University
stringclasses 19
values | Text
stringlengths 458
20.7k
|
|---|---|
Colorado School of Mines
|
7.3 REE Liberation
REE liberation was determined through the AMICS analysis as well. A particle was
considered to be liberated if more than 50% of its surface area was composed of a REE mineral,
in this case bastnäsite, parasite, and cerianite. Fifty percent was selected as the minimum threshold
for considering a particle liberated based on a recommendation of Dr. Paul Miranda for flotation
work. Liberation can provide insights into the efficacy of flotation collectors. Table 7.7, Table 7.8,
and Table 7.9 show the REE liberation for the ore sample, collector 2 locked cycle testing products,
and collector 5 locked cycle testing products, respectively. While liberation was reported for
bastnäsite, parasite, and cerianite, only liberation for bastnäsite will be discussed because it
represents the majority of REE bearing minerals in this material.
Table 7.7 - REE mineral liberation for the ore sample
Mineral Liberation
Bastnäsite 78.62
Parisite 64.55
Cerianite 81.89
Table 7.8 - REE mineral liberation for the collector 2 locked cycle flotation products
Mineral SFT RCFC CFT SFC RCFT
Bastnäsite 50.36 95.28 73.73 68.2 18.2
Parisite 18.52 52.03 53.73 28.67 4.14
Cerianite 61.64 53.07 62.33 70.26 10.58
93
|
Colorado School of Mines
|
Table 7.9 - REE Mineral liberation for the collector 5 locked cycle flotation products
Mineral SFT RCFC CFT SFC RCFT
Bastnäsite 75.68 59.81 52.55 60.47 77.61
Parisite 70.14 29.32 47.45 51.43 56.13
Cerianite 49.20 58.04 35.27 49.2 60.81
The liberation values presented above represent the percentage of the total surface area of
that mineral which was liberated based on the above description. Based on this description, 78.62%
of the total surface area of bastnäsite present in the ore was at a liberation of at least 50%.
Liberation can be compared within individual stages as well as for end of locked cycle test
products.
7.3.1 Scavenger Flotation Comparison
Collector 2 scavenger flotation showed 50.36% bastnäsite liberation for the tailings and
68.2% liberation in the concentrates. This means that the process is not perfect in its separation
but there is still an upgrade in the quality of the material exiting in the concentrate of the scavenger
flotation stage. Collector 5 scavenger flotation showed 75.68% bastnäsite liberation for the tailings
and 60.47% liberation in the concentrate. This indicates that collector 5 is not selective towards
bastnäsite as the liberation is decreasing in the concentrate versus the tailings.
7.3.2 Recleaner Flotation Comparison
The recleaner stage of locked cycle flotation provides another point of comparison between
collector 2 and collector 5. For collector 2, bastnäsite liberation in the tailings was reported at
18.2% and liberation in the concentrate in the concentrate was 95.28%. The large difference
between the liberation of the tailings versus the concentrate indicates that the collector 2 separation
is very efficient. Collector 5 recleaner bastnäsite liberation returned 77.61% in the tailings and
94
|
Colorado School of Mines
|
59.81% liberation in the concentrate. This indicates that the collector 5 separation is not efficient.
Particles with significant liberation are being left behind in the tailings of the recleaner stage.
7.3.3 Summary
The liberation data presented provides insight into the performance of the collectors tested
through locked cycle flotation. Through the analysis of the liberation data, it is shown that collector
5 leaves behind well liberated materials in flotation. This can be said because the liberation of the
tailings is higher than that of the concentrate. For collector 2, this is not the case. Collector 2 returns
increased liberation rates in the concentrates for both stages tested. Overall, the liberation data
further confirms the strong separation of collector 2 while also confirming the lack of flotation
strength for collector 5.
7.4 AMICS Mineralogy Images
Middlings materials have been identified as a point of losses for a significant portion of the
materials in the locked cycle flotation circuits. Using the identified backscatter images that were
provided by the AMICS analysis, it is possible to get an idea of the makeup of the materials in the
middlings portion of the flotation tests. Determining whether materials in the middlings are
composed of locked particles or liberated particles allows for better planning for future flotation
flowsheets and allows a better determination of whether a regrind circuit should be included to
further increase liberation. The middlings materials from locked cycle flotation include the cleaner
flotation tailings (CFT), scavenger flotation concentrate (SFC), and recleaner flotation tailings
(RCFT).
Figure 7.4 shows the identified image from AMICS for the CFT stage of flotation with
collector 2. This stage showed 73% liberation. From this image, several bastnäsite particles (red)
are fully liberated. Several particles of bastnäsite are also locked within larger particles of calcite
95
|
Colorado School of Mines
|
CHAPTER 8: CONTACT ANGLE STUDIES
Contact angle studies are a way to measure the interaction between air bubbles in a flotation
cell and the collector adsorbed to the surface of the minerals of interest. The contact angle refers
to the angle of contact between the flat surface of a smooth mineral and an air bubble applied to
its surface. It is possible to see the effect that different collectors have on the surface of minerals
of interest using these techniques. For more information on sample preparation and testing
procedures, refer to Chapter 3.
8.1 Bastnäsite
8.1.1 No Collector
The no collector case was used as a control case for the contact angle studies of Bastnäsite.
Figure 8.1 shows the mean contact angle vs time for Bastnäsite with no collector. The graph begins
with a horizontal line that is representative of the approaching contact angle for the experiment.
The line then has a constant slope up until the point where the bubble detached from the surface
of the Bastnäsite mineral. This range is the receding contact angle range. The bubble did eventually
detach from the surface of the mineral unlike some of the other tests. The ultimate contact angle
for this experiment was determined to be 9.54°. Figure 8.2 shows an image of the bubble in contact
with the surface of the Bastnäsite mineral without the presence of any collector.
98
|
Colorado School of Mines
|
8.1.2 Collector 2 Flotation Concentration
The conditions used in this contact angle experiment matched those used in flotation
experiments using Collector 2. The concentration of collector in solution was 0.0075M, the pH
was set to 8.5 using soda ash as the modifier, and the whole solution was raised to 80° C. The
polished mineral was then submerged in the solution for 10 minutes to match the conditioning time
used in batch flotation.
Figure 8.3 shows the mean contact angle versus time for a polished Bastnäsite mineral
surface in collector 2 solution. The graph can be split into four discreet parts. From 1.25 seconds
to 2 seconds are the approaching contact angle. From 2 to 3.25 seconds is the bubble held in place
by the syringe against the surface of the mineral. The range between 3.25 and 4.25 seconds is the
receding contact angle where the syringe is being pulled away from the surface and the rest of the
figure shows the period where the bubble rests on the surface without any syringe contact.
Based on these zones, the contact angle for this experiment was taken from the zone where
the bubble has no contact with the syringe. This point would be the most like what would be found
in a flotation experiment. The mean contact angle at this point was determined to be 47.97°. Figure
8.4 shows the image of an air bubble attached to the surface of Bastnäsite after the syringe was
removed from it.
100
|
Colorado School of Mines
|
8.1.3 Collector 5 Flotation Concentration
The conditions used in this contact angle experiment matched those used in flotation
experiments using Collector 5. The concentration of collector in solution was 0.001M, the pH was
set to 9 using soda ash as the modifier, and the whole solution was raised to 80° C. The polished
mineral was then submerged in the solution for 10 minutes to match the conditioning time used in
batch flotation.
Figure 8.5 shows the mean contact angle versus time for a polished Bastnäsite mineral
surface in Collector 5 solution. This figure can be split into three distinct sections. From 1 to 1.75
seconds is the approaching contact angle zone, the zone where the bubble was being lowered to
the surface of the mineral. The next section is from 1.75 to 3.25 seconds where the bubble was in
equilibrium with the surface and the syringe, this is the zone from which the contact angle was
taken. The final zone from 3.25 seconds to the end of the figure is the receding contact angle. This
zone describes the area in which the bubble is being pulled away from the surface of the mineral
until it eventually pulled off completely.
Based on this figure, the contact angle for the surface of Bastnäsite with Collector 5 was
determined to be 29.98°. Figure 8.6 shows an image of an air bubble in contact with the surface of
Bastnäsite submerged in Collector 5 solution.
102
|
Colorado School of Mines
|
8.1.4 Bastnäsite Contact Angle Summary
In summary, effects of collectors on the surface of a polished Bastnäsite mineral were
noticed. Table 8.1 shows the three different conditions as well as their mean contact angles. Based
on the contact angle experiments, Collector 2 has the best performance. The higher the contact
angle, the more hydrophobic a collector makes the surface of the mineral. The contact angle
experiments support the results from flotation in which Collector 2 performed more strongly than
Collector 5.
Table 8.1 - Summary of Bastnäsite contact angle experiment results for the three conditions
tested
Mean Contact
Condition
Angle (degrees)
No Collector 9.54
Collector 2 47.97
Collector 5 29.98
8.2 Calcite
8.2.1 No Collector
Figure 8.7 shows the mean contact angle versus time for Calcite without any collector
present. The software for this experiment returned erratic results. Because of this, a polynomial
trendline was added to the figure to show the average contact angle versus time. From the data
gathered from this experiment, the mean contact angle for an air bubble in contact with Calcite
without any collector present was determined to be 14.07°. Figure 8.8 shows an image of an air
bubble in contact with the surface of Calcite under the described conditions.
104
|
Colorado School of Mines
|
CHAPTER 9: ECONOMICS
One of the primary goals of this research was to further test collectors that showed promise
for the flotation of basntäsite from both a technical and economic perspective. Because these
collectors were previously only tested using single stage bench flotation, locked cycle flotation
testing was performed to get a better sense of how these collectors might perform in a full-scale
plant environment. This chapter covers the economic analysis for a flotation plant using the
collectors tested within this study and comparing them to the process which is in place and used
at Mountain Pass. Two values of ore and grade were considered for this assessment. These
included a conservative case using the grade and recovery from the concentrate production method
and a best case using the highest grade and recovery reported from the locked cycle testing.
9.1 Assumptions
Several assumptions must be made in the calculation of economics for this project. The
first assumption made for this economic model is the production from the mine. The production at
Mountain Pass mine is roughly 1800 tonnes per day of materials from the mine going into the
flotation plant for processing. For this economic model, the daily processing plant feed rate of
material was assumed to be 2000 tonnes per day. This value was chosen because it closely matches
the actual production of the mine and is also a production point for which many models regarding
operating costs and capital costs exist. CostMine 2013 was used for the calculation of operating
and capital costs involved for the mining and flotation [57]. Because an outdated version of
CostMine was used, the values needed to be updated to current costs. For this, the CostMine
indexes document was used [58].
109
|
Colorado School of Mines
|
Another assumption made in this economic study was that all the products from the
flotation mill would be sold as purified oxide versions of their respective metals. While this is not
how the materials are produced and sold at the Mountain Pass mine, this allows for a simplified
comparison of the results of the economic studies.
This economics assessment assumes that all hydrochloric acid is consumed and there is no
recycle of acid. Acid consumption is a major area of focus for this flotation work because higher
selectivity against calcite can greatly decrease the processing cost in downstream stages.
The final assumption that was made for this study is regarding the price for collector 2 and
collector 5. Collector 2, while commercially available, is only sold in small quantities for
laboratory research purposes. This makes the price per kilogram higher than the price would be if
it was produced on a bulk scale for industrial purposes. For the purpose of this study, a quote for
the price of hydroxamate acid was used in place of the price for collector 2 [44]. Hydroxamate
acid is a major constituent of collector 2 and can therefore be used a cost estimate point. Collector
5 is not commercially available and therefore a price must be estimated based on the methods for
synthesis of this collector. A price for production of collector 5 was provided by Marshallton Labs
(the synthesizer of collector 5 for this study) at a value of approximately $120,000 per metric
tonne.
9.2 Costs
Several costs are considered for the economics model. The costs can be split into two major
categories, capital costs and operating costs. Capital costs are incurred right at the start of the
project and include things such as building costs, equipment costs, construction labor, etc. Table
9.1 shows the capital costs associated with the flotation mill.
110
|
Colorado School of Mines
|
Table 9.12 - Collector 5 economic assessment results
Collector 5 Results
NPV: ($947,342,000)
PBP Days: N/A
IRR: N/A
Table 9.13 - MP economic assessment result
MP Results
NPV: $3,261,470,000
PBP days: 63
IRR: 683%
Of the three conditions tested, the poorest performer was collector 5. Collector 5 performed
poorly in the flotation studies and the poor performance is shown in the economic study as well.
The poor recovery of REEs in flotation with collector 5 leads to less material which can be sold.
Poor grade and recovery from flotation paired with the most expensive reagent in collector 5 leads
to the collector never being able to generate enough revenue to repay the investment in the mine
and flotation plant. Flotation with collector 5 leads to approximately a negative $947 million net
present value over ten years.
Of the remaining two scenarios tested, the economics of the mountain pass flotation
scenario proved to be more favorable given the flowsheet tested with collector 2. The collector 2
flowsheet returned a positive net present value and payback period of 111 days. The Mountain
Pass scenario returned a net present value of approximately $1.2 billion higher than that of the
119
|
Colorado School of Mines
|
collector 2 test with a payback period of 63 days. The economics of collector 2 could be improved
through the addition of a regrind or hydro cyclone circuit.
9.4 Best Case
9.4.1 Value per Tonne of Ore
Table 9.14 shows the metal oxide prices, grades, and recoveries used for the calculation of
revenue for the economic study and Table 9.15 shows the value per tonne of ore. The grade and
recoveries for collector 2 and collector 5 were taken from the experimental locked cycle flotation
data (Chapter 5), specifically the highest reported grade and recovery for each collector tested.
Grade and recovery for the Mountain Pass scenario were assumed to be 60% REO and 60%
recovery [44].
Table 9.14 - Metal oxide price, elemental grade, and recoveries for collector 2, collector 5, and
the Mountain Pass scenarios
Price Recovery REO Grade
Metal
Collector Collector Collector Collector
$/kg MP MP
2 5 2 5
Cerium 1.65 32.85% 6.78% 29.46%
Lanthanum 1.68 22.21% 4.58% 19.92%
74.7% 41.3% 60.00%
Neodymium 41.90 8.03% 1.66% 7.20%
Praseodymium 45.84 2.88% 0.59% 2.58%
Total 66.90% 13.80% 60.00%
120
|
Colorado School of Mines
|
Table 9.21 - Collector 5 economic assessment results
Collector 5 Results
NPV: ($718,149,000)
PBP Days: N/A
IRR: N/A
Table 9.22 - MP economic assessment result
MP Results
NPV: $3,261,470,000
PBP days: 63
IRR: 683%
Once again, considering the best-case scenario, collector 5 is unable to return a profit and
returns a negative $718 million net present value. The best-case scenario of flotation results
positively benefits collector 2 allowing it to perform more favorably than the Mountain Pass
scenario. The collector 2 best-case returned a net present value of approximately $4.6 billion
dollars, approximately $1.4 billion higher than the net present value of the Mountain Pass case.
9.5 Discussion of Results
Several factors are kept consistent between the different scenarios that were tested. These
factors include the mine, mill, and electrical operating and capital costs. This means that the
differences between the scenarios come down to the costs of reagents and the value of the ore as a
function of the recovery and grade of the flotation. Flotation with collector 5 returned the worst
grade and recovery and had the highest reagent cost by nearly five times. Together, these factors
125
|
Colorado School of Mines
|
CHAPTER 10: SUMMARY OF RESULTS AND DISCUSSION
10.1 Summary of Results
A rare earth element locked cycle flotation study was performed to determine the efficacy
of collectors that were tested previously only in single stage flotation tests. Performing locked
cycle flotation allows a reagent scheme and flowsheet to be tested in a way which simulates the
continuous flotation of a full-scale flotation plant. Locked cycle testing can give a better idea of
the performance of a collector if it were to be used in a real flotation plant. The flowsheet used for
locked cycle flotation testing can be found in Chapter 2.
For locked cycle testing to be performed, recovery curves first needed to be generated to
understand the optimal time of flotation for each stage of the flotation flowsheet. The recovery
curves were repeated for each stage of flotation for both collectors tested. The results for each
recovery curve and the corresponding optimal flotation time used in locked cycle testing can be
found in Chapter 4.
Detailed information from the locked cycle testing experiments can be found in Chapter 5.
Collector 2 locked cycle flotation returned a REO grade between 58.5% and 66.9% with a recovery
between 42.8% and 74.7% while rejecting 78% of the calcite from the feed. Collector 5 locked
cycle flotation returned a REO grade between 13.2% and 13.8% and a recovery between 26.6%
and 41.3% while rejecting 9% of the calcite from the feed. The rejection of calcite is an important
consideration because it affects the downstream reagent consumption in the leaching step of the
rare earth element processing.
The results from the recovery curves and locked cycle testing were plugged into
JKSimFloat to model the flotation testing at a full 2000 tonnes per day scale. More detailed
127
|
Colorado School of Mines
|
information about the setup for the simulation can be found in Chapter 6. Collector 2 JKSimFloat
simulation returned a REO grade of 71.8% and a recovery of 56.8%. Collector 5 JKSimFloat
simulation returned a REO grade of 33.7% and a recovery of 71.7%.
AMICS analysis was performed on the final products produced through locked cycle
testing. The AMICS elemental analysis data was correlated with the data measured at the Colorado
School of Mines. Modal mineralogy assessment determined that the primary rare earth element
mineral was bastnäsite with smaller amounts of cerianite, parasite, and allanite. AMICS analysis
also returned values for liberation. Collector 2 locked cycle testing concentrate returned a
liberation of 95.28% and the tailings returned a liberation of 50.36%. Collector 5 locked cycle
testing concentrate returned a liberation of 60.47% and the tailings returned a liberation of 75.68%.
More information about the AMICS analysis can be found in Chapter 7.
To better understand the disparity in performance between collector 2 and collector 5 in
locked cycle flotation testing, contact angle test work was also performed. Performing contact
angle test work allows for comparisons to be made regarding the applied hydrophobicity of a
collector to the surface of a mineral. For a mineral to float through the froth flotation process, it
must become hydrophobic through the application of a collector. Contact angle testing allows for
this hydrophobicity to be measured. More information about the method for contact angle testing
can be found in Chapter 8. Contact angle tests were performed on bastnäsite and calcite mineral
samples under conditions mimicking the flotation conditions of collector 2 and collector 5 locked
cycle flotation as well as a collector free control. For bastnäsite contact angles, the no collector,
collector 2, and collector 5 mean contact angles are 9.54°, 47.97°, and 29.98°, respectively. Calcite
contact angles for no collector, collector 2, and collector 5 are 14.07°, 4.8°, and 10.08°,
respectively.
128
|
Colorado School of Mines
|
A comparative economic analysis of a flotation mill was performed comparing flotation
with collector 2, collector 5, and the flotation scheme that was employed at the Mountain Pass
mine in California until operations were ceased due to bankruptcy. Directly comparing the
economics allows the full impact of a reagent scheme to be seen. It is important to consider the
impact the cost of a reagent can have on the economics of an operation. Similarly, more efficient
calcite rejection in this material leads to large downstream cost saving. More information about
the economic analysis can be found in Chapter 9. Flotation under the current system at Mountain
Pass leads to a net present value of $3.26 billion with an internal rate of return of 683%. Flotation
under the collector 2 reagent scheme returned a net present value of $2.02 billion with an internal
rate of return of 429%. Finally, flotation with collector 5 returned a negative net present value of
$948 million.
10.2 Discussion of Results
Of all of the collectors tested in previous work by Dylan Everly [44], collector 2 and
collector 5 were chosen to perform extensive locked cycle flotation work. Collector 2 was
primarily chosen because of its strong performance in both grade and recovery while maintaining
a very high calcite rejection. Collector 5 was chosen due to its very high recovery values in
microflotation despite poorer grade results.
Collector 2 performed as well in the locked cycle flotation testing as in the single stage
batch flotation tested previously. Locked cycle testing increased the grade of the material that was
produced at the expense of some of the recovery. The rejection of calcite also was maintained from
the single stage flotation study to the locked cycle testing. The high efficiency of the calcite
rejection is also visible in the contact angle studies performed with collector 2. Collector 2 showed
both the highest contact angle on bastnäsite and the lowest contact angle on calcite. This closely
129
|
Colorado School of Mines
|
matches the results from flotation studies which showed the highest grade and recovery of
bastnäsite and the best rejection of calcite of the collectors tested.
Collector 5 performed similarly in locked cycle testing to its performance in single stage
bench flotation tests performed previously. Collector 5 showed almost no rejection of calcite and
very little upgrade of grade over the head grade of rare earth elements in the ore. Collector 5 floated
all the materials in the ore with little selectivity for REEs, thus leading to a poor grade and a
recovery of approximately 50%. This result is confirmed by contact angle testing with collector 5
where the contact angle, while higher than the no collector case, was not as high as the contact
angle for the collector 2 case. This implies that the collector does not make the bastnäsite minerals
as effectively hydrophobic as the collector 2 reagent. Calcite contact angle also showed a higher
contact angle than collector 2 meaning that collector 5 did not as effectively suppress calcite as
compared to collector 2.
Poor performance in grade and recovery, poor calcite rejection, and high reagent pricing
led to collector 5 performing poorly in economic analysis as well. Flotation with collector 5 never
produced enough sellable material to return a profit under the conditions tested. Collector 2
returned a positive net present value but did not perform strongly enough to beat the scheme
employed at Mountain Pass before their bankruptcy.
This study did not investigate the effect of middlings materials on the flotation response.
