University
stringclasses 19
values | Text
stringlengths 458
20.7k
|
|---|---|
Chalmers University of Technology | C DIMENSIONING
C Dimensioning
Thecompleteflowsheetandcomponentlistforeachcaseispresentedonthefollowing
pages. In the component list the characteristic dimensions or properties of the
equipment is shown, and they are obtained from the from the following sources:
• The height of the columns are chosen in the simulations in order to fulfils the
85% capture rate of CO and requirements on H S reduction. The packed
2 2
height of the column is assumed to be 70% of the total height (Jilvero, 2014).
The diameter is also obtained from the simulations using the flooding ap-
proach. The diameter has been adjusted so that the largest fluxes are 70% of
those causing flooding.
• ThePinchanalysisdeterminedthecapacityandthenumberofheatexchangers.
The area was then calculated by assuming an overall heat transfer coefficient
using data in Sinnott & Towler (2009).
• The sizes of both vertical and horizontal flashes were estimated by calculating
the required vapour disengagement space for complete separation of vapour
and liquid fractions. See Sinnott & Towler (2009) for further information
about calculation procedure.
• The capacity of the compressors are obtained from Aspen Plus.
• The capacity of the pumps are calculated in Aspen Plus. However, the main-
tenancepumpscapacityhasbeencalculatedbyassumingtheyshouldovercome
a elevation difference of the same hight as the highest column and the pressure
drop caused by the first absorber in each case. See Sinnott & Towler (2009)
for further information about calculation procedure.
• The Buffer tanks are dimensioned for a storage capacity of at least 10% of the
liquid holdup volume in the adjacent process equipment (Jilvero, 2014). Tanks
for make-up chemical should have storage capacity of at least the amount
needed for 10days of production. The minimum tank size is 10m3 (Sinnott &
Towler, 2009).
69 |
Chalmers University of Technology | C DIMENSIONING
Table 29: Equipment list presenting all units and characteristic dimensions for the BLGCC-
S-CCS-case. The total height of the columns is provided together with the packing height in
parenthesis.
Column Type Pressure [bar] Temperature [◦C] Heigth [m] Diameter [m]
ABS-1 Norton IMTP: 2 IN 52.0 15 - 35 28.6 (20.0) 1.90
ABS-2 Norton IMTP: 2 IN 52.0 15 - 18 45.7 (32.0) 3.45
H2S-CONC Norton IMTP: 2 IN 52.0 96 - 104 37.1 (26.0) 1.25
STR-1 Norton IMTP: 2 IN 2.0 30 - 206 7.1 (5.0) 1.70
Heat exch. Type Pressure [bar] Temperature [◦C] Capacity [kW] Area [m]
HEX-1 Shell and Tube 1.0 - 52.0 10 - 95 1695 211
HEX-2 Shell and Tube 2.0 - 52.0 35 - 206 8707 526
HEX-3 Shell and Tube 1.0 - 52.0 15 - 104 410 48
HEX-4 Reboiler 2.0 52.0 206 256 122 22
HEX-5 Reboiler 2.0 - 11.5 170 - 206 12886 1184
HEX-6 Shell and Tube 1.0 - 2.0 10 - 96 129 4
HEX-7 Shell and Tube 1.0 - 2.0 10 - 129 12784 863
HEX-8 Shell and Tube 1.0 - 52.0 10 - 18 276 166
HEX-9 Shell and Tube 1.0 - 52.0 10 - 221 708 77
HEX-10 Shell and Tube 1.0 - 52.0 10 - 19 3047 1600
HEX-11 Shell and Tube 1.0 - 4.1 10 - 118 544 70
HEX-12 Shell and Tube 1.0 - 11.0 10 - 127 620 76
HEX-13 Shell and Tube 1.0 - 29.0 10 - 128 689 84
HEX-14 Shell and Tube 1.0 - 80.0 10 - 128 1594 193
Flash Type Pressure [bar] Temperature [◦C] Heigth [m] Diameter [m]
FLASH-1 Horizontal 14.0 20 - 20 19.16 6.39
FLASH-2 Horizontal 6.2 20 - 20 19.15 6.38
FLASH-3 Horizontal 1.5 17 - 17 19.04 6.35
74 |
Chalmers University of Technology | D OVERALL ENERGY AND MASS BALANCE
D Overall energy and mass balance
The calculations of the overall energy and mass balance has used the work of Pet-
tersson (2014) as a starting point. Pettersson has in turn based her calculation
on public data published by Skogsindustrins milj¨odatabas in 2010. From this data
Pettersson calculated various energy flows for e.g. bark and black liquor. The pulp
production in 2010 were 409kADt/year but in 2014 the pulp production rate has
increased to 422kADt/year. Hence, the values provided by Pettersson (2014) has
been linearly scaled to match the new production rate. By doing this calculation
the overall balance for the RB-M-NC-case is obtained.
For the RB-M-CCS-case knowledge of the steam cycle is needed. The ratio
between LP-steam and MP-steam use is assumed to be 2.7 for the pulp mill without
capture (Pettersson & Harvey, 2012). The utility demand results from the MEA
process simulations are added to the steam use of the pulp mill without capture.
To balance the energy demand of the pulp mill the steam cycle is simulated. These
simulations determines the solid wood fuel need and also the electricity production
of the pulp mill.
For the BLGCC-S-scenario the simulations was carried out using the black li-
quor mass flow obtained from SCA. However, by comparison with the black liquor
mass flow reported by Pettersson (2014) it was concluded that the data provided by
SCA was extracted when the pulp mill was running at part load. Hence, the results
from the simulations was linearly scaled to the correct sizes. The BLGCC-S-scenario
includes a number of additional processes, see Figure 9, their utility demand is ob-
tained from the following sources and scaled to match the black liquor mass flow.
The utility consumption of the gasifier, gas cooler and the air separation unit has
been extracted from Consonni et al. (2009). The steam demand for the Claus and
SCOT processes are extracted from Consonni et al. (2009) as well but, the electri-
city consumption is taken from Field & Brasington (2011a). The gas turbine and
the HRSG are modelled and the produced utilities are incorporated in the overall
balance. To balance the energy demand the steam cycle is modelled.
In the BLGMF-R-scenario, the sources mentioned in the previous paragraph is
used as well. For the DME synthesis and separation, process data is extracted from
Gadek et al. (2013) and Ohno et al. (2006). Following the same principle as the
other scenarios, the steam cycle is modelled to balance the energy demand.
Forallcasesthedistrictheatingisassumedtobeconstant. Thereasonisthatthe
heat used for the generation of district heating originates from the pulping process
and this is unaltered for every case. Only the recovery of the cooking chemicals is
changed, hence it is reasonable to assume the district heating is constant.
Other parameters needed for the calculations are obtained from the following
sources:
• The pulp mill is assumed to operate 355days/year with an availability of 98%
which means the operating time is 8350h/year. (Ekbom et al., 2005)
85 |
Chalmers University of Technology | D OVERALL ENERGY AND MASS BALANCE
• The biofuel boiler is assumed to have an efficiency of 90%. (Pettersson, 2014)
• When applying black liquor gasification the causticization load increases and
consequently so does the fuel need in the lime kiln. A 25% increase in fossil
fuel usage in the lime kiln is assumed. (Ekbom et al., 2005)
• Polysulfide cooking is assumed to increase the pulp yield by 4%. (C¸¨opu¨r,
2007)
• The fossil fuel used in the pulp mill is assumed to be Fuel Oil No. 6 with an
emission factor of 71.2tCO /TJ. (The Climate Registry, 2013)
2
• Bark combusted in the biofuel boiler is assumed to originate from spruce and
pine. The emission factor is 83.9tCO /TJ. (U.S. Environmental Protection
2
Agency, 2003)
• The emission factor for solid wood fuel is assumed to be 88.9tCO /TJ. (The
2
Climate Registry, 2013)
• The substituted electricity when the pulp mills sells their electricity surplus
is assumed to be a Nordic electricity mix which emits 258g/kWh. (M¨olndal
Energi AB, 2012)
• In the scenario where DME is produced it substitutes diesel which emits
160gCO eqv/km. (Salomonsson, 2013)
2
• Electricity is estimated to be sold at 490SEK/MWh, this price includes profit
from electricity certificates. (Swedish Energy Agency (2012); European Com-
mission (2013))
• Electricity is estimated to be bought at 590SEK/MWh. (European Commis-
sion, 2013)
• The price of cooling water is assumed to be 0.4SEK/m3, which corresponds
to 69SEK/MWh if the water is utilized between 10-15◦C. (Heat and Power
Technology, 2013)
• Thecostofsolidwoodfuelisabout200SEK/MWh, undertheassumptionthat
wood chips from branches and treetops are used. (Swedish Energy Agency,
2014)
86 |
Chalmers University of Technology | E MODELLING OF GAS TURBINE
E Modelling of gas turbine
The flowsheet of the gas turbine is shown in Figure 41. The model is based on the
work of Field & Brasington (2011b). Information about the property method and
components used in the simulations is from Aspen Plus help documentation.
Figure 41: Aspen Plus flowsheet of the gas turbine used for the BLGCC-S-scenario.
The PR-BM (Peng Robinson cubic equation of state with the Boston-Mathias
alpha function) is used as property method, since this mehtod is recommended for
gas-processing of light gases such as CO and H .
2 2
The syngas exiting the Selexol process is first heated (HEAT-1) to 240◦C and
then expanded from 52bar to 31.7bar. The EXPANDER is modelled using the
”Compr”block with the turbine setting checked. A isentropic efficiency of 0.8 and
a mechanical efficiency of 0.98 is used. The syngas enters the combustion cham-
ber together with compressed air at 16.2bar. A ”Compr”block with an isentropic
efficiency of 0.865 and a mechanical efficiency of 0.98 is used to compress the air.
The combustion chamber is modelled using a ”RGibbs” block. This block models
chemical equilibrium by minimizing Gibbs free energy and taking atom balance con-
straints into account. A pressure drop over the combustion chamber of 0.58bar is
assumed. The amount of air injected to the combustion chamber is adjusted so that
the outlet temperature is 1300◦C. Higher inlet temperatures to the turbine would
be possible if advanced material or thermal barrier coatings are used (Khartchenko,
1998). The turbine is modelled using the ”Compr” block and the outlet pressure
is specified to 1.05bar. A isentropic efficiency and mechanical efficiency of 0.898
respectively 0.988, has been used.
87 |
Chalmers University of Technology | F MODELLING OF HRSG
F Modelling of HRSG
The flowsheet of the HRSG (heat recovery steam generator) is shown in Figure
42. The model is based on the guidelines found in Khartchenko (1998) regarding
the design of a dual pressure HRSG. Information about the property method and
components used in the simulations is from Aspen Plus help documentation. The
PR-BM (Peng Robinson cubic equation of state with the Boston-Mathias alpha
function) is used as property method.
Figure 42: Aspen Plus flowsheet of the HRSG used for the BLGCC-S-scenario.
