the chemical treatments were decreased when applied on the notch area. 8.55 and 1.95% mass loss were obtained for AB+CP and AB+ECP samples respectively considering initial mass of the samples (b). The diameter of the notch root in AB sample was reduced from 8.13 mm to 7.92 mm in AB+CP sample due to high degradation. On the other hand, after applying ECP as conformed by the low percentage of mass loss, the diameter of the notch root reduced to 8.03 mm, which is very close the as-designed dimension.Rotating bending fatigue tests at fixed stress amplitude of 110 MPa were carried out on AB, AB+CP and AB+ECP series. Three samples were used for each set and the obtained fatigue lives are reported in c. The results indicate considerable fatigue life improvement after chemical treatments. Average fatigue life of AB sample with about 1.27 × 104 cycles, increased up to about 1.41 × 105 and 2.15 × 105 cycles in AB+CP and AB+ECP samples, respectively. The results indicate that among the two considered chemical treatments, ECP is more effective than CP in terms of fatigue life improvement of LPBF AlSi10Mg samples. Compared to the AB sample, the notch root diameter is reduced by 2.5 and 1.2% in AB+CP and AB+ECP samples, respectively. The notch root diameter in the AB+CP sample is lower than the diameter in the as-designed geometry. This can result in further stress concentration in the notch area. In addition, AB+CP samples exhibited higher surface roughness compared to the AB+ECP series. Therefore, AB+ECP samples with lower surface roughness and less stress concentration in the notch root could reach higher fatigue life with respect to the AB+CP series. Hence, in the next step, the ECP treatment (15 V and 240 s) was applied to the heat treated and shot peened samples to also evaluate the effect of hybrid surface treatments on the fatigue performance of the samples. represents the initial states of surface morphologies of shot peened sample as well as the heat treated one in the cylindrical part of the samples. Considering AB and AB+HT series (HT has no effect on surface morphology variation) with very high surface irregularities, it can be observed that in the shot peened samples surface imperfections and irregularities are remarkably reduced and the surface morphologies are highly modified to become more regular compared to the initial AB state. Surface morphology modification of LPBF AlSi10Mg after applying SP was also investigated by Uzan et al. b presents the SEM surface morphologies of the samples initially heat treated and shot peened samples after applying hybrid treatments with the final step of ECP. The results indicate high efficiency of ECP on modifying the surface morphologies. The previously formed dimples caused by the shot impacts were almost completely removed through the application of the chemical treatment. In addition, in both AB and AB+HT samples, the traces of pitting corrosions could be observed in some parts. However, in the shot peened samples, the higher surface hardness led to a higher resistance to pitting corrosions represents the surface roughness parameters for all treated samples and compares them with the initial as-built state. Considering the effects of single step treatments, the AB and AB+HT samples represented the identical surface roughness of 4.48 μm and 4.43 μm, respectively in terms of Ra, both were reduced after ECP to 2.96 μm and increased after SP1 and SP2 to 4.61 and 8.60 μm, respectively. SP2 which uses harder and smaller shots enhanced the roughness more than SP1, despite its lower Almen intensity.In the hybrid treatments obtained by combination of HT and SP, the microstructural homogenization and improved ductility after the heat treatment, led to the formation of deeper dimples by the impacting shots during SP. Hence, AB+HT + SP1 and AB+HT + SP2 samples showed slightly higher surface roughness of 6.67 and 8.91 μm in terms of Ra respectively, compared to the not heat treated counterparts.Contrastingly, due to the material removal nature of ECP, the samples subjected to hybrid treatments with the final step of ECP showed considerable roughness reduction representing Ra values of 3.29, 3.44, 2.61, 4.12 and 4.21 μm for AB+SP1 + ECP, AB+SP2 + ECP, AB+HT + ECP, AB+HT + SP1 + ECP and AB+HT + SP2 + ECP samples, respectively. The final material removal step was able to reduce the surface roughness compared to the as-built condition in the AB+SP2 and AB+HT + SP2 samples, which had relatively high surface roughness before chemical treatment.b depicts the measured mass loss after applying ECP on heat treated and shot peened samples. AB+HT + ECP and AB+ECP with 1.85 and 1.95% show the highest mass loss among the considered samples. The heat treated samples, on the other hand, show lower rate of material removal due to their different material properties. illustrates the SEM micrographs of notch area before and after applying ECP on samples treated by SP, HT and their combinations. Application of ECP mostly removed the dimples generated by SP treatments leading to a homogenous surface with relatively low roughness, especially in AB+HT + SP1 + ECP and AB+HT + SP2 + ECP samples. Also, in AB+HT sample, the initial surface irregularities such as partially melted and unmelted powders were mainly removed. reveals the measured notch root diameter and mass loss of the samples after performing HT, SP1, SP2, ECP and their combinations on notched fatigue samples. It can be seen that due to the low rate of material removal (maximum of 1.95% for AB+ECP) after ECP, the notch root diameters were reduced to around 8 mm in all samples but not lower. The results of the fatigue tests performed at a stress level of 110 MPa are shown in b and the average fatigue lives and the corresponding fatigue life improvements are presented in . The results indicate notable fatigue behavior improvements after applying hybrid post-treatments including HT, SP and ECP together. The effect of sequential application of the post-treatments can be analyzed by comparing the average fatigue life of AB sample that is improved up to 2, 23 and 44 times in AB+HT, AB+HT + SP2 and AB+HT + SP2 + ECP samples, respectively.Each of the applied treatments has a distinct contribution to the fatigue life improvement. HT improves the fatigue life through microstructural homogenization and increasing ductility. SP contributes by three major effects including surface layer grain refinement and hardening, inducing considerable compressive residual stresses and also modifying the surface morphology to a more homogenous state. Furthermore, ECP enhances the fatigue life due to the remarkable roughness reduction and altering the surface morphology to a smoother surface with almost no imperfections. Dealing with single treatments, SP (SP2 more than SP1) has the highest performance followed by ECP and HT on average fatigue life improvement. Accordingly, a hybrid treatment with the sequential application of HT, SP and ECP, can exploit all the positive effects simultaneously and highly improve the fatigue life of AM samples with irregular geometries.Recent studies have revealed that although theoretically the notch root accounts for the highest stress concentration in the component, it is not always the fatigue failure site. Fatigue fracture has been reported to mainly initiate on the downward face of notch due to the high surface roughness and the presence of potential crack nucleation sites The relative height of failure site was calculated for all sets as an average of 15 measurements per each set. The h0value was considered as 4.8 μm for all samples as extracted from the initial design. presents the variations of the fracture surface location in different treated sets compared to the AB (with h=hAB) and AB+HT (with h=hAB+HT). The schematic illustration of failure site definition and the determined results in terms of relative height of notch for all sets in the same notch acuity of ξ=0.3 are depicted in b. The results indicate that, the post-treatments clearly alter the location of fracture site and shift it closer to the notch root direction. Considering the sole effects of applied post-treatments, SP2, ECP, HT and SP1 have the highest efficiency in shifting the fracture site closer to the notch root, respectively. The hybrid treatments are even more effective with the AB+HT + SP2 + ECP sample representing the smallest distance between the fracture site and the notch root.The applied post-treatments changed the microstructural, surface and mechanical properties as well as the fatigue behavior. Sole effects of each post-treatment on microstructural characteristic, ductility, hardness, residual stresses, surface morphology, roughness and fatigue behavior are summarized in . It can be observed that, all the considered treatments led to fatigue life improvement. SP2 and SP1 have almost similar contributions on hardening and inducing compressive residual stresses. The AB+SP2 sample reached to the higher fatigue life despite its higher roughness compared to AB+SP1 series; this could be caused by its more homogenous surface morphology compared to the AB+SP1 samples. These results indicate that besides the surface roughness reduction (which can be obtained by ECP), surface morphology (that can be described by roughness parameters) has considerable influence on fatigue behavior of the notched samples. Therefore, as expected applying hybrid treatment, which includes both these beneficial features leads to a more remarkable fatigue life improvement (as obtained for AB+HT + SP2 + ECP sample).Besides improving the mechanical performance of the LPBF material, producing the final part in an economic manner is also quite substantial. Two different analyses including sole cost (money-time) analysis and cost-performance analysis were performed. In these analyses, time, costs, and especially performance of the effect of post-treatment on fatigue life improvement were considered as the key variable parameters. Each parameter was rated into five comparative levels of very low, low, moderate, high and very high. Scores of 1–5 were assigned to each level for developing a semi-quantitative matrix according to the common approach used in cost and risk analysis c represents the assigned scores to each post-treatment and the obtained final score and ranks. The comparative levels of time and costs corresponding to each parameter were estimated based on the authors' experience. The levels of performance were assigned based on the experimental fatigue test results. The obtained scores of money-time analysis were used as cost parameter in the cost-performance analysis. Considering both economical and fatigue improvement factors, it can be observed that applying sole SP can be the best option with first rank followed by ECP and hybrid treatment of HT + SP + ECP in the second and third ranks, respectively. The hybrid treatment of SP + ECP commonly had the fourth rank.It should be mentioned that the presented results in this part are to compare the performance of the applied treatments only in economic terms, regardless their eventual contribution to fatigue performance. However, as it was mentioned before, the single treatments of SP, ECP and HT have the highest performance on average fatigue life improvement, respectively.LPBF materials show numerous imperfections in terms of microstructure and surface quality in the as-built state; these could lead deteriorate mechanical performance especially regarding fatigue behavior. Post-treatments can play substantial role to overcome these defects. In the present study, the effects of four different post-treatments including heat treatment, mechanical surface treatment and chemical surface treatment as well as their combination as hybrid treatments were investigated on microstructural and mechanical properties of LPBF V-notched AlSi10Mg samples. The effects of heat treatment, two different shot peening treatments (with different Almen intensity, shot diameter and shot hardness) as well as two processes of chemical polishing and electro-chemical polishing were studied, through experimental analyses. Based on the obtained results the following conclusions can be drawn:The surface morphology imperfections were more notable on the downward skin of the notch.Inhomogeneous microstructure observed in the as-built state was considerably modified after the heat treatment. This modification improved the ductility of the material up to 5 times. Furthermore, the tensile residual stresses were relaxed after heat treatment.Surface layer grain refinement was obtained by shot peening treatments. This led to increased microhardness on the top surface up to 13%. Considerable compressive residual stresses were induced in the surface layer up to about −175 MPa with a corresponding depth of 200 μm.Microhardness and residual stress measurements on shot peened samples revealed that the higher Almen intensity led to higher surface hardness (122 vs. 117 Hv after shot peening treatment with steel and ceramic shots, respectively) and higher on-surface compressive residual stresses (−83 vs. -65 MPa after shot peening treatment with steel and ceramic shots, respectively).Shot peening increased the surface roughness; however, it had a significant effect on homogenizing the surface morphology and obtaining more regular surfaces.Ceramic shots with smaller diameter and higher hardness compared to the larger steel shots were more efficient in modifying the surface morphology (8.60 μm after shot peening with ceramic shots and 4.61 μm for shot peening with the steel shots vs. 4.48 μm for the as-built condition in terms of Ra).Chemical treatments efficiently removed the surface irregularities of the as-built material and reduced the surface roughness from 4.48 μm to 3.19 and 2.97 μm in terms of Ra after chemical and electro-chemical polishing, respectively.Fatigue life was increased up to about 17 and 11 times after electrochemical and chemical polishing, respectively.All the applied post-treatments enhanced the fatigue performance of the notched samples. Considering single post-treatments, shot peening with ceramic shot was found to be the most efficient followed by electro-chemical polishing and heat treatment with life cycles increased up to 167, 17 and 2 times higher than as-built condition, respectively.The hybrid treatments led to additional fatigue life improvement. Combination of heat treatment + shot peening with ceramic shots + electro-chemical polishing had the highest effects on fatigue life improvement leading to 414 times higher fatigue life compared to the as-built state.Considering the cost-performance analyses with respect to the time, money and performance, to obtain the highest fatigue efficiency in the V-notched AlSi10Mg samples in an economical manner, shot peening can be the best option ranked first followed by electro-chemical polishing in the second rank.Erfan Maleki: Conceptualization, Investigation, Formal analysis, Validation, Visualization, Writing – original draft. Sara Bagherifard: Conceptualization, Methodology, Writing – review & editing, Supervision. Farshad Sabouri: Investigation. Michele Bandini: Resources. Mario Guagliano: Resources, Supervision.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.Flexural behavior of HSFRC with low reinforcement ratiosTwo-point loading tests were conducted in order to examine the flexural behavior of high strength steel-fiber reinforced concrete (HSFRC) elements with a minimum amount of reinforcement. Eleven beam specimens were cast and tested. They included high strength concrete (HSC) beams with and without steel fibers, which were given at a constant volumetric ratio of 0.75%, as well as control normal strength concrete (NSC) beams (without fibers). It was found that the addition of fibers to flexural members with a minimum longitudinal reinforcement caused a more brittle behavior compared to the same specimens that did not include fibers. This result suggests that the minimum longitudinal reinforcement ratio in flexural HSFRC members should be higher than in conventionally reinforced members (i.e., without fibers) in order to achieve sufficient ductility. The moment capacities of the beams that were tested were compared to several available theoretical models. For the HSFRC specimens, the predictions according to Lim et al.’s model and to the ACI544 model, which was originally developed for NSC and was adapted here for HSC, were closest to the current experimental results.In the last two decades, concrete technology has enabled the manufacturing of concrete that has a uniaxial compression strength of about (or more than) 100 MPa, implementing conventional mixing technologies. This concrete has been commonly referred to as high strength concrete (HSC). Subsequently, study of the structural behavior of HSC has increased during this period together with the research and development on the material properties (e.g., Research on the structural behavior of reinforced HSC members that include steel fibers (or high strength steel-fiber reinforced concrete, HSFRC) has been relatively limited. It has been established that steel fibers affect the mechanical properties of NSC (e.g., Following its study, the use of FRC or SFRC has thus been, to a relatively large extent, incorporated into design guidelines, as outlined by Fischer This paper describes and discusses tests of lightly reinforced HSC and HSFRC beams. These tests are part of a comprehensive research program aimed at the study of HSFRC structural elements and the definition of tools for their static design. This part of the study of lightly reinforced beams involved the prediction of their ultimate moment capacity and the issue of the minimum flexural reinforcement required for HSFRC. The current experimental results and observations show that, for HSFRC, an increase in the amount of flexure minimum reinforcement is required. The prediction of HSFRC moment capacity is examined by comparison of the maximum loads that were measured in the current experiments with the above models.Two-point loading tests of HSC beam specimens and of NSC control specimens, which were designed to fail in flexure, were conducted. The test plan included lightly reinforced beam specimens with and without steel fibers. The specimens’ details are given in , which also show the test set-ups. The specimens were marked () N/H#-F2-0/1, according to the concrete strength and the use of fibers, where “N” and “H” denote NSC and HSC and “0” and “1” denote specimens without and with fibers. The beams had a 200 mm×300 mm cross-section and longitudinal reinforcement that consisted of 8 mm deformed steel bars with reinforcement ratios ρf of 0.28% and 0.56% for the HSC specimens and 0.18% for the NSC control specimens (see ). The concrete cover was 15 mm in all specimens.All the beams had a span of 3.5 m. The test plan consisted of two phases: the specimens of the first phase (specimens No. 1–6, ) had a shear span of 1.5 m. Following an interesting and somewhat surprising phenomenon that was observed in the first phase (as described later in the text), additional tests were designed and performed. The second-phase specimens included beams with a shear span of 1.5 m and with a higher amount of reinforcement (ρf=0.56%; specimens No. 7–8, ) and with shear spans of 1.0 m and 1.25 m and ρf of 0.28% (specimens 9–11, The tests were performed under a stroke control of the testing machine. The stroke rates were 0.5, 0.2 to 0.25, 0.5 to 0.7 and 1 mm/min within mid-point deflections of 0–3 to 10, 3 to 10–30, 30–50 to 80 and 50 to 80–150 mm (respectively). These rates were varied according to the (pre-test evaluation of the) specimen’s mid-point deflection, where the relatively faster stroke rates were applied during the relatively small and large deflections and the lower rate was applied during yielding of the reinforcing steel (depending on the specimen’s stiffness; see also Measurements included the total load, the deflections at mid-span and under the loads and the longitudinal strains in the constant moment region of the beam. The longitudinal compressive strain at mid-span was measured with strain gages that were glued to the top of the beam, while strains at the mid-height and bottom of the beam were measured with LVDTs that were attached to both sides of the beam (). Tests No. 9–11 (of the second phase, ) included additional top strain gages and lateral and vertical LVDTs in order to measure the distribution (and magnitude) of these specimens’ curvatures along their center part (The HSC and NSC mixes included the following ingredients: 1082 kg/m3 basalt and 872 kg/m3 dolomite aggregates (respectively) of 22 mm maximum size, 461 kg/m3 cement type CEM I-52.5N and 246 kg/m3 cement type CEM I-42.5N (respectively) and ordinary sand (704 kg/m3 and 1110 kg/m3, respectively). The HSC and NSC water–cement ratios (w/c) were 0.35 and 0.61 (respectively) and their dosages of super plasticizer (SP — type “Rheobuild 2000”) were 14.9 l/m3 and 4.8 l/m3 (respectively). Additionally, the HSC mixes included 65 kg/m3 of silica fume. All the mixes were prepared in a forced pan mixer. The concrete compressive strengths at 28 days are given in Hooked-end steel fibers of two different lengths were used (Dramix RC-65/60-BN and RC-65/35-BN) at a constant ratio of 0.75% by volume (60 kg/m3 mixture). They had similar aspect ratios (67 and 64) and, according to the manufacturer, these fibers have a minimum tensile strength of 1000 MPa. The two types of the fibers that were used were part of the overall test program (which also included toughness tests as described by The 8 mm deformed steel bars in specimens No. 1–8 () had a yield and ultimate stresses of 480 MPa and 720 MPa (respectively) and an average maximum elongation of 12%. In specimens No. 9–11, the 8 mm bars had a yield and ultimate stresses of 616 MPa and 734 MPa (respectively) and an average maximum elongation of 5.5%.The specimens’ moment–mid-span deflection curves are shown in and their modes of failure and maximum capacities are given in . A flexural mode of failure, characterized by flexural cracks, was observed. However, failure occurred in the flexure HSC specimens that included fibers when one of the 8 mm longitudinal steel bars ruptured (specimens No. 5–6, ). It was also observed that, at failure of the beam specimens, the fibers were pulled out (rather than ruptured). The tests of specimens H4-F2-0 (which did not include fibers) were stopped when the stroke of the testing machine reached its full range (150 mm) without concrete crushing or rupture of the reinforcement. Note that, due to the curvature of the beam, the “mid-point displacement” shown in is actually larger than the stroke. The current test results of HSC beams with minimal longitudinal reinforcement ratio yielded higher structural displacement ductility of the specimens without steel fibers than those with fibers (‘ductility’ refers here to the deflection at failure-to-deflection at yield ratio). As can be seen in , the H4-F2-0 type specimens without fibers had a lower yield moment at a mid-point deflection of about 5 mm. However, these specimens reached their ultimate state at a mid-point deflection of more than 140 mm, as compared to (only) 50 mm in the H5-F2-1_35/60 type specimens (No. 5 and 6, ) that included steel fibers (and had the same amount of longitudinal reinforcement, ρf in ). That is, the specimens without the fibers had a displacement ductility ratio of 30 (150/5) while the specimens with the fibers had a ductility ratio of (only) 6.3 (50/8). The deflections at yield (the denominators) in these calculations of the ductility ratios correspond to measured yield strains of 0.0024 (480/200 000, where 200 000 MPa is the steel’s Young’s modulus). These strains were measured by the horizontal LVDTs located at both sides of the beam, at the level of the tension steel (see The decreased ductility of the lightly reinforced specimens that included fibers is also shown in which shows pictures of specimens at the end of the test. It was observed that at the beginning of the tests, under relatively low loading levels, there were several cracks that started to develop in these specimens’ constant moment region, between the two loads. However, as can be seen in , while in the beams that did not include fibers (the control NSC and the HSC beams) several, well distributed flexural cracks continued to develop when the deflection of the beam was increased, there was a single major crack that developed