The flowsheet proposed did not include a regrind circuit or hydrocyclone circuit to increase the
rate of recovery and grade. Because middling materials were not recovered in either the
concentrate or tailings, calculations from locked cycle testing have a significant amount of
variance. To combat this variance for economic calculations, the results from the concentrate
production method of calculation were used. The concentrate production method is the most
130
|
Colorado School of Mines
|
accurate for recovery values when the locked cycle test does not come to a steady state. This
method does not overstate the production from the locked cycle test because it assumes the
concentrates produced are known and the tailings are calculated based on the feed and concentrate
masses and assays.
10.3 Recommendations for Future Work
This study did not include evaluations of additions to a flotation cycle such as regrind
circuits or hydrocyclone separators. Locked cycle testing showed that there is a significant portion
of the material which was not liberated well enough to report to either the concentrates or tailings,
leading to an increasing load of circulating middling materials. These middling materials could be
recovered to increase the grade and recovery of the flotation flowsheet if they are processed
through some method. AMICS mineralogy also indicated a significant portion of fine gangue
materials that are driving down the grade of the concentrates. Future work should investigate these
additions.
Future work should also include achieving a better understanding of the binding
mechanism differences between collector 2 and collector 5 which allowed collector 2 to drastically
outperform collector 5 in flotation testing. Understanding the fundamentals behind the
mechanisms for binding and inducing hydrophobicity may allow for the design of even better
performing collectors than collector 2.
Finally, additional work the field of collector design and synthesis could further improve
the grade and recovery of rare earth oxides from bastnäsite ore flotation. One of the collectors
which showed promise from previous research could not be tested as a part of this work because
synthesis in a quantity necessary for locked cycle testing was not possible. It is possible that other
collectors exist which could further improve the efficiency of flotation.
131
|
Colorado School of Mines
|
Table D.1 - Correlation between AMICS analysis and CSM XRF analysis for the ore sample
Cerium Lanthanum Neodymium Praseodymium
Sample CSM AMICS CSM AMICS CSM AMICS CSM AMICS
Ore 3.68 4.08 2.48 2.75 0.88 0.87 0.32 0.5
Table D.2 - Correlation between AMICS analysis and CSM XRF analysis for the ore sample
REE Minerals Calcium Barium
Sample CSM AMICS CSM AMICS CSM AMICS
Ore 7.36 8.31 12.09 14.47 13.36 14.5
Table D.3 - Correlation between AMICS analysis and CSM XRF analysis for the collector 2
locked cycle flotation products
Cerium Lanthanum Neodymium Praseodymium
Sample CSM AMICS CSM AMICS CSM AMICS CSM AMICS
SFT 6 1.07 1.68 1.4 1.07 0.29 0.37 0.11 0.17
RCFC 6 25.21 26.42 18.87 18.53 6.93 8.13 2.26 3.51
CFT 6 13.85 14.5 11.36 9.97 3.5 3.62 1.28 1.62
SFC 6 13.74 14.02 10.18 10.07 3.52 2.87 1.23 1.28
RCFT 6 20.48 17.78 9.98 12.47 5.16 4.73 1.85 2.09
Table D.4 - Correlation between AMICS analysis and CSM XRF analysis for the collector 2
locked cycle flotation products
REE Minerals Calcium Barium
Sample CSM AMICS CSM AMICS CSM AMICS
SFT 6 2.87 3.23 11.29 16.46 14.7 15.02
RCFC 6 53.27 55.47 2.67 2.84 1.74 1.88
CFT 6 29.98 29.72 9.88 11.39 6.21 5.42
SFC 6 28.68 29.13 10.65 12.59 3.53 3.5
RCFT 6 37.47 37.05 5.43 5.46 3.31 3.41
150
|
Colorado School of Mines
|
ABSTRACT
The mining industry strongly depends on milling operations to promote mineral liberation in
extracted ore by reducing particle size. The purpose of this project was to assess the effect of niobium
alloying and microstructure on the performance of grinding media via Nb alloying. If beneficial to wear
performance, a cost reduction in grinding media use for mining applications may result.
Experimental laboratory-prepared heats were produced via vacuum induction melting industrial
bar stock with a base composition of Fe-1.0C-0.96Mn-0.22Si-0.26Cu-0.11Ni-0.50Cr-0.005V-0.012Nb
and additional incremental alloying of Nb. Four different ingots were cast with Nb contents of 0.01, 0.25,
0.5, and 1.0 wt pct. Light optical microscopy (LOM), scanning electron microscopy (SEM), bulk
hardness measurements, and X-ray diffraction (XRD) were conducted to characterize the lab processed
material. SEM revealed eutectic niobium carbide (NbC) networks that were effectively broken up and
distributed within the microstructure during hot rolling.
Dilatometry was performed on industrial bar stock and hot rolled Nb-added laboratory prepared
materials in order to design a heat treatment to replicate current grinding ball properties. The martensite
start (M) temperature was found to increase with increasing Nb content from about 187 °C to 216 °C for
s
the 0.01 and 1.0 wt pct Nb alloys, respectively. An increase in hardness was observed as a function of Nb
content, with a total increase from 63.3 ± 0.5 to 65.4 ± 0.2 HRC for the 0.01 and 1.0 wt pct Nb alloys,
respectively. This trend is attributed to the increasing volume fraction of the hard NbC phase. All samples
had a retained austenite content of approximately 20 ± 5 vol pct with no clear trend between austenite
content and Nb alloying. Volume fraction of eutectic-containing NbC was measured using SEM
micrographs and ImageJ® image analysis software. An increase in the eutectic volume fraction of
approximately four times was observed between the 0.01 and 1.0 wt pct Nb alloys. However, an increase
in NbC content led to a decrease in solute C in the matrix, possibly decreasing the hardness of the
martensite and an increase in M temperature.
s
Bond abrasion and dry sand/rubber wheel (DSRW) wear tests were performed to evaluate wear
resistance as a function of Nb alloying. Bond abrasion testing was performed using both iron and copper
ores. Bond abrasion results revealed that there was no discernable increase in wear resistance with
increasing Nb content. DSRW results showed, however, a clear decrease in wear with an increase in Nb
alloying with a 65 pct decrease in mass loss observed between the 0.01 and 1.0 wt pct Nb alloys.
iii
|
Colorado School of Mines
|
ACKNOWLEDGEMENTS
I would like to first extend my gratitude to my primary advisor, Dr. Emmanuel De Moor, for all
of the time, mentorship, jokes, and effort he brought to this project. I would also like to thank my
co-advisor Dr. John Speer for all of the constructive criticism and thoughtful metallurgy knowledge
imparted during my thesis writing process. I want to thank my committee members Dr. Kip Findley and
Professor Erik Spiller for their input and involvement on my project.
This project would not be possible without the support of the Advanced Steel Processing and
Products Research Center (ASPPRC) including all the sponsors, faculty, and students. Specifically, I want
to thank John Heerema and the Gerdau team for supplying all my material, as well as for providing
funding, support, feedback, and overall trust in me to take on this project. I also want to thank Alea
Pommer from Gerdau and Matt Enloe from CBMM for all of their input and participation.
I want to thank the students and faculty within the Colorado School of Mines MME department
for the wealth of knowledge I have received in my time here. Specifically, thank you to Dr. Amy Clarke,
Dr. Kester Clarke, Dr. Kip Findley, Dr. John Speer, Prof. Erik Spiller, and Dr. Gerald Bourne. I want to
thank Melissa Thrun for first being my friend, but also always being willing to answer any metallurgy
questions I come to her with. Thank you to Diptak Bhattacharya, Chris Finfrock, and Trevor Ballard for
all the technical conversations we have shared. I also want to thank Brady McBride and Tomás Scuseria
for the endless coffee runs to keep me awake.
My love for metallurgy began during my time at the University of Wisconsin – Madison. I want
to thank Dr. Susan Babcock for providing a strong female mentor to look up to. I also want to thank
Dr. Michael Arnold for mentoring my first research experience. I would also like to thank Dr. John
Perepezko and Dr. Kyle Metzloff for their involvement in my education. Last, I want to thank my friend
and fellow UW-MSE alum Emily Proehl for her encouragement, numerous late nights, and endless laughs
to help me survive my undergraduate experience.
Last, I want to acknowledge my additional friends and family. Thank you to my core friends
Hannah, Heather, Leah, and Katelyn for being constants in my life since high school. I also want to thank
Anyka, Summer, Gracie, Claire, and Harlie for introducing me to climbing and indulging in my love for
the sport. I want to thank Drew for providing me love, support, and home cooked meals during my thesis
writing process. Thank you to my sister Shelby for being my best friend since the day I was born and
being my number one fan. Last, I cannot express my gratitude for my mom enough. My mom is the
strongest and most loving woman I have ever met, and I would not be the person I am today without her.
xvi
|
Colorado School of Mines
|
CHAPTER 1
INTRODUCTION
1.1 Problem Statement and Industrial Relevance
The goal of this project is to assess the effect of carbides and microstructure on performance of
grinding media via alloying with niobium (Nb). Performance was evaluated via Bond impact-abrasion
wear testing and dry sand/rubber wheel scratch testing. If successful, a cost reduction in grinding media
use for mining applications may result. Materials resistant to abrasive wear are also attractive in the
agricultural, earth moving, excavation, and transportation industries.
Grinding mills represent a significant portion of both capital and operating expenses in nearly all
mining operations. Figure 1.1 represents a breakdown of energy use based on different equipment
categories in U.S. mining. It is clear that grinding accounts for the largest percentage of energy
consumption at 40 pct, followed by materials handling combined between electric and diesel equipment at
21 pct [1]. As shown in Figure 1.2, grinding can further be broken into 3 main costs: energy, liners and
grinding media [2]. The grinding media affect the entire milling process cost (energy, liner, labor costs) in
primary/secondary ball mills, as well as semi-autogenous grinding (SAG) mills. Ball mills employ cast or
forged steel balls as grinding media, whereas SAG is accomplished by using a combination of grinding
media and large pieces of ore. In the case of fully autogenous grinding (AG), no grinding media are used,
and only large crush feed is used to grind a given ore. Pebble mills are a type of ball mill that are more
specific to secondary grinding operations. More discussion on mill types will be included in Section 2.2.
5%
Handling
10%
4%
40% 4%
2%
2%
4%
6%
17% 6%
Handling
Figure 1.1 Energy usage in U.S. mining operations broken into different equipment categories. Values
for most categories are estimated based on electricity usage [1].
1
|
Colorado School of Mines
|
CHAPTER 2
LITERATURE REVIEW
2.1 Introduction
This chapter presents a literature review of milling and grinding; common wear mechanisms
associated with grinding; the effect of microstructure, hardness, and impact toughness on abrasive wear
performance; types of grinding balls; comminution theory; and the role of Nb alloying in grinding media.
2.2 Milling and Grinding
The mining industry strongly depends on milling operations to promote mineral liberation in
extracted ore by reducing particle size [3, 4]. According to the Society of Mining, Metallurgy and
Exploration’s (SME) Mining Engineer Handbook, comminution, or the reduction of ore, is accomplished
through a number of mechanisms including explosive disintegration, crushing by compression, impact,
attrition or rubbing, and shearing or cutting [5]. Primary size reduction of ores is accomplished during
mining; the next stage of liberation consists of compression of individual rocks ranging in size from
several feet to a few inches [5]. Size reduction is known to be an inefficient process when comparing the
input energy with the increase in surface energy of the material subjected to comminution.
Grinding is a specific kind of comminution concerned with achieving the optimum size of
particles in regard to technical and economic considerations [5]. Most commonly, this process occurs
within a tumble mill, which is composed of a number of elements that work together to create a
combination of impact and abrasive actions to grind a given ore [6]. Commercial grinding equipment used
today consists of various modifications of a basic cylindrical or cylindrical-conical shell rotating about a
horizontal axis and charged with some type of grinding media [7]. The primary purpose of a tumble mill
is to liberate individual minerals trapped in rock crystals, thereby preparing the mineral for subsequent
enrichment in the form of particulate separation [8].
Compared to crushing, grinding is a relatively random process, and the degree of grinding is
affected by the probability of the ore entering a zone between grinding media and the probability of some
contact occurring after entry. Such contact can include impact/compression generated from forces applied
approximately normal to the particle surface, chipping due to oblique forces, or abrasion as a result of
forces acting parallel to particle surfaces [8]. Figure 2.1 depicts a typical charge zone within a tumble
mill. As the mill rotates, the grinding media, ore, and water (if present) mix and the media grind the ore
by any of the aforementioned mechanisms, depending on mill size/speed and liner design. At low speeds,
the mill creates a grinding zone where a cascading action leads to grinding media sliding over each other,
4
|
Colorado School of Mines
|
breaking material trapped between them. At higher speeds, grinding media are thrown clear of the charge,
causing comminution through low-energy impact in what is known as the tumbling zone. At maximum
speeds, a crushing zone may be present, where balls in flight crush material in high-energy impacts [6, 7].
Tumbling
Zone Falling Balls
Balls
Grinding
Zone
CrushingZone
Figure 2.1 Typical charge profile inside of a tumbling mill, schematically showing the tumbling,
grinding and crushing zones [6].
Tumble mills can further be defined by the grinding media used for comminution. Rod mills,
which use cylindrical shells with a length 1.33 to 2 times their diameter, are generally uncommon [7].
Ball mills, on the other hand, are the most widely used and the focus of this project. Ball mills employ
cast or forged steel balls as grinding media and range from about 5 - 60 mm in diameter. Ball mills can
either be primary or secondary in nature, where ore comes either directly from the mill feed or already
ground rock is broken down further to a critical size. Tumble mills of this type can produce a large
tonnage of fine material [7]. Autogenous or semi-autogenous (AG/SAG) mills are of interest in milling
because they reduce grinding media consumption by grinding with large pieces of ore, rather than steel
grinding media, but are generally less common than ball mills. AG milling can be accomplished by using
coarse crushed feed or run-of-the-mill material, sized crushed feed, or, in the case of SAG milling,
crushed feed and a small charge of large-diameter balls. Pebble mills are similar to ball mills but have a
greater length to diameter ratio.
Present comminution practices are inefficient because methods of breaking rock are still
fundamentally primitive. Methods of breaking down ore include striking (impact), squeezing
(compression) and rubbing (abrasion or shear) [9]. All methods apply deformation forces that exceed the
fracture strength of the ore and cause failure along zones of weakness. The necessary input of stress
imparts strain energy into the deformed rock; when this stress is removed by fracture or failure, the strain
5
|
Colorado School of Mines
|
energy is released as heat. It has been found that 95 to 99 pct of the imparted energy input is released as
heat, and only about 1 to 5 pct is used in the formation of new surface area [9].
2.3 Types of Grinding Balls
Two general types of grinding media are used in ball mills: high Cr white cast iron (WCI) and
forged high carbon low alloy steels (HCS). Forged grinding media are generally produced from
continuously cast steel billets that are heated, cut, forged into a ball, and subsequently heat treated. They
generally have a microstructure consisting of martensite and retained austenite with carbon levels
between 0.8 - 1.0 wt pct. Forged grinding media have superior abrasion resistance, while possessing little
corrosion resistance. While the properties of HCS grinding media will be discussed in detail in relation to
their microstructure and mechanical properties in Section 2.6, it is also important to understand alternative
high-Cr cast grinding media.
WCI grinding balls have a high carbon content in excess of HCS C levels, which leads to a
microstructure consisting of retained austenite and tempered martensite with inter-dendritic eutectic Cr
carbides [10]. The superior abrasion resistance of high Cr cast irons is believed to be a direct result of
their microstructure, namely these hard carbides, the Cr content in the matrix, and the ball surface
hardness [10]. Figure 2.2 shows the bulk microstructure of a WCI grinding ball containing 9.2 wt pct Cr
and 2.8 wt pct C consisting of a predominantly austenitic matrix with darker primary eutectic Cr
carbides [11]. If these steels are heat treated further, secondary Cr carbides begin to form in the
matrix [10].
Figure 2.2 Optical micrograph of a WCI grinding ball containing 9.2 wt pct Cr, primarily composed of
dendritic austenite (lighter regions) and inter-dendritic Cr carbides [11].
The high carbon levels of WCI grinding media, usually between 2 - 3 wt pct, give the balls their
increased hardness values, generally ranging between 52 - 65 HRC [10]. However, bulk hardness alone
6
|
Colorado School of Mines
|
does not determine the wear resistance of the material as wear resistance is also influenced by
microstructural features such as inclusions, carbides, retained austenite, and matrix structure.
Cementite (Fe C) is the predominant carbide present in low alloy steels, with a hardness between
3
800 - 1200 HV [10]. If the Cr content exceeds 10 wt pct, another carbide, (FeCr) C , forms with a
7 3
hardness between 1200 - 1800 HV – substantially harder than that of the martensitic or austenitic
matrix [10]. As the Cr content is increased beyond this, the carbide fraction increases, and the wear
performance is enhanced. Alternatively, these brittle carbide networks may lead to fracture initiation sites
and lower the fracture toughness of the grinding balls.
2.4 Wear Mechanisms
Wear in grinding media is defined as the progressive loss of material from a solid body due to its
contact and movement against a surface. Wear lowers the operational efficiency of the mining process,
leading to a major source of cost. To reduce steel consumption, it is necessary to understand wear
mechanisms that are active during the grinding process, as well as the susceptibility of the materials to
these mechanisms [3]. Total media wear and factors influencing grinding ball wear during grinding are
most commonly discussed across literature with respect to abrasion, impact and corrosion [6, 12, 13].
Abrasion is the direct removal of material from the grinding media surface due to contact with
hard particles or protuberances and is the most common wear mechanism and cause of metal loss in
grinding [3, 13, 14]. In both wet and dry grinding, abrasion can consume up to 2 kg of grinding media per
ton of ore milled [15]. Abrasion is further divided into two additional categories: two-body abrasion and
three-body abrasion. Two-body abrasion results when a rough surface or fixed abrasive particle slides
across another to remove material. Three-body abrasion, on the other hand, occurs when particles are
loose and mobile during interactions with wear surfaces. Three-body abrasion is said to be closed when
the loose abrasive particles are trapped between two sliding or rolling surfaces, and open when the two
surfaces are far apart or only one surface is involved in the wear process [16].
Upon changing from dry to wet grinding operations, there may be a large increase in wear
attributed to corrosion. The corrosive wear mechanism of grinding media is the least investigated or
understood due to its mechanical-electrochemical nature. Metal wear in wet versus dry ball milling of
various abrasion indices, used to quantify the abrasivity of a given ore, is displayed in Figure 2.3. The
difference in wear rate is attributed to a combination of corrosion and slurry rheology, or the flow
behavior [17]. Some sources report that corrosion is responsible for 50 - 80 pct of total wear, while others
such as Azizi et al. suggest that corrosion plays a smaller role in the consumption of wet grinding media,
attributing only 27 pct of the total wear to corrosion [13, 18, 19]. Corrosive wear, whether it is a
significant part of total wear or not, causes metal loss due to (electro)chemical reactions. It is a result of
7
|
Colorado School of Mines
|
galvanic corrosion between the grinding media and the ore, accelerated corrosion caused by abrasion,
and/or corrosion pitting [18]. Corrosive wear is beyond the scope of this project and only dry abrasion
testing was pursued.
Figure 2.3 Metal wear in dry and wet ball milling for ores of various abrasion indices [17].
In grinding balls with increased hardness, abrasive wear is generally decreased and spalling due
to impact loading can occur instead. Upon collision of two balls, the sudden release of energy causes
waves to propagate through the balls radially from the point of contact, creating a spalling point where the
waves exceed the fracture strength of the material [15]. Impact wear is proportional to the weight of the
ball and the amount of surface hardening [13].
High stress grinding abrasion produces practically all the wear on grinding balls in ball mill
units [3, 13, 14] and, therefore, will be the focus of this project.
2.5 Abrasive Wear Mechanisms
Abrasive modes of wear are a significant factor in the loss of usefulness of materials. There are
several factors that affect abrasive wear, some of which are outlined in Table 2.1.
Table 2.1 Factors Influencing Abrasive Wear Behavior [20]
Property Factors
Particle size, particle shape, hardness, yield strength, fracture properties,
Abrasive Material
concentration
Force/impact level, velocity, impact/impingement angle, sliding/rolling,
Contact Conditions
temperature, wet/dry, pH
Hardness, yield strength, elastic modulus, ductility, toughness, work-hardening
Wear Material
characteristics, fracture toughness, microstructure, corrosion resistance
8
|
Colorado School of Mines
|
The definition of abrasion is broad in nature, so abrasive wear can be further understood with
respect to ways it can manifest in grinding media. Hawk et al. from the National Energy Technology
Laboratory (NETL) outlined four different manifestations of abrasion in metals [20]:
• Gouging abrasion occurs when abrasive particles indent and move over the wear surfaces
under high stress levels. Gouging involves both cutting and tearing types of wear and the
wear surface is both plastically deformed and work hardened by abrasive forces.
• High stress, or grinding, abrasion occurs when abrasive particles are compressed between
two solid surfaces, such as grinding balls in a ball mill. A high contact pressure produces
indentations and scratches on the wear surfaces while fracturing and pulverizing the
abrasive ore. Grinding abrasion produces wear through a combination of cutting, plastic
deformation, microscopic surface fracture, tearing, and/or spalling.
• Low stress, or scratching, abrasion occurs when lightly loaded abrasive materials impinge
on, and move across, a wear surface. Scratching abrasion produces wear at a low rate by
cutting and plowing the surface.
• Sliding abrasion can either be two- or three-body abrasion where relative motion between
the abrasive particle and the wear surface is 100 pct sliding in nature.
Once it is understood how abrasion presents itself, it is then necessary to understand the
mechanisms under which abrasive wear works. The majority of the factors that affect abrasive wear are
related to the mechanical properties of the grinding media because abrasive wear is predominantly a
mechanical process [20]. The following mechanisms, shown in Figure 2.4, are used to describe abrasive
wear in materials [20]:
• Plowing occurs when a material is displaced to the side, away from the wear particle,
resulting in the formation of grooves that do not directly remove material from the wear
surface. The displaced material forms ridges adjacent to grooves, which may be removed
by subsequent contact with abrasive particles.