The flue gas exiting the gas turbine flows through a number of heat exchangers
and leaves at a temperature of 154◦C. All heat exchangers are modelled using the
”HeatX” block which is set to counter-current and shortcut calculation. The heat
supplied by the flue gases is used to generate HP- and MP-steam. Feed-water of
140◦Cand11.5barenterstheMP-ECONunitwhereitispreheatedtothesaturation
temperature of 186◦C. This stream is then split into two, where 9% enters the MP-
EVAP where MP-steam is produced. The rest is pumped to a pressure of 105bar
and is heated to the saturation temperature of 314◦C in HP-ECON. After being
evaporated in HP-EVAP the steam is superheated to a temperature of 514◦C in
HP-SUPER. As can be seen in Figure 42 the steam streams have been manually
divided to ease convergence. The properties of the streams is copied by a ”Transfer”
block in these cases.
The temperature-heat flux diagrams of the HRSG for the two BLGCC-S-cases
are found in Figure 43 and 44. They are almost identical with the exception that the
BLGCC-S-NC-case have slightly higher flue gas temperatures, because of a higher
flue gas flow, and thereby generates a small additional amount of MP-steam.
89 |
Chalmers University of Technology | G MODELLING OF STEAM CYCLE
G Modelling of steam cycle
The steam cycle is modelled in Aspen Plus to balance the energy demand of the pulp
mill. Figure 45 presents the Aspen Plus flowsheet of the steam cycle. Information
about the property method and components used in the simulations is from Aspen
Plus help documentation. The PR-BM (Peng Robinson cubic equation of state with
the Boston-Mathias alpha function) is used as property method.
Figure 45: Aspen Plus flowsheet of the steam cycle.
The pressure levels in the steam cycle is identical to those present in SCAs pulp
mill (SCA O¨strands Massafabrik, 2006). The following three levels are used:
• LP-steam at 3.5bar and 139◦C
• MP-steam at 11.5bar and 186◦C
• HP-steam at 105bar and 314◦C
For the BLGCC-S-scenario the two streams exciting the HRSG, ”HP-HRSGI”
and ”MP-HRSGI”, enters the steam cycle at their respective pressure level. For the
two other scenarios these streams are non-existent. Feed-water is pumped to 105bar
and then enters the unit named BOILER, where HP-steam is generated. This is the
boifuel boiler of the plant and it is modelled as a ”Heater”block. The HP-steam is
expanded in a high pressure turbine, HP-TURB, which is modelled by a ”Compr”
block set to turbine calculations. A isentropic efficiency and mechanical efficiency
91 |
Chalmers University of Technology | G MODELLING OF STEAM CYCLE
of 0.871 respectively 0.98, is used. MP-steam exits the high pressure turbine and is
divided into two streams. One stream goes to the cooler MP-CON which represents
the MP-steam need of the pulp mill. It then passes a second unit in which the
temperature is lowered to the LP-steam temperature. The heat released is used to
coverapartoftheLP-steamdemand. Theotherstreamentersasecondturbine, LP-
TURB, which expands the MP-steam to LP-steam. A isentropic efficiency of 0.871
and a mechanical efficiency of 0.98 is used. The LP-steam then passes a cooler,
LP-CON, which represent the LP-steam need of the pulp mill. The condensate from
LP-CON is pumped to 11.5bar and then rejoined with the MP-steam condensate in
a deaerator. The loop is now closed as this mixture is used as feed-water for both
the biofuel boiler and the HRSG.
Two design specifications are used in the simulations. The fist one adjusts the
split fraction in SPLIT-2 so that enough MP-steam enters MP-CON to satisfy the
demand. The other design specification adjusts the amount of heat added in the
BOILER unit so that the LP-steam demand is satisfied. From the heat added in
the BOILER unit it is possible to calculate the amount of solid wood fuel that is
needed.
92 |
Chalmers University of Technology | H REFRIGERATION CYCLE
H Refrigeration cycle
Where temperatures below 15◦C is needed, cooling water cannot be used to provide
cooling. Instead a refrigeration cycle has to be utilised. A simple refrigeration cycle
is shown in Figure 46.
Q
COND
4 3
Condenser
Valve Compressor W
Evaporator
1 2
Q
EVAP
Figure 46: Schematic representation of a simple refrigeration cycle. Adapted from R. Smith
(2005).
The refrigeration cycle is basically a heat pump which extracts heat from a low
temperature and rejects it at a higher temperature. A refrigerant fluid is used as a
medium for this transport. The refrigerant is evaporated (1-2) at a low temperature
and thereby provides cooling to the process stream. By compressing (2-3) the refri-
gerant the condensation temperature increases. Thereby, cooling water can be used
as a heat sink in the condenser (3-4). The refrigerant is thereafter throttled (4-1) by
a valve to complete the cycle. The temperature-entropy and the pressure-enthalpy
diagrams for the refrigeration cycle are illustrated in Figure 47.
3
T P QCOND
Heat rejected to condenser
Condenser
4
4 3
2 1 2
1 Evaporator Refrigeration Compressor
duty work
QEVAP W
S H
(a) Temperature-entropy diagram (b) Pressure-enthalpy diagram
Figure 47: Diagrams showing the characteristics of a simple refrigeration cycle. The numbering
corresponded to that found in Figure 46. Adapted from R. Smith (2005).
93 |
Chalmers University of Technology | H REFRIGERATION CYCLE
From the simulations the refrigeration duty is obtained and to calculate both the
compressor work and the cooling water duty, the coefficient of performance can be
used. The ideal coefficient of performance is defined in Equation 3.
Q T
EVAP EVAP
COP = = (3)
Ideal
W T −T
COND EVAP
The actual performance of a simple refrigeration cycle is typically 0.6 of the ideal
performance (R. Smith, 2005). Simple refrigeration cycles can be used to provide
cooling down to −40◦C. However, the Rectisol process requires cooling as low as
−50◦C. Hence, more complex cycles are needed such as multistage compression
cycles or cascade cycles. These are more efficient than simple cycles and therefore a
value of 0.75 is used instead. By multiplying the coefficient of performance by 0.75
and rearranging Equation 3, the required compressor work can be calculated.
Q (T −T )
EVAP COND EVAP
W = (4)
0.75T
EVAP
As seen in Figure 47, the heat rejected at the condenser to the cooling water is
equal to the sum of the compressor work and the refrigeration duty.
Q = W +Q (5)
COND EVAP
94 |
Chalmers University of Technology | Maintenance and operational patterns in thermal power generation
Methodology development for maintenance cost estimation
JOHAN EGELAND
WILLIAM FLODIN
Department of Department of Space, Earth and Environment
Chalmers University of Technology
Abstract
With increasing shares of solar and wind generation in the power system, the un-
derstanding of power generation system as we know it today, needs to be reassessed.
Wind and solar power, known as the non-dispatchable generation, are likely to im-
pact the net load on the grid and dispatchable power production, typically thermal
power plants, will face more frequent starts and stops and volatile operational pat-
terns. This is expected to cause tougher conditions for the plant and result in more
critical wear on the equipment, which must be considered and accounted for.
This master’s thesis develops a method to evaluate maintenance costs depending on
the operational pattern - in terms of the number and types of start-ups - of a steam
cycle. The work focus on the steam turbine rotor but could be applied also to other
critical components of the plant. The method includes a rotor model to perform
transient simulations of the rotor temperature during start-ups. The rotor tempera-
ture was then related to thermal stresses and life expenditure of the rotor, which are
typical input parameters to estimate maintenance costs from an LCC-perspective.
The method may be applied to evaluate the influence of energy system scenarios,
process designs, and maintenance policies.
To exemplify, this work evaluates the maintenance cost of the rotor for six defined
scenarios, four maintenance policies, i.e. failure-based, time-based, condition-based
and opportunity-based maintenance, and three types of labour services, internal,
external and contract service. The examples illustrate how the maintenance costs
mayberelatedtothenumberofstartsandthattheproposedmethodmaybeprovide
initial support to the initial discussions on the maintenance of thermal power plant
components.
Keywords: Maintenance, Production, Life Cycle Cost, LCC, Power generation,
Transient operation, Steam turbine, Flexible operation
v |
Chalmers University of Technology | 1
Introduction
Theavailabilityofelectricityindevelopedcountrieshasbecomeanobviousresource,
constantly available in our power outlets. This requires reliable energy production,
where it is made sure that the producing equipment can provide its function when-
ever demanded. This emphasize the need for proper maintenance.
This thesis studies the relationship between maintenance costs and operational pat-
terns at thermal power plants. A background is following along with the aim of the
thesis, delimitations and a specification of issues under investigation.
1.1 Background
The energy market and its affiliated energy systems have seen rapid development
in recent years towards an increased share of non-dispatchable renewable energy
sources such as wind and solar power generation [1]. This trend has initiated dis-
cussions on the role of dispatchable large scale thermal power generators as well as
the consumers in the system [2]. As some electricity markets has been deregulated
and opened up for private actors, the competitive landscape and the actual power
generation is changing.
On the traditional energy market, there was a distinct classification of producer
and consumer, connected on a centralized network. Today, the consumer could also
be the producer, by selling redundant self-produced electricity to the market [2].
Along with a varying electricity demand from the customers and the varying power
generation of non-dispatchable energy sources, this is creating strong irregularities
and discontinuous operational patterns at the power generation plants [3]. The
production must also constantly be matched with the demand as a fundamental
requirement [2]. This is relatable to a chase strategy typically known in capacity
planning within production environments [4], with the difference that the electricity
market is changing the demand in real-time. As changes in the load appear, the pro-
duction must follow and deliver what is being demanded. Without the continuous
supply and availability of electricity in our outlets, many of our societal functions
would not function properly. It is thus important to maintain and care for the
equipment making the power generation, as well as heat generation possible and its
future functioning in the society.
With a varying load in the power system, the operational pattern of power units
must be varying as well, as overproduction could impact on profits in terms of un-
necessary usage of resources as well as the fact that the power grid cannot handle a
1 |
Chalmers University of Technology | 1. Introduction
significant overproduction either, and the plant could even be paid to decrease pro-
duction. The varying operating cycles are followed by an increase of start-and-stops,
severe transients and possible longer downtimes of the thermal power plants. This
is often raised as a concern from the industry, which is generally used to continu-
ous operation patterns [5]. This type of changed production pattern is expected to
impact on the lifetime of critical components in the equipment due to an increased
probability of failure, caused by mechanisms such as creep and thermal-mechanical
fatigue loadings [6], which might lead to additional unplanned maintenance costs
and unavailability [3].
As industrial maintenance practices, as well as research in maintenance of produc-
tion systems, has undergone advancement the last years, it has now moved to be an
important part in business strategies [7]. As an example, in manufacturing, main-
tenance has become a central core of the production system [8], as research shows
that optimization of maintenance can increase the performance of the production
systems [9] and gain economical benefits [7]. By regarding a power plant as a pro-
duction system, that is transforming an input into an output, the same principles
of maintenance are possible to apply.
With the possibility of highly volatile electricity prices on the future electricity
market, it is believed that processes that can work efficiently during transient or
intermittent operations are needed and thus the maintenance perspective cannot be
ignored. Vishnu and Regikumar [10] are stating that the maintenance cost can reach
as high as 40% of the total operational cost of a plant, which could be impacting on
possible profits. This project aims at estimating and understanding the relationship
between maintenance costs and operational patterns in thermal power plants, and
act as a support in the discussions. The true understanding of maintenance in a
production system could help companies to stay competitive on their market.