in the beams with the fibers (H5-F2-1_60, ), which led to rupture of a reinforcing bar at a relatively low mid-span deflection.This unexpected result led to the design of further tests (phase II in ) with an aim of examining the (negative) effect of the fibers on lightly reinforced beams: Specimens No. 7 and 8 (H8-F2-1, ) were designed with an increased amount of longitudinal reinforcement (ρf=0.56%, ) to verify that this effect decreases as the reinforcement ratio increases. Specimens No. 9–11 () were designed with the same minimal, 0.28% reinforcement ratio to verify the phenomenon of decreased structural ductility of lightly reinforced HSFRC beams at different shear spans (i.e., at a=1.0 and 1.25 m, The deflections at ultimate state of the H8-F2-1 specimens were 99 mm and 115 mm () which, together with the deflections at yield, correspond to an average ductility ratio of 8.6 (107/12.5). Specimens No. 9–11 (), which had the same amount of minimum flexural steel (0.28%) but lower shear spans, exhibited the same phenomenon of decreased displacement ductility when steel fibers are applied.The phenomenon, which was observed in this series of tests, of decreased ductility of lightly reinforced beams that include fibers is demonstrated in , which shows the moment–deflection curves of the specimens with longitudinal reinforcement ratio ρf of 0.28%, with and without fibers. It can be seen in the figure that, at ultimate state, the specimens without the fibers reached deflections that were about three times larger than those of the same specimen type (that is, that were made of the same concrete mixture and had the same reinforcement ratio and static scheme), which included fibers.It was observed in the first phase of the tests that the reduced ductility ratios of the lightly reinforced beams that included fibers was associated with the crack distribution of these specimens (as described above and seen in ), which caused the development of a relatively small region with higher longitudinal steel strain. This eventually caused rupture of one of the reinforcing bars. Therefore, the additional tests of beams with a 0.28% longitudinal reinforcement (No. 9–11, ) included an increased number of (horizontal and vertical) displacement transducers, as shown in . The aim of this transducers’ setup was to obtain measurements of the curvature magnitude and distribution along the central part of the beam in order to verify the correlation, which was indicated in the previous tests, between the decreased ductility and the application of fibers in lightly reinforced beams. The plots of curvatures in of specimens No. 9–11 show their distribution along the beam, where the coordinate origin was set at the beam’s center and normalized with respect to (L−2a)/2 (L is the 3.5 m span and a is the shear span, ). In order to compare curvatures from specimens with different shear spans (and different constant moment regions) the curvatures were normalized with respect to the beams’ dimensions and so they were multiplied by (L−2a)⋅a/L. The values of the curvatures were obtained from the horizontal measurements as follows: shows that top and bottom strains were measured along the central part of specimens No. 9-11. The top strains, εTOP, were read directly from strain gages and the bottom strains, εBOT, were calculated from measurements of the LVDTs that were located at the tension steel level. Hence the nominal vertical distance from the top to the bottom strain readings was the effective depth d (the calculations employed the actual distance d∗, which varied according to the actual final location of the bottom transducers). Thus, the average curvature φ at the middle of each horizontal LVDT (located below the middle of the strain gage that was glued to the top of the beam, that the curvature of the specimen without fibers (H4-F2-0_4) was not only larger than that of the specimens with the fibers (H5-F2-1_35_3&4) but was better distributed along the beam. The curvature distribution along the specimens that included fibers supports the visual observations from the tests. That is, it shows a localization of the so-called plastic hinge within a limited length of the beam, which is equal to about (L−2a)/2, compared to at least a twice larger zone in the beam that did not include fibers (The result described above of reduced displacement ductility in lightly reinforced beams that include steel fibers is somewhat surprising, yet consistent — it was obtained in all the HSC specimens that included fibers. An increased bond of the longitudinal rebars caused by the confining action of the steel fibers may provide an explanation of this phenomenon of localization (as described in the above section). Note that, for the same amount of fibers and a higher amount of longitudinal steel, the relative confining effect of the fibers is expected to be reduced, thus reducing the localization effect and increasing the ductility, as observed in the H8-F2 specimens (Commonly, the minimum area of the flexural reinforcement is required to prevent sudden failure under service loads (at the moment of concrete rupture). The result of the current tests of reduced ductility in lightly reinforced SFRC beams suggests that the presence of fibers in flexure members that include a relatively low (or minimum) amount of longitudinal reinforcement may cause a decrease in their structural ductility and therefore, in addition to the criterion of the minimum moment bearing, this inferior structural behavior should also be considered. Hence, the minimum longitudinal reinforcement ratio in flexural HSFRC members should be higher than in conventionally reinforced members (i.e., without fibers) in order to achieve sufficient ductility. It is interesting to note that Fantilli et al. This criterion of a sufficient ductility ratio can be related to the length of the cracked region of the beam (’length of the plastic hinge’) — relative to its span. As indicated above, this length was observed to be longer in the beam specimens without fibers and exhibited larger mid-point deflections (i.e., higher ductility). The requirement for minimum flexure reinforcement for HSFRC members to satisfy this criterion can be formulated from moment–curvature relationship calculations, considering cracked and uncracked zones along their spans. Thus, the H8-F2-1 type specimens, which included steel fibers, had a higher amount of flexure reinforcement (ρf=0.56%, ), a corresponding higher moment capacity (as expected) but also higher ductility compared to the H5-F2-1 type specimens (). Note that the relatively large mid-point displacement of the H8-F2-1 specimens indicates (together with the yield point displacement) their higher ductility, according to the above definition of ductility, as well as ductility that relates to the beam’s energy absorption (i.e., defined by the area under the moment–deflection curve).The maximum values of the measured load Pmax () were compared with their corresponding values from several theoretical models that predict the moment capacity of steel-fiber reinforced concrete (where the relation M=Pa was applied). It is noted that, in order to compare the theoretical predictions to the experimental results, the stress of the reinforcing steel fsk is taken in the following models as equal to the steel’s ultimate strength (given above, in Section 2.3 on Materials). The following models were used to calculate the moment capacity of the beams that were tested: According to the ACI 544 committee where d is the effective depth (273 mm, see ), x is the depth of a rectangular stress block (found by equating the internal tension and compression forces) and σt is the tensile stress of the fibrous concrete, given by where lf/df, ρf, and FBE are the steel fiber’s aspect ratio, percent by volume and bond efficiency factor (respectively). FBE varies from 1.0 to 1.2, depending on the fiber characteristics ) is the distance from the extreme compression fiber to the top of the tensile stress block of fibrous concrete, given by: , y is the distance from the extreme compression fiber to the neutral axis (where x=β1y, and the coefficient β1 is equal to 0.85 and 0.65 for concrete strengths of 30 MPa and 100 MPa, respectively), εf is the fiber’s tensile strain which corresponds to the fiber stress at pullout σf (i.e., εf is equal to σf/Es with Es=200000MPa). Assuming that, on average, pullout is half of the fiber’s length (lf) (which also takes into account the anchorage effect of the fiber’s hooked ends) yields: Here, the fiber-HSC bond strength τf is taken as 4.15 MPa (with the coefficient 0.00772) incorporates a 2.3 MPa bond strength of NSC (and see also Imam et al.’s model in the following text)., which was developed for NSC, refers to σt which incorporates a factor for a typical bond stress of the fiber’s τf that was taken as 2.3 MPa (for NSC where the fibers factor F is given by the following expression , Vf is the fiber’s volume fraction (which is Vf=ρf/100) and that the fiber bond efficiency factor ηf in this model is equal to 1.0 for hooked-end steel fibers with the heights e and x calculated from Eq. and from equilibrium (respectively), according to the above assumptions.The model that was proposed by Lim et al.), where h is the height of the cross-section and x is the height of the equivalent concrete stress block. Another assumption of this model is that the concrete compressive stress fcyl is multiplied by 0.9 (vs. 0.85 in the ACI544 model). Thus, the moment capacity is given by: where the fibrous concrete tensile stress σt depends on the fiber’s dimensions, mechanical properties and efficiency. For rounded fibers with a diameter df, the stress σt is given by: depends on the fiber’s ‘critical length’ (e.g., see has been adjusted here for the higher fiber-HSC bond strength of 4.15 MPa ), similarly to Imam et al.’s model (Eqs. ). Thus this multiplication is equal to 0.00772/2.3×4.15=0.014.The test results were also compared to the capacity that is predicted by the ACI318-02 formulation The theoretical moment and corresponding load bearing capacities of the beams that were tested in this work are given in . It is noted that calculations of the moment capacity in all the theoretical models that were selected are based on the classical compressive stress block plus the additional capacity that is contributed by the steel fibers. Hence, it can be seen that, for the specimens that did not include fibers, these models yield the same prediction (c shows that, for the beams that included steel fibers, on average the Imam et al. model yielded un-conservative predictions, and the predictions according to the models of ACI544 and Lim et al. were closest to the current experimental results. As expected, the ACI318 model, which does not take into account the fiber’s contribution, yielded average conservative predictions. Note that the relatively un-conservative results that were obtained are because all the predictions according to the above models were made with the reinforcing steel’s ultimate strength rather than its lower yield strength (see Section Two-point loading tests of HSFRC specimens with low amounts of flexural reinforcement were performed. The test plan included specimens with and without steel fibers and beams with a relatively increased reinforcement ratio.The test results showed consistent reduced displacement ductility of the lightly reinforced beams that included steel fibers compared to similar specimens that did not include fibers. Higher structural ductility was obtained in beam specimens with fibers, in which the amount of flexural reinforcement was increased. Hence, these results show that the minimum longitudinal reinforcement ratio in flexural HSFRC members should be higher than in conventionally reinforced members (i.e., without fibers) in order to achieve sufficient ductility.Comparison of the measured moment capacity with theoretical models showed for the beams that included steel fibers that, on average, the predictions according to Lim et al.’s model and to the ACI544 model, which was originally developed for NSC and was adapted here for HSC, were closest to the current experimental results. For the specimens that did not include fibers, these models yield the same predictions. As expected, the ACI318 model, which does not take into account the fiber’s contribution, yielded average conservative predictions.Pan-African granite emplacement mechanisms in the Eastern Desert, EgyptLes plutons granitiques pan-africains tardifs de Fawakhir, Um Had et Um Effein dans le centre du Désert Oriental égyptien sont des petits corps allongés à circulaires formés de syéno- et monzogranites roses et de monzodiorites grises. Les observations structurales aux contacts intrusifs de ces plutons indiquent qu'il s'agit de lames intrusives marginales subhorizontales, correspondant à des injections de magmas le long de structures planes préexistantes incluant foliations minérales et failles de charriage. L'espace nécessaire à la mise en place de ces plutons granitiques a été accomodé par le bombement des roches environnantes encaissantes dont les contacts extérieurs sont dans le style des laccolites. D'une manière générale, le pluton granitique d'Um Had possède une forme phacolitique contrôlée par un couloir cisaillant mylonitique bombé et à pente sud, qui sépare les roches gneissiques (préservées comme couer du pluton) des unités charriées peu métamorphiques. Les caractéristiques structurales de ces plutons sont en accord avec une mise en place dans la croûte la plus supérieure, à une profondeur où σ3 est vertical. Cette mise en place s'est probablement produite après une transpression NE-SO plutôt que ans un environnement.One way coupled fluid flow-stress modelling3D geomechanical modelling for CO2 geologic storage in the Dogger carbonates of the Paris BasinCO2 injection into a depleted hydrocarbon field or aquifer may give rise to a variety of coupled physical and chemical processes. During CO2 injection, the increase in pore pressure can induce reservoir expansion. As a result the in situ stress field may change in and around the reservoir. The geomechanical behaviour induced by oil production followed by CO2 injections into an oil field reservoir in the Paris Basin has been numerically modelled. This paper deals with an evaluation of the induced deformations and in situ stress changes, and their potential effects on faults, using a 3D geomechanical model. The geomechanical analysis of the reservoir–caprock system was carried out as a feasibility study using pressure information in a “one way” coupling, where pressures issued from reservoir simulations were integrated as input for a geomechanical model. The results show that under specific assumptions the mechanical effects of CO2 injection do not affect the mechanical stability of the reservoir–caprock system. The ground vertical movement at the surface ranges from −2 mm during oil production to +2.5 mm during CO2 injection. Furthermore, the changes in in situ stresses predicted under specific assumptions by geomechanical modelling are not significant enough to jeopardize the mechanical stability of the reservoir and caprock. The stress changes issued from the 3D geomechanical modelling are also combined with a Mohr–Coulomb analysis to determine the fault slip tendency. By integrating the stress changes issued from the geomechanical modelling into the fault stability analysis, the critical pore pressure for fault reactivation is higher than calculated for the fault stability analysis considering constant horizontal stresses.One way coupled fluid flow-stress modellingmeaneffectivestress=(σ′11+σ′22+σ′33)/3 (Pa)Kronecker symbol = 0 if i |
≠ |
j, =1 if i |
= |
jAlthough most of the electricity produced in France is of nuclear origin, an increasing concern with respect to climate change has emerged among public authorities. The ANR-Geocarbone PICOREF project (), supported by the French National Agency of Research, aims to evaluate a site for CO2 storage. In the framework of the GESTCO European project, two deep formations of the Paris Basin, namely Dogger and Keuper, have been identified as relevant candidates for CO2 storage (). In the South-East of Paris, the previously mentioned formations are well known within both the petroleum and geothermal industries. As there are industrial sources of CO2 present in the area, a set of oil field structures were described as valid potential sites for a pilot project. In the Saint-Martin de Bossenay (SMB) oil field, several aspects of the storage technology could be investigated and in particular, the geomechanical behaviour of the CO2 storage.The underground storage of CO2 is liable to cause deformation of the reservoir and surrounding formations, induce ground movements and trigger microseismic events. The objective of the current work is to understand the evolution of the mechanical behaviour of the reservoir–caprock system under oil production followed by CO2 storage. In addition, the stress changes in and around the reservoir, as well as the induced ground movement are investigated.This paper reviews the set-up of the geomechanical model developed over a stratigraphic column for the Saint-Martin de Bossenay oil field, located in a carbonate reservoir of the Paris Basin. A geomechanical model was built using a compilation of relevant information about in situ stresses and rock mechanical properties based on geomechanical studies, and geological, geophysical and reservoir engineering models. The model also describes faulted reservoir structure and the gridding of the overburden environment. With this 3D geomechanical model, numerical modelling was performed to predict the induced deformations and in situ stress changes due to oil production and a CO2 injection scenario, which maximises the stored CO2 amount. The approach for this modelling implies a “one-way” coupling of flow and stress models. Finally, the modelled mechanical effects were analysed in terms of reservoir–caprock formation stability and, preliminary analyses of fault stability in the reservoir are presented.To solve the coupling between fluid flow and geomechanical problems, different approaches can be used (). The fully coupled and partially coupled approaches can be used to solve the stress-dependent reservoir problem.The fully coupled approach simultaneously solves the whole set of equations that govern the hydromechanical problem. This approach leads to consistent descriptions, however, the hydraulic or geomechanical mechanisms are often simplified by comparison with conventional uncoupled reservoir and geomechanical approaches.The partially coupled approach is based on an external coupling between conventional reservoir and geomechanical simulators. The stress and flow equations are solved separately for each step, however information is transferred between the reservoir and geomechanical simulators. This approach has the advantage of being flexible, relatively small in terms of CPU time, and benefits from the latest developments in physics and numerical techniques for both reservoir and geomechanical softwares. Different coupling levels can be achieved for the partially coupled methodology. The partial coupling is described “explicit” if the methodology is only performed once for each step and “iterative” if the methodology is repeated until convergence of the stress and fluid flow on the basis of a common porous volume.“One way coupling” is the simplest partially coupled approach in which the pore pressure history issued from a conventional reservoir simulation is introduced as input into the geomechanical equilibrium equation. In practice, the pore pressure computed by reservoir simulation is introduced in poroelasticity equations to deduce stress and deformation. This coupling is easy to implement and still includes interesting physics (In this paper, we focus on the effect of oil production and particularly CO2 injection with low pore pressure gradient. When pore pressure variation is low, the effect of deformation on porosity is expected to be insignificant and therefore a fully coupled model is not necessary. Hence, the one way coupling technique has been chosen (The thermal effects due to the injection of cool CO2 have been neglected in this work as well as geochemical effects on the rock stiffness.The Saint-Martin de Bossenay (SMB) field is located in a carbonate reservoir. Oil is trapped in the uppermost limestone unit of the Dogger Formation, at a depth around 1450 m. The initial pressure and temperature of the reservoir are around 14.5 MPa and 65 °C, respectively. The reservoir is sealed by a Marly caprock of Callovo-Oxfordien age (). The reservoir itself is subdivided into four reservoir limestone units: A and B called Dalle Nacrée, and C and D, called Comblanchien. A very compact limestone layer with very low permeability separates units C and D. These reservoir units range in thickness from 2 to 10 m for units A, B and C, and from 35 to 40 m for unit D. The average porosity ranges from 7 to 15% and permeability ranges from 10 to 100 mD. A major normal fault of the Paris Basin, the Saint-Martin de Bossenay fault, crosses the formations 2 km east of the field structural axis. It does not intersect the anticlinal structure and consequently should not be reached by stored CO2 during injection. Nevertheless, its geometry is represented in the grid block built to simulate potential mechanical effects due to the heterogeneity of mechanical properties. Smaller faults are not accounted for in the grid block.A reservoir simulation taking into account oil production and CO2 injection was performed with a reservoir simulator developed by IFP (), and based on a finite volume reservoir simulator. The reservoir model aims to provide pore pressure variations during a scenario of oil production followed by a scenario of CO2 injection. As discussed previously, the reservoir simulation does not provide any information about the mechanical behaviour, therefore, a coupling between fluid flow and stress modelling must be considered. For mechanical simulations, the reservoir part must be embedded by surrounding formations (overburden, underburden and sideburden) to account for the added influence of adjacent geologic structures. Such a model requires the definition of the geomechanical domain, constitutive laws and associated rock properties for the reservoir and the surrounding formations, and initial and boundary conditions. The 3D geomechanical modelling was performed with a general finite element program (Abaqus Software).The dimensions of the grid block used by reservoir engineers for the reservoir simulations are 19,900, 38,900 and around 60 m in the X, Y, Z directions, respectively. This model was built with 38 × 60 × 4 grid cells. The oil reservoir, 3900, 6900 and 60 m in the X, Y, Z directions is located in the centre (). Lateral extensions are necessary for better flow boundary conditions (). The reservoir model gridding used Cartesian cells and was not constrained by the main fault, while the geomechanical model required the definition of finite element connectivity with a continuous mesh taking into account the main fault path. In order to simplify the geomechanical model, only the regional normal fault was considered to constrain the model gridding.To build the entire geometry of the geomechanical model, a geological modelling tool (GOCAD) was used, along with the available data: position of seismic horizons at well SMB17, four maps issued from 1980 2D seismic interpretation and two sonic logs (). The overburden structure was split into 6 layers and the underburden into 5 layers according to the lithology and P-wave impedance variations (the impedance is the product of the P velocity and the density issued from log data). The reservoir model was used to define 6 surfaces: top of unit A, B, C, base of unit C, top of unit D and base of unit D. This leads to a model with a total of 17 horizon surfaces (), which are cut by the main fault trace. With the geometry of the 3D block, a grid of the entire structure was defined by the modelling tool in corner point geometry (b). This mesh with corner point geometry was not suitable for a finite element program because it requires finite element connectivity. For this, an interface between the geological modelling tool (GOCAD) and the finite element program (Abaqus) was developed. This leads to a geomechanical model split into two parts: an eastern part with 39 × 60 × 15 grid cells and a western part