• Cutting is the process under which material is separated from the surface in the form of
primary debris, with little or no material displaced to the side in the form of grooves.
• Fragmentation occurs when material is separated from a surface by a cutting process and
the indenting abrasive causes localized fracture to the material.
Understanding primary abrasion mechanisms and how they produce wear in materials allows for
better design of the most effective grinding media. The plowing and cutting mechanisms predominantly
involve plastic deformation whereas fragmentation involves fracture. For this reason, materials that
exhibit high fracture resistance and ductility with relatively low yield strength are more likely to be
abraded by plowing or cutting. Materials with high yield strength and low ductility and fracture resistance
9
|
Colorado School of Mines
|
abrade primarily through fragmentation. It is necessary to understand the effects of microstructural
development and mechanical properties to reduce wear and steel consumption of grinding media.
(a) (b) (c)
Figure 2.4 Material removal due to contact with abrasive materials via (a) plowing, (b) cutting, and (c)
cracking (fragmentation) mechanisms.
2.6 Steel Properties that Affect Wear
The following section outlines various steel properties that affect the wear response of forged
HSC grinding media, including mechanical properties such as hardness and impact toughness, as well as
wear resistance as a function of different microstructures.
2.6.1 Mechanical Properties
Hardness in steels is defined as the material’s resistance to penetration, therefore making it an
important parameter in determining the resistance of the material to abrasion. Wear resistance generally
increases with surface hardness of grinding balls due to a reduction in the depth of penetration by the
abrasive ore [15, 20]. A key factor influencing the hardness of a material is chemical composition; in
particular, the carbon (C) content.
Alloying, as well as processing parameters, allow for specific control over hardness and
microstructure of grinding media. Choice of material for grinding balls must take into account resistance
to wear, as well as balancing conflicting characteristics such as high hardness and adequate ductility [3].
The most common steel used commercially for grinding media is a high carbon low alloy (HCLA) steel,
which is the hardest, strongest, and least ductile in its hardened, tempered form [3]. HCLA steels have a
microstructure composed of martensite and retained austenite. High C contents are effective in increasing
the hardenability, or the ability to form the hard martensite phase, of a material and decreasing wear.
Additional elements such as manganese (Mn), chromium (Cr), molybdenum (Mo), copper (Cu),
nickel (Ni), vanadium (V), phosphorous (P) and sulfur (S) are all common in grinding media. The specific
role of these various elements is outlined in Table 2.2.
10
|
Colorado School of Mines
|
Table 2.2 The Effect of Alloying Elements on the Properties of Steel
Element Effect
C Austenite stabilizer
Mn Solid solution strengthening, austenite stabilizer
Cr Solid solution strengthening, corrosion resistance
Mo Pearlite suppressant, promotes bainite, ferrite stabilizer
Cu Corrosion resistance, residual
Ni Impact strength, austenite stabilizer
V Precipitation strengthening, ferrite stabilizer
P Embrittlement, residual
S Embrittlement, residual
It is interesting to note that alloying for corrosion resistance with high Cr content has a relatively
small effect on overall wear resistance. As shown in Figure 2.5, the only significant trend noticed is
between hardness and metal loss, not corrosion resistance (reflected via stainless steel) and metal loss.
When hardness is held constant for a medium carbon steel alloy and a high Cr stainless steel, wear rate is
nearly identical. However, when hardness is varied over stainless steel grinding balls with constant Cr
content through heat treatment, large deviations in wear rate are observed. Additionally, for a high C, high
strength steel, wear rate is at a minimum. Alloying and heat treating are effective in reducing metal loss
even in wet grinding environments by increasing hardness and martensite content [14].
Figure 2.5 Effect of alloying for corrosion versus for hardness on the wear rate of grinding balls [14].
11
|
Colorado School of Mines
|
In addition to hardness, mechanical properties such as plastic flow, tensile strength, work
hardening rate, ductility, and fracture toughness all play vital roles in influencing the predominant wear
mechanism occurring during grinding. An increase in hardness of grinding media is accompanied by a
reduction in ductility, changing the wear mechanism from plowing and/or cutting to fragmentation, and
potentially risking brittle fracture [15, 20]. In the absence of fragmentation, variations in the plastic flow
behavior of the material influence the predominant wear mechanism occurring [20]. The plastic flow
behavior can be characterized by the steel’s yield strength; an increase in yield strength favors a cutting
mechanism and a decrease leads to plowing [20]. Additionally, an increase in tensile strength and work
hardening rate will aid in increasing wear resistance as this is typically associated with an increase in
material hardness. However, an increase in tensile strength may decrease fracture toughness, and promote
wear. Overall, in order to reduce weight loss during grinding, it is generally accepted that a high value of
hardness helps reduce plowing and high toughness is required to inhibit spalling [21].
2.6.2 Microstructure
Differences in grinding media microstructure show a corresponding difference in wear resistance
and could lead to a change of up to 28 pct in mass loss during grinding [15]. The mechanical properties
discussed regarding wear are highly dependent on the microstructure of the grinding media. When steels
with different or mixed microstructures are compared, hardness can no longer be the sole criterion under
which wear resistance is evaluated because different microstructures of equivalent hardness show
different wear rates [13]. Figure 2.6 outlines trends reported between microstructure and hardness on the
high-stress abrasion resistance of steels.
3.0
BainiticSteels
(Austempered)
Austenitic
Alloys
Quenched and
Tempered Steels
2.0
Increasing
alloy
content
Cold work
Pearlitic Steels
1.0
100 200 300 400 500 600 700 800 900 1000
Hardness (HV)
Figure 2.6 Effect of microstructure and hardness on the abrasion resistance of steels in relation to alloy
content and cold work.
12
)0.1
=
eF
eruP(
ecnatsiseR
raeW
evitaleR
tnetnoc
nobrac
gnisaercnI
|
Colorado School of Mines
|
While steels with pearlitic microstructures possess some impact toughness, they exhibit lower
hardness compared to quench and tempered steels, which leads to increased consumption rates when used
in grinding applications. Pearlitic wear resistance is maximized with an extremely fine lamellar structure
achieved through judicious heat treatment. High C martensitic steels offer the best combination of
maximum hardness and wear resistance. The dependence of wear resistance on the hardness of the
material is most evident in quenched and tempered steels, as well as in bainitic steels. Hardness is
maximized through a combination of cold work (work hardening) and alloy content. Microstructures
containing a combination of martensite, retained austenite, and Cr carbides showed minimum wear at
certain levels of retained austenite [15]. Steels containing retained austenite exhibit superior wear
resistance due to the transformation properties of this phase. For constant hardness, increasing retained
austenite volume fraction improves wear resistance [21].
The ductility resulting from microstructures containing retained austenite exists due to the
transformation induced plasticity (TRIP) effect. When a material, such as a grinding ball, is deformed
plastically due to contact with other grinding balls and ore, the retained austenite transforms into the hard
martensite phase [22]. The introduction of this harder phase results in an increase in the work hardening
rate relative to the normal work hardening experienced in metals due to dislocation interactions. Materials
with increased work hardening rate exhibit greater levels of strain before reaching the ultimate tensile
strength, improving the overall ductility of the material [22].
Figure 2.7 shows the overall relationship between grinding ball microstructure, hardness, and, in
turn, wear rate. Hardness is plotted against a ball use factor, which is a normalized value based on ball
consumption during testing. A larger ball use factor indicates a lower wear resistance.
Figure 2.7 Effect of microstructure on hardness and ball use factor of grinding balls [14].
13
|
Colorado School of Mines
|
2.6.3 Impact Toughness
The impact toughness of steel grinding media is largely dependent on chemical composition and
microstructure and must be considered with respect to its role in the grinding media of interest in this
project. The repeated impact of grinding media in a ball mill may cause one or more of the following:
• plastic deformation of the ball surface,
• small flaking from the surface of the ball (micro-spalling),
• large pieces spalling from the surface of the ball (macro-spalling),
• ball fracture through the center or through a major portion of the cross-section [23].
While it was previously discussed that structures such as pearlite possess low hardness compared
to martensite, and high carbon steels have increased hardness at the expense of toughness, there are other
factors to consider when determining impact wear performance in grinding media.
In a ball drop test performed by Chenje et al., three balls of varying composition including a low
alloy steel (LAS), eutectoid steel (ES), medium Cr cast iron (MC), semi-steel (SS), and unalloyed cast
iron (UCI), all 60 mm in diameter, were dropped onto an Mn-steel anvil from a height of 6 m until
fracture [24]. Results from this test are displayed in Figure 2.8 with ball chemistries and microstructures
outlined in Table 2.3. Grinding balls were tested in both the as-cast (AC) and heat-treated (HT)
conditions.
LAS
ACES
HTES
ACSS
HTSS
HTMC
UCI
Figure 2.8 Effect of carbon content of grinding media type on toughness as indicated by the drop
count [24].
The heat treatment for the ES, MC, and SS consisted of a 3 hr austenitization at 750 ºC, 1050 ºC,
and 850 ºC, respectively, followed by a slow air quench. After this heat treatment, the microstructure of
the ES remained approximately unchanged from the AC condition, whereas previously coarse and
14
|
Colorado School of Mines
|
elongated Cr-carbides were refined in the HTMC alloy. The HTSS saw a similar Fe-carbide refinement
post heat treatment.
Table 2.3 Chemistry (wt pct) and Microstructure of Grinding Media Used in Chenje el al. Ball Drop
Test [24].
Type of Ball Acronym C Mn Si Cr S P Microstructural Features
Low alloy steel LAS 0.50 1.07 0.43 1.15 0.02 0.02 Very fine pearlite
AC eutectoid steel ACES 0.85 0.65 0.69 <0.15 0.01 0.01 Fine pearlite
HT eutectoid steel HTES 0.85 0.65 0.69 <0.15 0.01 0.01 Fine pearlite
AC medium Cr Primary Fe, Cr carbides
ACMC 3.00 0.55 0.62 17.81 0.04 0.02
cast iron in a pearlite matrix
HT medium Cr Primary Fe, Cr carbides
HTMC 3.00 0.55 0.62 17.81 0.04 0.02
cast iron in a pearlite matrix
Pearlite, Widmanstatten
AC semi-steel ACSS 2.18 0.68 0.87 <0.15 0.02 0.02
cementite
Pearlite, Widmanstatten
HT semi-steel HTSS 2.18 0.68 0.87 <0.15 0.02 0.02
cementite
Unalloyed cast Pearlite, massive
UCI 3.00 0.55 0.62 0.58 0.02 0.02
iron cementite
The ES and LAS with lower C contents performed exceptionally well. After 3,000 drops, the
eutectoid steel in both the AC and HT conditions had not fractured. There was, however, extensive
spalling on the surface of the balls, especially in the AC condition. The UCI and ACSS, a type of cast iron
produced with scrap steel, balls, however, fractured at relatively low drop counts without much spalling
or flaking [24]. The carbide phase is a major constituent of the microstructure of these balls, which is
characteristically hard, brittle, and exists in continuous networks that greatly facilitate crack propagation.
The continuous cementite phase provides a path of least resistance for crack propagation after a crack has
initiated [23]. As a result, these balls are likely to fracture under high impact conditions. Similar behavior
is observed in high Cr-cast grinding media with networks of primary Cr carbide [23]. The impact
resistance with respect to microstructures containing carbide networks can be improved through heat
treatment that leads to a more discrete carbide distribution [23, 24].
T. Noguchi et al. performed a similar ball drop test using Ni hardened and high Cr steel grinding
balls 50 mm in diameter [25]. The balls were dropped from heights of 1 m to 6 m in 1 m increments onto
a Ni-hardened steel anvil until spalling breakage occurred. The microstructure of the Ni hardened
grinding balls consisted of primary austenite, lower bainite, and eutectic ledeburite, and the high Cr
samples consisted of martensite and eutectic Cr carbides. The results from this test are displayed in
Figure 2.9 with ball chemistries included in Table 2.4.
15
|
Colorado School of Mines
|
Figure 2.9 Number of drops to failure as a function of drop height for Ni hardened and high Cr grinding
balls [25].
Table 2.4 Chemical Composition (wt pct) and Hardness of Grinding Media Used in
Noguchi et al. Ball Drop Test [25].
Grinding Ball C Si Mn P S Ni Cr Hardness (HRC)
Ni Hardened 2.82 0.5 0.52 0.021 0.010 4.6 1.8 45.5
High Cr 2.71 0.87 0.83 0.017 0.022 - 18.8 40.4
While both grinding balls had carbon levels comparable to those that performed poorly in the
previous study, both were able to withstand similar impact conditions up to a much higher drop count
before failure. This is likely to do with the superior martensite-retained austenite microstructures of these
grinding balls, compared to the lower hardness pearlitic grinding balls tested by Chenje et al.
It is also of importance to compare toughness behavior of forged versus cast grinding media. For
both cast and forged conditions, pre-existing defects can act as stress concentration sites leading to
fracture upon impact. In the case of cast balls, flaws such as shrinkage cavities and porosity will lead to
premature failure during mill operation [23]. For forged grinding media, defects such as quench cracks,
gas and shrinkage cavities, and centerline shrinkage must be avoided during continuous casting. During
the forging process, it is possible for the material to fold into itself and cause a forge lap, which can act as
a fracture initiation site [23]. Additional flaws in grinding media that will reduce the impact toughness of
the steel include flats (flat spots resulting from the metal not filling the die completely), soft spots (non-
uniform surface hardness due to inclusions), and cuts (physical damage to the grinding ball surface) that
will create stress concentrators and spalling points [23].
One final aspect of the role of impact toughness in grinding balls is in the case of large diameter
ball mills. In such mills, larger balls create more severe impact conditions which give rise to increased
spalling. This is also coupled with the fact that forged grinding balls especially show an undesirable, steep
16
|
Colorado School of Mines
|
hardness gradient due to macro-segregation of C between the surface and the center of the ball [17, 23].
Figure 2.10 shows the variation of C segregation from the surface toward the center of a 127 mm (5 in)
grinding ball, presumably produced during the preceding continuous casting operation [17]. Such
segregation is detrimental since it can give rise to considerable differences in austenite transformation
products during heat treatment and also provide increased thermal stress after the martensite
transformation [17]. This combination of factors can lead to a decrease in impact toughness of larger
diameter grinding media.
1.02
1.00
0.98 Center
of Ball
0.96
0.94
0.92
0.90
7 14 21 28 35 42 49 56 63
Distance from the Surface Towards Center (mm)
Figure 2.10 Macro-segregation ratio of carbon from the surface toward the center of a 127 mm (5 in)
diameter grinding ball [17].
As with elements such as Mn, P, and S, carbon is more soluble in liquid steel than it is in solid
steel. Therefore, during casting, as the outside of the ingot begins to solidify, it rejects carbon toward the
liquid center and the center will be more rich in carbon compared to the surface. The macro-segregation
ratio is determined by dividing the carbon content at a given distance along the radius of the ball by the
bulk carbon content, C .
o
2.7 Comminution Theory
Once general mechanical properties of alloys used in grinding media are understood, it is then
imperative to quantify their abrasivity with respect to comminution, or the reduction in size of a material.
It is difficult to predict the abrasive properties of rocks, minerals, and ores because rocks belonging to the
same geological classification vary widely from one site to another, or even to a considerable degree from
different locations within the same mine [26]. There is no accepted scientific theory that is fundamental to
comminution which can define how a given ore will affect grinding media due to this variability from one
rock to another, although many attempts have been made. However, it is essential that the grinding
17
)
C/C(
oitaRnoitagergesorcaM
0
|
Colorado School of Mines
|
behavior be known before the mill sizes are selected when designing a plant. Therefore, the following
properties are often determined to characterize an ore prior to grinding in a ball mill.
2.7.1 Work Index
Comminution theory is concerned with the balance between input energy and the product particle
size from a given feed size. The “Work Index” method, commonly referred to as the “Bond” method after
American mining engineer F.C. Bond, established a consistent common factor (the work index) used to
determine crushing and grinding mill sizes [27]. The work index is the total work input, in kWh per short
ton, required to reduce a material from theoretically infinitely sized feed to 80 pct passing 100 μm,
meaning that 80 pct of the output material passes through a 100 μm mesh [27, 28]. If the diameter of the
feed in μm is F, the diameter of the product in μm is P, and the work input in kWh per short ton is W, the
following equation is used to calculate the work index, W [27, 28]:
i
W
W=
i 10 10 (2.1)
-
√P √F
The work index is a comminution parameter which expresses the resistance of the material to
crushing and grinding, allowing for the proper designation of crusher or grinding mill unit size [27, 28].
2.7.2 Crushability
Another parameter that is essential to mill design is the quantification of rock crushability,
determined using a crushability test. During this test, pieces of broken rock are mounted between two
opposing equal 13.6 kg (30 lb) weights which swing on wheels [27]. The wheels are released, and the
weights strike simultaneously on opposite sides of the smallest measured dimension of the rock. The
height of the weights (h ), and thus the height of the fall, is incrementally increased until the rock breaks.
o
A schematic of this test set-up is shown in Figure 2.11.
If the impact crushing energy in ft-lb per in of rock thickness is designated as C and specific
gravity is S , the work index is found from the average of 10 breaks using the following equation [27]:
g
C
W=2.59 (2.2)
i S
g
After crushing tests, the crushing index and particle size of the rocks can be determined by dry
sieving the material from the collection box. The crushability index is an experimental parameter that
18
|
Colorado School of Mines
|
describes the resistance to crushing by indirectly measuring the force exerted on the rock crusher (falling
weight) against the fractured rock [30]. Crushed rocks are sieved through mesh of known size
(e.g., 9.5 mm) and the percentage of passing material is described as the crushability index (CI). Similar
to the work index, the CI does not provide a definitive quantity, but rather a consistent value to compare
different ores.
Figure 2.11 Schematic of the pendulum test set-up used to determine rock crushability. Image highlights
the falling hammers with respect to the placement of the ore [29].
2.7.3 Grindability
A ball mill grindability test is a method in which a specific ball mill’s work index is calculated.
This test represents a small-scale laboratory or bench-style test which is often necessary to measure the
resistance of a material, whether it be the ore or the grinding media, to grinding [9].
The standard feed for a grindability test is prepared by crushing ore to nearly 100 pct passing a
6 mesh (3.36 mm) sieve, or wire frame [27]. The mill itself is 30.5 cm x 30.5 cm with a smooth interior
lining and a 10.2 cm x 20.3 cm door for charging by hand. The mill generally includes a revolution
counter and runs at 70 rpm. The grinding media charge consists of 285 steel balls weighing a total of
20.125 kg, made up of the following sizes and quantities of grinding balls [27]:
• 43 balls 3.7 cm in diameter
• 67 balls 3.0 cm in diameter
• 10 balls 2.5 cm in diameter
• 71 balls 1.9 cm in diameter
• 94 balls 1.5 cm in diameter
After the first grinding period of 100 revolutions, the mill is emptied and cleaned, the ball charge
is separated out, and the remaining material is screened on sieves [27]. The undersize, or passing, ore is
19
|
Colorado School of Mines
|
weighed, and fresh unsegregated feed is added to the oversize until the original charge weight is reached.
Based on the weight of the undersize collected, the number of revolutions necessary to produce 250 pct
circulating load is calculated by determining the mass flow returning to the mill after screening, expressed
as a percentage of the mass flow of new feed. The grinding period cycles are repeated until the net mass
of the undersize product per mill revolution is no longer increasing. The undersize product and the
circulating net grams per revolution (Gbp) is the ball mill grindability [27].
Further, if P is defined as the size of the mesh in microns of the sieve tested, the ball mill work
1
index is calculated using the following equation [27]:
44.5 10 10
W= ×(Gpb)0.82$ - % (2.3)
i P0.23 √P √F
1
2.7.4 Bond Abrasion Test
The Bond abrasion test, named after F.C. Bond, provides a means for comparing abrasive
properties of one rock or mineral with another, or one metal alloy with another, under identical milling
conditions [26, 27, 31]. This procedure does not attempt to measure abrasion or abrasion resistance in
absolute terms, but rather allows one to juxtapose rocks and minerals of well-known behavior in parallel
with unknown materials. Abrasion properties of ore, as well as abrasion resistance of metal alloys used in
grinding, are important factors in mill design as they both vary the amount of metal worn away during
grinding.
The test is set up by first charging the drum with 400 g of ore screened to pass a 1.905 cm2 mesh
and be retained upon a 1.27 cm2 mesh [9, 26, 31]. A flat paddle that measures 7.62 x 2.54 x 0.64 cm made
from SAE 4325 Cr-Ni-Mo steel hardened to 500 Brinell is weighed to 0.1 mg and inserted 2.54 cm of its
length into a slot in the rotor [9, 26, 31]. This steel paddle acts as an impeller, with 12.9 cm2 of its surface
exposed to abrasion. The inside of the drum is lined with steel mesh to provide a rough surface for
continuously picking up the ore. The impeller rotates rapidly at 632 rpm as the ore particles fall
throughout the drum, which is revolving at a slower speed of 70 rpm, striking the paddle at a relatively
high velocity [9, 31]. At the tip of the impeller, the rotation radius is 10.8 cm and the linear speed is
approximately 430 cm · min-1, giving way to sufficient impact between the ore and the steel impeller. A
typical Bond abrasion drum and mesh liner is shown in Figure 2.12.