Research on maintenance and operational patterns is also believed to support the
operator experience by the giving of a support basis for decisions on maintenance
actions. With the possibility of making economical maintenance decisions in aspects
of lifetime extension, it could be possible to extend the lifetime of machinery and
minimize costs. As this could help to decrease used parts and resources, it could
further decrease environmental impact and support efficient energy production in
general, making our future more sustainable.
For the reader interested in a more comprehensive background to the thesis, it is
referredtolifetimeassessmentresearchbyBanaszkiewicz,Morozetal.,Benatoetal.,
Venkatesh et al., Musyafa et al. and Grosso et al. [11]–[16]. And for topics related
to steam turbine costs on maintenance, it is referred to Keatley et al., Stoppato et
al., Aminov et al. and Kumar et al. [3], [17]–[19].
2 |
Chalmers University of Technology | 1. Introduction
1.2 Aim and purpose
Theaimoftheprojectistodevelopacostassessmentmethodologyforhowthemain-
tenance cost varies based on the number of starts of a thermal power plant. This
will be done by combining theory and methods from the fields of energy technology
and production engineering, that will help to simulate, estimate and understand the
relationshipbetweenmaintenanceandoperationalpatternsatthermalpowerplants.
The relation is assessed in terms of life-cycle cost (LCC) for different maintenance
policies and services related to defined scenarios.
The purpose of the development of the methodology is sought to be the beginning
of a supporting tool that can help initial discussions of maintenance strategies at
thermal power plants. With further development and accurate data, it is thought
that the methodology and tool could help to enable cost efficiency in thermal power
production in long-term perspectives and support the operator experience in main-
tenance decisions.
1.3 Delimitations
The principal delimitation in the project is the focus on the steam turbine rotor and
the exposure to fatigue due to thermal stresses. The methodology development is
sought to become as general as possible to fit different steam turbines, even though
the methodology will be tested using process data for a specific steam turbine along
with wide assumptions of other necessary data. Other, eventually important and
impacting methodology data, such as plant dynamic performance, plant size, de-
pendable equipment, efficiency increasing processes, market conditions, production
disturbances and precise design parameters are not extensively considered.
Further, the focus on the rotor as a component is based on literature review findings,
and will not rely on any project-performed root cause analysis or failure mode effect
analysis among other critical component analysis tools, as the project is limited in
the availability of time and access to steam turbine equipment.
1.4 Specification of problem
A closer specification of the issue under investigation is summarized as follows:
• Identify previous work and related concepts that have connected maintenance
with operational patterns on steam turbines.
• Create a rotor model that is considering temperature differences in the ro-
tor, relating it to thermal stresses and lifetime, eventually acting as input to
maintenance cost calculations.
• Gatherqualitativeinputfromtheindustrytoachieveaconfirmationonproper
methodology development and results.
• Establish scenarios for which maintenance approaches and the relationship
between maintenance and operational patterns may be assessed.
3 |
Chalmers University of Technology | 2
Theory on
Production and Energy Systems
This chapter on production and energy systems is a theoretical framework covering
themostessentialtheoreticalpartsofthethesis, whichseekstoprovideenoughback-
ground for understanding the thesis contents without a previous expertise knowl-
edge.
Considering a production system, it is a concept involving all the relations and in-
teractions between different actors and components in a system with a purpose to
produce an output [8]. Hence, the general function of a production system can be
described as a transformation process of input to output and could be the creating
of goods and/or services. With this definition, a production system could be an
energy system, that would describe the relations between production, distribution
and usage of energy. Further, energy production is typically categorized as a process
production. This means that the input is undergoing a physical-chemical process
transformed to output, that is hard to restore back into its original raw material.
Anoverallgoalinproductionsystemsisoftentoachieveashighproductionefficiency
and performance as possible. The performance is then often measured by the help of
key performance indicators (KPIs). In order for the production system to function
properly over time and sustain high efficiency and performance, the system must un-
dergotechnicalandadministrativeactions, knownastheconceptofmaintenance[8].
Maintenance has for a long time been seen as a necessary evil in production systems
but has evolved to be regarded as a strategic concern [7]. Maintenance can also be
considered in a life-cycle perspective stretching from initial planning of the system
to its phase-out [8]. By considering maintenance as a strategic element it is possible
to accomplish business objectives and with maintenance as a core in the production,
it is possible to increase the availability of the resources and produce more when
required. As production systems themselves have developed over time and become
more advanced and technology-intensive, so has the maintenance management ap-
proach followed as well, where the maintenance manager could be confronted with
highly technical and demanding questions.
5 |
Chalmers University of Technology | 2. Theory on Production and Energy Systems
The industrialized world as know today, is currently on the edge of facing a fourth
industrial revolution within production, in the industry referred to as the concept
of "Industry 4.0". This concept is emphasizing digital technology and built upon
advanced computer science, information and communication technology [20]. Big
amounts of data are expected to be constantly gathered from the production sys-
tem and analyzed by artificial intelligence resulting in suggested conclusions. An
example of this could be computed conclusions predicting suitable opportunities for
carrying out preventive maintenance before a component or system fails.
As an energy system is a concept for all interacting parts in a cycling system related
to energy, including production, distribution and usage of the energy, some parts are
essential to be explained to understand their relationships. The following sections
are explaining those parts that are affiliated with the content of the thesis and
eventually followed by explanations on typical production concepts that also are
present in energy systems. The following theory sections are starting on higher
system-levelsandaresequentiallybeingnarrowedtowardsmorespecificanddetailed
topics.
2.1 Electrical grid
The electrical grid is the physical network that allows the distribution of electricity
from producers to consumers. From the power generating plants, transmission lines
arecarryinghigh-voltageelectricityandiseventuallybeingstep-downinvoltageand
distributed to consumers by local network lines. As electricity is difficult to store in
large amounts to a reasonable cost with current technology, the production of elec-
tricity must constantly meet the demand. If not, there is a risk of a power outage.
The production versus demand is being controlled by monitoring the frequency of
the grid. If the consumption is lower than production, the frequency increases, and
if consumption is higher than production the frequency decreases [21]. Balancing
control is hence an important function in a modern energy system and is commonly
a dedicated task to a special control center, known as balance responsibility [22].
This could be related to a strict production chase strategy known from supply and
operations planning [4].
With different types of power generating plants connected to the grid, the sources
are sometimes divided into the dispatchable and the non-dispatchable generation of
power sources. The dispatchable generation refers to the type of plants for which
its production is possible to start and stop whenever wanted, for example nuclear,
thermal and large hydro-electric dams. Wind and solar power are referred to as
the non-dispatchable generation since their primary energy sources are varying in
availability due to the impact of weather conditions. Their power production is only
available based on wind and sun light.
6 |
Chalmers University of Technology | 2. Theory on Production and Energy Systems
2.2 Electricity market
The electricity market refers to the market that is constituting the relations of the
actors involved in the energy system focusing on trading electricity. These involved
actors are in general producers, distributors and consumers [2], all connected on the
grid. It is also common that there is more than one actor-type on the market, for
instance, several producers. This is allowing competition. And as the market is
determined by the geographical constraints, it is also possible to let the competition
stretch over national borders [23]. Also, if major maintenance actions are needed in
the system of certain plants, the market and the balance responsible authorities can
still manage the production and distribution by creating new running strategies for
the other power plants on the grid.
The varying demand and production on the grid are what is contributing to a vary-
ing electricity price and fuel costs on deregulated markets. Depending on how the
market is divided, there could be varying prices across a geographical area. In
Nordic countries, electricity is traded on a common market platform and actors are
allowed to buy and sell electricity across national borders [24]. Producers are often
closing their selling price 24-hours ahead of production, while consumers are paying
a volatile tariff or a fixed price per consumed unit of electricity.
By the increasing availability and possibility of efficient power generation of renew-
able energy, such as wind and solar energy, the electricity market is evolving. One
distinct difference, as distributed generation is becoming encouraged, is the diffusing
line between producer and consumer [2]. As private home-owners with solar panels
or wind generators get the possibility of selling redundant electricity to the market,
they become producers. In the case of non-beneficial weather, the home-owner could
instead act as a consumer by buying electricity from the market for private usage.
This is not only requiring development of the whole energy system but also con-
tributes to the need of reliable and quick responding dispatchable power generation
plants that can handle these varying demand patterns.
2.3 Electrical power demand and load types
The production of electricity is often measured in watts, while the individual con-
sumption of electricity is measured by watt-hours, often expressed with a prefix.
The electricity demand is in general classified in load types related to consumer
usage patterns. Two commonly known loads types are base load and peak load.
7 |
Chalmers University of Technology | 2. Theory on Production and Energy Systems
Total demand
Net load
Load [GW] Solar
Wind
35
30
25
Peak load
20
15
Base load
10
5
Time
of day
12 AM 6AM 12 PM 6PM 12 AM
Figure 2.1: Schematic illustration of a total demand curve, base load, peak load
and net load, based on Denholm et al. [25].
Figure 2.1 presents the relation of base and peak load, where base load is referring to
the electricity usage being the minimum load demanded by society. A good exam-
ple of this is a night-time where most people are asleep and only essential electrical
appliances are turned on. This type of production is often covered by the kind of
power plants that do not change their output quickly, for example, nuclear plants
and large coal plants. Suddenly, a big increase could appear in the demand but
eventually soon diminishes, which is a peak load. An example of this could be the
simultaneous usage of microwave-ovens in half-time breaks during major national
sports games on television. These loads could typically be covered by the dispatch-
able generation that can respond quickly to increased demand. The load types are
therefore sometimes related to social patterns in society.
As the electrical grid could be connected with wind and solar power generation, this
is also introducing net load, also present in Figure 2.1. This type of load is defined
as the total power demand minus wind and solar power generation [26]. In such
systems, the net load must be covered by the help of dispatchable energy sources
that can respond quickly to the demand.
Some power plants have historically been used for base load operations but are
likely to face more cyclic operations due to more frequent peaks in the net load.
Figure 2.1 shows how decreasing solar production combined with increased demand
during the evening creates a ramping need of production that can be solved with
flexible dispatchable power generation. This will in many cases fall upon power
plants already in use, with configurations that can be ramped, for example, steam
turbines. Figure 2.1 may be an extreme case scenario, but it could be close to
reality in the near future for developed countries with a high development towards
renewable energy production [25]. This requires the systems to be reliable and
available when needed, emphasizing the need for proper maintenance.
8 |
Chalmers University of Technology | 2. Theory on Production and Energy Systems
2.4 Maintenance classification
Business strategy
Maintenance strategy
General decision structure with
Concepts a set of policies and actions
”The logic and recipe”
The triggering mechanism for
Policies the actions with a set of rules
”How the actions are triggered”
The actual task to carry out
Actions
or Intervention with the machine
”What to do?”
Figure 2.2: Hierarchical maintenance levels, based on Pintelon [7].