with 17 × 60 × 15 grid cells. The total dimensions are 19,900, 38,900 and 3000 m in the X, Y, and Z direction, respectively.In the geomechanical model, the fault is seen as an interface element between eastern and western parts. From one side to the other, the stresses and displacements of the fault plane are continuous. This fault representation is simplified, but still considers a discontinuous mesh due to the fault throw that ranges from 50 to 100 m. Using cohesive elements would have been more accurate to take into account the gouge zone material; however, cohesive elements combined with an interface element consume too much CPU time. Moreover, the pore pressure variation is not supposed to reach the fault zone so the stress changes should be small near this region, therefore using an interface element is thought adequate for this study. To investigate this point, complementary 2D simulations were carried out in order to compare the results with interface and cohesive elements in our case study.The material constitutive laws (elasticity) and properties were related to the rock types present in the geomechanical domain. In the overburden and underburden, 11 mechanical rock types were assigned per layer with respect to the geological description and one mechanical rock type was assigned for the reservoir. To assign poromechanical properties to each mechanical rock type, field data was used. Additional core measurements are preferred whenever possible to constrain the poromechanical properties for the considered field. Nevertheless, some correlations can be used for rocks if cores are not available. P velocity and density logs were used to infer the poromechanical properties (). The S velocity log was not available for this field but using data acquired from a field very close, to the North of Saint-Martin de Bossenay the same ratio Vp/Vs on the P velocity data () was applied to extrapolate the S data to the SMB field.P and S velocities and density logs were used to compute undrained dynamic elastic moduli:Eudyn=ρVS23VP2−4VS2VP2−VS2andνudyn=VP2−2VS22(VP2−VS2),Static elastic properties were then estimated from these dynamic elastic properties. This methodology was previously used for the geomechanical modelling of an underground gas storage reservoir () and follows two steps. First, to obtain drained dynamic elastic moduli from undrained dynamic elastic moduli, a saturation correction must be applied. For this, the Biot-Gassman's equation is commonly used (Kddyn=Kudyn−Kfl(1−Kudyn/Kmat)2(1−Kfl/Kmat)ϕ−Kfl/Kmat(1−Kudyn/Kmat)withKudyn=Eudyn3(1−2νu),Before gas injection, the saturating fluid in and around the reservoir is water, therefore the fluid bulk modulus (Kf) is equal to 2250 MPa. The Voigt–Reuss–Hill's model () is useful to estimate matrix bulk modulus (Kmat) when mineral content is available. If mineral content is not available, matrix bulk modulus is extrapolated knowing global lithology and clay volume.The drained static Young's modulus Edstat was then calculated from drained dynamic bulk modulus Kddyn using empirical relationships (). These relationships were established by analyzing static Young's moduli Edstat as a function of dynamic Young's moduli Eddyn from laboratory data. Two types of rocks were distinguished: “soft rocks” for which Edstat<15GPa and “hard rocks” for which Edstat>15GPa and for each category empirical relationships were given:forsoftrocks:Edstat=0.4145Eddyn−1.0595(GPa)andforhardrocks:Edstat=1.1530Eddyn−15.1970(GPa)With this drained static Young's modulus, the drained static bulk modulus is given byStudies on a variety of rock types show static values of Young's moduli are often lower than dynamic determinations for sedimentary rocks (). Static value of Poisson's ratio may be equal or higher than dynamic one. Unfortunately, the difference is neither systematic nor consistent with the rock or investigator, so it was assumed in our study that the static Poisson's ratio would be equal to the dynamic one.This approach gives an approximate value for Young's modulus and Poisson's ratio. Poromechanical properties for all rock zones estimated by the previously explained methodology are summarized in A failure criterion defines a domain in the stress space outside of which the rock cannot withstand the load. Rock failure strongly depends on the nature of the rock, its initial porosity and the loading path followed. Failure criterion for limestones is commonly addressed by Mohr–Coulomb's approach:where τ and σ′ are the shear stress and normal effective stress on the physical plane through which material failure occurs, c′ and φ′ being the cohesion and the angle of internal friction of the material. Using deviatoric stress (q) and mean effective stress (p′), Mohr–Coulomb's criterion is also expressed asSome empirical correlations for the cohesion and the angle of internal friction were derived from in-house IFP carbonate data and literature data (). c′ appears to follow an exponential trend with porosity while φ’ trend is linear:with ϕ expressed in %, c0 |
= 40.3 MPa, αc |
= 0.054, αf |
= −0.893° and βf |
= 49°. showed that there is a variation interval for c′ and φ′ due to quality of regression such asc′min=c′1.7andc′max=c′×1.7φ′min=φ′−7°andφ′max=φ′+7°For the Dogger limestone with 15% porosity, we obtain the following values for c′ and φ′:c′=16.1MPawithc′min=9.5MPaandc′max=27.4MPa,φ′=33.8°withφ′max=28.8°andφ′max=40.8°.We select the failure criterion that provides the weaker uniaxial compressive strength (Rc), given byand Rcmin |
= 30.8 MPa obtained for c′ = 9.5 MPa and φ′ = 40.8°. With this selected failure criterion, A |
= 18.2 MPa and B |
= 1.66.The failure criterion for the Marly caprock is issued from Bure laboratory site in the east of the Paris Basin (). Some triaxial measurements were performed on Callovo-Oxfordian samples from Marly Formations and the cohesion was determined to be c′ = 9 MPa and angle of internal friction to be φ′ = 20°.The information available about in situ stress for the Saint-Martin de Bossenay field comes from 1960 and 1980 data and is sparse and some is of questionable quality. To characterize the in situ state of stress, we determined the orientation of the three principal stresses and the magnitude of minimum horizontal and vertical stresses. An estimation of the maximum horizontal stress magnitude was made by using literature data from the same basin and formations.This paper assumed that one of the principal stresses has a vertical orientation such that the magnitude of vertical stress is determined by the weight of the overlying rock layers. We obtained the magnitude of vertical stress by integration of density logs. To determine the direction of the maximum horizontal stress, breakouts from wells SMB18 and SMB17 were analysed. These symmetric spalled regions are formed at various depths on the wellbore wall where the compressive stress concentration exceeds the shear strength of the rock. The axis defined by the breakout corresponds to the azimuth of the minimum horizontal stress. High-resolution Dipmeter Tool (HDT) and Stratigraphic High resolution Dipmeter Tool (SHDT) data were available to identify the breakouts. The azimuth of the maximum horizontal stress was found to be N 150° which is the orientation of the maximum horizontal stress at other sites in the Paris Basin (To determine the magnitude of the minimum horizontal stress, stimulation tests with or without hydraulic fracturing were analysed in 9 wells. shows the pressure measured during stimulation versus flow rate. The values are normalised by the vertical stress magnitude. The maximum injection pressure is considered as the upper limit for in situ minimum horizontal stress. The total stress ratio is varying between 0.6 and 0.85 and consequently, for geomechanical modelling σh/σv |
= 0.7 was chosen.According to the data obtained on the Bure site located 100 km east, the anisotropy of horizontal stresses in the eastern part of the Paris Basin can be very high and dependant on the lithology (). The ratio σH/σh in the Dogger layers at a depth around 500 m can take values from 1.5 to 2. Generally the stress anisotropy tends to decrease with depth for the same formation. In the absence of direct or indirect measurements on the SMB site, it was assumed that the magnitude of the maximum horizontal stress is close to the vertical stress, which is in agreement with the condition of strike-slip faulting. In order to constrain the magnitude of the maximum horizontal stress, it would be necessary to perform minifrac tests and anisotropy logging () however, we can consider the data sufficient for this first simulation.As the lateral boundaries of the geomechanical domain were set far from the reservoir zone, the normal displacements applied to the lateral boundaries (sideburden) were fixed to zero. The vertical displacement at the bottom of the geomechanical model was also fixed to zero.A computed initialization was performed to reach a mechanical equilibrium between the applied boundary conditions and the initial state of stress in the structure.The pore pressure change was computed by reservoir simulation performed at IFP with the following history (). The Saint-Martin de Bossenay field was in production from 1958 to 1995 and abandoned until 2006. The field was reconsidered and the production started again with the same wells. The period of production from 1958 to 1995 and subsequent abandonment were accounted for in reservoir simulations for history matching. Scenarios of CO2 injection combined with oil production were then simulated. In one of these chosen scenarios, 5 years of oil depletion and 10 years of CO2 injection into the reservoir formation were simulated with an additional oil production during CO2 storage (however, optimization of CO2 injection for EOR was not considered). In this simulation, CO2 was injected at injector wells SMB15 and SMB17 into the unit D of the reservoir formation and oil was produced during 5 years at 12 producer wells. The pore pressure variation at injector wells is given in , and the pore pressure evolution at the top of reservoir unit D during these 15 years of exploitation is given in At injector well SMB17, the initial pressure was 14.7 MPa and at the end of oil production (1799 days ≈ 5 years), the decrease in pressure was around 1.8 MPa. CO2 injection began at 1919 days (≈5.5 years) and was simulated during 10 years until 5511 days (≈15 years), in order to increase stored volume. The additional oil production was stopped according to a well closure criterion given by each producer well gas oil ratio (GOR). With this simulation, the additional oil produced was 81,000 m3. At the end of CO2 injection (5511 days ≈ 15 years), the stored volume was 3.52 Mt and the pore pressure reached around 18 MPa. The maximum pore pressure variations at the SMB17 and SMB15 well locations were respectively 4 and 4.7 MPa during the CO2 injection phase.One of the major purposes of this study was to investigate the stress changes in and around the reservoir and particularly the induced ground movement in this inhabited zone. Saint-Martin de Bossenay village is located in the northern part of the field close to the SMB17 injector well. With this feasibility study we wanted to evaluate ground movement. Under our rock properties assumption, behaviour laws and in situ stresses, the 3D geomechanical modelling addresses the deformation evolution of the structure from the reservoir to the surface. The 3D geomechanical analysis of the system was carried out using the pressure history using one-way coupling. shows the pore pressure changes at the top of unit D for six different time periods (t |
= 0 day, t |
= 1615 days ≈ 4.5 years, t |
= 1799 days ≈ 5 years (end of depletion), t |
= 1919 days ≈ 5.5 years (start of CO2 injection), t |
= 2467 days ≈ 7 years and t |
= 5511 days ≈ 15 years (end of CO2 injection)). presents maps of the vertical displacement variations at the top of reservoir and at the surface during the CO2 injection phase (between times 1799 days ≈ 5 years and 5511 days ≈ 15 years). These variations were obtained by subtracting the final vertical displacement from vertical displacement at the end of depletion. We observe that the injection of CO2 during 10 years induces a maximum vertical displacement of approximately +6 mm at the reservoir top (To illustrate the ground movement history during depletion and CO2 injection, the vertical displacement at the surface is also drawn along a north to south line (). During oil production (between times 0 and 1799 days ≈ 5 years), as the pore pressure decreases, the subsidence reaches a value of −2 mm. During CO2 injection (between times 1799 ≈ 5 years days and 5511 days ≈ 15 years), as the pore pressure increases, the vertical displacement at the surface moves from −2 mm to reach a maximal value of +2.5 mm. During the CO2 injection phase, the total vertical displacement variation is +4.5 mm because the reservoir production and CO2 injection displacements are associated with elastic strain. shows the vertical (top figures) and minimum horizontal effective stresses (bottom figures) at the top of unit D for the end of production when the pore pressure is at its minimum value, and for the end of CO2 injection when the pore pressure is at its maximum value. The changes in in situ stresses are not necessarily correlated with the pore pressure changes () because of the heterogeneity of the mechanical properties of the different layers and the influence of the structural shape of the reservoir. This mechanism was well studied for the depletion of reservoirs (). The stress-depletion response indicates that the in situ stress field may not remain constant during fluid production but may rather evolve in time and space, controlled by the evolution of pore pressure and by the field structural geometry. The magnitude of these stress-depletion response (pore pressure stress coupling ratio expressed as the ratio between the variation of total horizontal stress and the variation of pressure, β |
= (Δσh/Δpp) depends on geological conditions and vary between 0.5 and 1.2 (Analytical techniques may be used to estimate the magnitude of poroelastic stressing at local scale. It follows from linear poroelactivity that a reservoir will behave under uniaxial strain conditions such that the pore pressure stress-coupling ratio equals:where b is the Biot's coefficient (or effective stress parameter); b is usually assumed to be 1, but for sandstone this is not always the case (Substituting values for Biot's coefficient (b |
= 1) and Poisson's ratio (ν |
= 0.29), Eq. indicates that β would be approximately 0.59 for Saint-Martin de Bossenay field. Eq. can be used for analytical estimates of the pore pressure stress coupling ratio but the difference in the numerical approach is that we computed the spatial evolution of effective stresses including field geometry.As an illustration, stress changes were analysed for one grid element in the reservoir unit D, at the SMB17 well location, when the expected increase and decrease in effective stresses are the highest. At this element, during the depletion period, the pore pressure decrease is Δpp |
= −1.8 MPa and the pore pressure/stress-coupling ratio was found around β |
≈ 0.57. During CO2 injection, the pore pressure increase is Δpp |
= +4 MPa and the pore pressure/stress-coupling ratio was the same, because of the linear elasticity assumption. During these two phases (depletion and injection), the total vertical stresses are almost constant within the reservoir far from the model boundaries. In our case study, in this particular location, the analytical approaches provide a good approximation but it is not always the case if the development of the horizontal stress is controlled by nearby faults or if the rock mechanical behaviour is not elastic (To analyse the effective stress path, results are plotted on a Mohr diagram (σ′, the normal effective stress and τ, the shear stress) using the failure properties c′ = 9.5 MPa and φ′ = 41° as explained previously (Reservoir depletion causes an increase in the effective stresses throughout the reservoir. The increase in the effective vertical stress is higher than the increase in the effective horizontal stresses. Consequently, the shear stress throughout the reservoir also increases, as illustrated by the effective stress path plotted in (from “Initial state” circle to “End of depletion” circle). However, at the same time, the stress path moves away from the estimated Dogger limestone failure line, meaning that the stress development is not critical.In contrast with the depletion period, the CO2 injection period causes a decrease in the effective stresses throughout the reservoir and the shear stress throughout the reservoir also decreases as illustrated by the effective stress path plotted in (from “End of depletion” circle to “End of CO2 injection” circle). The stress path moves towards the estimated Dogger limestone failure line but the stress development does not become critical for the stability of the reservoir formation.Previously, we have seen that under a linear elasticity assumption the changes in in situ stresses are not critical for the reservoir and caprock formation stability at the well location. Another geomechanical factor affecting injection of CO2 is fault reactivation (). Inducing slip on an inactive fault provides a possible path for leakage. Slip will occur on a fault when the maximum shear stress acting in the fault plane exceeds the shear strength of the fault. The major SMB fault is located away from the reservoir but there are several other minor and subseismic faults in the field. For these minor faults, a fault stability analysis was conducted.The risk of fault reactivation was calculated using the frictional failure law described by the Mohr–Coulomb criterion (Eq. (13)) (). This criterion is used to characterize the fault strength, where τ is the critical shear stress for slip to occur, c′ is the fault cohesion, μ is the fault friction coefficient (μ |
= tan |
φ′, where φ′ is the fault friction angle), σ’ is the normal effective stress and pp is the pore pressure in the fault plane. The magnitude of the normal stress across the fault and the shear stress were calculated through 3D relationships established by a change of Cartesian reference system from the stress tensor across any fault.The faults are often assumed to be cohesionless and the friction coefficient is typically in the range of μ |
= 0.6–0.85 (). Lower friction values are found in faults that contain clay minerals (Computations of pore pressure levels required to cause faulting were conducted for the Saint-Martin de Bossenay field using the stress results from the 3D geomechanical modelling at SMB17 well location for any fault dip and azimuth. The aim of this simple analysis was to show the critical pore pressure for all possible fault orientations, in an area of the reservoir where induced stress changes were expected to be the highest. The total vertical and the maximum horizontal stresses are approximately 31 MPa, while the minimum horizontal stress is 20.4 MPa at a pore pressure of 12.9 MPa at the end of depletion ( “End of depletion” circle). We determined from the geomechanical modelling an approximately linear pore pressure/stress coupling ratio of β |
= 0.57 for both horizontal directions. The Mohr–Coulomb criterion was calculated for μ |
= 0.6 and a fault cohesion assumed to be 0. illustrates the fault orientations (azimuth and dip) and associated changes in pore pressure that could lead to fault reactivation. Faults are plotted in lower hemisphere stereonets as poles to planes. The fault reactivation propensity was computed first assuming constant horizontal stresses (β |
= 0) (a) and then taking into account the pore pressure/stress coupling ratio β |
= 0.57 (b). In these critical pore pressure estimates, we dealt with shear slip occurring along a fault as a result of increased fluid pressure (β |
= 0) and as a result of poroelastic stressing (β |
= 0.57). Hot colours indicate that smaller increases in pore pressure could result in slip. The pore pressure required to reactivate the fault is different in both cases. When the total horizontal stresses are assumed constant, the pore pressure increase required to reactivate fault is Δpp |
= 2.5 MPa while the pore pressure increase is Δpp |
= 13 MPa when the pore pressure/stress coupling ratio is β |
= 0.57.The methodology presented in this paper shows the complexity of the 3D geomechanical-modelling set-up. Input parameters and the identification of proper physical mechanisms are important issues to evaluate mechanical behaviour. Nevertheless, the lack of accurate geomechanical data makes this feasibility study quite simple: we performed a 3D geomechanical model analysis assuming linear elasticity, laterally homogeneous rock properties and a coarse model, using available data. However, the results of this modelling give a scale of sizes of the potential geomechanical effects due to CO2 injection.Only one set of parameters was tried: those described in this paper. Since the geomechanical parameters are not accurately constrained, it could have been interesting to perform a sensitivity analysis in order to obtain a range of variation in deformation and stresses. However, this kind of analysis is CPU consuming because several hours are necessary to run this previous case with a parallel version of Abaqus. Furthermore, to constrain poromechanical properties, porosity and facies description could have been considered in order to account for possible heterogeneity. Nevertheless, this must be handled with precaution because a huge lateral variation of elastic properties from a finite element to another could induce some stress concentrations due to numerical resolutions.The reservoir–caprock formation stability and the fault stability analysis might be improved by studying more than one measuring point. To investigate further, more accurate data, dedicated to a potential CO2 storage project must be acquired and applied to the methodology.A geomechanical research program is required to ensure that injection related fluid pressure increases do not affect the mechanical stability of reservoirs and caprocks. Thus, in order to understand the evolution of the mechanical behaviour of the reservoir–caprock system, a 3D geomechanical model was developed over a stratigraphic column of a site located in the Paris Basin. This model required the determination of the in situ stresses and knowledge of the rock mechanical properties. In depleted oil and gas fields, geomechanical modelling is needed to update the changes in in situ stresses due to oil production.In this paper, the stresses and strains in the subsurface, caused by oil production and CO2 injection, were computed in a 3D numerical geomechanical-modelling analysis of the Saint-Martin de Bossenay field in the Paris Basin based on the available geological, geomechanical, geophysical and reservoir engineering data. This study illustrates a practical application of geomechanical numerical modelling tools for predicting the mechanical impact of a realistic CO2 injection scenario into a geological reservoir of the Paris Basin.The results show some realistic stress and strain variations that can be compared to those obtained in studies on underground storage of natural gas (). The results also indicate that the shear stress decreases in the reservoir during injection and the mechanical effects induced by the pore pressure increase do not affect the mechanical stability of the reservoir and caprock.Experimental research on shape and size distribution of biomass particle► We researched the size distribution and shape of four kinds of biomass particles. ► Cumulative mass fraction has a distinct linear relationship with particle size. ► Different kinds of biomass particles are coincident well in size distributions. ► The aspect ratio of biomass particles decreases to 2.50 with the decrease of particle size.Particle shape and particle size distribution are both important factors which could influence the physical properties of granular materials. Due to the high content of cellulose, hemicellulose and lignin, biomass material is anisotropic in spatial structure, which induces the evident difference in mechanical property in different directions. In this paper, we studied the particle shape of four kinds of biomass quantitatively by images analysis, and researched the particle size distributions by sieving method. The