This rotative abrasion is continued for 15 min, after which the drum is emptied, cleaned, and
another 400 g of ore is added. Each 15 min segment constitutes one-quarter of a complete test, the total
being 1600 g of ore run over a period of 1 hr. After the completed 1 hr run the paddle is removed,
washed, and cleaned of all rock particles, dried, and reweighed to the nearest tenth of a mg. The weight
20
|
Colorado School of Mines
|
loss in grams by the steel impeller after impacting four successive 400 g batches of ore for 15 min each is
called the abrasion index, designated by A. Because ore only strikes the paddle on one face, the paddle
i
can then be reversed for a second test. If desired, the kg loss per kWh is calculated through the following
relationship:
kg & = 1.04 × A (2.4)
kWh i
Figure 2.12 Image of Bond abrasion wear tester drum lined with steel mesh liner at the Colorado School
of Mines.
Table 2.5 presents average abrasion indices of various common minerals from 125 tests [31].
Ores such as alumina, taconite, and quartzite have relatively high A values and are likely to cause the
i
most metal loss due to abrasion, whereas limestone, sulfur, and shale have comparatively low abrasion
indices and metal wear is not expected to be as severe.
The reason behind running the abrasion test in four segments is to ensure a consistent average
weight loss evaluation for the paddle. The first 15 min interval tends to give disproportionately high
abrasive losses due to the lack of work hardening early in the wear test. This phenomenon is particularly
noticeable when the impeller is made from high manganese steels [26].
The Bond abrasion test can also be used to evaluate the abrasion resistance of different metals and
alloys used in grinding. The abrasion losses of the steel can be compared with that of the standard
SAE 4325 paddle, using a constant rock material for testing [32].
2.8 Niobium Alloying of Grinding Media
This section provides an overview of important concepts related to alloying grinding media with
Nb to produce eutectic niobium carbides (NbC). Topics such as solubility of Nb as well as the formation
21
|
Colorado School of Mines
|
and characteristics of NbC are discussed in relation to how they affect wear performance of grinding
balls.
Table 2.5 Average Abrasion Indices for Various Common Ores
Ore No. Tested A Ore No. Tested A
i i
Sulfur 1 0.0001 Pumice 1 0.1187
Shale 2 0.0060 Chrome Ore 1 0.1200
Cement Clinker 2 0.0109 Quartz 7 0.1831
Slag 1 0.0179 Gold Ore 2 0.2000
Nickel Ore 2 0.0215 Magnetite 2 0.2517
Limestone 19 0.0256 Gravel 2 0.3051
Perlite 2 0.0452 Coke 1 0.3095
Iron Ore 4 0.0770 Granite 11 0.3937
Copper Ore 12 0.0950 Alumina 6 0.6447
Hematite 3 0.0952 Taconite 7 0.6837
Manganese Ore 1 0.1133 Quartzite 3 0.6905
2.8.1 Characteristics of NbC
The benefit of adding significant amounts of Nb to high carbon steels with the purpose of
improving wear performance can first be understood in regard to the solubility of NbC in austenite (γ).
The extent to which elements can be maintained in solid solution in γ is governed by the appropriate
solubility product, commonly used to predict the temperature at which precipitates begin to form upon
cooling [33, 34]. The extent of solubility of Nb and C in γ, when in equilibrium with stoichiometric NbC,
is expressed through the following relationship:
[C] + [Nb] = NbC (2.5)
γ γ
where [C] and [Nb] are the concentrations (in wt pct) of carbon and niobium in austenite, respectively.
γ γ
Thermodynamic equilibrium (assuming dilute solutions) can be described by the solubility
product, K:
K = [C] [Nb] (2.6)
γ γ
The temperature dependence of the solubility product is frequently given in the following form:
22
|
Colorado School of Mines
|
A
logK= - +B (2.7)
T
where A and B are constants and T is absolute temperature [33]. Values for A and B vary across literature,
however values of 7020 K and 2.81, respectively, given by Turkdogan were used in this review for the
formation of NbC in γ [33].
Solubility isotherms are an essential tool in understanding whether or not a precipitate will exist
in austenite. Figure 2.13 shows solubility curves for common nitrides and carbides in austenite with
solubility product, K, values plotted over a range of temperatures. The lower the solubility curve in
Figure 2.13, the lower the solubility product for a given temperature, meaning that less solute remains in
solid solution and a higher amount is undissolved. For relatively low alloying, precipitates such as
titanium nitride (TiN) and boron nitride (BN) form relatively easily over the entire range of temperatures
due to their low equilibrium solubility products. For steel grades of interest, discussed in Chapter 3.2,
however, it is assumed that neither titanium (Ti) nor nitrogen (N) exist in notable quantities, and the only
relationships of interest are those for Nb(C,N) and vanadium carbide (VC). Figure 2.13 clearly
demonstrates that NbC is a stronger carbide former compared to VC.
Figure 2.13 Solubility products of common carbides and nitrides in austenite as a function of
temperature [33].
It was reported by Mesquita and Jansto that at a temperature of 1200 ºC and composition of
1 wt pct C, in order to form 1 vol pct undissolved carbide, 6.7 wt pct tungsten (W) or 1.6 wt pct
vanadium (V) would be needed, compared to only 0.82 wt pct Nb [35]. The addition of larger amounts of
alloying elements such as W and V may lead to higher material costs. When these large, undissolved,
23
|
Colorado School of Mines
|
extremely hard precipitates are dispersed in a martensitic matrix, they increase the local hardness
properties and may improve the overall wear resistance of grinding media.
Nb is a relatively strong carbide former and it is expected that carbides will begin forming at high
temperatures and, in turn, become coarse. While this is generally undesired behavior in some steels, these
large NbC precipitates are the reason significant Nb alloying is attractive due to their hardness.
Figure 2.14 shows the range of hardness values that NbC is able to achieve. While martensite cannot be
hardened past values around 1000 HV around C levels of 1 wt pct, NbC has been reported to reach
hardness values up to 3000 HV [36]. The only carbides able to achieve comparable hardness values are
TiC, VC and tungsten carbide (W C) [36].
2
2.8.2 Eutectic Carbide Formation and Properties
Due to the significant alloying up to 1 wt pct Nb performed in this study, NbC is stable up to
significantly high temperatures and no longer forms within austenite, but rather in a coupled growth
mechanism with austenite, forming eutectic-structured carbide networks during solidification. Such
carbides are currently utilized in high speed tool steels which are a family of alloys used in the production
of cutting tools. High speed tool steels are comprised of the following two features:
(1) The alloys belong to the Fe-C-X multicomponent system, where X represents a group of
alloying elements such as Cr, W, Mo, Nb, and/or V [37].
(2) The alloys are characterized by the capacity to preserve high hardness and wear resistance at
high temperatures, resulting from cutting metal at high speed [37 – 39].
Figure 2.14 Hardness value ranges in HV for common carbides and nitrides [36].
24
|
Colorado School of Mines
|
The relevant link between high speed tool steels and grinding media comes from the source of
wear resistance for high speed tool steels. While it has not been extensively reviewed in the application of
grinding balls, the wear properties of these tool steels are provided by the presence of blocky, hard
eutectic carbides distributed in the matrix [38, 40, 41]. The active wear mechanisms will likely differ
between tool steels (cutting/plowing mechanism) and grinding media (abrasive/corrosive/impact);
however, the overall mechanical properties of high speed tool steels are determined by the size and
distribution of eutectic carbides. Such properties are closely related to chemical composition and the
casting of the ingot, and can be applied to novel grinding media chemistries [39, 40, 42]. The wear rate
can show up to a 40 pct improvement as the eutectic carbides are refined in size and more uniformly
distributed in the martensitic matrix [42].
In view of the significant degree of hardening and presence of interconnected, brittle eutectic
carbides, high speed tool steels commonly have a limited reserve of plasticity and deformability [43]. In
addition to wear resistance, a smaller and more uniform distribution of eutectic carbides aids in the
margin of plasticity of the steel. The most effective means of reducing size of carbides is through hot
deformation, where greater deformation results in smaller carbides [43]. Additionally, high speed tool
steels have shown a reduction in carbide size following homogenization of the ingots through high
temperature heating ahead of hot plastic deformation [43].
The pseudobinary Fe-6W-5Mo-4Cr-2V-C (wt pct) diagram shown in Figure 2.15 helps to
describe the solidification process of a large group of high speed tool steels with a typical carbon content
between 0.8 and 1.2 wt pct. Figure 2.15 shows the solidification process begins with the precipitation of
δ-ferrite from the liquid. In the composition region of the Fe-Cr-W-Mo-V-C system of interest, C has a
significant effect on the liquidus temperature. It is seen that an increase in C content is associated with a
decrease in liquidus temperature and significant partitioning of C between dendritic δ-ferrite and the
liquid. This is coupled with the fact that C is the sole γ-stabilizing alloying element with low solubility in
ferrite [37].
Upon further decrease in temperature and continued C segregation to the liquid during the
formation of δ-ferrite, the alloy undergoes a peritectic reaction, designated by ‘P’ in Figure 2.15, in which
the γ phase precipitates from δ-ferrite and liquid at the δ/liquid interface through the reaction L + δ → γ.
The temperature of the onset of the peritectic reaction, T , is again influenced by composition, as well as
p
cooling rate. As previously discussed, C is an γ-stabilizing element, working to increase T by enlarging
p
the γ phase field. Elements such as Nb, V, Mo, and W are all ferrite stabilizers, enlarging the ferrite
stability field and decreasing T [37].
p
The next process in the solidification sequence is the formation of eutectic carbides through the
eutectic reaction L → γ + C. The residual inter-dendritic liquid, now enriched with C and carbide-forming
25
|
Colorado School of Mines
|
elements, decomposes through different eutectic reactions as temperature is decreased within the γ + C
two phase eutectic region [37]. This reaction leads to the formation of up to three different carbide
structures (denoted as C in Figure 2.15): HCP M C, FCC MC, and/or FCC M C [37]. The resultant
2 6
microstructure contains a continuous network of inter-dendritic carbides.
δ
δ δ
δ
P
δ δ
Figure 2.15 Pseudobinary Fe-C section for an Fe-6W-5Mo-4Cr-2V-C (wt pct) high speed tool steel
alloy [37]. ‘P’ designates the peritectic reaction of δ + L → γ.
The phase fractions will continue to change as temperature decreases further until residual
δ-ferrite in the dendrite cores transforms to austenite and carbide. This happens at almost the same
temperature as the eutectic reactions, yielding a morphology that bears resemblance to a lamellar
structure [41]. Eventually, the austenite matrix will decompose at lower temperatures.
The M C eutectic has a three dimensional structure characterized by a central platelet of M C
6 6
carbide, from which arise secondary platelets of M C, separated from each other by austenite at high
6
temperatures [37]. This morphology can be seen in Figure 2.16 in a Fe-1.0C-4Cr-6W-5Mo-2V (wt pct)
steel. Cross-sections seen in Figure 2.16 reveal lamellar platelets, commonly described as a
‘fishbone’-like morphology. The M C eutectic morphology is not influenced by chemical composition
6
nor cooling rate, and forms primarily due to the presence of Mo, W, and Fe [37, 38]. This carbide has a
complex FCC crystalline structure and its hardness is reported to be approximately 1500 HV [37].
The MC eutectic has three different morphologies depending on cooling rate: divorced, irregular,
and regular. All three morphologies are shown in Figure 2.17. Figure 2.17(a) shows the divorced MC
structure in a M2 steel with a composition of Fe-1.8C-0.5Si-0.53Mn-0.53 Ni-4.12 Cr-5.86Mo-6.02V
(wt pct). This morphology consists of eutectic carbides that occur as isolated crystals in the shape of disks
with an average diameter around 3-5 nm and a thickness of 1-2 nm [38]. Figure 2.17(b) is classified as
irregular MC eutectic in the same M2 steel cooled at a slower rate. This structure is described to be a
26
|
Colorado School of Mines
|
poorly coupled eutectic, in which the austenite and carbide phase are not fully paired and the MC carbide
has a petal-like morphology [37, 42]. Complex regular MC is presented in Figure 2.17(c) in the M2 tool
steel cooled at a faster rate, characterized by a coupled austenite and carbide eutectic, and the MC carbide
has a branched petal-like morphology [37, 42].
(a) (b)
Figure 2.16 Typical morphology of M C eutectic in a Fe-1.0C-4Cr-6W-5Mo-2V (wt pct) tool steel.
6
Images taken using (a) light optical microscopy and (b) scanning electron microscopy [37].
Irregular MC Complex
Regular MC
(a) (b) (c)
Figure 2.17 Representative morphologies of (a) divorced, (b) irregular, and (c) complex regular MC
eutectic structures in a tool steel with a composition of Fe-1.8C-0.5Si-0.53Mn-0.53Ni-4.12Cr-5.86Mo-
6.02V (wt pct) [37, 42].
The type of MC eutectic formed is dependent on the difference between the peritectic temperature
and the temperature of eutectic formation; the larger the difference, the greater the tendency to form a
divorced structure [37]. The higher the cooling rate, however, the higher the tendency to form a less
coupled, irregular MC eutectic. An MC-type eutectic forms primarily due to V or Nb additions and has an
FCC structure and hardness values up to 3000 HV [37, 38, 42]. In Nb-alloyed high speed tool steels, V
27
|
Colorado School of Mines
|
and the other weaker carbide formers are partially replaced by Nb. The addition of an element that is a
strong carbide former, however, as well as high additions of carbon, can bring about a shift in the MC
precipitation temperature above the primary δ-ferrite, forming NbC carbides directly from the bulk liquid
phase [37]. These are regarded as primary, rather than eutectic carbides. Primary monocarbide MC
nucleates before the onset of the eutectic ledeburite crystallization, especially with high Nb alloying in
high C steels [41]. MC nucleates heterogeneously, often on platelets of Al O , as the basal plane offers
2 3
epitaxy conditions for the (111) plane of the MC structure [41]. Depending on the existence of
heterogeneous nucleation sites, crystallization of MC may occur at temperatures above that of δ-ferrite
nucleation, which then nucleates on MC crystals. When NbC crystallization precedes δ nucleation,
primary crystals are pushed ahead in front of δ dendrites. When these crystals exist in inter-dendritic
puddles, they can trigger the crystallization of highly regular eutectic MC rods, which take on the
orientation of the (100) growth directions of the primary MC crystals [41]. However, the degree of
heterogeneous nucleation can move the first formation of primary MC to much higher temperatures, with
a variation of approximately 100 ºC, which can totally suppress the formation of MC eutectic, as all Nb is
already tied up in primary NbC [37, 41].
Similar to MC eutectic, M C has two different variations in morphology: irregular and complex
2
regular. Unlike both M C and MC, however, M C eutectic is a meta-stable constituent and tends to break
6 2
down in a certain range of temperatures. Both M C eutectic morphologies are shown in Figure 2.18.
2
(a) (b)
Figure 2.18 The three-dimensional morphology of M C carbides in a M2 steel with a compositon of Fe-
2
0.9C-0.2Mn-5.9W-4.8Mo-4.0Cr-2.08V (wt pct) showing (a) irregular M C and (b) complex regular
2
M C [39].
2
The first M C morphology identified is irregular M C eutectic in a M2 tool steel, with similarities
2 2
to irregular MC eutectic. This structure is characterized by a ragged boundary that does not clearly outline
28
|
Colorado School of Mines
|
the interface between the eutectic and the matrix, suggesting limited coupling between the growth of the
eutectic phases [37, 39]. This morphology is promoted by either a low cooling rate or high V content.
Unlike the irregular structure, complex regular eutectic is characterized by a smooth boundary that clearly
outlines the interface between the matrix and the eutectic colony [37, 39]. M C carbides can contain any
2
of the principal carbide-forming elements in high speed steels, but are generally rich in Mo. They have a
hexagonal structure with hardness values around 2000 HV [37, 42]. M C eutectic is metastable and
2
reportedly decomposes according to the reaction M C + γ-Fe → M C + MC when heated between
2 6
900 - 1150 °C [37 – 39].
2.8.3 Eutectic versus Primary Nb Carbides
As previously discussed, carbides can precipitate in one of two forms depending on when they
form during the solidification process. If their formation occurs during the eutectic reaction L → γ + C,
where C can be carbides of the form M C, M C, or MC, they are eutectic carbides. While alloying
6 2
elements such as Mo, W, and Cr aid in the formation of M C and M C eutectic carbides, MC-type
6 2
carbides form primarily due to the presence of V and/or Nb. The remainder of this section focuses on
Nb-containing carbides. One example of a NbC eutectic carbide network in a tool steel is shown in
Figure 2.19.
Figure 2.19 Eutectic carbide structure in a Fe-2.2C-0.63Si-0.28Mn-19.15Cr-0.53Mo-2.0Nb (wt pct)
steel. Micrograph taken using SEM [36].
An important microstructural aspect of eutectic carbide networks is their distribution within the
austenite/martensite matrix. Coarse aggregates of carbides or extremely large carbides are not desired
since they leave part of the matrix disproportionally softer and susceptible to wear [35]. After forming
29
|
Colorado School of Mines
|
through rolling or forging, these eutectic carbide networks are broken up, and the carbides tend to be
better distributed in the microstructure.
The introduction of eutectic carbides in grinding media has not been extensively researched to
date; however, the wear resistance of high speed tool steels currently relies on the presence of these hard,
blocky carbides [38, 40, 41]. Alloys used in high speed steels are designed in such a way that they are
able to maintain high hardness and wear resistance at high temperatures that result from cutting metal at
high speeds. As these correspond somewhat to the properties desired of grinding balls in ball mills, the
success of NbC eutectic carbides in high speed steels emphasizes their potential in other wear prone
products.
As increasing amounts of Nb are added to a steel, carbides begin to form directly in the liquid,
growing freely since they do not have to respect a cooperative growth mechanism with austenite [35].
Such carbides are referred to as primary carbides and they tend to be large with cuboidal or octahedral
morphologies [35]. Forming at such high temperatures, primary carbides reportedly aid in refining the
microstructure through a combination of various mechanisms [44]. Additionally, primary NbC particles
are reported to have a very similar specific weight to liquid steel, negating gravitational segregation and
allowing for a more homogeneous dispersion [44].
The benefits provided by primary NbC must also be balanced with common issues that are
simultaneously present. With such large sizes of these carbides and often high volumes, they are hard to
break down during hot deformation and may remain large and undispersed in final microstructures. This
can also lead to a decrease in grindability and toughness as they act as stress concentration sites [35]. At a
certain point, primary carbides become undesirable. Recently, research has focused on intermediate Nb
amounts with an aim to optimize both wear resistance and toughness by forming eutectic rather than
primary carbides, with alloying commonly between 0.15 - 1.2 wt pct Nb [35].
Jarreta and Wright studied the presence of primary NbC and how it aided in the wear behavior of
slurry pump impellers in a mine in Brazil [44]. Slurry pumps are centrifugal pumps which have impellers
that are subjected to a considerable amount of wear. In a field study, the lifetime of current impellers was
tested against a patented material which used an addition of 15 wt pct NbC under identical operating
conditions [44]. With wear measured by weight loss of the impellers, it was found that the steel alloyed
with 15 wt pct NbC nearly tripled the impeller service life. Weight loss data are presented below in
Figure 2.20. Process variables including slurry feed rate, dry ore feed rate, working time, and motor
current were monitored online, and the test was paused every 500 hr to weigh and evaluate the impellers.
The end of service life was determined by the loss in yield of production due to loss in the pumping
power of the impeller [44].
30
|
Colorado School of Mines
|
A05 +
15 wt.% NbC
A05
Figure 2.20 A field study by Jarreta and Wright showing an increase in more than 3 times the service life
of a slurry pump impeller due to the addition of 15 wt pct NbC [44].
The abrasion mechanism acting on the slurry pump impellers was considered to be low stress
sliding/scratching abrasion (LSSA) with a high volume of abrasive particles [44]. The abrasion takes
place when these hard particles move across the impeller surfaces at different directions and velocities,
penetrating the surfaces and removing material. The amount of material lost is dependent on the ore itself
(velocity, hardness, edge sharpness, angle of indentation, and size), as well as the microstructure of the
impeller. In the A05 steel impeller, the presence of Cr carbides was able to partially improve the
impeller’s resistance to abrasion, however the wear resistance was improved remarkably in the patented
alloy due to the introduction of a high volume percent of hard, tough, homogeneously dispersed NbC
particles [44].
Figure 2.21 shows two microstructures containing primary NbC particles in the matrix. The first
is a cast steel with 2.9 wt pct C and 4.0 wt pct Nb and the second is a Fe-NbC steel similar to that used in
the slurry pump impeller study in which 10 - 20 μm NbC carbides are dispersed in an iron matrix; NbC
particles were directly added to the molten steel [44].
For the steel shown in Figure 2.21(a), the NbC (darker particles) precipitated at higher
temperatures in the melt prior to the precipitation of δ-ferrite. They were able to form well defined
crystallographic cubic or octahedral structures distributed randomly throughout the microstructure. The
largest carbides grew to a size on the order of 50 μm, which allowed for microhardness testing that
revealed hardness values around 2556 HV [36]. The lighter raised areas are eutectic Cr carbides.
The microstructure displayed in Figure 2.21(b) was developed through the addition of NbC
particles directly to the melt, rather than through precipitation as a result of excess C and Nb alloying.
One method of achieving such a microstructure was patented by K. Dolman in which scrap steel and/or
scrap WCI, high carbon FeCr, FeNb, and a source of free C such as graphite are held molten long enough
31
|
Colorado School of Mines
|
to dissolve the carbon and homogenize the melt to produce the desired level of NbC in a Fe matrix [45].
The carbides are then either added directly to the melt, crushed to fill a core-wire consumable, or
incorporated through laser cladding or plasma transferred arc (PTA) [44, 45]. Resulting microstructures
consist of NbC that are globular and not interconnected [44].
(a) (b)
Figure 2.21 SEM images of primary NbC in (a) an alloy with 2.9 wt pct C and 4.0 wt pct Nb [36] and
(b) an Fe-NbC steel [44].