The previously explained electrical grid and power plants, eventually all require
some kind of maintenance over time, either physically or more abstractly with soft-
ware updates or changes. The view on maintenance as well as its importance differ
from case to case, but must certainly be considered in a producing system. Large
companies often develop an overall business strategy defining their short and long
termgoalsandaroadpathfortheiroperations. Ifthecompanyisagoods-producing
company, the business strategy is often followed by a manufacturing strategy defin-
ing production related questions. And occasionally, there could also be a defined
approach on maintenance included in the manufacturing strategy, here referred to
as the maintenance strategy.
Different terminology is often used in industry when considering maintenance issues.
The following classification of three hierarchical maintenance levels is based on Pin-
telon [7]. The three considered maintenance levels are the terms concept, policy and
action where their relation to each other is illustrated in Figure 2.2.
2.4.1 Maintenance concepts
A maintenance concept refers to a set of maintenance policies and actions on lower
operational levels and creates the general structure and logic planning of mainte-
nance [7]. In other words, a concept is a recipe or model on the holistic of view on
maintenance. In research aiming at reviewing and categorizing maintenance con-
cepts, 24conceptswereidentified[27]. Asitwasconcludedthatsomeofthemtended
to present slightly similar things but in their own approach, it could make it hard
9 |
Chalmers University of Technology | 2. Theory on Production and Energy Systems
to decide suitable maintenance concepts for a particular businesses. With certain
prerequisites and requirements, some companies also define their own concepts [7].
Some of the most well-known concepts are life-cycle costing (LCC), total productive
maintenance and reliability-centered maintenance.
2.4.2 Maintenance policies
A maintenance policy is a set of rules that determine the triggering mechanisms for
maintenance actions [7]. Table 2.1 is presenting four maintenance policies, which
each are triggering maintenance actions.
Table 2.1: Description of four maintenance policies [7].
Policy Description
FBM Failure-based maintenance
Apolicythatis triggeringmaintenanceactions oncea failure orbreak-
down has occurred.
TBM Time-based maintenance
Maintenance actions are triggered based on a set time or usage inter-
val, for instance hours. The triggered actions aim at preventing failure
to occur.
CBM Condition-based maintenance
Suitable system parameters are being monitored, such as vibrations,
oil viscosity, temperatures, pressures, flow rates, acoustics etc. and
are compared with predetermined values. If the measured parameter
is exceeding the set limit, typically lower than the limit of failure, a
preventive maintenance action is triggered.
OBM Opportunity-based maintenance
Maintenance actions are carried out once an opportunity arises. This
is a typical policy to apply on non-critical or long-life components. Or
in cases where it is simply not possible to interact with the component
without extensive work. These opportunities can though arise non-
periodically.
2.4.3 Maintenance actions
A maintenance action is the actual maintenance intervention or elementary task
that is being carried out by a technician on a particular component or machine.
The action can be defined as either a corrective action or a precautionary action [7].
A corrective action is an activity repairing or restoring an item back to a functioning
condition and occurs after the item has failed or has broken down. A precautionary
action is an activity that has a goal of diminish or avoid a possible failure to occur.
It is assumed that precautionary actions are cheaper than corrective ones [7].
10 |
Chalmers University of Technology | 2. Theory on Production and Energy Systems
2.5 Maintenance as a service
As the view on maintenance has evolved over time, maintenance activities have
gradually been integrated into product-service system (PSS) solutions [28]. It is now
not uncommon that customers are not only offered a product but instead a function
where maintenance often is included. Examples of PSS-solutions are station-based
car-sharingservices,cityrentalbikesandjetenginespowered-by-the-hour. Severalof
thesesolutionsenableremotemaintenancemonitoringwithdata-drivenmaintenance
approaches in order to make sure its function can be provided. This product-service
system is also appearing in the power generation industry [29]. Three types of
service approaches are hereby introduced. These are internal crew, external service
and external service with a framework contract.
2.5.1 Internal crew
Maintenance by an internal crew is defined as maintenance that is being performed
by internal employees of the machine user [30]. They generally have their own facili-
ties on-site and an internal system for receiving, prioritizing and planning upcoming
work.
2.5.2 External service
Externalserviceismaintenancecarriedoutbyanexternalservicecompany, thatwill
perform maintenance upon the customer’s request [30]. The external maintenance
company typically have their facilities located on a remote location and apply a rate
when service is performed.
2.5.3 External service with framework contract
External service with a framework contract refers to maintenance carried out by
an external company that has written a contract with the customer on how and
when maintenance will be performed [30]. This type of contract can entail either
a fixed or varying cost. This type of external service is sometimes provided by the
original equipment manufacturer (OEM) and could then be providing up-time or
power-by hour service [28]. This could also allow the OEM to gather data on how
their machines are operated and use the data for future research and development.
Another approach from the OEM could be to only sell their products using this
PSS-approach. External service based on framework contracts is common in the
turbine industry [31]–[33].
11 |
Chalmers University of Technology | 2. Theory on Production and Energy Systems
2.6 Dependability
Dependability
Maintenance
Reliability Maintainability
support
Figure 2.3: Factors affecting the dependability of the system, adapted from Bell-
gran and Safsten [8].
The term dependability intends to describe a system’s ability to perform when re-
quired [34] and considering production systems, dependability is often including the
concepts of reliability, maintainability and maintenance support [8]. The relation of
them is illustrated in Figure 2.3 for which explanations follow.
2.6.1 Reliability
Reliability is a system’s ability to perform as required without the occurrences of
failure, under given conditions and time intervals and is measured as the probabil-
ity of survival at a specific point of time, noted as R(t) [34]. The given conditions
refer to aspects that could impact on the reliability, for example, operation mode,
stress level, temperature, pressure and previous maintenance actions. What is be-
ing considered as the system is defined by the user and could be a machine itself, a
sub-system of the machine or just one component in the machine.
The reliability of an item generally follows a certain underlying statistical distribu-
tion, where common distributions are exponential, log-normal and Weibull distri-
butions. For mechanical components exposed to fatigue, the Weibull distribution
is generally known as a suitable lifetime distribution [30], with the two-parameter
probability density function (pdf):
!β−1
f(t) =
β t e−( ηt)β
(2.1)
η η
where: t = Specified period of time
β = Shape parameter of the distribution, also known as the slope
η = Scale parameter of the distribution, also known as the characteristic life
Equation(2.2)isdescribinghowtocalculatereliabilityusingatwo-parameterWeibull
distribution. Depending on the parameter values of the Weibull distribution, differ-
ent life behaviours can be described [30].
(cid:16) (cid:17)β
− t
R(t) = e η (2.2)
12 |
Chalmers University of Technology | 2. Theory on Production and Energy Systems
If integrating the reliability from 0 to ∞, it is possible to obtain the Mean Time
To Failure (MTTF), a common maintenance KPI, stating the average time between
consecutive failures.
Z ∞
MTTF = R(t) dt (2.3)
0
2.6.2 Maintainability
The maintainability of a system refers to the ability of an item to be restored to its
operational function within a given period of time and given conditions [30]. The
given conditions in terms of maintainability include accessibility of the component,
needed maintenance procedures and necessary maintenance resources [34].
AcommonKPIaffiliatedwithmaintainabilityistheMeanTimeToRepair(MTTR),
which is a measure stating the average time it takes to restore a failed item. The
measure is closely related to MTTF and the availability A, of a considered system,
where the availability is defined as an item’s ability to be in a performing state
when required [34]. Equation (2.4) describes how the availability of a system can
be calculated based on known values of MTTF and MTTR.
MTTF
A = (2.4)
MTTF+MTTR
2.6.3 Maintenance support
Maintenance support is the providing of resources such as labour, equipment, ma-
terials, spare parts, facilities and instructions that are needed in order to maintain
an item [34]. The maintenance support concerns questions that are closely related
to the type of chosen maintenance service.
13 |
Chalmers University of Technology | 2. Theory on Production and Energy Systems
2.7 Power generation by steam turbines
The electrical grid can be provided with electricity by the help of various kinds of
power generation plants as described in section 2.1. An example of such plants are
thermal power plants using steam turbines for converting the thermal energy into
electricity.
Steam
Steam turbine
G
Generator
Steam generator
Fuel Condenser
Water
Water pump
Figure 2.4: A basic schematic illustration of a steam cycle.
Figure 2.4 illustrates a basic steam cycle producing electricity. Starting with a
suitably selected fuel for the specific power plant, the fuel is combusted and starts
to heat up water-filled pipes lead through the steam generator. After the steam
leaves the steam generator it is being delivered, highly pressurized, to the steam
turbine. The steam expands in the the steam turbine and converts thermal energy
to mechanical energy by flowing through its turbine blades making the rotor spin,
generating electricity by the help of the connected power generator [35]. With the
turbine blades placed in a sequence of different sizes, it is referred to as the number
of stages.
After the steam turbine, the steam condensates back to water. In order to increase
the efficiency of the plant, condensation could be made by heating water used in
a district heating system. This makes the plant a so called combined heat and
power plant. Lastly, the water is pumped back to the steam generator, repeating
the principle of the steam cycle.
2.7.1 Steam turbine fundamentals
A steam turbine allows conversion of thermal energy to mechanical energy due to
the enthalpy of the steam. When modelling a steam turbine for computational cal-
culations, it is possible to calculate the enthalpy after the turbine module using the
enthalpy before the turbine, the machine efficiency and the nominal values [36]. The
mechanical power produced by the turbine can then be computed and transferred
14 |
Chalmers University of Technology | 2. Theory on Production and Energy Systems
to a power generator model.
The steam turbine model may describe the turbine either with a single turbine
section or with many sections in order to consider the steam extractions, which is
done for increased efficiency of the cycle. The pressure change is however defined as
a function of the nominal mass flow rate, the nominal pressure at the turbine inlet
and the nominal pressure at the turbine outlet [36].
v
u (cid:16) (cid:17)n+1
p u u 1− pout n
m˙ = m˙ int u p inl (2.5)
nom p int,nomu t 1−(cid:16) pout,nom(cid:17)n+ n1
p
inl,nom
with:
ln pout
n = p inl (2.6)
(cid:16) (cid:17) (cid:16) (cid:17)
ln pout −ln Tout
p T
inl inl
where: m˙ = Mass flow rate [kg/s]
m˙ = Nominal mass flow rate [kg/s]
nom
p = Pressure at inlet [Pa]
inl
p = Pressure at outlet [Pa]
out
T = Temperature at inlet [K]
inl
T = Temperature at outlet [K]
out
The steam turbine expansion is not totally isentropic nor isothermal and is therefore
dependent on the isentropic efficency. The efficiency is generally assumed to be
constant for every section of the turbine due to simplicity. Isentropic efficiency of a
steam turbine is defined as [36]:
Real turbine work ∆h h −h
turb in out
η = = = (2.7)
T
Isentropic turbine work ∆h h −h
is in out,is
The isentropic enthalpy at the outlet can further be described as a function of the
pressure. The shaft work for a section is therefore:
∆h = η (h −h ) (2.8)
turb T in out,is
And the total power output for a turbine with k number of sections is:
i=k
X
P = ∆h m˙ (2.9)
el turb,i i
i=1
Transient heat transfer
The Dittus-Boetler Equation (2.10) for turbulent pipe flow can be used as an ap-
proximation to calculate the the Nusselt number (Nu) inside a steam turbine as a
function of the Reynolds number (Re), Prandtl numer (Pr) and a correction factor
K. These are then used to calculate the convective heat transfer coefficient (HTC)
[13].