results showed that biomass particles have large aspect ratio which decreases with particle size. In the larger size, aspect ratios of different kinds of biomass are obviously different. But all decrease to 2.50 with the decrease of particle size. Moreover, the cumulative mass fraction has a distinct linear relationship with particle size, and different kinds of biomass particles are coincident well in size distributions.As the fossil energy is non-renewable and the environment pollution is more and more serious, it is necessary to develop and utilize the renewable and cleaning energy. As a renewable energy, biomass which can be converted into gaseous Due to the high content of cellulose, hemicellulose and lignin, the shape of biomass particle is extremely irregular. Thus the influence of biomass particle shape cannot be ignored in particle transport, mixing and fluidization. Furthermore, various particle shapes result in different particle specific surface area, which are essential to the processes of heat and mass transfer Particle size distribution plays significant roles in flow ability and other properties, and even a small change in particle size can cause significant alterations in the resulting flow ability Considering the importance of particle shape and size distribution in the utilization of biomass, the particles shape and particle size distribution characteristics for different kinds of biomass material were studied. This work would conduce to the research on the related physical properties of biomass particles. This paper is organized as follows. After this introduction, experimental materials and methods are shown in Section , experimental results of particle size distribution and particle shape are discussed. Finally, conclusions are given in Section Four different kinds of biomass materials, pine, beanstalk, rice straw and reed were chosen in this study, as shown in . Pine was from one furniture factory and the others were obtained from the suburb of Shanghai in China. Before being ground, pine, beanstalk, rice straw and reed were pretreated in size and the average particle sizes fed into the pulverizer are 60 mm × 10 mm, 100 mm × 7 mm, 100 mm × 6 mm and 100 mm × 10 mm, respectively. All these biomass materials were just air-dried under the sun and the moisture contents (wet basis), as shown in , were measured at 105 °C by an infrared moisture meter MA150 which is produced by Sartorius of Germany.Biomass materials were ground in a pulverizer FW177 for 3 min, which is a batch knife mill and produced by the TEST Corporation of China. And the obtained biomass particles were sieved into five different particle size ranges: 83–106 μm, 106–150 μm, 150–180 μm, 180–300 μm and 300–425 μm. The particle mass in every size range was weighted to obtain the particle size distribution. For each kind of biomass material, the same experiment was carried out for three times and the average value was considered as the final result.In each size range, 300 biomass particles were chosen randomly, and their images with 300 times magnification were obtained by a microscope Nikon E200. The Image J software was used to measure these images to get the true length and width of each particle. The measurement of length and width of biomass particle, as shown in , is according to the Heywood’s description By measuring particle images photographed by the microscope, the average widths, average lengths and average aspect ratios of biomass particles in five different size ranges were obtained, as shown in , where the average A was not the ratio of average length to average width but the average value of 300 particles’ A., the average widths of biomass particles are all in the ranges of sieve sizes in spite of the size range of 83–106 μm, but the average lengths almost all exceed the sieve sizes. It indicates that the traditional sieve method would not present the real sizes of biomass particles well owing to the needle shape. In the size range of 83–106 μm, the average widths are all smaller than the minimum sieve size 83 μm, which is largely due to that the powder agglomeration in small size partly affects the sieving process.Compared with coal particle which would approximately be spherical, biomass particles have large aspect ratio and needle-like shape. As shown in , the average aspect ratio of different kinds of biomass particles exhibits the same trend that the average aspect ratios reduce with the decrease of particle size. And the discrepancy in aspect ratio between different kinds of biomass particles also decreases with particle size. In the size range of 300–425 μm, the average aspect ratios of pine, beanstalk, rice straw and reed are 3.01, 7.51, 9.31 and 9.98, respectively. But all decrease to 2.50 with the decrease of particle size.In this study, it was found that the breakup of biomass particles was directional obviously, which would conduce to forming larger aspect ratio. And such directional breakup would be related with the anisotropic structure of biomass material. is the scanning electron microscopy (SEM) images of biomass particles with 500 times magnification in the size range of 83–106 μm. It can be found that there are very obvious grains on the rupture sections of biomass particles. During plants growing, some tubular structures form, such as vessel and sieve tube which transport water and inorganic salts to help plants growing. After plants die, these tubular structures will form these grains to bring on the anisotropy of biomass material in structure and strength.Furthermore, it is also found that the average aspect ratios of pine particles are smaller than the other three biomass materials in the larger size ranges, but this difference decreases with particle size. Such difference would be related to organization structures which are dependent on the whole growth of plants. Beanstalk, rice straw and reed are all annual herbaceous plants. In the whole growth process, the division direction of stem cell is mainly along the growth direction, which causes the difference of cells bond strength in different directions. The vascular bundles in stem are mainly concentrated in the outer epidermis , it could be found that there are obvious dissimilarities in shapes before biomass materials being ground, but the microstructure of different biomass materials does not show evident difference. With the decrease of particle size, the structure difference of biomass particle in two directions would reduce, as well as the structure difference of different biomass materials, since the skeleton structures are all cellulose in the microcosmic point of view. Such decreases in structure difference could explain the experiment results that the average aspect ratios of different biomass particles reduce with the decrease of particle size and are almost equal to 2.5 in the size range of 83–106 μm.The mass fraction of biomass particles and the maximum difference between the average value and experimental results in each sieving size range are both shown in . In this paper, cumulative mass fraction w was used to present particle size distributions. And particle sizes were dealt with dimensionless by defining D as the ratio of particle size d to the average particle size da to enhance the universality of the experiment results, as shown in the following equation:where d is the upper limit of sieving size, da is calculated as follows:where mi is the mass of particles in the size range i, di is the average of the upper limit and lower limit of the size range i, and m is the total mass of particles., for biomass particles, a good linear relationship between cumulative mass fraction and particle size can be found, as shown in Eq. , with R (correlation coefficient) >0.99.Obviously, the particle size distributions of different kinds of biomass present a good coincidence, though these biomass materials are extremely different in shapes before being ground, as shown in As a comparison, Beisu coal, which is a widely used bituminous coal in China for entrained flow gasification, was prepared two particle samples. One was prepared in the lab and the other was obtained from a coal gasification factory. The size distributions of coal particles were measured by a particle size analyzer Malvern Mastersizer 2000. As shown in , particle size distributions of Beisu coal are similar to the usual form of particle size distribution. With the decrease of particle size, cumulative mass fraction presents an s-type change. And compared with the gently changing at the beginning and end of the curve, the decline rate is relative fast in the middle part of curve. However, the size distribution curve of biomass particles is almost a straight line. It is difficult to rule out the possibility that the change rate of cumulative mass fraction with particle size will tend to be slower in the smaller or larger particle size. But the facts that different biomass materials display a good coincidence in particle size distribution and cumulative mass fraction has a distinct linear relationship with particle size in such wide size range are unexpected and worth studying.The difference in particle size distribution between biomass and coal would be correlative with the process of particle breakup, which might be influenced by the properties of material. Because of the anisotropy in spatial structure, biomass particles present needle shape. In addition, the pulverizer FW177 used in this study is a knife mill and the grinding structure is shown in . Both these two reasons would make the breakup of biomass particle be a process existing directional difference. Differently, Coal is a brittle and rock-like material. Brace et al. Particle shape and particle size distribution are both important factors which could influence the physical properties of granular materials. In this paper, particle shapes of four kinds of biomass were characterized quantitatively with the parameter – aspect ratio and particle size distributions were also studied. Based on the results presented in the survey, the following conclusions were drawn:Due to the anisotropy of biomass materials, biomass particles have needle-like shape and large aspect ratio. In the larger size, aspect ratio of different kinds of biomass particles is greatly different. But all the average aspect ratios decrease to the same value 2.50, which is largely due to the decrease of difference in structure with particle size.Particle size distributions of biomass materials present a good coincidence, and cumulative mass fraction has a good linear relationship with particle size in a wide size range. Compared with coal particles, the difference in particle size distribution would be mainly caused by the mechanism difference of particle breakup which is largely influenced by structure characteristic of material.PVC/rice straw/SDBS-modified graphene oxide sustainable Nanocomposites: Melt mixing process and electrical insulation characteristicsThe incorporation of nanomaterials in the mixing process of polymers can effectively improve the melt mixing rheology of composites. This fact can be employed to control the morphology of sustainable composites comprised of incompatible ingredients such as natural fibers and polymers. The conventional solution to the incompatibility of natural-fiber fillers and polymers is to surface modify the natural filler. In this work, the role of graphene oxide and surface-modified graphene oxide nanosheets on the mixing rheology of a PVC/rice straw system are studied and impacts these nanosheets have on the viscosity of this sustainable system are used to reduce the morphological imperfections. The introduction of either graphene oxide or sodium dodecylbenzenesulfonate (SDBS)-modified graphene oxide nanosheets increased the melt mixing viscosity of PVC/rice straw compounds. Such a change in the melt mixing rheology of compounds resulted in an improvement in the morphological and structural characteristics. When only 1 phr SDBS-modified graphene oxide nanosheets were introduced in a PVC system filled with 15 phr rice straw fibers, more than 23 and 41% enhancement in the tensile ultimate strength of the system is achieved, compared to the neat PVC and the PVC/15 phr rice straw composite, respectively. Moreover, dynamic-mechanical results confirm that the incorporation of 1 phr SDBS-modified graphene oxide nanosheets resulted in more than 7% reduction in the damping peak intensity of the system, suggesting the formation of interactions between system ingredients. The electrical conductivity of the fabricated sustainable PVC/rice straw/SDBS-modified graphene oxide nanocomposite is in the range of 10−9–10−5 S/m, suggesting it can be a good candidate for insulating applications.Poly (vinyl chloride) (PVC) is one of the major thermoplastic polymers with a wide range of applications in construction, clothing, automobile, and packaging Calcium carbonate is a low-cost inorganic filler with a well-established performance as a PVC compounding ingredient. Calcium carbonate can be considered as the most common filler for different industrialized PVC formulations Cellulose (~35%), hemicellulose (~25%) and lignin (~11%) are the main biopolymers in the structure of RS fibers The surface modification of RS fibers with silanes mostly involves the employment of an environmentally harmful solvent for the mixing process, which should be removed after the modification process (through a drying step) With its superior mechanical and electrical properties, undoubtedly, graphene can be considered as the most interesting nanomaterial of our time Compared to the pristine graphene, GO nanosheets can be easily distributed and dispersed in a medium; either the medium is a polymeric matrix or a proper solvent The incorporation of nanomaterials in the compounding process of polymers can lead to a change in the rheology of the process ). GO and SDBS-modified GO nanosheets were nanomaterials used in this work to control the mixing viscosity of PVC/RS compounds, to fabricate sustainable composites with enhanced mechanical and structural properties. This is the first study on the role of SDBS-modified GO nanosheets (GOSDBS) on the melt mixing process and properties of PVC compounds. Moreover, to the best of our knowledge, this is the first systematic study on reducing morphological imperfections of natural-fiber-filled polymeric systems through controlling the compounding rheology. Mechanical, structural and morphological studies confirmed that such a strategy could be an effective approach to fabricate high-performance sustainable plasticized PVC composites with potential applications in electrical industries.PVC suspension powder (LS100E suspension resin, K-Value: 67) was from LG Chem. RS fibers were provided from a local market in the north of Iran (Lahijan local farms; an agricultural area in the north of Iran). All PVC additives were of commercial grades and used as received. Graphite powder was from LECO Co. Sulfuric acid (98%, H2SO4), potassium permanganate (KMnO4), hydrogen peroxide (30%, H2O2) and hydrogen chloride (37%, HCl) were from Merck Chemicals Co. Sodium dodecylbenzenesulfonate (SDBS) powder was from Daejung Chemicals and Metals.GO nanosheets were prepared from graphite as the precursor using a modified Hummers method GO nanosheets were surface modified by adding 0.5 wt% of SDBS to the solution (1 mg/mL) followed by a sonication step to prepare GOSDBS nanosheets Samples were fabricated through dry blending and melt compounding of ingredients listed in using an internal mixer. The internal mixer used here for the fabrication of all samples was a W 50 EHT Brabender mixer operated at a rotational rate of 60 rpm, a fill factor of 0.7 and a mixing temperature of 175 °C according to ASTM D 2538. Different PVC formulations are available for a variety of applications. As for the insulation application in the cable industry, two main additives with high contents are plasticizer and filler Atomic force microscopy (AFM) results were obtained using a Dualscope DS 95–200, DME atomic force microscope. A solution of GOSDBS nanosheets was placed on a freshly cleaved mica surface and then was dried in a vacuum oven to prepare samples for the AFM test. The scan rate used for topographic imaging was 20 μm/s on the non-contact mode and the microscope was equipped with a rectangular cantilever (cantilever coating: aluminum). X-ray powder diffraction (XRD) patterns were conducted on an Inel Equinox 3000 diffractometer equipped with a Cu kα radiation source (λ = 1.54056 Å) operated at 40 kV and 30 mA (step size: 0.02°/s). Transmission electron microscopy (TEM) results were provided using a Philips CM10 electron microscope at the voltage of 100 kV. Fourier transform infrared (FTIR) spectra were obtained using a Nicolet IR100 spectrometer through the spectral range from 4000 to 400 cm−1 according to the ASTM E 1252-98.Torque (Γ) vs. time (t) graphs (fusion curves) were obtained using the batch mixer through the mixing process of each sample. These results can be used for rheological studies through flow patterns close to the flow pattern expected for an industrial Banbury mixer where η is the viscosity (Pa·s), Γ is the torque value (N·m), N is the rotational speed used for the fabrication of samples Scanning electron microscopy (SEM) images were obtained using a Tescan VEGA-II microscope equipped with an energy beam of 20 kV. Samples were fractured in liquid nitrogen before the test and fracture cross-sections were used for the test. Dumbbell-shaped specimens were prepared for tensile tests. Mechanical properties were conducted on a HIWA 200 universal testing machine with a crosshead speed of 5 mm/min (according to ASTM D638). True (Cauchy) stress and true (Hencky) strain curves were obtained according to the literature Dynamic mechanical analysis (DMA) was conducted on a TTDMA analyzer with a heating rate of 5 °C/min, a constant frequency of 1 Hz and in the tension mode. Electrical characteristics were recorded on a GW INSTEK LCR-8101G LCR meter at room temperature and in the frequency-sweep AC mode according to ASTM D 150–98. AC and DC voltage breakdown tests were conducted using a KvTester ZC-524 withstand voltage equipment at a frequency of 50 Hz and under air and silicon oil atmospheres. Thermogravimetric analysis (TGA) and differential scanning calorimetry (DSC) results were recorded on a Mettler Toledo TGA/DSC 1 analyzer under N2 atmosphere with a heating rate of 10 °C/min.AFM method was employed to investigate the thickness of GOSDBS nanosheets used here for the fabrication of nanocomposites, as shown in . GOSDBS nanosheets were of bi- and few-layers in thickness, with a thickness range of 2–4 nm, as observed in multiple height profiles of different topographic points in . Based on the average thickness of a single GO nanosheet (~1.5 nm a), one can postulate that GOSDBS nanosheets used here for the fabrication of nanocomposites were bilayer and few-layers in thickness.Structural characteristics of GOSDBS nanosheets were further studied using XRD, TEM and FTIR methods, as shown in . No sign of the graphite characteristic peak (2θ ~ 26°), nor the GO characteristic peak (2θ ~ 11°) were observed in the XRD pattern of GOSDBS nanosheets. It suggests that the surface treatment of GO nanosheets with SDBS as the surfactant results in an enhancement in the exfoliation state of nanosheets. This is in close accordance with the TEM result presented in b, where a GOSDBS monolayer nanosheet can be observed. Moreover, the folding of nanosheets observed in b suggests that the thickness range of GOSDBS nanosheets in can be partially related to the folded nanosheets, which can have a thickness of 2–4 nm.FTIR method was employed to investigate the nature of interactions between GO nanosheets and SDBS macromolecules, as shown in bonding) between hydroxyl groups on basal plane and edges of GO nanosheets and CH3 end groups of SDBS can be considered as the most probable mechanism. As shown in d, the presence of different oxygen-containing functional groups on the basal plane and edges of GO nanosheets has been discussed in models and experimental results in the literature A comparison between the intensity of bands related to the hydrophilic moiety of SDBS molecules and the intensity of the same bands in the FTIR spectrum of GO nanosheets (inset in c) suggests that the intensity of these peaks is much higher in the case of GOSDBS nanosheets. In addition, the weak band observed at 1324 cm−1 was related to the asymmetrical bending of C–H bonds in the structure of SDBS molecules. The relatively low intensity of this peak in comparison with the hydrophilic bands of SDBS molecules can be considered as a confirmation for the suggested interaction mechanism between GO nanosheets and SDBS molecules, shown schematically in e. Moreover, the shoulder of the band at 1637 cm−1, which refers to the overlap of this band and a peak at 1600 cm−1, can be attributed to the CC in-plane stretching vibrations of the GO backbone The rheological behavior of samples through the melt mixing process was studied using torque vs. time curves of samples as well as the viscosity of samples calculated using Eq. . These results were used to investigate the role of different reinforcing agents on the rheology of the PVC resin as well as the successful mixing of ingredients. Torque vs. mixing time of PVC was a typical curve of polymeric compounds with an increasing trend by the introduction of the polymer due to the resistance of polymeric resin to the free rotation of the blades When RS fibers were added to the mixing system, an increase in the maximum torque required for the melt mixing of ingredients increased. Moreover, the observed torque in the steady-state region was higher in the case of the PVC/RS composite, compared to the original PVC sample. This indicates an increase in the viscosity of the system with the incorporation of RS fibers. Viscosity values were calculated according to equations b. The viscosity of the original PVC sample was ~1340 Pa·s, which is close to the reported viscosity of the PVC in the literature at a similar shear rate value (~1800 Pa·s at 160 °C and a shear rate of 60 s−1Compared to the neat PVC sample, the initial maximum torque was observed at the later stages of the mixing process. As discussed, the rising trend of the torque at this stage is related to the resistance of the PVC resin to the blade rotation. When RS fibers were incorporated, more time was required to overcome the polymer resistance, as the incorporation of RS fibers increases the viscosity of the compound.The maximum viscosity was observed when only 1 phr GO nanosheets were incorporated in the PVC/RS mixing system. Due to the two-dimensional platy form of GO nanosheets, the incorporation of these nanosheets, even at very low loading contents of 1 phr, can result in a clear change in the rheological properties of polymeric compounds XRD results were used to investigate the structural characteristics of fabricated samples, as shown in . The Gaussian fitting equation was used for the calculation of full width at half maximum (FWHM) values in this work, including results reported in . Four main peaks were distinguishable in XRD patterns of PVC/RS, PVC/RS/GO and PVC/RS/GOSDBS composites at 2θ values around 14.8°, 17.7°, 19.5° and 26.2° (indicated with *, ⁑, , respectively). The first peak in all XRD patterns of RS-filled composites can be assigned to the crystalline form of cellulose in the structure of RS fibers. From the structural point of view, RS fibers consist of cellulose, hemicellulose, lignin and residual ash d, two main crystalline peaks were observed in the XRD pattern of RS fibers. Peaks observed at 2θ values around 16° and 22° in d indicate the Type-I crystalline structure of the cellulose (Iβ polymorph) in original RS fibers. The observed shift in the position of cellulose-I peaks to lower 2θ values as well as the observed change in the intensity ratio of these peaks indicates structural changes of the cellulose crystalline form into a more stable Type-II (or Type III) polymorph through the melt mixing process Since cellulose cannot be melted (due to the strong hydrogen bondings between adjacent chains in the structure of this polymer), when cellulose is heated up, it only can degrade (rather than melt or plasticize) , calcium stearate and calcium carbonate are two ingredients used as the external lubricant and filler in the formulation of the PVC, respectively. Calcium stearate is known