2.8.4 Effect of Nb on Hardenability
Nb is reported to have an influence on the hardenability and martensite start (M) temperature in
s
steel as discussed below. Literature presented here discusses the effect of Nb micro-alloying in low
carbon steels; however, concepts presented in these studies may be applicable to behavior seen in
materials of interest to this report.
Capdevila et al. studied the influence of Nb on the M temperature for samples with three different carbon
s
contents and increasing Nb alloying [46]. The results are presented in Figure 2.22 . The increase in M
s
observed as Nb is increased indicates that martensite begins forming at higher temperatures. This increase
is more pronounced as C concentration is increased from 0.1 to 0.8 wt pct. Nb is a strong carbide-forming
element, which, in turn, can influence the activity of C in solid solution. Such interactions between C and
carbide forming elements may remove C from solution, weakening the role of carbon on hardness, and
increase the M temperature. These interactions are stronger as C content is increased.
s
J.M. Gray reported that the effect of a 0.05 wt pct Nb-alloyed steel on hardenability was
comparable to alloying with 1 wt pct Mn, becoming even more pronounced at faster cooling rates [47].
Similarly, P. Yan and H. Bhadeshia observed a depression of the γ to ferrite (α) transformation
temperature and an intrinsic improvement of hardenability in the presence of dissolved Nb [48]. Three
predominant factors are often attributed to this increase in hardenability: an increase in austenite grain
32
|
Colorado School of Mines
|
size, increase in concentration of Nb in solution, and a decrease in fraction of NbC in the austenite. In
Nb-containing “control-rolled” steels, Nb additions accelerate austenite decomposition because of the
strain energy stored in the deformed austenite [49], but this mechanism is not applicable to the heat
treatments discussed here.
Figure 2.22 Effect of Nb on the M temperature for carbon contents of 0.1, 0.4, and 0.8 wt pct and a
s
constant prior austenite grain size of 20 μm [46].
The hardenability increase in Nb-alloyed steels is not independent of the prior austenite
grain (PAG) size, as an increase in grain size leads to a reduction in heterogeneous nucleation sites and
ferrite/pearlite formation is suppressed. An increase in M is observed as PAG size increases with higher
s
austenitization temperatures. In Nb micro-alloyed steel, grain growth is limited until a temperature of
carbide dissolution is reached. After this temperature, the grain boundary pinning effect vanishes as the
particles dissolve completely, leading to rapid austenite grain growth [50]. However, it was observed by
Fossaert et al. that this increase in hardenability occurs over a range of very little austenite grain growth
and that once the critical dissolution temperature is reached, martensite transformation is no longer
affected by grain size [50]. This indicates that the dominant factor affecting the hardenability is not the
austenite grain size.
Solute drag and “solute drag-like effects” are commonly believed to enhance hardenability in
Nb-alloyed steels. However, in a tracer diffusion study performed by DeArdo et al., the interdiffusion
coefficient of Nb in γ is significantly larger than the self-diffusion coefficient in iron (Fe) [51]. If this can
be extrapolated to an α/γ interface, then significant drag effects due to Nb diffusion at the interface are
less likely given that this is a diffusional transformation and Fe atoms must diffuse for the transformation
to take place [51]. For that reason, the effect of Nb on hardenability was not attributed to solute-drag
behavior.
33
|
Colorado School of Mines
|
Given that PAG size and solute drag may not account for the increase in hardenability observed,
the role of Nb in steel can best be understood when considering the retardation effect Nb has on γ
decomposition to α. The rate of γ to α transformation was reported by P. Yan et al. to be the slowest in a
0.028 wt pct Nb alloyed steel compared with lower concentrations [48]. This suppression effect was
greater for the diffusional transformation to allotriomorphic α than in the displacive transformations to
bainite and/or martensite [48]. This effect is more pronounced when all the Nb is in solution, and
hardenability begins to decrease once carbide precipitates can form in γ prior to the α transformation.
Additionally, with higher Nb content, Nb(C,N) readily precipitates upon cooling, promoting α formation
by generating more nucleation sites; albeit Nb precipitation in recrystallized γ is relatively sluggish [50].
The final mechanism by which Nb is believed to influence hardenability in steel is due to
segregation of Nb to austenite grain boundaries. Nb segregates to γ boundaries in order to reduce its
lattice strain energy, and, in turn, lowers the γ/γ interfacial energy per unit area, making the boundaries
less potent heterogeneous nucleation sites [48, 50]. Of the common micro-alloying elements, Nb is
reported as interacting the most strongly with γ boundaries because Nb atoms have a large misfit within
the Fe lattice, making γ boundaries attractive sites.
34
|
Colorado School of Mines
|
CHAPTER 3
EXPERIMENTAL PROCEDURES
3.1 Introduction
This chapter summarizes the materials, equipment, and characterization methods employed in the
present work. The chemical compositions of the industrial and laboratory prepared materials studied are
provided. Estimations of critical transformation temperatures of each material are determined using both
empirical calculations as well as Thermo-Calc® thermodynamic modeling software. Dilatometry was used
to validate these temperatures and to aid in heat treatment design, with a goal of matching laboratory
prepared material microstructure with industrially produced material. Characterization was carried out
using light optical microscopy (LOM), scanning electron microscopy (SEM), and X-ray
diffraction (XRD). Mechanical test methods used in this study include macro- and micro-hardness testing,
Bond abrasion wear testing, and dry sand/rubber wheel testing. All error analysis provided is the standard
deviation of the data collected.
3.2 Materials
This section details the materials studied. High carbon industrial bar stock and industrial, forged
grinding balls, as well as laboratory cast and hot rolled plates, were investigated. High Cr cast iron
grinding media are investigated for the limited purpose of a comparison to forged HCS grinding media;
however, the focus of this project is on high carbon forged grinding balls.
3.2.1 Industrial Bar Stock and Grinding Balls
The 3.5 and 5.08 cm bar stock listed in Table 3.1 were provided by Gerdau Long Bar Product,
North America, to be used as reference material for mechanical properties and microstructure. The 3.5 cm
and 5.08 cm diameter round bars were 7.24 m and 7.47 m in length, respectively. Chemical composition
of the grinding balls was not provided; however, they are expected to match the chemistry of the bar stock
as they are cut and forged from the same material.
Table 3.1 Industrial Bar Stock Chemical Composition (wt pct)
Bar Diameter C Nb Mn Cr Cu Si Ni Mo S P Sn V
3.5 cm 0.98 0.012 0.96 0.44 0.26 0.24 0.12 0.024 0.021 0.014 0.010 0.005
5.08 cm 0.99 0.010 0.96 0.57 0.26 0.20 0.10 0.026 0.020 0.010 0.013 0.004
35
|
Colorado School of Mines
|
3.2.2 Laboratory Prepared Heats
A set of laboratory prepared heats, listed in Table 3.2, were produced by vacuum induction
melting the industrial bar stock with incremental alloying of Nb up to 1 wt pct. Two ingots of each
chemistry were cast and designated by A or B. Materials identified as ‘HR’ are the same ingots that were
laboratory hot rolled from about 12.7 to 0.85 cm thickness. Only the chemistry from one hot rolled plate
of each alloy was provided. NP indicates that the composition was not provided.
Table 3.2 Laboratory Prepared Material Chemical Composition (wt pct)
Material C Nb Mn Cr Cu Si Ni Mo S P Sn V
0.01 Nb - A 1.04 0.009 0.97 0.61 0.29 0.20 0.15 0.033 0.021 0.006 0.009 0.005
0.01 Nb - B 1.01 0.010 0.95 0.60 0.29 0.20 0.14 0.032 0.023 0.006 0.009 0.002
0.01 Nb - HR 1.01 0.007 0.98 0.60 NP 0.19 NP NP 0.017 0.008 0.009 NP
0.25 Nb - A 1.00 0.285 0.92 0.61 0.28 0.22 0.14 0.032 0.018 0.005 0.008 0.002
0.25 Nb - B 1.00 0.294 0.88 0.59 0.27 0.23 0.14 0.032 0.019 0.004 0.008 0.003
0.25 Nb - HR 1.00 0.260 0.98 0.60 NP 0.20 NP NP 0.018 0.008 0.010 NP
0.5 Nb - A 1.01 0.519 0.95 0.61 0.32 0.20 0.15 0.032 0.022 0.010 0.010 0.002
0.5 Nb - B 0.96 0.520 0.91 0.60 0.30 0.20 0.14 0.032 0.022 0.009 0.010 0.002
0.5 Nb - HR 0.96 0.500 0.96 0.58 NP 0.19 NP NP 0.016 0.009 0.011 NP
1.0 Nb - A 0.95 1.093 0.92 0.60 0.31 0.22 0.14 0.031 0.023 0.011 0.010 0.003
1.0 Nb - B 0.97 1.085 0.92 0.60 0.31 0.22 0.15 0.032 0.022 0.011 0.010 0.003
1.0 Nb - HR 0.96 1.022 0.94 0.60 NP 0.20 NP NP 0.020 0.012 0.010 NP
After each ingot was cast, slabs from the top and bottom were cut and removed for
microstructural analysis to better understand the solidification behavior of the material. The remaining
material was hot rolled. Figure 3.1 shows a schematic diagram of the lab cast ingots to show the geometry
of the slabs and the material that was used for hot rolling. Each ingot had dimensions of 30.5 cm in
height, 15.2 cm in width, and 12.7 cm in thickness at the top and 11.4 cm at the bottom. Each ingot had
2 - 3 slabs cut from the top and the bottom, ranging from about 1.25 to 2.5 cm thick.
3.3 Microstructural Characterization
This section discusses the microstructural characterization techniques and equipment used in the
present work including light optical microscopy (LOM), scanning electron microscopy (SEM), LECO®
carbon analysis, and X-ray diffraction (XRD).
36
|
Colorado School of Mines
|
Figure 3.1 Schematic of lab cast ingot displaying location of slabs cut for analysis and the section that
was hot rolled.
3.3.1 Light Optical Microscopy
LOM was carried out using an Olympus PMG3 microscope and an Olympus DSX100
Opto-Digital microscope. Samples were prepared for LOM by sequentially grinding the surface with 320,
400 and 600 grit silicon carbide abrasive paper, followed by a 9, 6, 3, and 1 micron diamond polish, and
finally a 0.5 micron colloidal silica polish. After each grinding and polishing step, samples were washed
with soap and dried with isopropyl alcohol. Samples were subsequently etched using a 2 pct Nital
solution consisting of 2 vol pct nitric acid (HNO ) in ethanol (C H OH) for roughly 15 seconds to reveal
3 2 5
the microstructure. LOM analysis was performed on all as-received and heat treated material.
3.3.2 Scanning Electron Microscopy
Field emission scanning electron microscopy (FESEM) was conducted using a JEOL®
JSM-7000F microscope on baseline bar stock and grinding ball samples, as well as the high chrome-cast
grinding ball. Environmental SEM (ESEM) was performed using a Quanta 600® environmental
microscope on the slabs from the as-cast ingots to analyze solidification behavior, with a schematic of
samples sectioned from each slab shown in Figure 3.2. SEM was additionally conducted using a
TESCAN® scanning electron microscope of the hot-rolled plates before and after salt pot heat treatments.
Secondary electron images (SEI) were recorded of as-cast pearlitic and heat-treated
martensite-retained austenite microstructures, and backscatter electron (BSE) images were taken of
niobium carbides (NbC) to provide compositional (atomic number) contrast. Polished and etched
metallographic samples were prepared for SEM examination using the same procedure as for LOM. The
same etchant was used with a slightly shorter etching time of approximately 10 seconds.
37
|
Colorado School of Mines
|
(c) (a)
(d)
(f) (e) (b)
Figure 3.2 Schematic of lab cast ingot slab displaying where samples were sectioned for analysis – (a)
9 cm, (b) 7 cm, (c) 5.7 cm, (d) 4.5 cm, (e) 3.5 cm from the center of the slab, as well as (f) directly at
the center of the slab.
Qualitative chemical microanalysis was employed in conjunction with the SEM, ESEM, and
FESEM via EDS (energy dispersive X-ray spectroscopy). EDS is an analytical technique used for
elemental characterization of a sample, relying on characteristic X-ray emission from a sample due to
excitation of its electrons. Point scans produce a spectrum with characteristic peaks useful in identifying
specific elements, used in the present work to identify microstructural components. EDS elemental
mapping was additionally employed to capture the signal of certain elements across an area of the sample.
3.3.3 LECO® Carbon Analysis
A LECO® carbon combustion analyzer was used to confirm whether significant carbon
segregation took place during ingot solidification. Slabs taken from both the top and bottom of each ingot
were machined such that pieces approximately 1 g in mass were cut sequentially from the surface of the
ingot to the center using a LECO® MSX saw. A schematic displaying how slabs were machined is shown
in Figure 3.3. Each sample was approximately 0.5 cm x 0.5 cm x 0.5 cm in dimensions.
The analysis consists of loading a pre-weighed sample into a chamber where it is combusted in a
stream of purified oxygen using radio frequency (RF) induction to heat and melt the sample. Carbon
present in the sample is oxidized to carbon dioxide (CO ) and swept by the oxygen carrier past a heated
2
catalyst where carbon monoxide (CO) is converted to CO and carbon is detected as CO by a pair of
2 2
non-dispersive infrared (NDIR) cells [52]. The analyzer reports carbon level as a weight pct based on the
initial weight of the sample.
38
|
Colorado School of Mines
|
Figure 3.3 Schematic of cast lab ingot slab showing sample machining for LECO® carbon analysis to
assess any carbon segregation from surface to center during solidification.
3.3.4 X-ray Diffraction
XRD was performed on grinding balls, dilatometry samples, and salt pot treated samples to
characterize the retained austenite fraction. Grinding ball samples were machined using a LECO® MSX
saw by cutting through the diameter of the ball and removing samples from both the surface and the
center for XRD analysis. Dilatometry samples were cut in half lengthwise using a LECO® MSX saw to
expose a flat surface from the center of the sample for analysis. Salt pot treated samples were cut
similarly to the grinding ball to again analyze a small sample from both the surface and the center of the
material.
Once sectioned, samples were chemically thinned in a solution of 10 parts de-ionized (DI) water,
10 parts hydrogen peroxide (H O ) and 2 parts hydrofluoric acid (HF) for 2 minutes. Selective thinning
2 2
was performed on only the surface of interest for XRD analysis by coating all other faces with liquid
electrical tape. The purpose of chemical thinning prior to XRD was to remove any surface oxides without
inducing surface damage via grinding which may potentially cause a portion of the retained austenite to
undergo transformation to martensite [53].
XRD was performed using a Siemens® X-ray diffractometer with a copper X-ray source at 35 kV
and 30 mA. Samples were scanned from a 2-theta value of 40º to 102º to reveal the {110}, {200}, {211},
and {220} ferrite peaks, and the {111}, {200}, {220}, and {311} austenite peaks [53]. Step size was set
to 0.05º and a 5 second dwell time was used. Diffraction data were uploaded into a JADE software
package to remove the Cu K peak, determine and remove the background, and smooth out the data from
α2
background noise.
The processed XRD data were then used to calculate phase fractions of austenite and ferrite
(martensite) using a peak integration method outlined by the Society of Automotive Engineers (SAE) [54]
and ASTM Standard E975-13 [55]. For numerous ferrite (α) and austenite (γ) peaks, the volume fraction
of austenite, V, is determined by averaging each measured integrated intensity to R-value ratio through
γ
the following equation:
39
|
Colorado School of Mines
|
1 ∑q I γj
q j=1R
γj (3.1)
V =
γ 1 ∑P I αi +1 ∑q I γj
P i=1R q j=1R
αi γj
where I and I are the integrated intensity of a given austenite and ferrite peak, respectively, q and P are
γ α
the number of ferrite and austenite peaks, respectively, and R is a parameter proportional to the theoretical
integrated intensity which depends on the interplanar spacing, Bragg angle (θ), crystal structure, and the
composition of the phase being measured. R is calculated via the following equations:
|F2| p(1+cos22θ)
R= e-2M (3.2)
v2 sin2θcosθ
Bsin2θ
(3.3)
M=
λ2
B=8π2μ2 (3.4)
s
where v is the volume of the unit cell, F is the structure factor, p is a multiplicity factor of the (hkl)
reflection, e-2M is the Debye-Waller factor, λ is the wavelength of the Cu source, and μ is the mean square
s
displacement of the atoms from their mean position in a direction perpendicular to the diffracting
plane [54, 55].
3.4 Hardness Testing
In order to assess the results of dilatometry and salt pot heat treatments, hardness readings were
taken of all experimental material and compared to industrial grinding balls. Rockwell macrohardness and
Vickers microhardness testing were performed both at the surface and center of each sample to correlate
relative wear resistance to hardness.
3.4.1 Rockwell Macrohardness
Rockwell macrohardness testing was performed in accordance with ASTM Standard E18-20
using the HRC scale [56]. A diamond intender tip was used to make contact with the sample surface at a
load of 150 kgf and a dwell time of 10 s. Hardness is correlated to the depth at which the indenter
penetrates the surface of the sample through the following equation:
40
|
Colorado School of Mines
|
h (3.5)
H = 100 -
0.002
where H is Rockwell hardness and h is indentation depth in mm. A standard test block with a hardness of
61.5 HRC was used before each set of measurements to validate the accuracy of the machine. Hardness
was determined by performing five indentations at each location, at least three indentation distances apart,
and averaging the final three values.
3.4.2 Vickers Microhardness
Vickers microhardness testing was performed in accordance with ASTM Standard E92-17 using a
LECO® AMH55 automated microhardness tester [57]. A load of 30 kgf and dwell time of 10 s was used
during testing. Vickers microhardness was used to estimate the depth of decarburization during salt pot
heat treatment by searching for any hardness drop near the surface of the sample. Samples were polished
to a 1 µm diamond finish prior to hardness testing. Single row arrays of 20 indents spaced 100 μm apart
were placed at each edge of the sample. A schematic of this analysis is shown in Figure 3.4.
Figure 3.4 Schematic diagrams of indent placements for Vickers microhardness analysis of
decarburization following heat treatments.
3.5 Heat Treatment Design
Once the microstructure and mechanical properties of the industrial grinding balls were
established, a heat treatment to replicate these properties in the experimental material was necessary.
Determination of critical temperatures such as A and M temperature were first estimated via empirical
cm s
calculations and Thermo-Calc® modeling. Quench temperatures resulting in retained austenite content
comparable to current grinding balls were then predicted using the Koisten-Marburger equation. Quench
temperature was the temperature to which samples were quenched following the austenitization step.
3.5.1 M Temperature Calculations
s
The M temperature is of interest in this project for the purpose of designing a heat treatment for
s
41
|
Colorado School of Mines
|
each experimental composition. M is most commonly modeled for hypoeutectoid steels and existing
s
formulae [58 – 63] are less accurate at the higher carbon levels seen in this study, so 2 different empirical
expressions for calculating M temperatures were explored – a parabolic model published by K.W.
s
Andrews [64] and a model based on the effect of alloying elements predicted by K. Ishida [65] which
takes into account the effect of Nb alloying. These formulae are presented in Equations 3.6 and 3.7,
respectively, where the composition of each element is in wt pct.
M (°C) = 512 - 453 C - 16.9 Ni + 15 Cr - 9.5 Mo + 217 C2 - 71.5 C ∙ Mn - 67.6 C ∙ Cr (3.6)
s
M (°C) = 545 - 330 C + 7 Co - 14 Cr - 13 Cu - 23 Mn - 5 Mo - 4 Nb - 13 Ni - 7 Si + 4 V (3.7)
s
Using the compositions in Table 3.1 and Table 3.2 and Equations 3.6 and 3.7, the M
s
temperatures shown in Table 3.3 were predicted for each material by taking the average composition from
each ingot and hot rolled plate of all experimental lab heats.
Table 3.3 M Temperature Predictions
s
M (ºC)
s
Equation
Bar Stock 0.01 Nb 0.25 Nb 0.5 Nb 1.0 Nb
Andrews 180 172 175 177 181
Ishida 183 174 177 182 187
3.5.2 Thermo-Calc® Model Predictions
Along with empirical temperature calculations, Thermo-Calc® software was used to aid in
predicting critical temperatures as well as phase stability during heat treatments with increasing Nb
alloying. A set of phase diagrams are displayed in Figure 3.5; the first is a simple Fe-Nb binary and the
second is a Fe-1.0C-Nb pseudobinary diagram. The third pseudobinary phase diagram is a more complex
Fe-1.0C-0.96Mn-0.44Cr-0.26Cu-0.24Si-0.12Ni-Nb system to incorporate the effect of the major alloying
elements.
Figure 3.5 can aid in determining equilibrium transformation temperatures such as the A , or the
1
α + Fe C (pearlite) solubility line, and the A , or the cementite solubility line. Based on the pseudobinary
3 cm
diagrams in Figure 3.5(b) and (c), Table 3.4 outlines the predicted A temperatures as a function of Nb
cm
alloying. The first set of predictions (PS1) only account for C and Nb alloying, whereas the PS2
predictions, like Figure 3.5(c), also include contributions from Mn, Cr, Cu, Si, and Ni. However, the
complexity and introduction of a number of new phases may affect the results as well. Overall, it is
expected that the A of the baseline material is approximately 726 ºC and is expected to be insensitive to
1
the Nb variations. The A of the baseline material is expected to be between 820 - 855 °C and decreases
cm
42
|
Colorado School of Mines
|
3.5.3 Quench Temperature Design
As previously discussed, industrial grinding balls have a specified volume fraction of retained
austenite that provides an attractive combination of surface hardness and toughness. Using the average M
s
temperature calculated in Equations 3.6 and 3.7 for the industrial bar stock and the Koistinen-Marburger
relationship [66] presented in Equation 3.8, the phase fraction of untransformed austenite as a function of
quench temperature can be predicted. This relationship is shown in Figure 3.6.
f = 1 -
e-1.1×10-2ΔT (3.8)
martensite
where ΔT is the difference between the M temperature and the quench temperature in ºC or K. The
s
following relationship is then used to determine the volume fraction of austenite at the quench
temperature, assuming that the matrix is made up entirely of martensite and untransformed austenite.
f + f = 1 (3.9)
austenite martensite
Figure 3.6 Koistinen-Marburger relationship between phase fraction of untransformed austenite and
quench temperature for industrial bar stock.