Nu = K ·Re0.8 ·Pr0.333 (2.10)
15 |
Chalmers University of Technology | 2. Theory on Production and Energy Systems
The HTC governs the temperature profile in the boundary layer between steam flow
and rotor surface. Since a steam turbine has a complex geometry, the Reynolds
number would be difficult to determine. Instead, the HTC can be expressed as a
function of HTC at nominal load [37], see Equation (2.11) . The equation is based
on the proportionality of the HTC in relation to the Reynolds number, which in
turn makes the proportional to the mass flow rate, m˙ 0.8.
HTC (cid:18) m˙ (cid:19)0.8
= (2.11)
HTC m˙
nom nom
HTC is one of the most critical variables in the calculations for transient conduction.
The calculation for radial temperature gradient in the rotor is solved by discretizing
the temperature into small time steps and introducing the non-dimensionalizing of
the governing equations [38]. The temperature difference between the rotor and
steam flow, θ, is divided by the maximum possible temperature difference θ .
i
θ T −T
θ∗ = = ∞ (2.12)
θ T −T
i m ∞
where: T = Steam temperature [°C]
∞
T = Initial metal temperature [°C]
m
The purpose of non-dimensionalizing the temperature difference is to express θ∗ as
a function of dimensionless numbers. These numbers are a spatial dimensionless
coordinate, a dimensionless time and the ratio of thermal resistance [39].
By definition θ∗ must be in the range 0 < θ∗ < 1. A spatial dimensionless coordinate
using the radius of a rotor, r , is defined as:
0
r
r∗ = (2.13)
r
0
The dimensionless time is also known as the Fourier number (Fo), which is defined
as:
α ·t
diff
Fo = (2.14)
r 2
0
where α is the thermal diffusivity of the rotor material, which combines thermal
conductivity with density and heat capacity:
λ
α = (2.15)
diff
C ·ρ
p
The ratio of thermal resistance, or the Biot number (Bi), compares the convective
heat transfer calculated in Equation (2.11) to the heat transfer within the body [38].
r ·HTC
0
Bi = (2.16)
λ
where λ is the thermal conductivity of the rotor material.
The exact solution for θ∗ can then be expressed as a function of r∗, Bi and Fo [39].
16 |
Chalmers University of Technology | 2. Theory on Production and Energy Systems
∞
θ∗ = f(r∗,Bi,Fo) = X C ·e(−ζn2Fo) ·J (ζ r∗) (2.17)
n 0 n
n=1
The Biot number is used to determine ζ and C which are roots to the Bessel
n n
function J . The first roots as functions of the Biot number are presented in Table
0
2.2. By solving Equation (2.17), the temperature gradient can be found at any
radial position in the component.
Table 2.2: Shortened table with coefficients for the one-term approximation to the
series solution for transient conduction, based on Özisik [38].
Bi ζ C
1 1
0.01 0.1412 1.0025
0.04 0.2814 1.0099
0.08 0.3960 1.0197
0.10 0.4417 1.0246
0.30 0.7465 1.0712
0.70 1.0873 1.1539
1 1.2558 1.2071
Start-up types and its conditions
When starting a steam turbine, there are typically three types of start-up scenarios
to consider; cold, warm and hot start-ups. Which type of start-up that is being
initiated, is generally defined by the turbine casing metal temperature measured at
the steam turbine just before the start-up begins.
By considering the rotor in the steam turbine, the type of start-up can be deter-
mined by the temperature difference between the rotor surface and the rotor mean
temperature, see Equation (2.18). The temperature differences further act as input
to the calculations of thermal stresses, see Equation (2.19).
2 Z r0
∆T = T −T = T − · r·T(r) dr (2.18)
S S r 2
0 0
where: r = Rotor radius [m]
0
r = Radial coordinate [m]
T = Temperature at the radial coordinate location [°C]
T = Rotor surface temperature [°C]
S
T = Mean rotor temperature [°C]
Thermal stresses due to start-ups
Starting up a turbine too quickly will most likely result in failures due to low-cycle
fatigue and thermal expansions. If wanted to reduce the risk of fatigue, the start-up
time must be limited with respect to thermal stresses [40]. A thermal stress module
is generally a part of the turbine control system, which is continuously evaluating
and monitoring the thermal stress at critical locations and helps to limit the tem-
peratures. These critical locations are often those with the largest wall thickness
17 |
Chalmers University of Technology | 2. Theory on Production and Energy Systems
causing the largest thermal differences [12]. The thermal expansions can further
cause glitches in the turbine, leading to vibrations and uneven spinning. In a worst-
case scenario, a blade can loosen and penetrate through the casing.
The rotor in the steam turbine can be approximated as a solid cylinder of infinite
length [41], being heated on the surface, the permissible stress can be calculated
by using Equation (2.19), where k is a stress concentration factor taking actual
c
geometrical change of the rotor in consideration [16] and the axial stress, σ , is
z
assumed to be the same as the hoop stress.
E ·α ·∆T
exp
σ = k · (2.19)
z c
1−ν
where: σ = Axial stress [Pa]
z
k = Stress concentration factor
c
E = Young’s modulus for given material [Pa]
α = Coefficient of thermal expansion [°C−1]
exp
∆T = Rotor temperature difference [°C]
ν = Poisson’s ratio [-]
Based on the calculated maximum stress of each start-up type, the number of cycles
to failure on the specific stress level can be obtained. This is typically done by the
help of an S-N curve for the specific component material that is stating the correla-
tion between stress level and its number of cycles to failure.
Equivalent operating hours
Depending of the type of start-up, the start types require different ramping times
in order to reach the full power output of the steam turbine without exceeding the
permissible thermal stress levels. This means that a cold start, requiring a long
start-up time and causes higher stress levels than a hot start, will tear more on the
turbine. Equivalent operating hours (EOH) is then introduced as a concept that
supports lifetime calculations of components exposed to thermal stress by translat-
ing operating hours on various stress levels to an equal and comparable amount of
time in relation to how they wear the turbine.
InordertocalculateatotalnumberofEOH,individualEOH-numbersforeachstart-
up type are required. Further, the expected lifespan τ of the rotor is required as
life
input to those calculations, beneficially based on the manufacturer’s specifications.
From Aminov et al. [18] the total EOH is described as:
k y
X X
EOH = a n + b ψ (2.20)
i i j j
i=1 j=1
where k is the number of start-up types, a is the time coefficient of EOH for the
i
i-th start type and n is the scenario input on the number of start-up types. The
i
second summation part of Equation (2.20) is considering load changes where y is
the number of operating regimes, b is a coefficient for the operating regime of the
i
18 |
Chalmers University of Technology | 2. Theory on Production and Energy Systems
j-th load and ψ is the time in operation for the j-th load [18].
j
As the start-up types are causing various damage to the rotor due to their stress
amplitudes, it is of interest to introduce a cumulative damage model. A commonly
known model is the Palmgren’s-Miners’ rule for fatigue damage in combination with
Robinson’s hypothesis considering creep [12], see Equation (2.21). When the dam-
ageability parameter Z is reaching 1, failure commonly occurs, but it should be
denoted that failure mechanisms can appear earlier than Z = 1. In Equation (2.21),
the Palmgren’s-Miners’ rule is denoted by Z and Robinson’s by Z .
F C
k n r t
X i X j
Z = Z +Z = + (2.21)
F C
N t
i=1 i j=1 Bj
where N is the number of cycles to failure for the i-th start type, t is the time of
i j
a load at a given temperature and magnitude and t is the time to rupture at the
Bj
j-th load [12]. As the thesis is only focusing on fatigue caused by thermal stresses
from start-ups, the aspects of load changes and creep, b and Z will not further be
i C
explained, but are important to keep in mind. From Equation (2.20) we recall the
time coefficient a , defined by Aminov et al. [18] as:
i
τ 1
life
a = · (2.22)
i
Z N
i
Lastly, the total EOH can be divided by the expected years in service Y and give
the number of EOH on a yearly basis.
2.7.2 Maintenance on steam turbines
A steam turbine consists of some major components that are crucial for the tur-
bine functioning and are supported by sub-systems allowing the turbine to actually
work. Examples of major components are the rotating discs equipped with blades,
the shafts, the rotor, bearings, seals and casings [42]. These components are in gen-
eral exposed to several possible failure mechanisms such as corrosion, creep, fatigue
and stress cracking. As failures typically occur on different components, it is often
suitable to apply separate strategies for maintenance depending on the type of ma-
chine part. From the OEM, the turbine is often stated to be monitored based on
some specific measures such as the speed, load, pressures, temperatures and possibly
also vibrations and expansions [42]. And it is common nowadays that the OEM is
offering full maintenance programs for certain components or even the whole plant
by signing service contracts [32], [33], [43].
Maintenance on steam turbines is typically carried out based on inspection intervals
with frequencies varying from daily, weekly, monthly and annually intervals. Com-
monly spoken outages, where the steam turbine is stopped for a time and inspection,
reparation and replacements are taking place, are known as minor and major out-
ages occurring every 3-5 years and 9-12 years respectively or every 25 000 EOH and
100 000 EOH respectively [42].
19 |
Chalmers University of Technology | 2. Theory on Production and Energy Systems
A minor inspection on the steam turbine is a periodic inspection, related to the
maintenance policy TBM, due to its occurrence every 25 000 EOH. During this
outage, the turbine casing, bearing pedestals and generator remain closed, while
inspection is performed, on for example, the control system, control valves and the
turbo-set [41]. It is also possible to include boroscopic inspections and diagnostic
measurements. A minor inspection could typically require 2-7 days to perform.
AmajoroutageisamoreextensiveTBM-approachandaninspectionoccurringevery
100 000 EOH, often requiring 40-60 days of outage [41]. During this inspection,
basically all components and systems are inspected visually, diagnostically or by
testing. Casings are opened up and worn out parts are repaired or replaced, and
tolerances are checked. Imperfections are reviewed and the aim is to restore the
steam turbine back to perfect shape. These operations often require a large group
of personnel where it is not unusual that a representative from the OEM is present
during the maintenance operation and restoration [32], [41].
2.8 Cost calculations on maintenance
Maintenance can be evaluated and calculated by using various approaches and per-
spectives, typically based on the chosen maintenance concept approach. One of
the presented evaluation models by Lundgren et al. [27] is the Life Cycle Costing
analysis (LCC) approach evaluating the impact of maintenance from cradle to grave
and can also be designed in such way that it helps to evaluate whether maintenance
should be performed using internal or external service [27]. With the increasing
demand and requirements on sustainable energy production processes, a cost anal-
ysis model spanning over the whole life cycle is a suitable evaluation model. The
LCC-analysis could not only help to minimize the cost of maintenance by evaluating
strategies but also be adjusted towards a life cycle profit model, that could consider
a loss of income due to maintenance issues [27].
Reina et al. [30] present a maintenance planning software with calculations on how
to calculate LCC maintenance costs and eventually help to design a maintenance
plan based on the minimum cost. The costs are further divided into four types
of maintenance policies; failure-based maintenance (FBM), time-based maintenance
(TBM), condition-based maintenance (CBM) and opportunity-based maintenance
(OBM). Each policy, are then evaluated for both internal, external and external
contract maintenance services.