to have a characteristic peak at 2θ values around 19-20° e. No significant change in the position of this peak was observed through the incorporation of RS fibers, GO nanosheets or GOSDBS nanosheets. Peaks related to the calcite polymorph of calcium carbonate were also indicated in SEM results were used to study the role of melt mixing rheology on the morphology of fabricated composites, as shown in . Due to the low compatibility between RS fibers and PVC, the incorporation of RS fibers into the polymeric system resulted in the poor bonding between fibers and PVC and consequently, voids and gaps formed around RS fibers in the PVC/RS composite. These voids, indicated with arrows in b, can act as stress concentration points in the structure of the composite and decrease mechanical properties of the RS-filled composite to values even lower than the neat PVC. As a result, a high degree of fiber pullout was observed in the fracture cross-section of the PVC/RS composite in b. The same behavior was reported in the literature for PVC/rice husk When GO or GOSDBS nanosheets were incorporated ( c and d), higher viscosity of the mixture through the melt mixing process increased the wetting of RS fibers with the PVC matrix and consequently, an enhancement in the mechanical strength of the system. It is well known that three factors affect the microstructure of a polymeric hybrid system: (i) interfacial forces, (ii) kinetic viscoelastic properties and (iii) the system composition The proposed mechanism for the observed role of viscosity on the wetting of RS with PVC is as follows: the compounding process of PVC and RS fibers includes a high-shear mixing stage followed by a relaxation stage (the consolidation stage) SEM images of the PVC/RS/GOSDBS nanocomposite suggested that GOSDBS nanosheets in the structure of the nanocomposite were in the form of stacked nanosheets. However, as there was no sign of GO or graphite characteristic peaks in the studied XRD pattern of the PVC/RS/GOSDBS nanocomposite, one can postulate that the intercalated morphology was the dominant morphological state of GOSDBS nanosheets in the structure of the PVC/RS/GOSDBS nanocomposite.TEM images of the PVC/RS/GOSDBS nanocomposite were used to further investigate the morphological state of nanosheets, as shown in . GOSDBS nanosheets were well distributed throughout the PVC matrix due to high shear through the melt mixing process. Moreover, the intercalated morphological state was observed in TEM images. Furthermore, TEM images suggested that, although calcium carbonate particles were agglomerated to form larger particles with a size in the range of ~ 300 nm, but were properly mixed with the matrix and were distributed throughout the nanocomposite.Mechanical properties of fabricated samples were studied using tensile stress–strain results, as shown in . The incorporation of RS fibers, without GO or GOSDBS nanosheets, resulted in a diminution of all mechanical properties, expect the modulus of elasticity (E, the modulus of elasticity values were measured from the Hookean region in each stress–strain curve according to ASTM D638, see a for the shift of stress–strain curves through the Hookean region). This behavior can be directly related to the incompatibility of RS fibers and the PVC matrix, which resulted in the formation of gaps and voids around fibers in the composite, as observed in b. The same behavior is reported in the literature for the mechanical properties of PVC/rice straw wood-plastic composites When GO or GOSDSBS nanosheets were incorporated, due to an improvement in the melt mixing rheology of the system () and a change in the wetting of RS fibers with the PVC matrix (), no stress concentration point was formed in the structure of the nanocomposite. Consequently, an enhancement in the ultimate tensile strength of PVC/RS/GO (54.323 ± 2.985 MPa) and PVC/RS/GOSDBS (58.434 ± 5.215 MPa) nanocomposites was observed in , compared to both PVC (47.268 ± 5.999 MPa) and PVC/RS (41.447 ± 5.901 MPa) samples. Moreover, the tensile modulus of GO and GOSDBS loaded samples were 52 and 54% higher than the PVC sample, respectively. It was expected for both PVC/RS/GO and PVC/RS/GOSDBS nanocomposites, as the incorporation of GO-based nanosheets (which can present a theoretical modulus of ~1.0 TPa Dynamic-mechanical results were used to further investigate the role of GO or GOSDBS nanosheets on mechanical properties and glass transition behavior of samples, as shown in . The dynamic performance of samples can be discussed through three different regions in : (i) temperatures lower than the glass-transition temperature (Tg) of PVC, (ii) through the glass-transition region (indicated with shades in a), and (iii) temperatures higher than the glass-transition region. At low temperatures, where the mobility of polymeric chains is restricted, one expects higher storage moduli for both GO and GOSDBS loaded nanocomposites (considering high modulus of nanosheets and static mechanical properties).Although the storage modulus of the PVC/RS/GO nanocomposite was slightly higher in this range (~1175 MPa at 30 °C), but the modulus of the GOSDBS loaded nanocomposite was even lower than the neat PVC sample (927 and 1150 MPa at 30 °C, respectively). However, the loss factor (tan δ: the ratio of the loss modulus, E“, to the storage modulus, E') of the PVC/RS/GOSDBS nanocomposite was lower than all other samples (see the intensity of tan δ peaks in b). Since the dynamic storage modulus represents the structural rigidity of a nanocomposite system As expected, the storage modulus of all samples decreased through the glass transition region, due to more movability of PVC chains. A slight increase in the storage modulus of the PVC/RS/GO nanocomposite was observed through the first stages of the glass transition region (50–60 °C). The same behavior was observed for PVC/GO nanocomposites in the literature A clear shift in the position of the glass-transition temperature of nanocomposites (the position of tan δ peaks in b) to lower temperatures was observed in dynamic-mechanical results. To verify this observation, Tg values of samples were also measured using DSC results and a similar behavior was observed ( c). Same behavior for different nanocomposites was observed in literature and a decrease in the intermolecular interaction between PVC macromolecules due to the dispersion of nanosheets It is noteworthy that the fitting results of tan δ peaks ( d) also revealed that the tan δ-temperature curves of both PVC/RS/GO and PVC/RS/GOSDBS samples contained two overlapped peaks: a small peak around the glass transition range of the pure PVC and the main peak at lower temperatures. Considering the low intensity of small peaks, one can relate these peaks to the part of PVC chains interacted with either GO or GOSDBS nanosheets. Interestingly, the nanosheet-related peak in the tan δ-temperature curve of the PVC/RS/GOSDBS was observed at higher temperatures than the glass transition peak of the PVC sample, indicating the formation of interfacial interactions between nanosheets and PVC macromolecules, as suggested repeatedly in the literature Another important factor regarding the dynamic properties of nanocomposites is the width of the tan δ peak. As shown in , the PVC/RS/GO nanocomposite presented the highest FWHM of tan δ peaks among all studied samples. The FWHM of this nanocomposite was about 10% higher than both PVC and PVC/RS samples. An increase in the width of the tan δ peak can be considered as the indicator of the increased volume of the matrix/fiber interface in composites To investigate the impact of GOSDBS nanosheets on the electrical conductivity of compounds, AC electrical conductivity (σAC) of the PVC/RS/GOSDBS nanocomposite was studied in details, as shown in . The AC electrical conductivity of pure PVC is in the range of 10−11–10−6 S/m in the frequency range of 1–106 Hz a) and much lower than the reported σAC for PVC/CNTs nanocomposites, which can be in the range of 10−3–10−2 S/m in the same frequency range (with 0.5 wt% CNTs loading) The dissipation factor (tan δ = ε“/ε', where ε” and ε' are dielectric losses and dielectric constant, respectively) can be considered as the part of electromagnetic energy dissipated in a structure b). Moreover, the tan δ of the PVC/RS/GOSDBS nanocomposite was about 0.036 at 1 kHz. Although these values are close to the dissipation factor of the pure PVC Although the incorporation of GOSDBS nanosheets can result in an enhancement in the carrier transport, due to the low electrical conductivity of both GO and GOSDBS nanosheets, the electrical conductivity of the GOSDBS-loaded nanocomposite should not increase significantly. Electrical conductivity is an important characteristic of fabricated nanocomposites, as one of the main applications of plasticized PVC compounds is in cable insulation. c and d presents a comparison between the voltage breakdown of the PVC/RS/GOSDBS nanocomposite and the PVC sample. Results in revealed that the breakdown behavior of the PVC/RS/GOSDBS nanocomposite was very close to the PVC sample. This behavior is in close accordance with the aforementioned electrical behavior of the PVC/RS/GOSDBS nanocomposite. Moreover, the observed minor reduction in the breakdown behavior of the PVC/RS/GOSDBS nanocomposite can be related to the slightly higher electrical conductivity of the nanocomposite, compared to the PVC sample. Such a behavior can be related to the fact that there are many functional groups on the basal plane and edges of GOSDBS nanosheets, which reduce the electrical conductivity of graphene nanosheets. As a result, GOSDBS nanosheets did not affect the electrical conductivity and voltage breakdown behavior of the system significantly. Consequently, the PVC/RS/GOSDBS nanocomposite can still be considered as a candidate for insulating applications in cable industries, as these functional groups act not only as sites for interactions between nanosheets and macromolecules but also as scattering points for charge carriers.FTIR results were used to further investigate the surface chemistry of fabricated samples, as shown in . Several ingredients were used here for the fabrication of samples (). Consequently, FTIR results presented in can represent a variety of components in the system. When only RS fibers were incorporated in the system as the filler ( b), no significant change was observed in the intensity of OH band related to the structure of cellulose, hemicellulose, and residual moisture, compared to the FTIR spectrum of PVC. Moreover, a slight increase in the intensity of bands assigned to νC-H and νCHCl of PVC Bands related to the functional groups of GO and GOSDBS nanosheets were distinguishable in the FTIR spectra of PVC/RS/GO and PVC/RS/GOSDBS nanocomposites, respectively. A clear increase in the intensity of the band assigned to the stretching vibration of the CO bond (~1720 cm−1) was observed with the incorporation of either GO or GOSDBS nanosheets. Moreover, the intensity of bands related to the stretching vibrations of COne another important characteristic of PVC-based systems for any application is the thermal stability and thermal degradation rate of such systems. TGA and differential thermogravimetry (DTG) results were used to investigate the thermal stability of samples, as shown in . PVC is known to exhibit a two-stage thermal degradation process, with the first stage at temperatures in the range of 250–300 °C and the second stage above 400 °C , not only this stage is the main thermal degradation stage of the PVC matrix, but also the fastest stage of this process. Through the second stage, the thermal cracking of these carbonaceous conjugated polyene sequences takes place b), not only the onset degradation temperature of the system shifted to lower temperatures (T2 = 238.23 °C and T5 = 253.87 °C), but also a clear reduction in the thermal degradation rate of the sample through the first stage was observed. This behavior can be related to the degradation of RS fibers, which is reported to start at onset temperatures close to the PVC (~227 °C a. No significant change in the mass loss or degradation rate of the system through the second degradation stage was observed with the incorporation of RS fibers (see When GO or GOSDBS nanosheets were incorporated ( c and d, respectively), a slight decrease in the degradation onset temperature of the system was observed (T2 = 231.83 °C and 237.57 °C for PVC/RS/GO and PVC/RS/GOSDBS, respectively). This can be related to the thermal conductivity of these nanosheets, which enhance heat transfer through the system. Moreover, the incorporation of GO and GOSDBS nanosheets resulted in a clear enhancement in the degradation rate of the system through the first degradation stage. However, due to the reasons discussed earlier, the impact of GOSDBS nanosheets on the degradation kinetics of the system was lower than GO nanosheets.In this study, the influences of GO-SDBS nanosheets on the melt mixing rheology of PVC/RS compounds were used to fabricate sustainable nanocomposites with enhanced mechanical properties. The melt mixing viscosity of PVC/RS compounds increased ~15% with the incorporation of 1 phr GOSDBS nanosheets. Morphological and structural studies confirmed the formation of structurally integrated PVC/RS/GOSDBS sustainable nanocomposites with the enhancement of matrix-filler bonding using this strategy. Moreover, mechanical studies revealed more than 41% enhancement in the tensile strength of the PVC/RS composite with the introduction of only 1 phr GOSDBS nanosheets. Such an improvement in the mechanical performance of the system was related to the strong bonding between PVC and reinforcing agents, as confirmed through dynamic-mechanical studies. In addition, the electrical conductivity of the fabricated PVC/RS/GOSDBS sustainable nanocomposite was in the range of 10−9–10−5 S/m in the frequency range of 10–106 Hz. These results suggest that fabricated PVC/RS/GOSDBS sustainable nanocomposite can be a good candidate for electrical insulation applications, where plasticized PVC compounds are applicable.Ahmad Allahbakhsh: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing - original draft, Writing - review & editing.Elevated temperature mechanical properties of Inconel 617 surface oxide using nanoindentationInconel 617 is a principal candidate material for helium gas cooled very-high-temperature reactors with outlet temperatures of 700–950 °C. Recent findings showed that this alloy develops unique surface oxide layers especially at high temperature (HT) helium environment with distinctive wear, friction and contact properties. This study investigates the elevated temperature mechanical properties of Inconel 617 top surface layers aged in HT helium environment. Nanoindentation technique is used to obtain load-displacement graphs of the alloy top surface oxide in temperatures ranging from 25 °C up to 600 °C. In addition, using finite element analysis along with an iterative regression technique, a semi-numerical method is developed to further measure and quantify the material parameters and, in particular, time-independent and creep characteristics of the oxide. While Young's modulus of the oxide is found to be relatively close to the bulk for the tested temperatures, the yield strength and hardness, in comparison to the bulk material, increase significantly as the material is oxidized after aging. The oxide exhibits significant softening as the temperature increases to 600 °C. Unlike the bulk material, diffusion through the grains is found to be the dominant creep mechanism for the oxide. Considerable difference between the mechanical properties of the oxide and the bulk material shows the need for accurate measurements of near surface mechanical properties, if reliable predictive contact and tribological models are sought at elevated temperatures.In view of design demands, providing materials that can withstand HT and harsh environments is crucial for reliable and effective operation of mechanical components in applications such as nuclear reactors, jet engines and gas turbines. HT superalloys, exhibiting excellent mechanical properties, resistance to thermal creep deformation and surface stability, are among the main solutions to the challenges associated with these applications. In particular, Inconel 617 superalloy (designated as IN617, hereinafter) is one of the principal candidates for next generation very high-temperature gas-cooled reactors (VHTR) and high temperature gas-cooled reactors (HTGR) with outlet temperatures up to 950 °C []. IN617 is developed to be utilized in reactor control rods, valves and valve seats. It is an austenitic solution superalloy containing nickel (Ni), chromium (Cr), cobalt (Co) and molybdenum (Mo), and has significant strength, good oxidation, corrosion and creep resistance, especially at elevated temperatures [The main coolant of HTGR/VHTR reactors is Helium (He) which contains low levels of impurities including H2, O2, H2O, CH4, CO, CO2 and N2 []. Due to the presence of these impurities, the metallic alloys can inevitably develop unique surface oxides that are highly controlled with temperature, oxygen partial pressure, carbon activity, and alloy composition. The presence of this oxide can significantly influence surface friction, wear and contact properties which in turn impact the performance of the components. For example, the precise motion of the control rods is mission-critical for reactors’ performance and safety. In fact, control rods surface can be affected by the HT contact creep and self-welding during long idle intervals resulting in unpredictable high static friction and often surface damage []. In general, contact, friction, wear and adhesion behavior of control rod surfaces, as well as other tribo-pairs in HTGR/VHTR, are governed by the oxide layer properties and it is thus imperative to understand and quantify their mechanical response [Instrumented nanoindentation is a common method of measuring mechanical properties of thin films [], with the ability of measuring several material time independent []. Interpreting and measuring mechanical properties at HT through instrumented nanoindentation is, however, challenging due to the temperature effect on the compliance of the system, thermal expansion of the tip, tip damage, as well as complexity of the contact at high temperatures. Hence, studies at HT and especially at the nano/micro-scale are scarce, with limited measured mechanical properties []. Gibson et al. used HT indentation to measure the effect of He implantation on tungsten hardness at temperatures up to 750 °C []. Wang et al. measured temperature-dependent mechanical properties of additively manufactured Inconel 718 up to 650 °C where HT nanoindentation was used to measure hardness, Young's modulus and creep stress exponent []. Zhang et al. reported elastic properties, hardness, indentation creep exponent, and thermal activation of bulk IN617 at HT. Applying high loads, Zhang's study was focused on the bulk material rather than the oxide layer, and more importantly, oxidation in their study, if any, formed in air for a short period of time. They observed a slight decrease in elastic modulus and significant reduction in hardness as temperature increased [Generally, constitutive equations of creep, based on uniaxial tests cannot be used for nanoindentation parameters, and vice versa, requiring systematic modifications. This is due to the complicated stress/strain status during indentation and contact evolution, resulting in inaccurate estimation of creep properties. However, theoretical modeling, numerical simulation, and experiments, when perform collectively, can resolve some of the issues, and more importantly can be used to measure time dependent mechanical properties at HT, by relating strain-stress during indentation creep to uniaxial tests []. The problem, nonetheless, gets more complex when plasticity and creep are both involved. Kang et al. reported successful hybrid experimental indentation at HT and finite element (FE) analysis to extract the elastic-plastic and creep parameters of P91 steel and XN40F []. Other works report on the iterative use of FE analysis to measure yield strength, work hardening rate and creep parameters []. These measurements are only applicable for specific creep mechanisms and should be implemented with care.The current study evaluates the HT mechanical properties of the surface oxide, formed after aging, of IN617 through nanoindentation. A combined experimental and iterative regression-based numerical analysis is employed to find the time independent properties, elastic modulus and yield strength of the oxide layer along with creep exponent, creep coefficient and activation energy. These properties are important for understanding the underlying contributors to contact behavior as well as friction behavior of IN617 at HT and are critical to develop predictive friction models.IN617 samples were received from Idaho National Laboratory (INL), Idaho, USA (see Ref. [] for the detailed chemical composition of the as-received alloy, provided by the manufacturer). The samples were mechanically polished to reduce the roughness and remove the initial oxide on the surface. The polished samples were then cleaned with acetone and iso-propyl alcohol in an ultrasonic cleaner, and dried. To simulate the effect of the He-cooled reactor environment, the samples were then placed in a 99.999%-pure He filled furnace at 950 °C and aged for 100 h. The samples were then cut, and the nanoindentation experiments were performed on the oxide from the cross section of the aged samples. Nanoindentation on the cross section was highly beneficial, compared to the top surface, due to high roughness of the oxide surface after aging, and also it prevented the ambiguity of having bulk substrate effect on the extracted property. For the cross-sectional sample preparation, a layer of Ni coating was deposited (using electrodeposition technique) on the samples’ surface to protect the oxide layer from delaminating during polishing.a, obtained by scanning probe microscopy (SPM) imaging, shows a representative topographical image of the cross section of the sample, where the indentations were performed. An area of approximately 8 mm2 of cross section with 1 mm thickness was cut and gradually polished to obtain a mirror polished surface. The sample cross section was polished first with 400, 600 and 800 grit SiC papers, followed by polishing using 0.3 μm alumina suspension and finally with 0.02 μm colloidal silica suspension, to ensure obtaining surface roughness values close to the minimum possible. The surface roughness was measured using SPM imaging of an area of 30 μm by 30 μm of the cross section with the nanoindenter Berkovich tip, and found to be around 40 nm (RMS). The imaging area included both oxide (approximately 6–10 μm thick) and bulk material (20–24 μm thick). However, the indentations were performed only in the clean areas that were carefully chosen by SPM scanning. Particularly, for the oxide indentation measurements, small areas of 1 μm by 1 μm were used with surface roughness values of less than 8 nm. It should be noted that at each temperature, consistent indentation measurements at multiple locations were observed, confirming the minimal effect of surface roughness on the measurements. Cross-section SEM image in b shows the oxide layer formed on bulk IN617. The oxide thickness of aged IN617 after 100 h of aging in He varied between 6 and 10 μm attesting to the fact that the contact mechanics of aged IN617 is influenced by the oxide properties.Nanoindentation experiments were conducted using a commercial nanoindenter (Triboindenter TI Premier, Bruker Inc.) shown in a. Indentation experiments were performed at room temperature (RT), 200 °C, 400 °C and 600 °C. b shows the detailed schematic of the indenter hot chamber (Bruker xSol® stage) heated by two sets of heating elements on the top and bottom of the sample holder. A constant flow of 95% Argon and 5% Hydrogen was introduced through the hot chamber to mitigate oxidation damage to the tip and any further oxidation to the sample. The gas flow also helped accelerate the process of achieving thermal equilibrium by convection heat transfer [In the current work, a Berkovich indenter probe made of diamond was used for the nanoindentation experiments up to 400 °C, whereas a Berkovich Cubic Boron Nitride (cBN) probe was used for 600 °C experiments. Before and