3.6 Dilatometry
Dilatometry was performed for the purpose of heat-treating bar stock and hot rolled plate samples
under controlled conditions and to experimentally validate critical transformation temperatures. The
dilatometry experiments were conducted using a TA Instruments HF Generator 750 and
DIL805L software. The dilatometer uses induction heating and either argon or helium cooling with the
44
|
Colorado School of Mines
|
temperature monitored via a platinum thermocouple welded to the center of the sample surface.
Cylindrical samples were machined 10 mm in length and 4 mm in diameter.
The A , A , and M temperatures were all measured using continuous heating and cooling in the
1 cm s
dilatometer to confirm calculations based on Equations 3.6 and 3.7. Samples were heated at 90 ºC/s under
vacuum to 850 ºC to measure the A and A . The A is determined as the temperature at which linear
c1 cm c1
expansion first deviates from linearity due to a volume contraction associated with austenite formation, as
FCC-austenite is a more dense phase compared to BCC-ferrite. Eventually, the curve reverses to a volume
expansion once the matrix is fully austenitic due to a continued increase in temperature. The A is
cm
defined as the point at which the thermal expansion again becomes linear and all the cementite is
dissolved. A and A temperatures are determined by extrapolating the linear portions of these thermal
c1 cm
expansion curves [67]. Example continuous heating dilatometry curves are shown in Figure 3.7(a) with
the A and A temperatures highlighted. While carbide (cementite) dissolution is usually considered to
c1 cm
be complete above A , in the present steels is should be recognized that NbC may remain undissolved
cm
above A .
cm
Upon continuous cooling, a linear contraction is observed initially. The A upon cooling
cm
corresponds to the onset of pro-eutectoid cementite formation. This temperature is identified by a slight
deviation in linear contraction. The A temperature corresponds to the onset of non-isothermal
r1
decomposition of austenite to pearlite. This temperature is determined similarly to the A and is
c1
identified by a transition from linear contraction to a short period of linear expansion. Once the
transformation of γ → α + FeC is complete, contraction is reinstated as the sample is cooled further. If
3
the quench temperature is sufficiently rapid, the A and A are surpassed without transformation, and a
r1 cm
sudden volume expansion is observed as the M temperature is reached. The critical temperature is
s
determined as the point at which linear contraction ends and expansion begins. Example cooling curves
are shown in Figure 3.7(b) and (c).
3.7 Salt Pot Heat Treatment
Once heat treatment parameters were confirmed with dilatometry, wear test samples were
processed via salt pot immersion as the samples were too large to be treated in the dilatometer. The
austenitization step was performed in a high temperature salt pot composed of sodium chloride and
potassium chloride, with an upper temperature limit of 900 ºC. Two different quench parameters were
tested – the first being a direct water quench and the second involved quenching to 150 ºC in the low
temperature salt pot. The low temperature salt is composed of sodium nitrate and potassium nitrate, with
150 ºC as the lower operating limit. A second quench temperature was explored to both mitigate quench
cracking if present in the samples quenched in water, as well as investigate differences in untransformed
45
|
Colorado School of Mines
|
To track heating and quench rates of the salt pot processing, one sample with dimensions of
25.4 mm x 72.6 mm x 8.47 mm was run with a thermocouple inserted into its center and temperature was
tracked during the heat treatment. The sample was heated to 850 ºC for 30 min, and one sample received a
water quench and the other sample was quenched to 150 ºC in the low temperature salt pot and held for
20 min. A schematic of time-temperature profiles is shown in Figure 3.8.
1800 s
1200 s
Figure 3.8 Proposed time-temperature profiles for water quench (dotted) and 150 ºC quench in the low
temperature salt pot (solid) for salt pot temperature tracking.
3.8 Bond Abrasion Testing
The Bond abrasion tester was used to determine resistance to wear as a function of Nb alloying.
There is no ASTM standard for this test, but testing is outlined in a series of publications by
F.C. Bond [9, 27, 31]. The test set-up is shown in Figure 3.9. The primary components of the wear tester
include inner and outer rotating shafts, and a rotating drum lined with mesh to promote lift of the abrasive
ore; all of which rotate in the same direction. The drum is 305 mm (12.0 in) in diameter and
102 mm (4.0 in) deep.
Wear resistance was measured for 2 different ores – an iron ore and a copper ore with abrasion
indices reported in literature of 0.0770 and 0.1475, respectively [31]. All ore was washed and dried prior
to testing, then sieved to pass a 1.905 cm2 mesh and be retained on a 1.270 cm2 mesh. Steel impeller
samples were machined according to the dimensions in Figure 3.10, heat treated, and surface ground to
0.4 μm (16 μin) roughness average (Ra).
Four samples of each alloy composition (0.01, 0.25, 0.5 and 1.0 Nb) were tested with the Fe ore,
and only one sample was tested with the Cu ore due to material limitations. Both faces of each sample
were tested, with the exception of a few samples that fractured during testing, for a total of 32 tests run
with Fe ore and 8 tests with Cu ore. Each sample was weighed to the nearest 0.1 mg prior to testing and
47
|
Colorado School of Mines
|
one sample was loaded into the same side of the inner rotor for each test, along with 400 g of ore. Tests
were conducted in 4 increments of 15 min (900 s) for a total test time of 60 min (3600 s). After each
15 min segment, the steel impellers were washed with soap and water, the water was displaced with
ethanol, and the samples were allowed to fully dry before again being weighed to the nearest 0.1 mg. All
ore was removed from the drum and 400 g of fresh abrasive was loaded for each interval. Abrasive prior
to testing is shown in Figure 3.11. Total mass loss was converted to volume loss using the material’s
density of 7.8 g cm-3.
(a) (b)
Figure 3.9 Bond abrasion machine set-up showing (a) the mechanical innerworkings including the
motor and gear box and (b) the rotating drum with 400 g abrasive ore and two steel samples
mounted.
After testing, one sample from each composition tested with Fe ore was sectioned approximately
1.3 cm (0.5 in) from the bottom of the sample and mounted on the 25.4 x 6.4 cm face such that wear
through the thickness at the sample surface could be evaluated via SEM. A schematic of a sectioned and
mounted sample is shown in Figure 3.12.
3.9 Dry Sand/Rubber Wheel Testing
Dry sand/rubber wheel (DSRW) testing is used to determine the resistance of metallic materials
to scratching abrasion, as outlined by ASTM standard G65. The abrasive is introduced between a rotating
rubber wheel with a chlorobutyl or neoprene rubber rim of a specified hardness. The test specimen is
pressed against the rotating rubber wheel at a specified force by means of a lever arm, while a controlled
flow of sand abrades the test surface. The rotation of the wheel is such that its contact face moves in the
direction of the sand flow. Specimens are weighed before and after the test and the loss in mass is
recorded and correlated to wear resistance.
48
|
Colorado School of Mines
|
Figure 3.12 Schematic of sectioning performed on Bond abrasion samples after testing to evaluate wear
at the sample surface with SEM.
Typical specimens used are rectangular in shape with dimensions of 25 x 76 mm (1 x 3 in) and
between 3.2 and 12.7 mm (0.12 and 0.50 in) in thickness. Samples used in this experiment were 6.35 mm
(0.25 in) thick. Samples were surface ground to a finish of 0.4 μm Ra prior to testing.
The test specimen were cleaned with solvent or cleaner and dried completely before weighing to
the nearest 0.1 mg. The specimens are then seated securely in the holder and the proper weights are added
to the lever arm to develop the appropriate force pressing the specimen against the wheel. The revolution
counter is set to the prescribed number of wheel revolutions. Once sand is flowing through the nozzle and
a proper uniform sand curtain is developed, the wheel rotation is begun, and the lever arm is immediately
lowered to allow the specimen to contact the wheel. After the test is complete, the specimen is removed
and weighed to the nearest 0.1 mg. The dwell time between testing should allow the rubber wheel to
return to room temperature. A minimum dwell time of 30 min was used for the duration of this
experimentation. Similar to the Bond abrasion samples, DSRW samples were sectioned after testing
according to Figure 3.12 to evaluate wear at the sample surface via SEM.
Depending on the severity of the testing desired, ASTM G65 outlines 5 different procedures for
the wear testing of samples. Procedure A was used for the testing performed in this report, with specific
parameters outlined in Table 3.5. This is a relatively severe test, which ranks metallic materials on a wide
volume loss scale from low to extreme abrasion resistance.
Table 3.5 Dry Sand/Rubber Wheel Test Parameters
Force N (lb) Wheel Revolutions Wheel Speed (RPM) Lineal Abrasion (m)
130 (30) 6000 200 ± 10 4309
50
|
Colorado School of Mines
|
CHAPTER 4
RESULTS
4.1 Introduction
This chapter presents the results of microstructural characterization via light optical
microscopy (LOM), scanning electron microscopy (SEM), and energy dispersive X-ray spectroscopy
(EDS); retained austenite assessment via X-ray diffraction (XRD); bulk and micro-hardness
measurements; and heat treatment design using dilatometry on industrially produced and laboratory
prepared materials. Finally, Bond abrasion and dry sand rubber wheel (DSRW) wear testing results on
heat treated laboratory prepared alloys are reported. These results will be discussed in more detail in
Chapter 5.
4.2 Industrial Bar Stock and Grinding Ball Characterization
This section outlines the industrially produced bar stock and forged grinding balls provided by
Gerdau Long Bar Product, North America, with respect to microstructure and mechanical properties. A
high Cr white cast iron (WCI) grinding ball is additionally examined.
4.2.1 Microstructural Characterization
The microstructure of the continuously cast, hot rolled bar stock is displayed in Figure 4.1. The
light optical micrograph in Figure 4.1(a) shows a primarily pearlitic matrix. At higher magnification using
SEM, Figure 4.1(b) reveals pro-eutectoid cementite having formed where the austenite grain boundaries
previously existed. Using a lever rule construction on a simple Fe-C binary phase diagram with a known
composition of 1 wt pct C, it can be estimated that the microstructure consists of 4 wt pct pro-eutectoid
cementite and 96 wt pct pearlite. These values may vary somewhat due to alloying.
Microstructures of the cut, forged, and industrially heat-treated grinding media are displayed in
Figure 4.2. Due to the hyper-eutectoid C level in the steel, when quenched from the austenitic regime
after the forging process, the microstructure of the grinding ball consists of plate martensite (darker
regions in Figure 4.2(a)) and retained austenite (lighter regions). At a lower magnification in LOM,
lighter bands are visible parallel to the rolling direction of the material in Figure 4.3. This is likely due to
manganese (Mn) segregation during casting, leading to Mn rich bands that may influence the chemical
etching response of the material. These Mn-rich bands may also increase austenite stability, leading to
51
|
Colorado School of Mines
|
This interpretation was supported via EDS results shown in Figure 4.5. A higher magnification
micrograph showing secondary carbides in the austenite is shown in Figure 4.6.
MnS γBands
Figure 4.3 Lower magnification light optical micrograph of a forged grinding ball showing plate
martensite, retained austenite, MnS inclusions and additional banding at the surface of the bar stock.
Sample was etched with 2 pct Nital.
(a) (b)
Figure 4.4 Microstructure of a high Cr white cast iron grinding ball using (a) LOM and (b) SEM at the
surface of the grinding ball. Samples were etched with 2 pct Nital.
4.2.2 Hardness
Bulk hardness measurements of the bar stock and both grinding balls were made using a
Rockwell hardness indenter at both the surface and center of all materials. These values are displayed in
Table 4.1. There is a slight increase in hardness from the surface to the center of the forged and WCI
grinding balls, as well as a pronounced increase in hardness from the as-cast bar stock to the heat-treated
grinding ball.
53
|
Colorado School of Mines
|
(a) (b) (c)
(d) (e) (f)
Figure 4.5 EDS maps taken of high Cr WCI grinding ball including (a), (d) SEIs of the area analyzed,
(b), (e) Cr-rich, and (c), (f) Fe- rich areas. The microstructure in (d) was analyzed at a higher
magnification to show smaller secondary carbides (color image – see PDF copy).
4.2.3 X-Ray Diffraction
X-ray diffraction was performed on the forged grinding media to quantify the volume pct of
retained austenite in the microstructure. Measurements were taken at both the surface and the center of
the grinding ball, and diffraction patterns are shown in Figure 4.7 with peak information listed in
Table 4.2. Using the peak integration method described in Section 3.3.4, it was determined that the
surface contained 30.9 vol pct retained austenite, while the center had 31.9 vol pct retained austenite.
4.3 As-Cast Laboratory Heat Characterization
The following section characterizes the as-cast laboratory prepared ingot slabs via LECO® C
analysis to assess carbon segregation and SEM to characterize the NbC distribution in the as-cast
condition.
4.3.1 LECO® Carbon Segregation Analysis
LECO® C analysis was performed on the as-cast laboratory prepared ingots to evaluate whether
significant C segregation occurred during the solidification process. C levels were measured in slabs from
both the top and the bottom of the ingot to account for temperature gradients from the surface to the
54
|
Colorado School of Mines
|
center, as well as from the bottom to the top, of each ingot. The results are presented in Figure 4.8, with
the center of the ingot on the left side of the graph and the surface at the right side. Despite some scatter
in the data, the results suggest a small reduction in the carbon concentration in the central region of the
slabs, due to C segregation during solidification of the ingot.
Figure 4.6 High magnification SEI of the Cr WCI grinding ball showing secondary Cr carbides in an
austenite matrix and cementite along deeply etched austenite grain boundaries. Sample was etched
using 2 pct Nital.
Table 4.1 Hardness Values for Industrial Bar Stock and Grinding Balls
Material Surface Hardness (HRC) Center Hardness (HRC)
Bar Stock 34.1 ± 0.7 33.4 ± 0.9
Forged Grinding Ball 60.8 ± 1.0 61.2 ± 0.9
Cr WCI Grinding Ball 61.4 ± 1.3 63.9 ± 1.6
Figure 4.7 Diffraction patterns obtained using XRD on the surface and center of a forged grinding ball.
Martensite (α) and austenite (γ) peaks are labeled.
55
|
Colorado School of Mines
|
Figure 4.9(a)-(b) appeared to consist almost exclusively of pearlite colonies. Some manganese
sulfide (MnS) inclusions are also present, confirmed using EDS. As Nb concentration increased, eutectic
carbide networks with a bright appearance became more visible along with the pearlite, as seen in
Figure 4.9(c)-(d).
MnS
(a) (b)
MnS NbC
NbC
(c) (d)
Figure 4.9 Light optical micrographs of the (a) 0.01, (b) 0.25, (c) 0.5, and (d) 1.0 wt pct Nb as-cast
slabs. Samples were etched with 2 pct Nital.
To further investigate these eutectic carbide networks, micrographs in Figure 4.10 were taken at
the center of the slab using an environmental SEM (ESEM) with a backscatter detector in sum mode to
provide a large collection angle and, thus, compositional contrast. With Nb having a higher atomic mass
compared to Fe, the signal of the Nb carbides appeared much brighter than the pearlitic matrix. These
micrographs emphasize the evolution of NbC networks in the ingots as Nb is increased from
microalloying levels of 0.01 wt pct, with no Nb signal detected, to 1.0 wt pct Nb. NbC takes on a eutectic
morphology as expected [37 – 43], indicating a coupled growth mechanism with austenite.
57
|
Colorado School of Mines
|
(a) (b)
(c) (d)
Figure 4.10 Scanning electron micrographs from the center of the (a) 0.01, (b) 0.25, (c) 0.5,
and (d) 1.0 wt pct Nb as-cast slabs. Samples were etched with 2 pct Nital.
This eutectic carbide structure can be understood by examining the Fe-Nb pseudo-binary diagram
at a composition of 1 wt pct C shown in Figure 3.5. As the ingot begins to solidify, the austenite (γ) phase
begins to form from the liquid in conjunction with a new phase, designated by FCC A1 #2. This phase is
a result of Nb alloying and is representative of the FCC NbC phase which forms at similar temperatures
as austenite, leading to the eutectic structure observed in Figure 4.10. The eutectic microstructure
continues to develop in the inderdendritic regions upon solidification, making it possible to discern
boundaries between NbC-containing networks.
Energy dispersive X-ray spectroscopy (EDS) point scans and maps of these samples confirmed
that the brighter microstructural constituents were rich in Nb and are assumed to be NbC based on
interpretation of the Fe-Nb pseudo-binary diagram in Figure 3.5. The EDS results are presented in
Figure 4.11. As suggested by the EDS map in Figure 4.11(b), almost all the Nb in each sample resides in
58
|
Colorado School of Mines
|
these carbides, with an undetectable amount remaining in the matrix due to low solubility of NbC in the
iron-rich phases.
(a) (b)
(c)
Figure 4.11 EDS scans showing the presence of NbC in the as-cast pearlite matrix. (a) includes a
secondary electron image using the field emission SEM (FESEM) of a carbide network, (b) shows an
EDX map with Nb signal in red, and (c) is an EDS point scan from one of the eutectic carbides (color
image – see PDF copy).
The area fraction of the eutectic constituent containing the carbide networks was evaluated by
tracing these networks and inputting simplified “micrographs” into ImageJ global thresholding software.
Area fraction was inferred to represent eutectic volume fraction. Ten micrographs were analyzed for each
alloy with an example from each chemistry shown in Figure 4.12. Figure 4.13 plots NbC-containing
eutectic constituent area coverage as a function of Nb content. As expected, as the Nb content is
increased, the area fraction of NbC-containing eutectic constituent increases.
59
|
Colorado School of Mines
|
Figure 4.13 Area fraction of NbC-containing eutectic constituent as a function of Nb (wt pct) in the as-
cast laboratory prepared material.
One final microstructural aspect of interest observed in the as-cast ingots is displayed in
Figure 4.14. The backscatter ESEM images were taken at both the center and the outer corner of each slab
to characterize differences in solidification behavior of the ingot within the mold. A finer microstructure
with smaller dendrites and smaller/finer interdendritic eutectic regions is evident at the outer corner in
comparison to the center, due to the greater cooling rate (faster solidification rate) near the surface.
If solidification temperature is increased, increased lamellar growth kinetics are expected to lead
to an increase in the spacing between NbC lamellae. At the surface of the ingot, the solidification rate is at
its highest with a relatively low temperature gradient, leading to columnar dendritic solidification of the
material. Nb partitions into the spaces between primary and secondary dendrite arms, leading to the
microstructures seen in Figure 4.14(a), (c), and (e). In the center of the ingot, however, the solidification
rate is decreased and a cellular grain structure is observed after solidification. As seen in
Figure 4.14(b), (d) and (f), the NbC-containing eutectic constituent again exists in the areas between
previously-formed cellular grains.
This difference in solidification behavior from the surface to the center of the ingot may lead to a
potential gradient of mechanical properties, such as hardness and toughness, from the surface to the center
of the ingot. Hardness measurements at both the surface and center of slabs from each chemistry were
made and reported in Table 4.3. These hardness measurements had considerable scatter; however, an
increase in hardness at the surface of the slab was observed for each alloy potentially due to the more
refined microstructure.
61
|
Colorado School of Mines
|
Table 4.3 As-Cast Laboratory Prepared Ingot Hardness Data
Nb (wt pct) Corner Hardness (HRC) Center Hardness (HRC)
0.01 34.6 ± 2.6 28.3 ± 0.7
0.25 37.0 ± 3.8 25.0 ± 2.2
0.5 37.4 ± 1.9 28.6 ± 4.2
1.0 35.8 ± 3.7 23.2 ± 2.3
4.4 Hot Rolled Laboratory Heat Characterization
This section compares the as-cast laboratory prepared ingot microstructure and carbide
distribution to the microstructure of the same ingots/plates after hot rolling. Reheating was conducted at
1085 ºC, with a reduction of the ingots from 12.7 cm to 0.85 cm in 7 passes. Figure 4.15 displays
backscatter SEM micrographs of the hot rolled plates to highlight the distribution of NbC. Again, the
microstructure is composed of NbC-containing eutectic surrounded by a pearlitic “matrix.” The carbide
networks were elongated in the rolling direction, with the carbides experiencing some deformation and
fragmentation. Dark elongated features were determined to be MnS using EDS.
A discernable difference noted between the as-cast structure in Figure 4.10 and the hot rolled
plates in Figure 4.15 is the distribution of carbides within the microstructure. In the as-cast condition,
NbC forms in eutectic networks between primary eutectic grains. For the hot rolled microstructure, these
networks were broken up and appear somewhat aligned in the rolling direction. Hot rolling was
performed at a temperature below that of the dissolution or melting temperature of the
carbides (~1500 ºC), ensuring that the carbides did not melt or dissolve, but rather the structure of the
eutectic networks was broken up. There was probably very limited plastic deformation experienced by the
(hard) carbides, but the hot rolling process produced a more distributed array of carbides without causing
major defects or complete fracture of the material. Some smaller voids were visible directly next to some
carbides reflecting strain incompatibility due to the variation in constituent flow stresses. A higher
magnification micrograph of an NbC network post hot-rolling is shown in Figure 4.16 with a small void
present directly to the left of the large NbC particle.