20 |
Chalmers University of Technology | 2. Theory on Production and Energy Systems
2.8.3 Condition-based maintenance
Condition-based maintenance can be calculated by using Equations (2.29)-(2.31) for
INT, EXT and FWC services [30]. When using a CBM-policy on a steam turbine
rotor, the machine must be equipped with sensors and other diagnostic tools that
can support the gathering of online performance data. The data must then be
analyzed, preferably in real time by a computer. By trend analysis and pre-set
limits, maintenance actions are triggered. This would mean that the policy could
have varying time intervals for which preventive maintenance tasks are carried out.
The aim is to predict failures to occur just in time before they happen.
L
INT
C = +TM ·C +C +F (2.29)
CBMINT
1/MTTF·0,75·p ·(1−p )
LP SP CBM
1 2
C = L ·TM +TM ·C +C +F (2.30)
CBMEXT EXT LP SP CBM
L /OH
FWC Y
C = +TM ·C +C (2.31)
CBMFWC
1/MTTF·0,75·p ·(1−p )
LP SP
1 2
2.8.4 Opportunity-based maintenance
Opportunity maintenance can be calculated by using Equations (2.32)-(2.34) for the
three labour services [30]. The OBM-policy is a suitable approach of maintenance
when there are opportunities arising periodically with enough time available for the
carryingthroughofpreventiveactions. Inthesteamturbinecase, theseopportunity-
windows could typically appear when the demand from the market is low and it can
becoveredbyotherpowerplants. Thesetimewindowscouldeventuallybepredicted
by using computers.
L
INT
C = +TM ·C +σ ·C +F (2.32)
OBMINT
1/t
LP SP SP OBM
C = L ·TM +TM ·C +σ ·C +F (2.33)
OBMEXT EXT LP SP SP OBM
L /OH
FWC Y
C = +TM ·C +σ ·C (2.34)
OBMFWC
1/t
LP FWC SP
22 |
Chalmers University of Technology | 3. Methodology
3.1 Development of methodology framework
Scenarios on
No. of start-ups
Transient simulation
Temperature differences Cost of FBM
Thermal stresses Stress-profiles Cost of CBM
LCC Results
Calculation of
Cycles to failure Cost of OBM
R(t), MTTF, MTTR
Calculation of
Calculation of EOH Cost of TBM
time coefficients
Steam turbine simulation EOH and Reliability Maintenance cost calculations Result presentation
Figure 3.1: Visual mapping of methodology framework.
Figure3.1illustratesthemappingofthemethodology, whichwasdevelopedbysplit-
ting it into four parts. These four parts are noted as the bottom elements in Figure
3.1 and are steam turbine simulation, reliability and EOH data, maintenance calcu-
lations and result presentation. The complete methodology development took place
in a Matlab-environment allowing construction of algorithms and mathematical
computations. The following sub-sections reflect the four presented elements of the
methodology.
The left top part of the figure was the starting point of the methodology, requesting
scenario input for the succeeding transient simulations of the steam turbine rotor.
From the simulation, the temperature differences were sought to be obtained for
which thermal stresses could be determined. Based on the calculated stress levels,
stress-profiles and the maximum number of cycles to failure were given.
When the numbers of cycles to failure and the stress-profiles were defined, calcula-
tions on time coefficients allowed calculations of EOH and reliability respectively.
The numbers of cycles are input to the LCC-calculations for assessing the FBM-,
TBM-, CBM- and OBM-policies for maintenance on a steam turbine rotor. Finally,
the cost model normalizes the results and present them in comparable numbers
(SEK/MWh).
The construction of Figure 3.1 and the methodology was conducted by perform-
ing literature searches on theory as well as studies on industry practice. Covered
24 |
Chalmers University of Technology | 3. Methodology
literature material were printed books, e-books, journal articles, web-pages, video
tutorials and grey literature. Grey literature included documents such as conference
abstracts, presentations, proceedings, regulatory data, unpublished data, govern-
ment publications and reports (such as white papers, working papers and internal
documentation). Major parts of the literature review were also serving as input for
the writing of the theoretical background.
3.1.1 Steam turbine simulation and thermal stresses
To investigate thermal stresses during a start-up, a thermo-mechanic stress model
of the rotor was made based on Can Gülen and Kim [39]. A one-dimensional model
of heat exchange in radial direction was adopted, where the rotor underwent a step-
change in surface temperature. The stress was calculated on the rotor surface by
applying the equations described in section 2.7.1. The thermal mass in the turbine
was modelled as a unit length cylinder with the radius r , a simplification described
0
by Grosso et al. [16]. The transient heat conduction was calculated by using Equa-
tion (2.17), where r∗, Bi and Fo varied both over time and radial location in the
rotor.
The steam properties at every time step were determined by a mathematical func-
tion which interpolated in steam tables, with pressure and temperature as the input.
This was needed for determination of HTC described in Equation (2.11). The Biot
number compared the convective heat transfer to the conductive (Eq. 2.16) and was
later used to determine the coefficients ζ and C , which solved the non-dimensional
n n
temperature difference θ∗ at every radial location at any time.
When the temperature gradient in the rotor was determined, the stress described
in Equation (2.19) was calculated. The only parameter in this expression that var-
ied over time was the rotor temperature difference ∆T, all other parameters were
assumed to be constant. The lowest values of the data on material propertieis were
chosen when running the transient simulation and calculations.
Based on the chosen material, a relationship between stress and number of cycles
to failure, also known as an S-N curve, was obtained from the material database
CES EduPack [44]. As CES EduPack provided the equation of the S-N curve, an
accurate calculation of the relationship with the available data was given as input,
for which an S-N curve could be constructed. The stress ratio included in the
equation was set to 0, which implied undirectional testing starting with zero stress
[45]. Byconsideringthemaximumstresslevels, theircorrespondingnumberofcycles
to failure, could be derived from the curve.
25 |
Chalmers University of Technology | 3. Methodology
3.1.2 Equivalent operating hours
The equivalent operating hours were calculated by following the approach presented
in the theory-section 2.7.1 based on Aminov et al. [18]. As the thesis only consid-
ered fatigue due to thermal stresses, variables involving creep were neglected. These
were b in Equation (2.20) and Z in Equation (2.21).
i C
With data assumptions on the expected lifespan of the rotor and expected years in
service, as well as scenario input on the number of ran cycles, the total number on
EOH was calculated by adding overall operating hours to Equation (2.20). Hence,
the equation was written as:
(cid:20) (cid:21)
EOH = OH + (a ·n )+(a ·n )+(a ·n )
cold cold warm warm hot hot
Lastly, the total EOH could be divided by the expected years in service, Y, to obtain
the yearly EOH. The calculated values were used as input to the computation of
maintenance costs.
3.1.3 Reliability due to fatigue
The maintenance cost calculations required values on reliability followed by MTTF
and MTTR in order to allow computation. As the rotor was exposed to three start-
up types, a stress model had to be designed considering varying stresses over a
period of time in order to give accurate results. Simple reliability calculations could
have been made by using only the i-th number of cycles to failure points from the
S-N curve, but would then have neglected the scenario input in the methodology,
which was an important factor.
By using the constructed S-N curve, followed by the thermal stress calculations, a
Weibull distribution fit was conducted on the data in the Matlab-environment.
The analysis gave the Weibull parameters β and η using a maximum likelihood esti-
mation. Thereafter, the parameters were used to calculate the individual reliability
of each start type over time using Equation (2.2).
As each start-up type corresponded to a certain reliability or probability of failure,
the three reliability functions were weighted according to the number of scenario
start types n . The cumulative reliability was then expressed as a weighted mean
i
value of each individual reliability curve according to Equation (3.1), where the
equation considered operational patterns and the damage caused by the start-up
types according to Palmgren’s-Miners’ rule (Eq. 2.21).
n R (t)+n R (t)+n R (t)
c c w w h h
R (t) = (3.1)
tot
n
tot
The damage or lifetime usage of the rotor could further be determined by using a
cumulative damage model. As the rotor is subject to damage due to both creep and
thermal fatigue, a summation method taking both measurements into consideration
is wise. Since the thesis only considered fatigue, creep was neglected. However, as
26 |
Chalmers University of Technology | 3. Methodology
Banaszkiewicz [12] shows, the creep life expenditure could consume 55% or more
of the rotor life, making creep an important factor to actually consider. Therefore,
it was decided to not let the scenario input to the methodology exceed a fatigue
damageability parameter of Z =35%. This also allowed a 10% safety margin until
F
the rotor would fail at Z = 1. The damageability parameter was calculated by
considering Z in Equation (2.21).
F
To obtain MTTF, the reliability was integrated over time as in Equation (2.3).
Thereafter, the correlation between MTTF, MTTR and availability in Equation
(2.4) were used to calculate MTTR. This required an assumption on the availability
of the steam turbine, and was set to a typcial number of 94% [46]. This meant that
6% of the cases when the steam turbine was wanted to be used, it was not possible
to use it due to a prohibiting event, for instance an unexpected failure. By rewriting
Equation (2.4), the MTTR was computed the following way:
1
MTTR = MTTF·( −1)
A
3.1.4 Maintenance cost calculations
The cost of maintenance was calculated and evaluated from an LCC-perspective for
the maintenance policies FBM, TBM, CBM and OBM with respect to the three
types of services INT, EXT and FWC. Equations (2.23)-(2.34) were used as the
cost models to calculate the maintenance costs. Input data to the calculations were
defined in section 3.2, along with computed numbers on reliability, MTTF, MTTR
and EOH.
Further, as the maintenance cost calculations on the TBM-policy required a time
intervalontheintendedinspectionintervals,thesewerecalculatedbyusingEquation
(3.2). It was determined to use the typical EOH-inspection interval in the industry
today,knownfromsection2.7.2,namelyevery25000-thEOH,inordertorecalculate
E to normal operating hours.
TBM
25 000 EOH
E = ·OH [h] (3.2)
TBM Y
EOH for each scenario
3.1.5 Result presentation
The last part of the methodology illustrates the element that would show and visu-
alize the results in a way that they would assist and support discussions. At first,
all calculated maintenance costs were normalized to SEK/MWh in order to create
reasonable numbers to compare. The normalization was done by taking each total
maintenance cost and divide it by the expected power output from a 150 MW steam
turbine during a 25-year period with 8000 yearly operating hours.
Thereafter, the maintenance costs were plotted in a 3D-plot with the maintenance
cost per produced MWh on the z-axis, the total number of start-ups in the specific
27 |
Chalmers University of Technology | 3. Methodology
scenario on the y-axis and the number of cold starts in the scenario on the x-axis.
This way of presenting the costs was believed to give an interesting indication of
how cold starts could impact on the maintenance cost, and how a shift of start type
could change the costs. The graphs were further recommended to also be supported
by tables containing numbers for easier interpretation.
3.2 Determination of data input
In order to compute and confirm the functioning of the developed methodology,
data had to be defined. This section presents the data that were needed in order
to perform the calculations. Each following subsection present the determined data
input for each part of the methodology framework along with explanations on how
the data was defined.