after every set of experiments, we also performed indentation on fused quartz. This was done to evaluate the consistency of load-displacement curves as well as measured the elastic modulus and hardness values to ensure the tips preserved their geometry at high temperatures. For further analysis, the tips were also checked by SEM for deformation, contamination, and degradation at each temperature. The risk of tip contamination for HT indentation is well acknowledged. However, here, no significant contamination or degradation was observed either using fused quartz indentation measurements or SEM imaging, even at 600 °C. A maximum load of 6 mN was used for experiments up to 400 °C, where a lower load of 500 μN was used for 600 °C due to softening of the oxide sample, and to achieve similar indentation depth range for all tests. The loading and holding times were selected after evaluating the nanoindentation test results for different holding stages, maximum loads and load patterns. Indentations were repeated at least 5 times at each temperature. The load was ramped from zero to a peak value at the rate of 0.5 mN/s before it was held at peak load for 100 s (for indentation creep measurements), and then it was unloaded at 1.0 mN/s. Generally for the creep tests, the loading rate is important due to its effect on creep during indentation and subsequent extraction of modulus and hardness []. Therefore, here, we kept the loading rate constant for all temperatures to achieve consistent results.To obtain an accurate result for elastic modulus and hardness at HT, the effects of machine compliance, thermal drift rate, and sample creep should be considered and excluded from the raw results. Oliver and Pharr's method was followed for the machine compliance calibration, which is based on iterative calculation of tip area function and frame compliance []. Based on this method, the machine compliance was calibrated by the nanoindenter software automatically. To have the compliance calibration as close as possible to the actual test, the fused quartz sample was placed within the heating chamber, where the sample for indentation was placed. Simultaneous calibration of area and compliance was performed by 100 indents within the load range of 5 mN to maximum system load capacity of 10 mN on the standard fused quartz. Also, additional 50 indents with the load range of 100 μN to 5 mN were performed to obtain accurate tip area function for shallow depth indentations. It should be noted that the indentation load was lower than the maximum calibration load. Moreover, the depth of the indentations performed for this study were well within the calibration depth range (200 nm). In addition, the drift rate was calculated by the machine integrated software at the beginning of each test. During unloading, the load was held at 10% of its peak for 120s to calculate the thermal drift rate from the fraction of displacement during holding time []. It should be mentioned that for HT indentations, the sample and the probe were held at HT for half an hour to achieve thermal stability. As mentioned, to reduce thermal drift and to achieve quick thermal equilibrium, cover gas was introduced to the hot chamber as the primary convective medium [The common technique to measure elastic modulus and hardness along with creep, also used in this research, is constant load and hold (CLH), which is loading to a maximum load and maintaining the load for a specific time. Phani et al. showed that this method of loading minimizes the influence of depth on hardness during creep tests and at the same time can cover a wide range of creep strain rates []. In fact, it is shown that the creep displacement during holding time in CLH covers a wide range of strain rates in a single test and is closer to uniaxial creep tests, in comparison to other loading patterns [Hardness was measured using the maximum load (Pmax) as well as the associated projected area A of the contact and is given by:Area A can be calculated based on contact depth, hc, using the following shape function:A(hc)=C0hc2+C1hc+C2hc1/2+C3hc1/4+C4hc1/8+C5hc1/16where Co represents a constant, which is a function of tip geometry, and for a perfect Berkovich tip equals to 24.5 and coefficients C1–C5 account for the geometrical imperfection of the tip, and can be estimated through tip area calibration, based on known properties of the fused quartz standard.Contact depth hc, was estimated using a load-displacement curve, and the following formula [where hmax is the maximum displacement at the end of the loading step, ε denotes the geometry correction factor which is 0.75 for a Berkovich tip and Su is the contact stiffness at the onset of unloading.Following the Oliver-Pharr method, the reduced elastic modulus Er can be calculated through measurement of the contact stiffness, Su , and the projected contact area [Using the reduced elastic modulus from Eq. , and known mechanical properties of the diamond tip (Ei,νi), the elastic modulus of the indented material can be calculated using:where, Es and νs are elastic modulus and Poisson's ratio of the oxide, respectively.Unlike hardness, to obtain accurate results for elastic modulus at HT, the Oliver-Pharr method should be modified. Feng et al. suggested a correction for the Oliver-Pharr method (Eq. ) to exclude the effect of creep on the elastic modulus by introducing a modified stiffness parameter, Sm, given by Ref. [where h˙hc represents the indenter creep rate at the end of the load hold stage, and P˙ is the unloading rate at the beginning of the unloading stage. Following Feng et al.‘s correction, in the current study, the unloading stiffness, Su, was replaced by Sm in Eqs. ) to reduce the effect of creep at HT. It should be noted that Feng et al. introduced a creep compliance factor, CP, only below which Eq. Here, for measuring the Young's modulus and hardness of IN617, the unloading rate of the indenter was kept constant at 1 mNs. Considering the indenter displacement rate at the end of the load holding stage for all tests and temperatures, the maximum creep factor was calculated to be C=0.027<0.1. Therefore, Eq. can be used to calculate the Young's modulus and the holding time used was sufficient, even at HT, to effectively reduce the influence of creep []. Su et al. also defined a threshold for minimum creep displacement during holding stage. It was referred to as “elastic transient” threshold, below which the elastic deformation can significantly influence steady-state creep strain rate. In other words, when elastic deformation below the indenter is significant, the strain propagation through unyielded material delays the steady-state creep. This is important since creep parameters are measured at this stage (a, holding time), and hence, to obtain an accurate measurement of the parameters, the elastic deformation should be surpassed to reduce its effect on creep parameters. Su et al. suggested the following minimum indentation depth beyond which the results are not affected by the elastic transient [where θ is the half-angle of the indenter (70.3° for Berkovich tip) []. The calculated elastic transient for the current indentation tests was between 3 to 15 nm, based on the maximum load and reduced elastic modulus for all tested temperatures. Herein, the measured creep displacements were higher than the elastic transient expecting minimal influence of elastic transient on creep displacement.Representative indentation load-displacement curves for the oxide on aged IN617 cross-section are presented in a at different temperatures, where creep displacement is shown during the holding period at maximum load (b). Thermal drift plots of nanoindentation tests are depicted for 400 °C and 600 °C as it is insignificant at lower temperatures (c). As seen, the thermal drift rate is 0.11–0.24 nm/s, which is relatively low and similar to what is reported in Ref. [] for fused silica. Surprisingly, it is noted that at a higher temperature of 600 °C the thermal drift is lower compared to 400 °C. It can be attributed to cBN material having lower thermal conductivity, ~700 W⁄(m°K), compared to diamond, 2000 W⁄(m°K) [An FE model for the indentation process associated with an iterative regression method was developed to determine the mechanical properties using experimental load-displacement results. The FE model was developed using ABAQUS []. It consisted of an indenter, simulating the same material and dimensions to the experimental indenter [] and a deformable sample modeled as an axisymmetric body (a). It is worth mentioning that the maximum load was sufficiently high for the indenter to indent the material with its pyramidal sides, so the indentation can be modeled with axisymmetric geometry. The rigid indenter was modeled as a conical indenter with half-included angle of 70.3 equivalent to the Berkovich tip used for the nanoindentation experiments. The deformable sample was modeled with fine elements near the contact (b) to capture the stress/strain field with high resolution. An incremental load was applied on the rigid body in three steps: loading, holding, and unloading, which was identical to the experimental steps.To capture the stress/strain variations during the analysis, and following the work of Kang et al. [], an elastic-plastic-viscoplastic model was selected. The plastic deformation was modeled by the work hardening (time-independent) relation as follows:where σp is the plastic stress, n1 represents the strain hardening exponent, Es is the elastic modulus of the oxide and σy and εp denote the yield strength and plastic strain, respectively. In addition, the following power-law creep with strain hardening was selected for the time-dependent creep analysis [where ε˙ is the steady-state creep rate, σ is the creep stress, n2 is the creep stress exponent, which can be related the to creep mechanism of the material, and m is the creep coefficient given by:where m'is a material constant, Q represents the activation energy, k is the universal gas constant, and, T is the thermodynamic temperature in units of Kelvin.A double non-linear regression (DNR) procedure was developed in MATLAB, using least-squares regression function (LASQNONLIN) to minimize the objective function F(x),defined as:where x is the desired material properties, and D(x)iFE,D(x)iexp are displacement functions (output) of the FE simulations, and experimental results, respectively, at each load increment (i). The analysis uses an initial estimation of material properties for unknown (time-dependent and independent) properties (x in Eq. ) and tries to fit the FE curves to the experimental load-displacement and creep-time results by systematically iterating through the input properties and minimizing the objective function.In the current study, variable x contained elastic, plastic, and viscoplastic properties of the desired material including yield strength (σy), work hardening (n1), creep coefficient (A), and creep exponent (n2), which were the output of the regression procedure (total variables: N = 4). To reduce computational time, the Poisson's ratio of the oxide was taken as 0.25, similar to what is reported for chromium oxide coatings [), the elastic modulus of aged IN617 was calculated and used together with the Poisson's ratio as input variables for the numerical calculations. shows representative FE simulation results fitted to the experimental data at 400 °C for load vs indentation depth and creep displacement vs time. The DNR process was utilized to fit the FE results to the load-displacement (a), where the time-independent properties were dominant. Then, we used the converged material properties (x) to fit the FE to the results of the experimental creep displacement-time, where creep properties were dominant (b). This dual procedure was repeated till both fitting processes converge. It is worth mentioning that the converged results were independent of the initial values. In addition, the FE simulation and the numerical algorithm were verified against available numerical results of Kang et al. []. To further validate the method, a nanoindentation experiment was performed on bulk IN617 at RT, and the extracted parameters were compared with available uniaxial data. lists the extracted parameters by DNR method and comparison with uniaxial tensile tests of IN617, available in literature.XRD results taken from the surface of the aged IN617 sample are shown in a. The presence of a complex compound of (Ni/Mn) Cr2O4 spinel is observed, specifying that oxide atoms diffuse deeper through the surface into the bulk and oxidize the nickel []. Also, the main component of the oxide is Cr2O3 which has a high hardness and yield strength []. More detailed elemental presence can be observed from the EDS results for the cross section of aged IN617, presented in b. It confirms the Cr-rich surface oxide on top of the IN617 and as a result, it is anticipated that the oxide behavior follows that of chromium oxide. Chromium oxide in comparison to cubic metal oxide is capable of retaining its strength and lattice structure at higher homogenous temperatures, as well as for longer durations, which is due to its close-packed hexagonal structure and restriction of slip systems. Generally, at the same stress and temperature levels, this slows creep deformation in comparison to cubic metal oxides []. However, surface diffusion of NiCr2O4 at HT and longer times could affect the deformation of the oxide [ shows the Young's modulus and hardness for the oxide layer at 25 °C and up to 600 °C along with data obtained by Zhang et al. [] for bulk IN617 at RT and 400 °C. They used high loads of 200 mN–400 mN on the surface after heating the sample in air for 8 h, and indented the surface down to a depth of 3μm, while the oxide was measured to be around 400 nm []. Therefore, Zhang et al. results represent the properties of the bulk IN617 at 25 °C and 400 °C []. Young's modulus of the oxide layer is 1.4 times higher than the reported values for bulk IN617 [], and close to Young's modulus of similar metal oxides (see ). Especially, it is close to the oxide main constituent (Cr2O3) elastic modulus. Hardness of the oxide layer is approximately 8 times higher compared to bulk IN617 at RT, which is similar to other metal oxides with similar crystal structure, e.g., Al2O3. At 400 °C, the oxide hardness decreases but it is 3 times higher than that of bulk IN617. The large drop in oxide hardness can be attributed to increasing softening effect and change of deformation mechanism []. This drop has also been observed in Bulk 617 material []. It is worth mentioning that following the HT tests and after the samples were cooled down to RT, hardness of the oxide was again measured, and hardness values close to the original value (prior to heating) were measured. Similar to hardness, it is believed that indentation for short period (1 h), up to 600 °C and under argon gas does not change the topography and roughness of the oxide already formed after 100 h of aging at 950 °C.It should be added that delamination can be an important concern for oxide and coating indentation tests, affecting the hardness values. However, here, indentations were performed on the cross section of the oxides and due to the loading direction and sample alignment, delamination was of minimal concern. In fact, for the selected loads and temperatures, no delamination was detected. In addition, as mentioned above, similar RT hardness values were measured before and after HT indentations, showing that hardness decrease at HT is due to material change, rather than delamination. shows RT experimental hardness and elastic modulus of aged IN617 through the section, at different distances from the surface. There is a sudden drop in hardness and, to a lesser extent, in elastic modulus when indentations pass the top oxide layer to the bulk material, as expected. Both measured hardness and elastic modulus for the bulk are fairly close to measured values reported by Zhang et al. [The yield strength of the oxide at RT is estimated through combined experimental and numerical analyses and is depicted in , along with the yield strength of bulk IN617, measured with uniaxial tests, as well as the yield strength of the oxides listed in . The oxide strength is significantly higher than the bulk, and closer to that of similar oxides, especially Cr2O3. However, the oxide strength is lower compared to Cr2O3, which can be attributed to the surface diffusion of NiCr2O4 discussed above. The oxide yield strength and strain hardening exponent are plotted vs. temperature in . As seen, both the yield strength and strain hardening exponent decrease with temperature rise, which can be due to the activation of slip systems []. It is noted that the strain hardening exponent decrease is relatively linear with temperature rise. Based on the results of Young's modulus (), yield strength, and strain hardening exponent () it can be concluded that plastic deformation rate is changing rapidly from 400 °C to 600 °C. Also depicted in a is the oxide hardness to yield strength ratio showing an increase as the temperature increases from RT to 600 °C, due to lower hardness reduction rate compared to yield strength. Similar behavior has been observed for Al2O3 up to 700 °C, as reported in Refs. [a presents a comparison of creep stress exponent of bulk IN617 (reported in Ref. []) and that of the oxide at the cross section of aged IN617, measured in the current study. Zhang et al. [] reported a range of 5–6 for the creep exponent. A creep exponent above 5 generally shows that the main mechanism of creep for bulk IN617 is dislocation climb, however, for the oxide layer, a lower creep exponent, n ≈ 2, observed here, suggests that creep deformation is dominated by diffusion through the grains []. Based on the deformation mechanism map introduced by Frost and Ashby for Cr2O3, diffusion through the grain is dominant for low stress and temperature (T<0.5Tm) conditions []. Here, it can be argued that due to low displacement rate during steady-state creep, and softening effect at HT, the equivalent stress drops, and steady-state creep happens by diffusion through boundaries.The activation energy, Q, can be calculated by the slope of the logarithm of creep strain vs 1/T, based on Eq. b are the experimental results with fairly constant slope over the tested temperatures, showing that the creep mechanism is not changing. The activation energy is calculated to be ̴ 23 kJ/mole for the oxide on the aged IN617. This activation value is much lower than 291 kJ/mole reported for bulk IN617 []. However, it is close to the values for boundary creep mechanism in which less energy is needed to move through the boundaries []. Similar activation energy (29 kJ/mole) calculated from creep coefficient, m is extracted using the DNR model analysis.It is worth emphasizing that IN617 is the principal candidate for heat exchangers, control rods, and valves to be used in Gen IV reactors in the near future. Our previous studies showed superior behavior of IN617, as compared to other candidates like Incoloy 800H []. Notwithstanding promising performance, multiscale qualitative and quantitate knowledge of IN617 material properties, especially when it comes to tribology and surface properties are not well-understood. In particular, the presence of the unique oxide can substantially affect surface friction, wear and contact properties, deviating significantly from the bulk material. This is especially important for engineering applications involving continues and intermittent operation of tribo-pairs at HT under air or other gaseous environments. Considerable difference between mechanical properties of the oxide and the bulk material shows the need for accurate measurements of near surface mechanical properties, if reliable predictive contact and tribological models are sought at elevated temperatures.Experimental nanoindentation experiments were carried out to investigate the mechanical response of Inconel 617 top layer oxide. Elevated temperature nanoindentation was performed on the cross section of the oxide layer following an aging process for 100 hrs at 950 °C in a furnace filled with 99.9999% He, simulating a new generation of He-cooled nuclear reactor environment. The experimental data was further analyzed with an iterative combined regression and FE simulations, extracting the time independent and dependent mechanical properties for the oxide, and comparing to that of the bulk Inconel 617. The main findings from this study are:XRD, SEM and EDS results of the oxide layer confirmed a chromium-rich oxide on the surface.Young's modulus of the oxide was found to be 1.4 times higher than bulk IN617.Unlike Young's modulus, the hardness of the oxide layer is significantly higher (6.5 times at RT and 3.15 times at 400 °C) compared to the bulk, and comparable to similar metal oxides, such as Al2O3 and Cr2O3. The hardness of the oxide decreases almost linearly as temperature increases from room temperature to 400 °C, where it drastically reduces from 400 °C to 600 °C showing excessive softening at 600 °C.The estimated yield strength of the oxide was found to be ~10 GPa at RT (close to Cr2O3 coating), and significantly reduces to 120 MPa at 600 °C.Based on the time-dependent measurements, it can be concluded that creep deformation of IN617 oxide at elevated temperatures is dependent on diffusion through grains, which is different from dislocation climb creep mechanism observed for bulk Inconel 617.In comparison to the bulk IN617 (with creep activation energy of 291 kJ/mole), lower creep activation energy (23 kJ/mole) was estimated for the oxide, which is in agreement with diffusion through grains mechanism.As previous studies showed, IN617 is a promising principal candidate for several mission-critical components in future nuclear reactors. Although material modification can be a path for the future, qualitative and quantitate knowledge of IN617 material and mechanical properties and especially its tribology and surface properties are not fully understood. This is especially important for engineering applications involving contacting pairs. These surface properties are essential for developing tribological models to predict behavior of contacting parts at higher temperature, especially after long dwell times for specific applications.Sepehr Salari: Methodology, Investigation, Formal analysis, Data curation, Visualization, Writing - original draft. Md Saifur Rahman: Methodology, Investigation, Formal analysis, Data curation, Visualization, Writing - review & editing. Andreas A. Polycarpou: Project administration, Conceptualization, Writing - review & editing, Supervision, Funding acquisition, Resources. Ali Beheshti: Conceptualization, Writing - review & editing, Supervision, Funding acquisition, Resources.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.The raw data required to reproduce these findings are available from the corresponding author on reasonable and timely request.The internal geometry of salt structures – A first look using 3D seismic data from the Zechstein of the NetherlandsWe present a first look at the large-scale, complexly folded and faulted internal structure of Zechstein salt bodies in NW Europe using 3D reflection seismic reflection data from two surveys on the Groningen High and the Cleaver Bank High. We focus on a relatively brittle, folded and boudinaged, claystone–carbonate–anhydrite layer (the Z3 stringer) enclosed in ductile salt. A first classification of the structures is presented and compared with observations from salt mines and analogue and numerical models.Z3 stringers not only are reservoirs for hydrocarbons but can also present a serious drilling problem in some areas. Results of this study could provide the basis for better prediction of zones of drilling problems. More generally, the techniques presented here can be used to predict the internal structure of salt bodies, to estimate the geometry of economic deposits of all kinds and locate zones suitable for storage caverns.Structures observed include an extensive network of zones with increased thickness of the stringer. These we infer to have formed by early diagenesis, karstification, gravitational sliding and associated local sedimentation. Later, this template was deformed into large-scale folds and boudins during salt tectonics. Salt flow was rarely plane strain, producing complex fold and boudin geometries. Deformation was further complicated by the stronger zones of increased thickness, which led to strongly non-cylindrical structures. We present some indications that the thicker zones also influence the locations of later suprasalt structures, suggesting a feedback between the early internal evolution of this salt giant and later salt tectonics.This study opens the possibility to study the internal structure of the Zechstein and other salt giants in 3D using this technique, exposing a previously poorly known structure which is comparable in size and complexity to the internal parts of some orogens.