A similar area fraction analysis as shown in Figure 4.12 was performed on the hot rolled plates,
with results of eutectic area fraction as a function of Nb content included in Figure 4.17. A similar trend
exists as with the as-cast slab carbide volume fraction analysis in Figure 4.13, with increased eutectic
fraction at higher Nb concentrations. However, rather than existing in discrete networks, there is a
somewhat more uniform distribution of carbides after rolling though they still exist in bands within the
microstructure.
63
|
Colorado School of Mines
|
Figure 4.17 Area fraction of NbC-containing constituent networks as a function of Nb (wt pct) in the
hot-rolled laboratory prepared material.
4.5 Industrial Bar Stock Dilatometry
This section outlines the results of a dilatometry study performed on samples machined from
industrially produced bar stock to evaluate phase transformation kinetics and identify process parameters
to duplicate the hardness of current grinding balls. Properties were evaluated via LOM, hardness testing,
and XRD to determine retained austenite volume fraction. The main variables evaluated in this study
included quench rate, austenitization time, and austenitization temperature.
The effect of quench rate on the properties of industrially produced bar stock was evaluated by
heating samples in the dilatometer past their A temperature of approximately 800 ºC, as confirmed by
cm
interpretation of the dilation curves in Figure 4.18, to 850 °C, holding isothermally for 300 s, and
quenching the samples at different rates ranging from -5 ºC·s-1 to -105 ºC·s-1. Samples were quenched
below their M temperature to 150 ºC and allowed to air cool to room temperature. Samples were
s
quenched to this intermediate temperature rather than directly to room temperature in order to mitigate
any potential quench cracking. There tended to be a continued deviation from linearity past the transition
from contraction to expansion upon heating observed in Figure 4.18, best seen in the differential change
in length plots. This is interpreted as being continued dissolution of cementite into austenite after all
pearlite has transformed. Once all cementite has dissolved, the slope is again linear and the microstructure
is assumed to be fully austenitic. Each sample was evaluated for hardness, critical transformation
temperatures (A , A , M), microstructure, and retained austenite content. Table 4.4 presents hardness
r1 cm s
values and retained austenite levels for fully martensitic samples as a function of quench rate.
Additionally, a continuous cooling transformation (CCT) diagram was constructed and displayed in
Figure 4.19.
65
|
Colorado School of Mines
|
(a) (b) (c)
Figure 4.18 Sample dilation as a function of temperature for industrial bar stock samples used to
determine average A temperature for three different samples. Differential change in length data is
cm
included to highlight changes in slope, corresponding to phase changes. Location of determined A is
cm
indicated by dashed lines.
Table 4.4 Bar Stock Dilatometry Results
Quench Rate (ºC·s-1) Hardness (HRC) Retained Austenite (vol pct)
-5 39.7 ± 0.8 N/A
-15 53.6 ± 5.1 N/A
-20 62.4 ± 0.6 22.4
-30 63.0 ± 0.3 25.1
-70 62.1 ± 0.2 26.1
-105 62.9 ± 0.5 23.7
The proeutectoid cementite + pearlite (C+P) bay was characterized both by analyzing dilatometry
curves for deviations in linearity during cooling, as well as visual confirmation via LOM. Micrographs
from each heat-treated sample are included in Figure 4.20. A lower magnification micrograph of the
sample quenched at -20 °C·s-1 is additionally included in Figure 4.21 to better show banding within the
microstructure. As seen by the hardness data in Table 4.4, those samples which contained some degree of
C+P correlated to lower hardness values compared to those that were fully martensitic. Pearlite is a
relatively soft microstructural constituent, so its presence is detrimental to the wear properties of steel
grinding media. For that reason, a quench rate greater than -30 °C·s-1 would be desired for this steel.
66
|
Colorado School of Mines
|
Figure 4.19 Continuous cooling transformation (CCT) diagram developed by dilatometry on
industrially produced bar stock with a 300 s austenitization at 850 °C, and varying quench rates. C+P
designates the proeutectoid cementite + pearlite bay and M designates the martensite start (M)
s
temperature.
The micrographs in Figure 4.20 can aid in understanding the effect of lower quench rates on the
microstructure. The slowest cooling rate represented by Figure 4.20(a) shows a fully pearlitic
microstructure, intermediate quench rates in Figure 4.20(b)-(c) show regions of martensite that formed
within the pearlite matrix, and fast quench rates in Figure 4.20(d)-(f) result in a fully martensitic
microstructure. Figure 4.21shows a lower magnification image of the microstructure following an
intermediate quench rate to highlight the “banding” of pearlite and martensite/austenite that was observed
in most samples. As previously discussed in Section 4.2.1, banding is commonly seen in steels with high
Mn. Mn, like C, tends to segregate to the interdendritic regions during solidification due to its low
solubility in austenite, leaving regions within the solidified product that are Mn rich, and others that are
Mn deficient. These regions are elongated during the hot rolling process of the bar Mn additionally is
known to be a strong austenite stabilizing element, locally increasing the hardenability. For that reason,
bands rich in Mn will more likely form martensite upon quenching, whereas the Mn poor regions are
more likely to transform to pearlite, as shown in Figure 4.21.
Mn segregation during casting is difficult to avoid due to the sluggish diffusion of Mn in
austenite. Therefore, it is unlikely that Mn diffusion during the austenitization step will be sufficient to
homogenize the Mn, and in turn the hardenability. W. Smith states that the diffusivity of Mn in FCC Fe at
1300 °C is 1.5 x 10-14 m2∙s-1 and at 400 °C is 1.5 x 10-15 m2∙s-1 [69]. Using the following diffusion distance
equation,
x = √Dt (4.1)
67
|
Colorado School of Mines
|
where x is the diffusion distance, D is the diffusivity of Mn in Fe, and t is the time in seconds, this gives a
range of 2.1 to 6.7 µm for the 300 s austenitization step. As these Mn bands tend to be a few hundred
microns wide, this diffusion distance is not sufficient to homogenize Mn within the microstructure.
-5 ºC·s-1 -15 ºC·s-1 -20 ºC·s-1
(a) (b) (c)
-30 ºC·s-1 -70 ºC·s-1 -105 ºC·s-1
(d) (e) (f)
Figure 4.20 LOM micrographs corresponding to dilatometry tests run on industrially produced bar stock
with an austenitization temperature of 850 ºC and hold time of 300 s. Each micrograph includes cooling
rate (in ºC·s-1). Samples were etched using 2 pct Nital.
C has a higher diffusivity in γ-Fe, reported as being 3.4 x 10-10 m2∙s-1 at 1300 °C and
1.25 x 10-16 m2∙s-1 at 400 °C [70]. This predicts diffusion distances ranging from 0.2 to 320 µm for a 300 s
austenitization heat treatment. Therefore, it is expected that C can homogenize on a local scale during the
austenitization and then redistribute during diffusional transformation.
The effect of quench rate on hardness and retained austenite content is plotted in Figure 4.22.
Essentially constant hardness values greater than 60 HRC can be attributed to a fully martensite plus
retained austenite matrix. Given that hardness values for quench rates greater than or equal to -20 °C·s-1
68
|
Colorado School of Mines
|
825 °C, held isothermally for 10 min (600 s), quenched to 60 °C, and slowly cooled to room temperature.
The quench rate was varied from -5 °C·s-1 to -75 °C·s-1. Separate CCT diagrams were constructed for each
chemistry and these are displayed in Figure 4.23. M, A , and A temperatures were determined by
s r1 cm
examining dilation curves as a function of temperature for a deviation in linearity, as outlined in
Figure 3.7. The most notable difference between CCT curves for different alloys is the position of the
proeutectoid cementite + pearlite bay, as well as the M temperature. M temperature was not affected
s s
substantially by quench rate and thus was averaged over all quench rates for each alloy. The average
M temperatures are presented in Table 4.5 with a slight increase in M temperature observed with an
s s
increase in Nb content.
(a) (b)
(c) (d)
Figure 4.23 Continuous cooling transformation (CCT) diagrams developed by dilatometry on
laboratory cast and hot rolled ingots containing (a) 0.01, (b) 0.25, (c) 0.5, and (d) 1.0 wt pct Nb. C + P
designates proeutectoid cementite + pearlite, and M designates martensite. Each sample was heated to
825 °C, held for 600 s, quenched to 60 °C, and slowly cooled to room temperature.
70
|
Colorado School of Mines
|
Table 4.5 Measured M Temperatures
s
Nb (wt pct) M (°C)
s
0.01 187 ± 10
0.25 198 ± 7
0.5 203 ± 3
1.0 216 ± 12
Each hot rolled plate dilatometry sample was additionally tested for hardness, with hardness as a
function of quench rate for each alloy displayed in Figure 4.24. Similar to the bar stock hardness results in
Figure 4.22, hardness was approximately constant at the more rapid quench rates above 15 °C·s-1 and
began to drop off at slower quench rates. This decrease in hardness is again attributed to the formation of
pearlite at slower cooling rates. The samples with 0.5 and 1.0 wt pct Nb achieved a higher hardness
plateau over a larger range of quench rates, which might be attributed to the shift in the C+P bay to longer
times with increasing Nb and a decrease in pearlite within the microstructure. An increase in Nb content
may also lead to a Zener pinning effect due to remaining solute Nb in the austenite. Zener pinning would
lead to a reduction in austenite grain size as the Nb atoms inhibit grain growth, and a resulting increase in
nucleation sites for pearlite to form. Grain refinement may result in an increase in the materials M
s
temperature and an increase in the volume fraction of martensite that forms in the microstructure.
However, due to the low solubility of Nb in austenite, there is likely minimal variation between alloys.
Additionally, as more Nb is added, solute C is consumed by the formation of NbC. If less C interstitial
sites are filled in the martensite, there will be less strain on the lattice, and the strength of the martensite
will decrease.
4.7 Salt Pot Heat Treatment
A series of heat treatments was performed on samples machined from the hot rolled plates having
dimensions of 7.62 x 2.54 x 0.85 cm (3 x 1 x 1/3 in) to match the sample dimensions for Bond abrasion
testing. The goal of this testing was to scale up the dilatometry heat treatments to larger samples and
compare the resulting hardness and microstructure to industrial grinding balls. These heat treating
parameters are outlined in Table 4.6 with austenitization temperatures and hold times, as well as
quenching parameters.
The first set of heat treatments (HT1) consisted of an austenitization at 850 °C for 600 s (10 min).
Both austenitization time and temperature were increased slightly from the values used in the dilatometry
study to account for the larger sample size and possible temperature gradients. One set of samples was
quenched to room temperature in water (WQ) and a second set were quenched directly into a salt pot set
to 150 °C, held for 60 s (1 min), and allowed to air cool to room temperature (150Q). This intermediate
71
|
Colorado School of Mines
|
quench temperature was explored due to the possibility of quench cracking when water quenching a high
carbon steel, as well as in an effort to increase the volume fraction of retained austenite, as it is possible
for C to partition to the austenite as the sample is held at a higher intermediate temperature and air cooled
to room temperature, stabilizing the untransformed austenite. Using Equation 4.1, this would predict a
diffusivity of C in austenite of 2.8 x 10-23 m2∙s-1 and a diffusion distance of 0.13 nm when held at 150 °C
for 20 min. Each sample was evaluated for hardness, retained austenite content, and microstructure, with
hardness and retained austenite results presented in Table 4.7 and plotted in Figure 4.25. Microstructures
are displayed in Figure 4.26 and Figure 4.27 for the WQ and 150Q samples, respectively. All data and
micrographs presented are taken from the near-surface of the samples.
Figure 4.24 Relationship between hardness and quench rate for hot-rolled dilatometry samples
compared for alloys with 0.01, 0.25, 0.5, and 1.0 wt pct Nb. Samples were heated to 825 °C, held for
600 s, quenched to 60 °C, and slowly cooled to room temperature.
Table 4.6 Salt Pot Heat Treatment Parameters
Heat Treatment Austenitization Austenitization Quench Hold
Quench Medium
ID Temperature (°C) Time (s) Time (s)
HT1-WQ 850 600 Water N/A
HT1-150Q 850 600 150 °C Salt Pot 60
HT2-WQ 875 1200 Water N/A
HT2-150Q 875 1200 150 °C Salt Pot 600
It is observed in Figure 4.26 that the matrix of the water quenched samples consists primarily of
retained austenite and martensite with “bright” NbC precipitates in elongated bands parallel to the rolling
direction. Additionally, smaller spherical precipitates are visible within the microstructure, especially in
the 0.01 wt pct Nb sample, where NbC is not expected. Rather, these particles are interpreted to be
spheroidized cementite that did not fully dissolve during the austenitization step.
72
|
Colorado School of Mines
|
Table 4.7 Measured Values of Hardness and Retained Austenite Volume Fraction
from HT1
Hardness (HRC) Retained Austenite (vol pct)
Nb (wt pct)
HT1-WQ HT1-150Q HT1-WQ HT1-150Q
0.01 62.5 ± 1.5 61.8 ± 1.1 23.5 22.1
0.25 62.8 ± 4.3 59.7 ± 5.0 20.8 21.0
0.5 65.7 ± 1.0 62.8 ± 1.7 22.6 22.5
1.0 64.2 ± 1.3 64.0 ± 5.0 19.4 19.4
(a) (b)
Figure 4.25 Relationship between (a) hardness and (b) retained austenite content as a function of Nb
content for the samples heat treated according to HT1-WQ and HT1-150Q.
In order to confirm whether or not the smaller spherical particles observed in Figure 4.26
and Figure 4.27 were NbC, EDS was performed on areas containing larger NbC precipitates surrounded
by smaller unknown particles. EDS results are displayed in Figure 4.28. Figure 4.28(a) shows an SEI
from the 0.01 wt pct Nb alloy with no Nb rich areas detected in Figure 4.28(c). These micrographs make
it unlikely that these small particles are NbC.
Figure 4.28(d) displays an SEI taken from the 1.0 wt pct Nb sample containing NbC, as
confirmed by the Fe-deficient and Nb-rich areas in Figure 4.28(e) and (f), respectively. Smaller particles
between these large carbides do not have a noticeable Nb signal, again suggesting a different constituent,
such as spheroidized cementite. However, without quantitative Fe and C analysis, this cannot be
confirmed. Undissolved cementite remaining in the microstructure is undesired as it leads to a decreased
amount of C in solution to aid in hardenability, as the C remains tied up in forming Fe carbide.
Additionally, this refinement will weaken the martensite in the microstructure as less interstitial sites are
73
|
Colorado School of Mines
|
(a) (b)
(c) (d)
Figure 4.27 Micrographs taken from the surface of the salt pot heat treated samples following the HT1-
150Q profile for the (a) 0.01, (b) 0.25, (c) 0.5, and (d) 1.0 wt pct Nb alloys. ‘C’ denotes the unknown
deeply etched constituent. Samples were etched using 2 pct Nital.
Figure 4.27 shows a similar microstructure composed of retained austenite, martensite, NbC, and
smaller particles assumed to be cementite. However, another deeply etched constituent is also present
(denoted by ‘C’ in Figure 4.27) which can be attributed to a slower cooling rate compared to the samples
quenched in water. This constituent is likely either pearlite or bainite; however, this cannot be confirmed
without the use of transmission electron microscopy (TEM). It is presumed, however, that this constituent
is softer and likely contributes to the hardness decrease seen in the HT1-150Q samples.
A second set of heat treatment profiles was run according to HT2-WQ and HT-150Q in Table 4.6
in an attempt to dissolve the remaining cementite and avoid the deep-etching constituent present in
Figure 4.27. Both austenitization time and temperature were increased to 1200 s (20 min) and 875 °C,
respectively. The holding time at 150 °C was additionally increased to 20 min to allow the entire sample
75
|
Colorado School of Mines
|
remains. The spherical particles in the microstructures of the 0.01 wt pct Nb alloys, shown in
Figure 4.30(a) and 4.31(a), are likely predominately cementite that still exists in bands across the
microstructure.
Figure 4.29 Relationship between hardness and Nb content for the samples heat treated according to
HT2-WQ and HT2-150Q.
Due to the promising results obtained from the HT2 profile, this heat treatment was further
validated by inserting a thermocouple into the center of a 1.0 wt pct Nb sample and collecting temperature
data during both the HT2-WQ and HT2-150Q treatments. The temperature is plotted as a function of time
in Figure 4.32.
These thermal histories reveal that both samples were heated above their A temperature to
cm
875 °C and quenched below their M temperature. The heating rates for the WQ and 150Q samples were
s
13.4 °C·s-1 and 20.7 °C·s-1, respectively. The water quench rate was 41.7 °C·s-1, and the quench rate into
the 150 °C salt pot was 15.4 °C·s-1. The sample air cooled to room temperature at 0.07 °C·s-1. It is also
important to keep in mind that these measurements were taken from the center of each sample, and the
surface is likely to experience more rapid heating and cooling.
Finally, the HT2-WQ samples were examined for signs of quench cracking due to the more
aggressive quench rate. The sample surfaces were cleaned of any remaining salt residue and lightly
ground with 1200 grit SiC paper to remove any scale. A light optical stereoscope was used to examine the
surfaces at increased magnification. Images of each heat treated alloy are displayed in Figure 4.33.
Evidence of surface cracking is present in all samples and several specimens fractured during preparation
for Bond abrasion testing. The surfaces of the HT2-150Q samples were smooth when the same
investigation was performed. For that reason, samples were heat treated according to HT2-150Q for all
wear testing.
77
|
Colorado School of Mines
|
(a) (b)
(c) (d)
Figure 4.31 Micrographs taken from the surface of the salt pot heat treated samples following the
HT2-150Q profile for the (a) 0.01, (b) 0.25, (c) 0.5, and (d) 1.0 wt pct Nb alloys. Samples were etched
using 1 pct Nital.
4.8.1 Bond Abrasion Wear Testing
Samples tested for Bond abrasion had microstructures corresponding to those presented in
Figure 4.31. The first set of abrasion tests were run with a Fe ore, which has an abrasion index reported in
literature to be approximately 0.077 [31]. The second set of abrasion tests were run using a Cu ore. Cu ore
is reported in literature as having an abrasion index of approximately 0.095 [31], and was used here with a
goal of being a more abrasive feed compared to the Fe ore to validate or compare any trends seen in the
first set of wear data.
A preliminary Bond abrasion test using Fe ore was performed on a set of samples that did not
have an automated surface grinding procedure applied, but rather were hand ground using 1200 grit SiC
paper to remove any scale. Wear results, as well as sample hardness values, are presented in Figure 4.34.
79
|
Colorado School of Mines
|
Table 4.9 includes all hardness and mass loss data. A higher mass loss indicates a lower resistance to
wear.
(a) (b)
Figure 4.32 Time-temperature profiles obtained with a thermocouple inserted into the center of a 1.0 wt
pct Nb sample during HT2-WQ and HT2-150Q profiles.
The results in Figure 4.34(a) show a sharp decline in mass loss between the 0.01 and 0.25 wt pct
Nb alloys, followed by a relatively constant wear rate for the 0.5 and 1.0 wt pct Nb alloys. Figure 4.34(b)
is included to highlight that wear rate was reasonably constant over the entire duration of the 60 min test.
Figure 4.34(c) shows that for this set of samples, hardness increased with increasing Nb, correlating to the
mass loss data as a function of Nb content.
The next set of Bond abrasion tests were run on samples that were surface ground at a machine
shop using a reciprocating surface grinding machine to a consistent finish of 0.4 µm (16 µin) Ra.
Approximately 4 samples were tested from each alloy on both sides of the paddle for a total of 8 tests per
alloy. Each sample was tested for retained austenite content and hardness prior to wear testing; these
measurements are included with the mass loss results in Table 4.10. Mass loss results as a function of Nb
alloying are presented in Figure 4.35. Additionally, hardness and retained austenite content are plotted
versus Nb content or mass loss in Figure 4.36. In contrast to Figure 4.34, these results do not indicate a
significant relationship within error between Nb concentration and Bond abrasive wear resistance, though
hardness increases with increasing Nb. Retained austenite content shows no influence of Nb content.
Further discussion on the discrepancy observed in the first set of data is included in Chapter 5.
Macroscopic imaging of samples following wear testing with Fe ore used a light optical
stereoscope, while microscopic imaging employed an environmental SEM. These images are displayed in
Figure 4.37- 4.39, with Figure 4.37 indicating regions that will be referred to as the “paddle face” and as
80
|
Colorado School of Mines
|
the “paddle end”. These figures suggest that a majority of the wear occurred on the paddle end, causing
the edges and corners to become rounded. Wear was also observed on the paddle faces of the samples.
There were no major cracks or fracture initiation sites visible, nor were there obvious difference in the
wear patterns between compositions. Overall, there were no obvious differences in overall appearance
that would suggest differences in wear mechanism as a function of Nb content.
(a) (b)
(c) (d)
Figure 4.33 Light optical micrographs of the surfaces of HT2-WQ samples with (a) 0.01, (b) 0.25, (c)
0.5, and (d) 1.0 wt pct Nb. Quench cracks are visible on all samples.
The micrographs in Figure 4.39 show linear scratch marks forming as a result of Fe ore moving
across the samples surface during testing. Figure 4.39(b) shows micro-fragmentation within one of the
scratch tracks. There are also bright protuberances that may be caused by small pieces of Fe ore becoming
embedded in the steel paddle upon impact, as well as ridges forming as a result of a plowing abrasion
mechanism. It is again not easy to discern different wear mechanisms that may be active as a function of
Nb alloying; rather, each sample demonstrates a similar, complex wear response to Bond abrasion testing.
81
|
Colorado School of Mines
|
(a)
(b)
Figure 4.37 Macroscopic images taken of (a) one worn paddle face and (b) the worn paddle end after
Bond abrasion testing both side faces with Fe ore.
One sample from each alloy was sectioned through its 2.54 x 0.635 cm cross section and mounted
such that the worn edge could be examined through the thickness of the sample. SEM micrographs of the
cross section of the worn face are included in Figure 4.40.