3.2.1 Material properties
The detailed properties on the material in steam turbines are often disclosed and
intellectual properties of the OEM. It is however commonly known that the rotor
material in a steam turbine is generally a steel alloy where the composition aims
to minimizing failures due to creep, fatigue, cracking, corrosion and erosion [47].
The chosen alloy for the thesis was a CrMoV-alloy (chromium alloy) suggested by
Osgerby [47] and also used in lifetime assessments by Banaszkiewicz [12]. This steel
alloy is suitable for structural parts in severe service conditions and is also known
as AISI 10 [44]. Table 3.1 presents the material properties on the alloy, that were
used in the transient start-up simulations and construction of the AISI 10 S-N curve,
where the values were obtained from the material database CES EduPack [44].
Table 3.1: Material properties for the chromium alloy AISI 10. Data was retrieved
from the material database CES EduPack [44].
Property Symbol Value Unit
Young’s modulus E 208 - 218 GPa
Yield strength σ 973 - 1740 MPa
YS
Tensile strength σ 1210 - 1980 MPa
TS
Elongation (cid:15) 6.6 - 15.5 % strain
Poisson’s ratio ν 0.285 - 0.295 -
Fatigue strength at 107 cycles - 292 - 522 MPa
Fatigue stress range - 235 - 648 MPa
Density ρ 7740 - 7890 kg/m3
Thermal conductivity λ 36.8 - 39.8 W/m.°C
Specific heat capacity C 292 - 522 J/kg.°C
p
Thermal expansion coefficient α 11.3 - 11.8 µ strain/°C
exp
28 |
Chalmers University of Technology | 3. Methodology
3.2.2 Steam turbine data
Three types of start-ups were simulated on a steam turbine. This created a need
for classifying what kind of start-up that was being run. The classification is of-
ten turbine-specific, but, usually, one categorizes the start-ups based on the metal
temperature in the rotor or more simplified after the number of stand-still duration
hours. For this thesis, Table 3.2 defined for which measured metal metal temper-
atures in the rotor that the respective start-up type belonged to. The start-up
conditions were defined accordingly to Grosso et al. [16] and Banszkiewicz [48].
The range of operating temperatures were confirmed by experienced people in the
industry based on conducted interviews [32] and concluded as suitable for a steam
turbine with a possible power output of approximately 150 MW.
Table 3.2: Start-up ramping conditions for cold, warm and hot steam turbine
start-ups.
Start type Metal temperature Stand-still duration Start-up time
i T - t
m i
Cold < 150 °C > 108 h 180 min
Warm 150 - 410 °C 16 - 108 h 105 min
Hot > 410 °C < 16 h 40 min
Cold start Warm start Hot start
100 100 100
50 50 50
0 0 0
0 50 100 150 0 50 100 0 20 40
100 100 100
50 50 50
0 0 0
0 50 100 150 0 50 100 0 20 40
600 600 600
500 500 500
400 400 400
0 50 100 150 0 50 100 0 20 40
Time [min] Time [min] Time [min]
Figure 3.2: Start-up curves for cold, warm and hot starts.
29
]s/gk[
wolf
ssaM
]rab[
erusserP
]C°[
erutarepmeT |
Chalmers University of Technology | 3. Methodology
The values in Table 3.2 were given as input to the transient simulations and calcula-
tions of the steam turbine, along with the needed start-ups curves presented visually
in Figure 3.2. The start-up curves described how each start-up was controlled by
keeping respect to permissible temperatures in the turbine, hence thermal stresses.
This was done by monitoring and controlling the inlet mass flow rate of steam, the
pressure and the temperature over time. At a real plant, this is monitored and
controlled by the use of a by-pass valve. Closer details on the specific turbine are
being disclosed.
Other important steam turbine data were the material properties on the rotor from
section 3.2.1, the dimensionless Biot numbers from Table 2.2 and an assumption on
the rotor radius. A rotor radius was defined as r = 0.3 m, a typical value according
0
to Can Gülen and Kim [39].
3.2.3 Equivalent operating hours data
The first step in the EOH calculations was to define the individual time coefficients
a for each start-up type. These calculations used numbers on the rotor lifespan
i
τ and the maximum number of cycles to failure of each start type, N . Data on
life i
the expected rotor lifespan was taken from Aminov et al. [18], while the numbers
of cycles to failure were given from the calculations on thermal stresses.
As the scope of the thesis only considered fatigue, no data was needed for creep
calculations. However, as the rotor should still consider the creep impact during its
lifetime, the damagebility parameter Z was set to 1 in Equation (2.21), for which
failure typically occurs after the exposure to both creep and fatigue [18]. Further,
for each start-up type i, a scenario input was required stating the expected number
of total starts during the lifetime n , as well as expected years in service, Y. The
i
number of start-ups are later defined in the scenario definitions, see section 3.3, and
the required data inputs for the EOH-calculations were:
Variables: τ = 200 000 [h]
life
Z = 1 [-]
Y = 25 [Years]
3.2.4 Maintenance cost data
The LCC-calculations on maintenance costs required numbers on calculated data
for reliability, MTTR, MTTF, EOH as well as input data on consuming costs and
time deviations. Table 3.3 is stating all the variables that were needed in the LCC-
calculations and the determined values. Argumentation on how they were defined
follow after the table.
30 |
Chalmers University of Technology | 3. Methodology
Table 3.3: Data input used for calculations of LCC maintenance costs.
Variable Value Unit Explanation
L 700 SEK/h Internal labour cost
INT
L 1400 SEK/h External labour cost
EXT
L 10 000 000 SEK External contract labour cost
FWC
F 137 000 SEK Fixed and consumable FBM cost
FBM
F 205 500 SEK Fixed and consumable TBM cost
TBM
F 274 000 SEK Fixed and consumable CBM cost
CBM
F 205 500 SEK Fixed and consumable OBM cost
OBM
C 5 000 000 SEK Cost of collateral damage
CD
C 30 000 SEK/h Cost of lost production
LP
C 5 000 000 SEK Cost of spare parts
SP
SDT 2160 h Delay time of spare part acquisition
MDT 1 h Mean delay time of internal response
int
MDT 168 h Mean delay time of external response
ext
σ 80 % Factor neglecting spare part change
SP
σ 75 % Warranty considerations
FWC
TM 0.75·MTTR h Time for maintenance intervention
ξ 100 % Interventions causing downtime
p 70 % Fraction of replacement decisions
1
p 5 % Fraction of unrecognized failures
2
Labour costs - L , L and L
INT EXT FWC
Two experienced internal workers were assumed to be working full time at the sce-
nario plant. They were assumed to be employed with a fixed monthly gross salary,
but together costing the plant owner 700 SEK per hour. For the external service, an
hourly fee of 1400 SEK/h was assumed to be billed to the plant for every MTTR-
hour when performing maintenance. And lastly, the external contract service was
expected to cost 10 000 000 SEK a year, as the contract service reduces the cost of
spare parts due to better warranty conditions (σ ) and do not have any spare
FWC
part delay times or affiliated fixed costs.
Fixed and consumable costs - F , F , F and F
FBM TBM CBM OBM
Keatly et al. [17] have created a methodology that forecasted the lifetime start costs
and considered long-term service agreement with minor and major overhaul costs
as their cost input. Kumar et al. [19] have on the other hand mapped the capital
and maintenance cost for different types of power plants which resulted in big cost
spreads among the same type of power output. The assumptions on costs for fixed
and consumable costs were hence a difficult one, as it is impacted by plant type, size,
configuration, age etc. [19]. However, an assumption had to be made in order to
make the computations possible. For the FBM-policy, fixed costs were expected to
be lower than the other policies as the policy is basically holding the spending until
something breaks down. When using the other maintenance policies, the costs were
expected to be higher as tools and consumables are continuously used due to its
preventive characteristics. Therefore, F was determined to be 1000 SEK/MW
FBM el
31 |
Chalmers University of Technology | 3. Methodology
for a 150 MW power plant. TBM and OBM occurring periodically were set to 1500
SEK/MW and for CBM it increased even more due to increasing consumables of
el
things such as lubrication, oil, seals etc. and was set to 2000 SEK/MW for the
el
same plant output. The costs were then multiplied with the expected yearly oper-
ating hours and expected yearly power output.
Consequential costs - C , C and C
CD LP SP
The cost of collateral damages, C , caused by failures was a difficult variable to es-
CD
timate since there are many circumstances impacting on how a failure could impact
dependable systems. Also, if a failure is detected early and an operator is shutting
down the process, the collateral damage may be less severe than a heavy failure
causing a ruined turbine. As these collateral damages could include everything from
broken bearings to broken turbine blades, the cost range is wide and was therefore
set to 5 000 000 SEK. The cost of lost production, C was calculated based on
LP
the possible total power output multiplied with the market price for electricity. It
was assumed that the producer was loosing 50% in revenue during unplanned dis-
ruptions while the other 50% were production costs. The current electricity price
was set by Nord Pool and approximated to be 400 SEK/MWh [49]. For a steam
turbine with a power output of 150 MW, C was therefore calculated to 30 000
LP
SEK/h. Another challenging variable to determine was the cost for spare parts,
C . When assumed that the steam turbine is running in such a way that the aim
SP
is to make it last the expected years in service, Y, the rotor must seldom be re-
placed. However, if replaced, the price could be several millions SEK depending on
turbine size and customization. Hence, the cost of for C was set to 5 000 000 SEK.
SP
Delay time variables - SDT, MDT and MDT
int ext
Asmoststeamturbinesaremoreorlesscustommade[32], itwasassumedthatthere
is rather limited availability to spare parts on the market. Long transports could
also have an impact on the delay time. The spare part delay time for an unforeseen
error SDT was therefore set to 3 months, 2160 h. Important to denote is that this
delay time can vary from case to case as companies can work differently with spare
part management and keep higher or lower levels of parts in stock [50]. The mean
delay time internally, MDT , for the response of the internal maintenance crew
int
was set to 1h, considering that the workers sometime can be busy with another
broken component, but sometimes also faster and can respond immediately to the
call. Considering MDT , it was assumed that the service company was having a
ext
remote location off the plant and also having more plants to maintain. That caused
the mean delay time to be set to seven days.
Deviation parameters - σ and σ
SP FWC
The deviation parameters are coefficients that are paying respect to the fact that
not all maintenance interventions necessarily need a spare part change. In case of
a service contract with the OEM, the coefficient σ also considers that spare
FWC
parts may be covered by warranties [30]. In case of internal and external mainte-
nance the coefficient σ , was assumed to be 0.8, meaning 80% of all interventions
SP
needed a spare part change. When using external contract service with warranty
32 |
Chalmers University of Technology | 3. Methodology
consideration, another 5% of the spare part changes were assumed to be covered by
warranties. σ was then defined as 0.75.