► The internal structure of Zechstein salt deposits was mapped ► Different structures of several evolutionary stages are interpreted ► No conclusive evidence for significant gravity-induced sinking of stringers is found ► The internal geometry of salt structures results from non-plane strain deformation ► Early, internal structures seem to influence overburden sedimentation and deformationThe sedimentary basins of NW Europe are classic areas of salt tectonics (). The Dutch part of the Central European Basin contains five evaporite cycles of the Late Permian Zechstein Group (Z1–Z5, see: ), including a relatively brittle layer consisting of anhydrite, carbonate and clay (the “Z3 stringers”A large part of the world’s hydrocarbon reserve is associated with evaporitic deposits (), for example, in the Central European Basin, the Caspian Sea, the Gulf of Mexico, offshore Brazil, and the basins of the Middle East. Prediction of the thickness, porosity, geometry and fluid fill of stringers is of significant economic importance. In some settings in Europe as well as in Oman, stringers enclosed in the salt are hydrocarbon reservoirs (). Better understanding stringers in NW Europe can help the interpretation of the complex geometry and history of the hydrocarbon-bearing stringers in the Ara Salt in Oman.In addition, in most cases the Z3 stringer is considered a drilling hazard by operators in the Central European Basin. The Carbonate Member of Z3 stringer can be significantly overpressured, with pressures up to lithostatic (). Overpressures in stringers are difficult to predict, therefore when planning the well path, stringers are avoided where possible and not drilled when strongly folded and/or faulted. Zechstein salt is also used for different kinds of geological storage or solution mining () and prediction of internal structure is of major relevance in these fields (see In the literature, salt structures are typically shown in two strikingly different ways. In studies using 3D seismic and well data that focus on the subsalt or suprasalt sediments and are typically hydrocarbon-related, the evaporites are invariably shown as structureless bodies (for example: On the other hand, studies of the internal structure of salt are typically salt-mining or storage-related and are based on observations from mine galleries and borehole data (for example: ). These studies show the extremely complex internal geometry with less attention to the structure of the surrounding sediments.Detailed observations of salt mines and drill holes (with cm-to-m resolution) display a variety of deformation structures in the salt on a wide range of scales (). Observations (typically 2D to 3D in salt mines and 1D in storage or solution mining) include boudins and folds together with shear zones (). The folds have curved, open-to-isoclinal fold axes, and boudins from millimeter () are common. Cross-sections through the Zechstein in the Gorleben and Morsleben salt domes ( further describes several examples of fold structures with amplitudes over half the height of the salt structures.Field studies, from Iran and Oman, have also shown the internal complexities of surface-piercing salt domes, such as the distribution of different age salt, the position and internal deformation of solid inclusions, the microstructures and, by inference, the deformation mechanisms (among others: Numerical as well as analogue centrifuge and extrusion models of salt tectonics tend to assume relatively homogeneous rheological properties (although mechanical stratigraphy is used), and consequently produce relatively simple salt structures (for example: ). It must be noted, however, that most of these models do provide a way to study structural evolution due to deforming meshes or the use of multicoloured analogue materials. Despite the relatively simple rheological models, analogue and numerical models with mechanical stratigraphy have shown the complex deformation associated with (brittle) inclusions in deforming ductile media (see, for example: In this paper we aim to contribute to the understanding of the structural style and geometry of the Z3 stringer, by describing detailed interpretations of two 3D reflection seismic surveys from the Dutch onshore and offshore. We focus on thickness variations and structural style at the scale of 30 m–20 km, to draw inferences about sedimentary and diagenetic evolution and salt tectonic processes.In the Dutch subsurface, the Zechstein can be subdivided in a marine lower part (Z1–Z3) and a playa-type upper part (Z4 and Z5) with more clastic deposits (). Z1–Z3 follow the classic carbonate–evaporite cycle: claystone–carbonate–gypsum–halite–potassium and magnesium salts consecutively (). The cycles correspond to major transgressions from the Arctic and evaporation of seawater in the arid Southern Permian Basin (). In the northern Netherlands deposition was relatively continuous, but the sedimentary sequence has major periods of non-deposition in the south (Z1 halite is absent in both areas studied in this paper (). The Z2 Main Dolomite member and Z2 Basal Anhydrite Member (Stassfurt Formation) are deposited directly on top of the Z1 anhydrite in both study areas. Therefore, up to the top of Z2 anhydrite, all rocks are brittle and coupled to the Rotliegend and underlying basement. The top Z2 anhydrite reflector is therefore used in this study to define top of mechanical basement (). The Z2 halite reaches a primary thickness between 500 and 600 m in both study areas, which was later strongly modified by salt tectonics (). Near the top of Z2 locally (not in the study areas) thick deposits of potassium–magnesium salt layers are found (). The Z3 Leine Formation starts with an approximately 1 m thick grey shale (Grey Salt Clay) with a high Gamma Ray signal in wire line logs (). The overlying Platy Dolomite (“Plattendolomit”) reaches thicknesses of 30 m () on the shelf, considerably more than in the basin where it reduces to a few meters (). On top of the Platy Dolomite the Main Anhydrite (“Hauptanhydrit”) is found (). In the Dutch part of the basin the thickness of the Main Anhydrite increases from 3 m on the shelf to 45 m in the basin, with local excursions to 100 m and complex changes in thickness (). Local variations in the Z3 anhydrite thickness were first described by and were subsequently interpreted to result from a type of gypsum diapirism or sedimentary swells, although some were interpreted to be tectonic in nature (The Z3 Leine halite is overlain by two thick potassium–magnesium salt layers, with bischofite, kieserite, carnallite and sylvite (). Primary thickness of the Z3 halite is 200–300 m in the Groningen and about 100 m in the offshore study area (The Aller (Z4) and Ohre (Z5) formations consist of sabkha deposits and are quite thin () are not found in the Netherlands. The youngest Zechstein unit is the Upper Claystone Formation (ZUEC) which is between 10 and 50 m thick and is present in large portions of the Dutch subsurface (Zechstein carbonates in the Netherlands are subdivided in shelf, slope and basin facies (). The platform carbonates consist of shallow water deposits with occasional karst features (). The carbonate members are thickest in the slope, where individual reefs and off-platform highs as well as gravity flows are found (). The basin facies was deposited in water depths up to 200 m and can contain high Total Organic Carbon (The high acoustic impedance contrast between the stringer and surrounding halite makes the Z3 stringer image reasonably well in seismic data (). The stringer is visible as a pair of parallel loops in the relatively transparent halite. Imaging limitations are related to the frequency content and noise level of the seismic data, to the thickness of the layer and to the local high dip of the layer (). If the thickness of the stringer is below the tuning thickness of about 30–35 m, exact thickness determination from seismic data is not possible (and thickness will be overestimated). Sections of the stringer with thicknesses below ca. 10 m are not resolved. However, the majority of stringers in the study area are typically 40–50 m thick, with local excursions up to 150 m or more (see below). The internal structure of the stringer generally is not resolved, probably because of the low impedance contrast between anhydrite and carbonate.We studied stringers in two areas: the Groningen High and the Cleaver Bank High (The first study area is 20 × 30 km and is located on the Groningen High (NE Dutch onshore, ) contains one of the largest gas reservoirs of the world ( showed that salt tectonics in the nearby Ems Graben initiated during the Early Triassic with subsequent phases in the Late Triassic and the Upper Cretaceous/Lower Tertiary. There was little or no activity between the Jurassic to Lower Cretaceous. The stringer is clearly visible in the seismic volume (). The data from this area is part of a large, merged, 3D pre-stack depth migrated seismic dataset, provided by the Nederlandse Aardolie Maatschappij BV. (NAM, a Shell operated 50–50 joint venture with ExxonMobil.) The horizontal resolution and seismic positioning uncertainty are around 50–75 m, with the seismic bin size being 25 m. Vertical sampling is 2–4 ms. In total over 250 wells were drilled in the Groningen area.The second study area, a 20 × 20 km survey, is located in the Dutch offshore (), on the Cleaver Bank High (CBH), directly north of the Broad Fourteens basin, and is in close proximity to the Sole Pit Basin and the Central Graben. During the Late Carboniferous, Saalian tectonic phase, the CBH was tectonically active and uplifted. Between the Permian and the Middle Jurassic the CBH was quiet, but during the Late Jurassic to Early Cretaceous the CBH was uplifted and deeply eroded (). However, the study area is located outside the most important zone of erosion (Top Zechstein is dominated by an NNE striking salt wall with minor grabens on both sides in the suprasalt deposits. The salt wall has a listric fault and asymmetric graben above its crest (). The subsalt top Rotliegend horizon shows three fault trends (). The first, NW striking trend is associated with Variscan wrench faulting (). The second, NNE trend, is parallel to the salt structure and may be related to the Central Graben/Broad Fourteens Basin fault system (see ). The third trend is a minor, approximately E–W fault system, which is related to the northern border of the Broad Fourteens basin.Depth and thickness maps of the suprasalt deposits show that the faults parallel to the NW trend clearly influenced the younger deposits (cf. The seismic data used for this area were also provided by NAM. The horizontal resolution and seismic positioning uncertainty are typically around 50–75 m (as the seismic bin size is 25 m). Vertical sampling is 2–4 ms. Well control is provided by a dozen production and exploration wells.In addition to the interpretation of the Z3 stringer, subsalt and suprasalt reflectors and major faults were used to establish a geological framework. We used the seismic interpretation package Petrel 2005 and Petrel 2007 (Schlumberger).Starting from an existing interpretation (Joris Steenbrink, personal communication, 2007) in the Groningen area, the complex geometry of the stringers was studied by detailed manual interpretation. The stringers were interpreted on a 100 × 100 m seed grid, followed by seeded autotracking with a very low level of tolerance. This means in practice that the interpretation software only extends the manual interpretation a few lines away from the seed. In Groningen the Z3 stringer can be mapped with high confidence (). The occurrence of overlapping stringers in the offshore area () provides a challenge when interpreting horizons because of the limitation of a single z-value at any x–y location in the interpretation software.In both areas the stringer is locally less clearly defined or absent: either only one loop is visible or there is no reflector visible at all. A stringer was only interpreted in locations where both loops were visible. In the surface-building interpolation step this resulted in two continuous and overlapping surfaces in the areas without a reflector. In the Groningen area, we used the imaging resolution to define a threshold value (5 m) for the distance between these surfaces to locate “holes” in the stringer (Figs. a). The exact value of this threshold is not critical because the transition from seismically visible stringer to zones without visible stringer is abrupt. These holes correspond closely to the areas without well-defined stringer reflectors in cross-section (compare, for example, Figs. In the offshore area we completed two different interpretations of the stringers. First, stringers were interpreted as a horizon, with very low tolerance autotracking around the manual picks. Here we consistently chose the higher stringer in case of overlap. A continuous surface was thus created over gaps in the interpretation.In addition, we manually interpreted stringers using the “Fault Interpretation” routine of Petrel in cross-sections, at 100–200 m spacing to define a point cloud of 54 000 points. Although no realistic surface can be created in the interpretation software from these data points, this allowed interpretation of overlapping and near-vertical parts of the visible stringers and was used to study more complex 3D structures (see Section 4.3).The Base Zechstein map (top Z2 anhydrite, a) shows a number of basement faults with offsets up to 300 m, while the top Zechstein map (Although the original stratigraphic position of the Z3 stringer is about 200–300 m below top Zechstein, in its present geometry () its position varies from very close to top Zechstein, to very close to top Z2 anhydrite. In the central area (away from the salt pillows) the enveloping surface of the stringer is sub-horizontal and approximately in its original stratigraphic position (c). In what follows, we focus on areas with anomalously thick stringers, on areas where stringers are not visible, and on folds of different wavelengths, orientation and amplitudes (The observed stringers usually have a rather constant thickness of around 40 m, with areas of increased thickness (up to 150 m). We call these TZs (Thicker Zones; ). A stringer thickness distribution map is shown in Figs. a (areas with no visible stringers – as defined above – are in grey).We studied seven wells that penetrate the visible Zechstein stringer. In those wells that penetrate a TZ, the stringers are 80–125 m thick with two strong Gamma Ray peaks: one in the basal claystone and a second peak at the top. In all wells penetrating seismically visible stringers outside the TZ, stringer thickness is around 50 m with only one GR peak in the basal claystone.d), a clearly defined regional network of up to 400 m wide, long, branching TZ is seen throughout the area (red and yellow colours in Figs. e and f). These zones are curved and are connected at triple (rarely quadruple) junctions. TZs are synclinal (never anticlinal) in shape and lie below the stringers’ enveloping surface (Figs. In more detail, in cross-section, TZ can be (I) symmetric, with two overlapping top stringer reflectors (top Main Anhydrite) and top stringer slightly raised in the thickest section (a–c and e), or (II) asymmetric, with the stringer invisible on one side of the TZ (d–f). In the Type I TZs, structures that appear to be internal thrusts are interpreted (c), but it cannot be ruled out that this is an imaging effect.There is at best a weak correlation between Rotliegend faults and TZ. In the south of the study area TZ is only roughly parallel to the Rotliegend fault trend, while the NE–SW oriented TZ along the western edge of the survey is perpendicular to this trend (compare a and d). There is, however, a clear correlation between TZ and the topography of top Zechstein (compare b and e), as well as in local depressions of the stringer (e and f): TZs are dominantly found in areas where top Zechstein is deep.The enveloping surface of ZE3 forms 10 km-scale fold structures which are harmonic with the top Zechstein (). TZs are located in the synclinal hinges of these large-scale folds. In the area marked with a star in a), a TZ crosses an anticlinal structure. Here the amplitude of the anticline is much lower than in the rest of the fold.On a smaller (km) scale, fold structures also exist, most clearly expressed in areas away from TZ. The seismic section in b). A number of low wavelength (200 m) folds are observed in the stringer, while the Z2, Z4 and Z5 are not folded in the same way. Since the latter are mechanically coupled to the basement and top salt, respectively, this suggests that the folding of the Z3 solely results from movements in the salt. These folds have fold axes perpendicular to the axis of the diapir (c shows a similar structure of folds in an area with fold axes running up the crest of a salt pillow and wavelengths of a few 100 m. The observed fold axes in generally are in the dip direction of the Z3 enveloping surface.Areas with No Visible Stringer reflector – called ANVIS – are common in the Groningen study area. In some cases, TZ is spatially related to ANVIS in sub-parallel, occasionally en-echelon, zones (ANVIS, key 1 in a). In other areas, ANVIS have no apparent relationship with TZ but are related to the E–W striking salt diapir in the north of the area (key 2 in d). These ANVIS are clearly elongated and sub-parallel to the axis of the diapir on its southern side, most clearly expressed where no TZ is present. On the northern side of the diapir ANVIS are also present but here the presence of TZs do not allow a clear attribution to TZ or the diapir. The cross-section in e cuts through these gaps showing a series of asymmetric monoclines with lengths between 250 and 1000 m. The Z4 and Z5 reflectors in this area do not show the same structure, suggesting once again that the Z3 salts serve as a decoupling layer.In four wells penetrating ANVIS, two wells encountered no stringer. In one case a very thin (10–20 m) stringer was encountered. Finally, one well in an ANVIS penetrated a 100-m long, steeply dipping stringer section, this shows that the stringers’ visibility can be limited by their dip.In the east of the survey, the roughly N–S oriented TZ intersects the E–W diapir (Figs. d). Here the ANVIS which are sub-parallel to the diapir crest further west curve towards the diapir and the diapir is also narrower here. This may suggest that TZs are older than ANVIS.Compared to the Groningen area, the Z3 stringer in the offshore study area has a much more complex structure. In seismic cross-sections (d) the visible Z3 stringer is less continuous and varies strongly in depth over short distances. The stringer is more intensely folded and frequently offset vertically over more than half of the total Zechstein thickness. The Mesozoic tectonic inversion of the Broad Fourteens basin is manifested in several inverted basement blocks in the area (The individual interpreted stringer fragments in this area are relatively chaotic, with a weak trend of their long axis parallel to the axis of the salt wall in the east (c and d). Additional trends are recognized, for example, the alignment of stringer fragments along the NW trend of basement faults in the NW of the study area () and along the NW–SE trending faults in the overburden (The stringer surface which was created by horizon interpretation, autotracking and surface interpolation, is, as expected, quite similar to the point cloud interpretation which was created using the fault interpretation routine (in areas with no overlapping stringers – compare Figs. ). The point cloud, however, shows locally a high level of additional complexity (). TZs were not observed in this survey; where the stringer could be interpreted, the top and bottom reflectors were about 40 m apart. ANVIS occurred in a larger fraction of the area than in Groningen.At the regional scale, top Rotliegend shows several inverted blocks (a) and top Zechstein shows the NE–SW striking salt wall with two secondary, NW–SE striking graben structures which are all clearly associated with fault systems in the overburden (Figs. ). On average the stringer is several hundreds of meters higher on the west side of the salt wall than on the east side. The Z3 stringer is very close to the top of the inverted basement blocks in almost every case studied (see also . There are a number of clearly defined structural elements. Firstly, the enveloping surface of the stringers follows the large-scale salt wall (Figs. b). In the areas where the stringers can be autotracked and dip less than 30° (areas with bright colours and grey dots in ) the surface shows gentle, open folds (F in Figs. e) with a wavelength of around 400 m and amplitude less than 200 m. Fold axes are usually curved and do not show a clear preferred orientation. Second, the surface is frequently offset (O in ) along steep (steeper than 35°, grey in ) discontinuities. This offset can be more or less symmetric (graben like, G in ). The offsets can be not only small and become zero towards the tip of the steep zone, but also up to 1000 m vertically, offsetting the stringer across almost the whole Zechstein. In map view, these discontinuities have a peculiar, curved morphology (f) with no clearly defined preferred orientation. In the point cloud some of these offsets resemble ductile ruptures (c). Fold axes are often at right angles to these offsets.Stringers are usually not visible in the steeply dipping discontinuities described above (see ). Other ANVIS like those in the east of the survey () in the area have been described above. More striking are parts of the stringers which form complex, three dimensional, tight to isoclinally folded surfaces defined by the interpreted point cloud (). Visualization of these surfaces is difficult because of software limitations, and analysis of the structures was best done by interactively rotating the point cloud and defining the folded surfaces by visual inspection. In a, b and d some of these structural elements are shown as the raw point cloud together with a basic interpretation. It is striking that these structures have been only observed in direct proximity of inverted basement blocks. are rotated to get a better feel for the 3D structure are available on our website and YouTube Channel (). Here also a direct comparison between the horizon and fault interpretation of the offshore study area can be found.Both in the Groningen area and offshore study area, folds are observed at different scales and with different morphologies. The structures can be classified in the following four different groups: (1) diapir-scale folds, (2) “constriction” folds, (3) strongly non-cylindrical structures and (4) isoclinal folds inside the salt structures. We will now discuss the interpretation of these in detail. We note here that our descriptions follow the “classical” fold description methods, focusing on 3D extrapolations of essentially 2D observations (). Defining the folded surfaces using differential geometry (At the regional scale of the salt structures the Z3 stringer resembles the top Zechstein surface (). This produces large, regional folds in the stringer’s enveloping surface which are harmonic with the large-scale salt structures and are therefore interpreted to reflect the large-scale flow paths during salt tectonics. The other three types of structures are disharmonic to this movement and interpreted to result from local deviations from the relatively simple deformation pattern expected (see When the fold axes are steeply dipping, folds in the Groningen area () are interpreted as “constriction folds” similar to “curtain folds” (but the latter have vertical fold axes). These reflect simultaneous shortening and extension in the surrounding salt (cf. d) and implies that stringers underwent significant internal deformation and do not deform in a fully brittle fashion. Similar structures were described