Figure 4.40 demonstrates that a highly deformed layer formed at the surface of the samples due to
impact by Fe ore. Depending on whether the sample was impacted at a certain area as well as the degree
of the impact, the deformed layer thickness ranged from negligible to a few tens of micron. These impact
layers were relatively consistent in thickness across all alloys and the presence of NbC in this layer did
not seem to affect the response of the material to impact abrasion. This deformed layer induces increased
hardness at the surface of these samples, observed by the hardness measurements in Table 4.11 compared
to those in Table 4.10.
84
|
Colorado School of Mines
|
(a)
(b)
(c)
(d)
Figure 4.38 Montage of light optical micrographs taken of the paddle end of samples containing (a)
0.01, (b) 0.25, (c) 0.5, and (d) 1.0 wt pct Nb after Bond abrasion testing on both faces with Fe ore.
The final set of Bond abrasion tests was performed using a Cu ore that was expected to lead to a
higher wear rate compared to the Fe ore and confirm the data trend in either Figure 4.34(a) or Figure 4.35.
A single surface ground sample for each alloy was tested with the Cu ore on both sides for a total of 2
data points due to a limitation on machined, heat treated, and surface ground material. Wear data are
presented in Figure 4.41. Hardness is plotted versus both Nb content and mass loss in Figure 4.42. Table
4.12 summarizes the hardness and mass loss results.
The total mass loss measured from the samples tested with the Cu ore was less than that of the
samples tested with the Fe ore; however, the trend was similar to Figure 4.35 in that wear rate was
approximately constant, or even increased, with increasing Nb. There was no trend observed between
85
|
Colorado School of Mines
|
sample hardness and wear rate. The overall conclusion from the data obtained from Bond abrasion testing
is that increased Nb alloying did not substantially affect the Bond abrasion resistance.
(a) (b)
(c) (d)
Figure 4.39 Scanning electron micrographs taken of the worn surfaces of samples containing (a) 0.01,
(b) 0.25, (c) 0.5, and (d) 1.0 wt pct Nb after Bond abrasion testing on both faces with Fe ore.
4.8.2 Dry Sand/Rubber Wheel Wear Testing
Dry sand/rubber wheel (DSRW) testing was also explored. This procedure is typically used to
evaluate the scratching-abrasion properties of metallic materials. Two samples of each composition were
tested on each face for a total of 4 tests per alloy. All samples were measured for hardness prior to wear
testing with hardness values outlined in Table 4.13. Retained austenite content was not measured prior to
testing as it was assumed to match the values for the Bond abrasion samples since the heat treatment,
sample size, and surface preparation were identical. Samples were again evaluated for wear resistance in
terms of mass loss. Mass loss data are presented in Figure 4.43. Sample hardness is plotted vs. both Nb
86
|
Colorado School of Mines
|
sectioned through its 2.54 x 0.635 cm cross section and mounted such that the worn edge could be
examined through the thickness of the sample. SEM micrographs of the cross section of the worn face are
included in Figure 4.46.
Table 4.13 DSRW Sample Properties and Results
Nb (wt pct) Hardness (HRC) Mass Loss (g) Volume Loss (mm3)
0.01 63.2 ± 0.3 0.3843 ± 0.07 50.1 ± 9.6
0.25 64.6 ± 0.1 0.2429 ± 0.06 31.6 ± 8.2
0.5 65.0 ± 0.3 0.1771 ± 0.04 23.1 ± 4.5
1.0 65.8 ± 0.1 0.1363 ± 0.06 17.8 ± 7.6
Figure 4.46 reveals that little to no deformed layer formed at the surface of the DSRW samples,
in contrast to the cross-sectional micrographs from the Bond abrasion test in Figure 4.40. Furthermore, it
was commonly seen that the worn surface traced along NbC particles which appeared to hinder the
scratching abrasion mechanism - a good example of which is shown in Figure 4.46(d). Due to the absence
of impact at the surface of the DSRW sample surfaces, the deformed layer was negligible as deformation
is not as substantial in scratching abrasion environments, though it possibly still occurred to a lesser
degree. Overall, it can be concluded that increased Nb alloying is favorable in terms of
scratching-abrasion resistance as assessed by DSRW testing.
Figure 4.43 Dry sand/rubber wheel wear test results.
89
|
Colorado School of Mines
|
CHAPTER 5
DISCUSSION
5.1 Introduction
This chapter provides additional discussion of results presented in Chapter 4. Topics include a
comparison of measured grinding ball properties to those discussed in literature, microstructural
behaviors observed in both industrially produced and lab prepared material, as well as an evaluation of
wear results.
5.2 Grinding Ball Mechanical Properties
Grinding ball properties including hardness and retained austenite volume fraction are compared
to those observed in literature. Trends seen in the collected data are discussed.
5.2.1 Hardness
Hardness data are reported for various grinding media throughout literature. To verify the
properties measured in this study and for use as a baseline for experimental material, hardness values
were compared to those reported by Hawk et al. [20], Dodd et al. [14], and Rajagopal and Iwasaki [13]
for the bar stock, forged grinding ball, and Cr WCI grinding ball, respectively, in Table 5.1. All values
correlated well to ranges provided in literature, apart from the hardness of the pearlitic bar stock being
somewhat higher than was reported for a pearlitic microstructure. It is possible that the higher C content
of the bar stock, as well as refined pearlite spacing, increased the hardness beyond what is expected for
as-rolled bar stock. Hardness ranges reported for both the forged and Cr WCI grinding balls were similar.
Table 5.1 Hardness Measurements and Literature Values
Measured Hardness Reported Hardness
Material Reference
(HRC) Range (HRC)
Bar Stock 34.1 ± 0.7 20 - 28 [20]
Forged Grinding Ball 60.8 ± 1.0 60 - 65 [14]
Cr WCI Grinding Ball 61.4 ± 1.3 61.5 - 65 [13]
Both the forged and WCI grinding ball showed an increase in hardness from the surface to the
center of the ball. This is attributed to C segregation during solidification and C enrichment at the center,
increasing the hardness of the martensite.
92
|
Colorado School of Mines
|
The rapid, diffusionless microstructural transformation of FCC austenite to BCT martensite in the
forged grinding media nearly doubled the hardness values observed in the untreated bar stock. Having a
relatively high C content, the grinding balls experience a large lattice strain upon the formation of
martensite, leading to a high dislocation density and superior hardness due to an increased number of
C interstitial sites being filled. Krauss provided Figure 5.1 showing hardness measured as a function of C
for low alloy steels by a number of investigators [72]. The variation in hardness data from different
studies can be attributed to differences in microstructure such as austenite grain size, retained austenite
content, or other carbide phases present. Using the C content of 0.98 wt pct for the grinding balls
investigated in this study provided in Table 3.1, it is expected that the forged grinding balls after full
austenitizing and rapid quenching would have a hardness ranging from approximately 845 – 920 HV
(65.5 - 67.5 HRC). The measured surface hardness value of 60.8 ± 1.0 HRC falls below this range, which
may be due to the steel’s considerable retained austenite content. Krauss reported that some investigators
quenched their specimen in liquid nitrogen to reduce retained austenite, especially for higher C steels,
making it unlikely any of the microstructures studied contained a comparable retained austenite content.
Mola and Ren provide a linear approximation of austenite hardness in steel as a function of C given
by [73]:
HV = 140 + (48.6 ∙ C) (5.1)
γ
where C is the C content in wt pct. This approximates the hardness of the austenite in the grinding balls to
be 188 HV. If it is assumed that the microstructure contains 30 vol pct retained austenite as reported in
Table 4.2, this would predict a bulk hardness range of about 650 – 700 HV (56 – 60 HRC) based on the
following rule of mixtures calculation [74]:
HV = 0.3 ∙ HV + 0.7 ∙ HV (5.2)
bulk γ α'
where HV , HV, and HV are the hardness values of the bulk material, austenite, and martensite,
bulk γ α’
respectively. The slightly higher measured hardness value may be attributed to various responses to
alloying, as only C is considered in Equation 5.1 and Figure 5.1, and/or differences in heat treatment
processing.
The Cr carbides present in the microstructure of the WCI grinding ball increase the hardness to a
value slightly greater than that of the forged grinding ball. Cr carbides are reported as ranging in hardness
from about 1000 - 1750 HV [36]. Rajagopal and Iwasaki reported that WCI grinding media with
11.25 wt pct Cr and 2.8 wt pct C had a hardness value of 61.5 HRC [13]. As this is consistent with the
93
|
Colorado School of Mines
|
WCI grinding ball examined in this study, and the composition of the ball is not known precisely, it is
perhaps similar in composition to the sample examined by Rajagopal and Iwasaki.
Weight Percent Carbon
Figure 5.1 Hardness of martensitic microstructures as a function of alloy C content [72].
5.2.2 X-Ray Diffraction
Similar to the hardness data presented in Table 4.1, an increase in retained austenite from the
surface to the center of the forged grinding ball was observed in Table 4.2. This may again be related to
solute segregation such as C or Mn at the center of the grinding ball during solidification. As C content is
increased, the M temperature decreases, leading to a decrease in the amount of martensite and an
s
increased vol pct of retained austenite. Mn is a strong austenite stabilizing element, so an increased
amount of Mn at the center of the ball additionally stabilizes austenite, as observed in Table 4.2. This
effect, in combination with Mn segregation/banding after solidification, may contribute to the austenite
banding observed in Figure 4.3.
94
)VH(
ssendraH
)CRH(
ssendraH
|
Colorado School of Mines
|
Using the M temperature of 180 ºC predicted for the bar stock by Andrews [62] in Table 3.3, as
s
well as Equation 3.8 by Koistinen and Marburger [66] to model austenite phase fraction as a function of
quench temperature, it is expected that quenching the grinding balls to room temperature would result in
approximately 19.2 vol pct retained austenite. The higher austenite content of 31 vol pct measured
experimentally may be attributed to a heat treatment process in which the sample was quenched to an
intermediate elevated temperature below its M temperature, allowing some C to partition from the
s
martensite to the untransformed austenite, increasing its stability before reaching room temperature. The
measured M temperature determined using dilatometry was closer to 200 ºC possibly due to alloying not
s
included in Equation 3.6, which would further decrease the calculated retained austenite content to
15.4 vol pct if quenched to room temperature.
XRD diffraction patterns often contain broad peaks with substantial noise due to the high C
content and dislocation density. An increase in peak broadness may skew integration values if broadening
is not perfectly consistent among all peaks. It is likely the high signal-to-noise ratio leads to each
measured retained austenite value having a deviation of 5 - 10 vol pct. Additional retained austenite
analysis could employ electron backscatter diffraction (EBSD), but EBSD was beyond the planned scope
of this project.
5.3 Nb Carbide Solubility Calculations
To understand the precipitation of niobium carbide (NbC) in austenite, the solubility plot in
Figure 5.2 was prepared using Equations 2.5 - 2.7 [33]. Alloy compositions of 0.01, 0.25, 0.5, and
1.0 wt pct Nb and 1.0 wt pct C are highlighted and an equilibrium isotherm was constructed at 1085 °C
representing the temperature at which the ingots were hot rolled. All four chemical compositions reside
above the isotherm, indicating that the reaction of Nb + C → NbC will proceed and NbC will exist in
equilibrium with austenite. The slope of the dashed lines represents the stoichiometric Nb:C ratio,
determined using the atomic mass of each element, and these lines are included to show the extent to
which each element participates in forming carbides.
Figure 5.2 can be useful in predicting the amount of soluble C that is removed from the material
due to the formation of NbC. At compositions above the equilibrium solubility isotherm, NbC is insoluble
in austenite and Nb and C remain tied up in NbC. Below this line, however, the elements completely
dissolve into solute Nb and C within the matrix. The amount of each element removed from solution to
form NbC is found by tracing each composition back to the equilibrium solubility line using the
stoichiometric lines. This is schematically shown for the 0.5 wt pct Nb alloy in Figure 5.2(b). Being a
heavier element, it is observed that more Nb by wt pct is necessary to form NbC compared to C. Table 5.2
outlines the amount of C or Nb removed from the steel to form NbC and the amount remaining in
95
|
Colorado School of Mines
|
solution. This analysis demonstrates that with increasing Nb, more C is removed from the matrix to form
NbC, which may lead to a decrease in hardness and, in turn, wear resistance of the martensite matrix. The
concentration of Nb remaining in solution is very low at equilibrium for all of the steels.
(a) (b)
Figure 5.2 (a) Solubility isotherm describing the relationship between Nb and C (wt pct) at 1085 ºC in
austenite. The dashed lines represent stoichiometric amounts of 1.0 wt pct C and 0.01, 0.25, 0.5, and
1.0 wt pct Nb. (b) Schematic depicting how relative amounts of Nb and C removed from the matrix to
form NbC were determined for the 0.5 wt pct Nb alloy.
Table 5.2 Calculated Amounts of Carbon and Niobium in the Form of NbC (wt pct)
Nb C in NbC C in Austenite Nb in NbC Nb in Austenite
0.01 0.0007 0.9993 0.0056 0.0044
0.25 0.0341 0.9659 0.2455 0.0045
0.5 0.0640 0.9360 0.4953 0.0047
1.0 0.1286 0.8714 0.9950 0.0050
It is of importance to note that the constants provided by Turkdogan in Equation 2.7 may not be
valid for the C and Nb concentrations examined in this study. Turkdogan determined these constants
based on a thermodynamic model that assumes a dilute solution. However, this may be less accurate for
an alloy containing 1 wt pct C and greater than 0.25 wt pct Nb.
Solute C contents calculated in Table 5.2 were correlated to hardness values using Figure 5.1 and
Equation 5.1 for the martensite and austenite, respectively. For C contents above approximately
0.7 wt pct, Figure 5.1 has substantial scatter in the data and a median value was chosen for the hardness of
the martensite. Bulk retained austenite content was assumed to be 25 vol pct for all alloys, consistent with
measured values in Tables 4.7 and 4.10. The following rule of mixtures equation was employed to predict
bulk hardness:
96
|
Colorado School of Mines
|
HV = 0.25 ∙ HV + 0.75 ∙ HV (5.3)
bulk γ α'
Predicted austenite, martensite, and bulk hardness values are displayed in Table 5.3. A small decrease in
hardness might be expected when increasing the Nb content due to the reduced carbon concentration in
the martensite. This decrease may not be substantial, and the model does not account for any influence of
NbC on hardness.
Table 5.3 Estimated Hardness Values Based on Solubility Calculations
Alloy Solute C (wt pct) HV HV HV HRC
α’ γ Bulk Bulk
0.01 Nb 0.9993 881.5 188.6 708.2 64.1
0.25 Nb 0.9659 880.0 186.9 706.7 64.0
0.5 Nb 0.9360 880.0 185.5 706.4 64.0
1.0 Nb 0.8714 875.0 182.4 701.8 63.9
5.4 NbC Volume Fraction Evaluation
In Chapter 4.4, micrographs were evaluated for the area fraction of the microstructure that
consisted of NbC-containing constituent. This section evaluates the amount of NbC within this eutectic
constituent in the hot rolled material. These values are then used to update the rule of mixture for
hardness in Equation 5.3 to account for hardness provide by NbC in the microstructure.
The first analysis was performed using the solubility calculations in Section 5.3. This was
performed by directly adding the content by mass of both C and Nb that were determined to participate in
forming NbC. These values were then converted to volume fraction using the densities of 7.82 g∙cm-3 and
7.71 g∙cm-3 for austenite and NbC, respectively. These values are provided in Table 5.4.
A second analysis was performed using a single equilibrium calculation with Thermo-Calc®
thermodynamic software. This provides equilibrium fractions of the phases present given the
experimental lab prepared material chemical compositions in Table 3.2 at the hot rolling temperature of
1085 °C. These values are also provided in Table 5.4.
Table 5.4 Calculated NbC Volume Fraction Values
Volume Fraction (pct)
Nb (wt pct)
Solubility Calculations Thermo-Calc®
0.01 0.01 0.00
0.25 0.28 0.29
0.5 0.55 0.57
1.0 1.11 1.16
97
|
Colorado School of Mines
|
Volume fraction values from both methods of analysis are consistent. Using the values provided
by Thermo-Calc® as these take into account all alloying, not just Nb and C, as well as a hardness of
2500 HV for NbC [36, 37, 42, 75], the following rule of mixtures can be used to predict the hardness of
the hot rolled plates:
V V
HV = V ∙ 2500 + $0.25 - NbC % ∙ HV + $0.75 - NbC % ∙ HV (5.4)
bulk NbC 2 γ 2 α'
where V is the volume fraction of NbC from Table 5.4. Since it is not straightforward if the formation
NbC
of NbC is at the expense of the martensite or austenite, it was assumed that the formation is distributed
equally between the phases. Hardness values predicted using Equation 5.4 are provided in Table 5.5.
Table 5.5 Predicted Laboratory Prepared Material Hardness
Including NbC
Bulk Hardness
Nb (wt pct)
HV HRC
0.01 708.3 60.5
0.25 712.4 60.7
0.5 717.6 60.9
1.0 724.7 61.2
Hardness values in Table 5.5 remain to be lower than those measured of the heat treated samples;
however, these calculations aid in better predicting the effect NbC has on hardness. For each 0.25 wt pct
Nb added, an increase in approximately 0.2 HRC is expected to result.
5.5 Hot Rolled Plate Dilatometry
This section analyzes differences in the CCT curves displayed in Figure 4.23. All CCT diagrams
for each alloy were combined into Figure 5.3 to better highlight differences in transformation
temperatures as a function of Nb content. It is first observed that the C + P bay is shifted to the right as
Nb alloying is increased. For the samples with 0.01 and 0.25 wt pct Nb, a pearlite transformation was
seen in dilation data for both -10 °C·s-1 and -5 °C·s-1 quench rates, and no martensite formed during
the -5 °C·s-1 quench. For the alloys with 0.5 and 1.0 wt pct Nb, the pearlite bay only exists for the samples
quenched at -5 °C·s-1 and pearlite is not expected to form under typical water quenching conditions. This
behavior supports literature presented in Section 2.8.4 [46 – 48, 50, 51] stating that Nb is an austenite
stabilizing element, allowing austenite to remain stable in the microstructure at lower temperatures and
suppress the formation of the undesired pearlite constituent.
98
|
Colorado School of Mines
|
As previously discussed in Section 4.6, as Nb alloying is increased, an increase in M temperature
s
is observed in the hot rolled dilatometry samples. This trend can likely be attributed to the strong carbide
forming property of Nb, which removes C from the austenitic matrix to form NbC. This phenomenon is
discussed in detail in Section 5.3, with the estimated (calculated) amount of C removed from austenite
shown in Table 5.2. Using Equation 3.6, it is seen that M is expected to increase as C is decreased.
s
Combining the C in austenite values in Table 5.2 and the M temperature calculations in Equation 3.6, M
s s
temperature can be predicted as a function of Nb alloying. Calculated M temperatures for each alloy, as
s
well as those measured experimentally in the dilatometry study, are included in Table 5.6, and displayed
schematically in Figure 5.4 as a function of Nb content.
(a) (b)
Figure 5.3 (a) Continuous cooling transformation (CCT) diagram developed by a dilatometry study on
laboratory cast and hot rolled ingots containing 0.01 (black), 0.25 (blue), 0.5 (red), and 1.0 (green) wt
pct Nb. Samples were heated to 825 ºC, held for 600 s, quenched to 60 ºC, and slowly cooled to room
temperature. (b) Close up of M temperature for all alloys (color image – see PDF copy).
s
Table 5.6 Experimental and Calculated M Temperature
s
M Temperature (ºC)
s
Nb (wt pct)
Experimental Calculated
0.01 186.7 171.8
0.25 197.6 176.4
0.5 203.4 183.2
1.0 216.3 193.2
While the values found experimentally are slightly higher than those calculated, the slopes of the
trend lines between each data set are nearly identical, with an increase in M from the 0.01 Nb and 1.0 Nb
s
99
|
Colorado School of Mines
|
alloys being 29.6 °C and 21.4 °C for the experimental and calculated results, respectively. Discrepancies
in calculated versus experimental temperatures might stem from both experimental techniques and from
the empirical Equation 3.6, as M temperature calculations deviate throughout literature [58 – 65]. Often,
s
they are predicted to model lower C steels, unlike the alloys used in this study. Equation 3.6 also does not
consider the effects of Nb on M directly, but rather this effect is only incorporated via the decrease in C.
s
Additionally, if not all the cementite was able to dissolve during the austenitization step as was observed
with the salt pot treated samples in Section 4.7, the matrix will have less solute C and the M temperature
s
will increase slightly. Overall however, the shift in M temperature with Nb concentration seems to be
s
well explained by the removal of carbon from the austenite due to NbC formation.
Figure 5.4 Relationship between bulk Nb content and M temperature, found both experimentally and
s
calculated empirically. Linear fit trend line equations are included.
5.6 Salt Pot Heat Treatment Evaluation
This section discusses the hardness and microstructural results obtained during heat treatment
experimentation on samples in the salt pots. Samples tested were 7.62 x 2.54 x 0.85 cm to match the
dimensions necessary for Bond abrasion testing. Samples were machined from hot rolled plates of each
alloy, with heat treatment profiles listed in Table 4.6.
Table 4.7 demonstrated that for HT1 samples, hardness generally increased with an increase in
Nb, though this trend is minor in light of the range of uncertainty in the hardness measurements. In
addition, the samples that were quenched in water after HT1 tended to have higher hardness values than
those quenched to 150 °C possibly due to an increase in lattice strain attributed to a more aggressive
quench. However, hardness values of both 0.01 wt pct Nb samples were comparable to the hardness of
current grinding balls (60.8 ± 1.0 HRC), so both heat treatments can be considered “suitable” based on
hardness data alone.
100
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.