in many other studies in salt mines (see, for example: are very similar to the observed structure in Groningen directly south of the E–W diapir (here the pattern of holes in the stringer around the E–W diapir is similar to the pattern of ruptures in the analogue models, see also d). The open folds with sub-horizontal fold axes (Figs. ) in the Western Offshore study area can also be interpreted to result from a constrictional process as the stringer moved along with the salt into the salt structure (see Two types of non-cylindrical folds (see for definitions) are observed. The first type is illustrated in Figs. a with a relatively large anticlinal stringer fold, harmonic to top salt, which is bisected by a TZ at high angle to the fold axis. The axial surface of this fold is (approximately) planar, but the fold hinge is lower where the TZ crosses this structure. We infer that TZ had a mechanical effect which influenced the buckling process and locally led to a longer wavelength, lower amplitude fold structure (). Experimental and numerical modeling of the development of folds in layers with strongly variable initial thickness () is in agreement with this, but much more work is needed to allow a full interpretation.The second type of non-cylindrical folds is observed in the offshore area. Here large areas of low amplitude folding with non-cylindrical axial planes, but relatively horizontal fold hinges (). These are very similar to structures formed in deformation of a constricted layer () and constrictional plasticine experiments (Although reliable interpretation of steeply dipping, complex structures is difficult, it is tempting to interpret tight to isoclinal folds in the stringers in a number of areas in the Western Offshore ( closely resemble those documented in large salt structures by underground studies (cf. ), and we propose that the high quality of our seismic data allows the 3D mapping of these structures in our study areas. These folds show strongly curved axial planes and fold axes and are associated with ruptures in the stringer (see below). In the regions where these structures dominate, deformation at the scale of the salt dome was clearly much more heterogeneous.The thickness of the Z3 stringer in the Groningen area and in the Western Offshore is generally about 40 m. Similar Z3 units are between 30 and 50 m thick in Poland, UK and Germany (). This shows that the thickness of the undisturbed Z3 stringer is relatively constant over large areas of the Zechstein basin, but local deviations are present like in the Blijham area (south of the Dollard Bay, ) where the stringer is 80–90 m thick, perhaps related to the proximity of the Off-platform shoal (). Complex thickness variations are observed on the hydrocarbon field scale. Examples include the Vries area, south of Groningen (see for location) and several other locations in the Netherlands (Thickened zones (TZs) in this study were only observed in the Groningen study area and their increased thickness in seismic data is consistent with well data. They form a branching network (Figs. ) of approximately in 4–5 km wide synclines (never anticlines), where the TZ is about 400 m wide and 300 m deeper than the surrounding stringer (f). In synclinal areas between salt pillows where the present-day top Zechstein is deep, TZs are more frequent than in areas where this horizon is shallow (Figs. ). In cross-section at smaller scale, we observe internal overlapping reflectors and clear geometrical difference with respect to the surrounding thinner stringer (). In many cases, there is an ANVIS zone parallel to the TZ where the stringer is not visible.The pattern of TZ and the peculiar internal structure cannot be explained by salt tectonics and basement-related faulting alone. In the following section we present one possible interpretation of the events leading to the development of the observed structures, other interpretations are certainly possible. We assert the formation of these structures to be the product of sedimentary and diagenetic processes, modified by deformation in the flowing salt. Keeping in mind the fact that the seawater during deposition of the Z3 stringer (clay, carbonate, anhydrite) is undersaturated with respect to the underlying Z2 halite, it is clear that any pathway for circulation of this seawater into the underlying Z2 halite will lead to strong dissolution, a system of karst caves and collapse structures which enhance further dissolution. Several authors describe sinkholes and collapse breccias related to karst in Z1 platform carbonates (). The early evolution of salt giants is still enigmatic (), particularly with respect to karstification (). It is also possible that subtle (tectonic) movements during Zechstein times () played a role in the development of the regional depressions and the incipient rupture of the Z3 clay, allowing solution of the Z2 evaporites.Our preferred scenario involves the formation of 4–5 km wide depressions after deposition of the Grey Salt Clay (otherwise a thicker clay would later prevent groundwater circulation and formation of karst). In these depressions, during or after the deposition of the Z3 Anhydrite (but before the start of deposition of Z3 Halite from hypersaline brines), minor salt deformation led to the formation of an open fracture system in the Z3, allowing NaCl-undersaturated brines to circulate into the Z2 halite. This led to the formation of an extensive network of dissolution channels and collapse of overlying Z3. Sections of the Z3 stringer on the edges of the collapse zones ruptured and slid down the slope forming the ANVIS sub-parallel to TZ (). The slides collect at the base of the valleys resulting in the strongly deformed and tectonically thickened Z3 (TZ) observed in cross-sections (). The observed structures are more compatible with early sliding structures (The incursion of seawater into the Southern Permian Basin that is related to the deposition of the Z3 Stringer represents the most important Zechstein transgression and flooding of the Southern Permian Basin (). Little is known of the structures formed during the re-flooding of the centre of large salt basins, as most information is from the edge of the basins (). It is, however, easy to imagine currents (perhaps from the off-platform high; ) during Z3 carbonate deposition having formed erosional or karst structures when (halite-undersaturated) seawater was flowing over a halite substrate covered by a thin Grey Salt Clay layer. Sub-areal karst in salt forms the same structures as carbonate karst, but at faster rates (). In fact, the shape of the water-filled caves mapped by , including triple junctions, is remarkably similar to TZ in Groningen.The formation of thicker zones in a deforming layer has significant implications for the evolution during salt tectonic deformation. Since the stringer is more competent than the surrounding host, the thicker zones of these layers represent stronger zones in this layer and will therefore be more difficult to deform. This will, for example, lead to different dominant wavelengths in folds (). The thicker zones clearly influence the patterns of ANVIS and fold geometry in the study area (Figs. The density of natural halite is about 2200 kg/m3, anhydrite is about 2900 kg/m3, and dolomite is about 2850 kg/m3 (). This density difference makes the stringers negatively buoyant and they tend to sink if the rheology of the surrounding salt allows this to happen at geologically significant rates.The question of gravity-induced sinking of stringers has been a subject of much controversy (). Analogue and numerical models show that dense blocks in diapirs sink when the diapiric-rise velocity is not sufficiently high to keep the blocks in the diapir (). However, estimates of the in situ rheology of salt vary widely (cf. ), and consequently there is much uncertainty about the expected in situ sinking rates. It would therefore be useful to obtain additional constraints based on our results on this process. Diagnostic structures which allow separating salt tectonics-related and gravitational sinking-related processes are difficult to define but could consist of a correlation between vertical position of the stringer and stringer size (). Such a correlation is not apparent in our dataset. Keeping in mind that the major salt tectonic movements occurred before the end of the Cretaceous, the fact that many of our stringers are located high in the Zechstein (about 250 m from top Zechstein) can be used to calculate an upper bound of sinking velocity. Considering the case that these stringers were first ruptured into individual bodies, tectonically displaced to top Zechstein and then started to sink 65 Ma, the upper bound velocity is around 4 m/Myr. This is much lower than the rates suggested by some analogue and numerical models and is in agreement with the wide variation of salt rheologies used by the different studies, calling for more work to resolve this question.). High strain, folded intervals are often located close to less strongly deformed salt. This indicates that flow in evaporites can be strongly partitioned, possibly associated with the strong topography of the inverted basement blocks (compare with: ). Also the presence of shear zones in salt can add to the complexity (There are four possible explanations for the absence of a stringer in seismic sections. First, it may not have been deposited in this location. Second, the layer may be too thin to be imaged. Third, it may be discontinuous after being disrupted by tectonic deformation. Fourth, it may be in an orientation, which is too steep to produce a seismic reflection (). The first explanation is rejected considering regionally constant thickness of the Z3. The well data from the Groningen area which were available for this study (see above) do not allow a clear distinction between alternatives two, three and four. However, it is not possible to explain all ANVIS by steep orientation only, because they also occur in areas where the surrounding stringer is shallow-dipping and has no vertical offset.Keeping in mind that a full explanation of ANVIS in the Zechstein requires additional data from many wells and that in some ANVIS steeply dipping stringers were found by drilling, in what follows, we make the reasonable assumption that in areas where the stringers’ enveloping surface dips less than 30°, ANVIS correspond to the absence of a stringer due to tectonic disruption (boudinage). In areas where the stringers’ enveloping surface dips more than 30°, ANVIS can either correspond to the absence of a stringer due to tectonic disruption (boudinage or faulting) or to a tectonically tilted stringer which is not imaged seismically.We interpret some of the fracturing of the stringer in the data to result from boudinage. The best examples of boudinage in our study areas are shown in Figs. c, but the data quality does not allow a detailed description. We will compare our results with published example of boudinage in this section. The process of boudinage is the “disruption of layers, bodies or foliation planes within a rock mass in response to bulk extension along the enveloping surface” (), and a boudin can be described as “a fractured sheet of rock situated between non-fractured (…) rocks” (There is large variety of boudin geometries which can be subdivided both according to their kinematics as well as to the shape of the boudin blocks (see ). Boudinage of brittle inclusions in salt is described by several authors. ). The six published wells of the Gorleben salt structure () penetrate the Z3 stringer for a total of eight times in four wells. In three of these penetrations the stringer is interpreted to be disrupted, faulted or fractured, showing the high fracture density in the Z3 stringer.Studies of boudins to date have almost exclusively been in 2D profiles, and although there is a reasonable understanding of the range of structures occurring in profile view, there is a striking lack of understanding of the 3D morphology of boudins.A 3D exposure of pegmatite boudins in marble () shows boudins that are bound by a set of normal faults formed by reactivation of Mode-I fractures. The most extensive study to date of how boudins evolve in 3D is provided by the model experiments of Zulauf and coworkers (). These experiments employ plasticine and natural anhydrite as strong layers. Here the plasticine is sufficiently ductile to fold when shortened parallel to the layering but at the same time sufficiently brittle to rupture if extended parallel to the layers, much like natural anhydrite. The patterns of this “ductile rupture” are much more complicated than brittle fracture patterns. It is clear that the deviation from plane-strain deformation results in the formation of very complex structures.Since the thickness of the stringer in Groningen (Figs. e) is between 40 and 50 m, the observed boudins have unusually high aspect ratios, in the order of 5–20, much larger than the largest mean aspect ratios described by in siliciclastic and carbonate rocks. In the Kłodawa salt mine, aspect ratios of around 20 are observed in the larger boudins (). The exceptional torn halite boudins in a carnallite matrix () have aspect ratios in the order of 2.5–5 (assuming the view is perpendicular to the extension direction) and also the experimental D1-boudins of have aspect ratios around 5. Observations of boudins of sand in viscous putty () also indicate a high aspect ratio. On the other hand, produced experimental boudins of very brittle anhydrite in halite with aspect ratios of 1.5 ± 1.0. This difference in aspect ratio between the boudins of and “evaporitic” boudins may originate from the different rheologies, but more work is needed to resolve this question.The rheology of anhydrite at shallow depth in the crust is not well known (). We note that during shortening, anhydrite layers in salt domes commonly form concentric folds (see above), which means that the viscosity ratio is rather high, but at the same time boudins (brittle structures) are formed in extension. Corresponding to the present state of research (), our present preferred model is that rheology of anhydrite is controlled by pressure solution and is Newtonian in layer-parallel shortening. However, the high, near lithostatic, fluid pressures, commonly observed in stringers (), allow tensile failure in layer-parallel extension in the anhydrite encased in sealing salt. This model explains why the stringer forms ductile folds as well as brittle boudins at the same time. A similar model has already been proposed for pegmatites encased in marble (). Alternatively, brittle fracturing may be initiated in the underlying more brittle limestone and claystone (sensu Folds formed under plane strain can be used to approximate the relative viscosity of the two layers ( for non-Newtonian fluids. It is tempting to apply a similar method here and to compare, for example, this normalized arc length of folds to the aspect ratio of boudins to calculate the relative viscosity. However, conventional folding theory is based on plane-strain deformation and this is not the case in our field area, so a simple analysis may lead to errors (Stefan Schmalholz, personal communication). In future work we will attempt to constrain the relative rheologies of halite and anhydrite by 3D numerical modeling (In summary, this study has shown early karst-related thickness variations overprinted by complex folding and boudinage, producing the complicated present-day geometry of the Z3 stringer. Possible effects of early karst, diagenesis or gravitational deformation, probably augmented by a local increase in sedimentation of the stringer, formed a network of ruptures and slide-related folds. This resulted in a series of thicker zones in the stringer. During the Early Mesozoic salt flow, the salt moved into the narrow stems of salt walls and diapirs (d). This resulted in the coeval folding and boudinage in the stringer, leading to significantly more complex structures than in plane strain. This complexity is further increased by the thicker zones which are significantly stronger than the surrounding stringer and might form preferred instabilities for folding and boudinage. When the growth of salt structures was halted, the possible sinking of the broken anhydrite blocks added even further complexity, although we do not expect this sinking to have geologically significant velocities.A useful conceptual model for the displacement and deformation of stringers is that of passive and active processes. Passive means that the stringers are displaced by the flowing salt as passive objects and active means that the contrast in mechanical properties causes local instabilities, resulting in deformation, folding and boudinage of the stringers. It is also clear that the temporal evolution of the position or orientation of a stringer in a salt dome has a large effect on the strain history (An interesting observation of this study is the apparent consistent position of the TZ between salt pillows. This is surprising because the TZs are thought to be older than the main phase of salt tectonics and also much smaller than the first-order structures. If this correlation can be shown to be real by additional observations of the same correlation in different areas, it would point to a subtle feedback mechanism in which the early internal structure of the salt basin can have a large effect on the evolving structures. In previous work, the main drivers for this were thought to be deformation on basement faults, regional tilt of the basin and lateral differences in overburden stress () but not the internal structure. An interesting concept emerging from this project is then that active deformation of the stringers during the early life of the Zechstein salt giant can initiate feedback processes which control the evolution of the later instabilities and the growth of large-scale salt structures (). One mechanism for this could be that weak early deformation of TZ affects the topography of the top of the salt, creating a small topography which controls sediment architecture and evolution of overburden load (cf The Zechstein of the Central European Basin has a fascinating, complexly folded structure which is known only in a few cases based on mining data. The methods outlined in this study can be used at a regional scale because the Z3 reflector is present almost everywhere and can be mapped in 3D using seismic reflection data which are available over very large areas (Using the same 3D seismic data, the kinematics of the suprasalt sequence can be relatively accurately reconstructed using palinspastic reconstruction techniques (e.g. ). Rheology of the Zechstein salt is not completely understood but is constrained by a large amount of data. It is clear that even relatively pure halite can have strain rates variable by two orders of magnitude at the same differential stress and temperature () making the mechanical structure of the Zechstein strongly layered.This dataset can now be combined into geomechanical models starting from the reconstructed original structure of the Zechstein including stringers, with the palinspastically reconstructed kinematics of top salt as kinematic boundary conditions, and the final structure compared with the interpretation of the Z3 stringer from 3D seismic. In addition, differential stress can be measured by subgrain size piezometry in drill cores. This method therefore produces models which can be tested against observations and, if the test is passed, can predict the internal structure of the salt body in the whole volume.We mapped complex internal structure of salt domes using 3D seismic reflection data. This opens the possibility to study the internal structure of the Zechstein and other salt giants in 3D using this technique, exposing a previously poorly known structure which is comparable in size and complexity to the internal parts of some orogens.The evolutionary sequence of sedimentary and diagenetic processes, followed by deformation at different scales results in early thickness variations, overprinted by a range of fold and of boudin structures.Flow heterogeneities in salt caused by presence of thickness variations in stringers are interpreted to lead to subtle topography of top salt, which in turn influenced sedimentation and deformation of overburden.Our observations show no conclusive evidence for significant gravity-induced sinking of stringers.The methods used in this study can be combined with numerical modeling to predict the internal structure of salt bodies without extensive drilling or construction of galleries.Fracture mechanics and microstructure in NiTi shape memory alloysCrack extension under static loading in pseudoplastic and pseudoelastic binary NiTi shape memory alloy (SMA) compact tension (CT) specimens was examined. Two material compositions of 50.3 at.% Ni (martensitic/pseudoplastic) and 50.7 at.% Ni (austenitic/pseudoelastic) were investigated. The SMAs were characterized using differential scanning calorimetry to identify the phase transformation temperatures and tensile testing to characterize the stress–strain behavior. A miniature CT specimen was developed, which yields reliable critical fracture mechanics parameters. At 295 K, cracks propagate at similar stress intensities of 30±5MPam into martensite and pseudoelastic austenite. Integrating the miniature CT specimen into a small test device which can be fitted into a scanning electron microscope shows that this is due to cracks propagating into regions of detwinned martensite in both materials. Investigating a pseudoelastic miniature CT specimen in a synchrotron beam proves that martensite forms in front of the crack in the center of the CT specimen, i.e. under plane strain conditions.Microstructures and properties of NiTi shape memory alloys (SMAs) have been described in detail The elementary processes governing crack initiation and crack instability are not well understood. Thus, McKelvey and Ritchie The crack tip regions of NiTi SMA CT specimens have been studied using a variety of methods, including optical microscopy The present work has four aspects. First, it provides a comprehensive treatment of the fracture mechanics characteristics of three material states in NiTi: martensitic, pseudoelastic (austenitic, but prone to the formation of stress-induced martensite) and austenitic (without the potential to form stress-induced martensite) NiTi. Secondly, we compare and discuss the response of miniature CT specimens to mechanical loading and provide critical data for crack extension under static loading for all three material states. We then study the evolution of microstructures in front of cracks, which grow into martensitic and pseudoelastic NiTi using in situ experiments in a scanning electron microscope. Finally, we perform in situ synchrotron experiments to clarify whether stress-induced martensite forms in front of the central part of a crack in a thick pseudoelastic CT specimen (i.e. under plane strain conditions).Two NiTi SMAs with 50.3 at.% Ni (martensitic/pseudoplastic at room temperature) and 50.7 at.% Ni (austenitic/pseudoelastic at room temperature) were purchased from Memory Metalle, Weil am Rhein. Both alloys were subjected to thermomechanical treatments (forming and aging steps) including a final 6 min heat-treatment at 500 °C for the pseudoelastic alloy. Both materials were characterized using differential scanning calorimetry (DSC), uniaxial tensile testing, CT fracture testing, optical and scanning electron microscopy (SEM), and synchrotron diffraction.DSC was performed on both SMAs using a TA Instruments’ DSC 2920CE machine in a temperature range from 123 to 423 K at a heating/cooling rate of 10 K min−1, whereby specimens were held for 5 min at the maximum und minimum temperatures. The details of DSC testing have been described elsewhere a shows the DSC chart of the martensitic material investigated in the present study. When forward and reverse transformations represent one-step events, the characteristic transformation temperatures are defined as Ms (start of the martensitic transformation), Mf (temperature where the forward transformation is completed), As (start of the austenitic transformation) and Af (temperature where the reverse transformation is completed). For the martensitic alloy investigated in the present study, the characteristic temperatures were obtained as Ms |
= 317 K, Mf |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.