FWC
Time variables - TM, ξ, p and p
1 2
The TM-variable is the time needed to carry out a preventive maintenance task,
and can according to Reina et al. be estimated as 0.7-0.8 of the MTTR [30]. The
variable was hence determined as 75% of the MTTR. The ξ-variable is a coefficient
that considers that maintenance interventions can take place either by causing or
not causing downtime of the machine, in other words, a stop of the system [30]. It
was assumed that all of the failures were causing unplanned stoppages of the steam
turbine which needed interventions or tasks that could not be done without stopping
the machine. When using a CBM-policy, it must be taken into consideration that
the instruments cannot recognize all deviations and is hence creating monitoring
uncertainties. This introduced the variables p and p , where p is a coefficient
1 2 1
that considers the percentage of life the component in which it is possible to make
decisions about a components replacement [30]. This coefficient was assumed to be
high for a steam turbine, since they are often monitored by several instruments.
The coefficient p was set to 70%. The coefficient p is the small fraction of failures
1 2
that can occur without being recognized by the monitoring instruments [30]. As in
line with the previous argument and with instrument highly sensitive, this number
was assumed to be rather low and was set to 5%.
3.3 Scenarios and results analysis
In order to answer one of the objectives of the thesis, how the maintenance cost may
change with operational patterns, it was of interest to define scenarios reflecting the
expected future impact of increased shares of renewable energy sources [1]. The
scenarios were determined to reflect a base line scenario on the number of start-ups
and increases based on expected figures. All the defined scenario are presented in
Table 3.4.
Banaszkiewicz et al. analyzed the number of start-up at a 200 MW thermal power
plant during 43 387 real operating hours 267 start-ups occurred [11]. As they also
performed detailed analyzes on the casing temperature of the turbine, they were
also able to classify the start-ups into the start-up categories cold, warm and hot.
The numbers were taken for usage in this thesis and recalculated to match the de-
fined lifetime input τ . These number of starts were defined as scenario 1 (S1),
life
presented in Table 3.4.
In 2017, Schill et al. [51] analyzed how the start-up costs and the number of start-
ups of thermal power plants are expected to change with an increasing share of
non-dispatchable power generation. They also gave figures on the expected increas-
ing number of starts. It must be denoted that this analysis was made based on the
German market, where the primary product from the thermal plants is electricity
[32] and where the renewable generation is on the forefront since the country is
aiming at a drastic increase of renewable energy sources until 2050 [51]. With the
33 |
Chalmers University of Technology | 3. Methodology
number of starts in 2013 as a reference, to an expected scenario in 2020, Schill et
al. expected an increase of 4% in terms of number of starts. After another 10 years,
until 2030, an 81% increase was expected relative the case in 2013.
In this thesis, scenario 2 (S2) was then determined to represent an increase of 40%
as a case in between the 2013-case from Schill et al. [51] and their 81% expected
increase. Thereafter, scenario 3 (S3), was defined as a scenario with an 81% increase
relative S1. And as previously pointed out, it was of interest in the thesis to study
theimpactofcoldstartsandhowashiftfromacoldtohotstartwouldimpactonthe
maintenance cost. Hence, a fourth scenario (S4) was determined to match the exact
numberoftotalstartasinS3, butwithanotherfractionofcold, warmandhotstarts.
Lastly, another two scenarios were defined, scenario 5 (S5) and scenario 6 (S6). S5
was defined as one cold start every second month during a year and a hot start every
second day. These numbers are not based on any type of reference, but wanted to
be tested in relation to S6 to see how the methodology would respond when doing
half the cold starts from S5, but twice as many hot starts from S5.
Table 3.4: Scenario definitions in terms of number of start-ups.
Scenarios
Start type
S1 S2 S3 S4 S5 S6
i n n n n n n
i i i i i i
Cold 144 187 261 185 150 75
Warm 995 1294 1801 261 0 0
Hot 102 133 185 1801 4563 9125
Total 1241 1641 2247 2247 4713 9200
34 |
Chalmers University of Technology | 4. Results
4.1 Thermal stresses on the steam turbine rotor
The constructedrotor model computed thermal stresses with respect tothe dynamic
start-up conditions of the steam turbine. The results from the transient simulations
are presented in Table 4.1, as the maximum thermal stresses due to the temperature
gradient and when they occurred in time t after the start initiation. The last column
of the table is stating the maximum number of cycles to failure that was obtained
from the constructed S-N curve, see Figure 4.3.
Figure 4.1: The step changes of the temperature distributions of cold, warm and
hot start-ups. The distribution is shown for four points of the rotor.
Figure 4.1 presents the temperature distributions in four points of the rotor of a
cold, warm and hot start-up. The maximum temperature differences between the
rotor core and the surface caused the maximum thermal stresses in the rotor. For
the presented start-ups, the largest temperature difference between the core and the
surface occurred when performing a cold start. At t = 18 min, after start initiating
the resulting maximum stress was calculated to 421 MPa. How each of the start-up
stresses were varying in the rotor is shown in Figure 4.2. These calculated tempera-
ture differences and stress levels corresponded well to similar calculations performed
by Grosso et al. [16] and Casella and Pretolani [52].
36 |
Chalmers University of Technology | 4. Results
The constructed S-N curve of the alloy material was used to plot the calculated
stresses, σ , and hence obtain the maximum cycles to failure for each start type,
max
N . The obtained cycles are shown in Figure 4.3 as well as in Table 4.1. It can
i
further be concluded that 1 cold start, corresponds approximately to 10 warm or
100 hot starts. This is in line with what Venkatesh et al. [14] stated based on
experience, strengthening that the rotor model is likely to perform accurately.
When changing the scenario in the rotor model, no changes were done on the tem-
perature distributions, stresses or maximum cycles to failure. This was due to the
unchanged process parameter number that were not included in the scenarios. How-
ever, if letting the ramping times, temperatures or pressures in the steam turbine
vary, the resulting changes in the rotor model would create different results. This
opens up to discussions of a more advanced methodology.
Table 4.1: Results of the transient start-up simulations with the maximum tem-
perature differences, the maximum stresses and when they occurred in time after
start initiation.
Maximum Maximum Cycles to
Start type Time
temp. diff. stress failure
i ∆T σ t N
maxi maxi i
Cold 125 °C 421 MPa 120 min 885
Warm 108 °C 366 MPa 75 min 9786
Hot 80 °C 318 MPa 28 min 101 265
4.2 Equivalent operating hours calculations
The first step in the EOH-calculations was to determine the time coefficient a for
i
each start-up type. With the defined input data and the calculated results on N ,
i
the following time coefficients were obtained:
200 000 1
a = · = 226 h
cold
1 885
200 000 1
a = · = 20 h
warm
1 9786
200 000 1
a = · = 2 h
hot
1 101 265
These calculated numbers on EOH can be compared to the actual start times pre-
sented in Table 3.2. From a comparison, it can be concluded that the EOH-times
are significantly larger. The EOH-values were also confirmed in interviews to be of
reasonable size [32] compared to real data, but it must be denoted that the calcu-
lations are highly depended on specific steam turbines. The calculations of EOH
for each scenario resulted in a maximum number of hours of continuous operation,
which are presented in Table 4.2.
38 |
Chalmers University of Technology | 4. Results
Table 4.2: The operating hours for continuous operation for each scenario.
Scenarios
Variable
S1 S2 S3 S4 S5 S6
OH [h] 147 352 131 592 104 624 149 368 156 974 164 800
OH [h] 5 894 5 264 4 185 5 975 6 279 6 592
Y
The individual time coefficients remain the same throughout the methodology, and
hence the total EOH-values as well, unless no changes would have been made in
the transient simulations. In a real turbine, where the ramping conditions are easily
varied, the outcome would likely be different and could result in varying EOH-values
for each unique start.
4.3 Reliability analysis
TheresultsoftheperformedanalysisoftheWeibulldistributiongavetheparameters
presented in Table 4.3. It can be seen in the table that the scale parameters varys,
depending on the maximum number of cycles to failure N , but are close to similar
i
for the shape parameter β. This can be explained by the same type of failure
distribution, only having varying stress levels.
Table 4.3: Calculated Weibull distribution parameters of a cold, warm and hot
start up, based on their number of cycles to failure.
Start type Shape parameter Scale parameter
i β η
Cold 1.6262 489.2
Warm 1.6190 5399.8
Hot 1.6182 55866.1
39 |
Chalmers University of Technology | 4. Results
Reliability plot
100
Cold
90 Warm
Hot
80
70
60
50
40
30
20
10
0
100 101 102 103 104 105 106 107
Cycles [-]
Figure 4.4: Reliability curves for each start-up type.
With the reliability of each start-up type plotted individually shown in Figure 4.4, it
can be seen that the reliability is decreasing earlier for a cold start than for a warm
or hot start-up, for the same number on N. This is in line with the hypothesis of
shorter lifetime expenditure the rotor is suffering, due to the higher stress levels of a
cold start. As the rotor is impacted by different reliability characteristics, an overall
reliability was calculated by considering the scenario input as a weighted contribu-
tion. Considering the MTTF originating from the reliability, it was expressed in
terms of cycles, referring to start cycles until the rotor fails without a specification
on failure cause. Therefore, MTTF do not consider neither impact from stops or
continuous operation. The results on reliability for each scenario are presented in
Table 4.4.
Table 4.4: Numbers on reliability, MTTF and life usage of each scenario after the
total number of cycles.
Cycles No. of start-type Reliability MTTF Life usage
Scenario
n C W H R(t = n ) Cycles Z
tot tot F
1 1241 144 995 102 81.42 % 8 041 26.54 %
2 1641 187 1294 133 77.82 % 8 051 34.48 %
3 2247 261 1801 185 71.12 % 8 047 48.08 %
4 2247 185 261 1801 88.83 % 40 703 25.35 %
5 4713 150 0 4563 95.06 % 48 459 21.46 %
6 9200 75 0 9125 93.97 % 49 633 17.49 %
40
]%[
ytilibaileR |
Chalmers University of Technology | 4. Results
When analyzing the results in Table 4.4, starting with S1 to S4, one can see a de-
creasing trend in reliability from S1 to S2 to S3. In terms of MTTF, the explanation
on the varying results is caused by rounded values. Since MTTF is a measure de-
fined as the integral of R(t) from 0 to ∞, the difference lies in the shape of the curve
creating the integrated area. Rounding up or down the values from the scenario
input could hence impact on the difference in MTTF. The range from 8041 cycles
for S1 to 8051 cycles for S2, is therefore not considered as faulty. Looking at fatigue
life usage Z , the usage in S3 is exceeding the set threshold of Z = 35%. The
F F
chance of failure in combination with the discarded creep, is now significant and one
must suggestively investigate this scenario closer in more aspects than just fatigue.
In S4, when swapping numbers of starts from S3, the life usage drops to a lower
value than S1, due to its lesser number of warm starts and only slightly increased
number of cold starts. The reliability is higher than in any of S1, S2 or S3 and
the measure on MTTF has five folded to 40 703 cycles. This clearly shows how the
different types of starts is impacting on the rotor lifetime.
In case of S5 and S6, the reliability is dropping from 95.05% for S5 to 93.97% for S6.
Even though the number of cold starts is cut to half, one must keep in mind that
the total number of starts in the scenario has almost been doubled. If considering
S6 and stepping back in its scenario history, one would obtain a higher value of
reliability when N would be 4713. That explains the higher number on MTTF. In
termsoffatiguelifeusageZ , thelowervalueinS6issimplyduetotheshifttowards
F
more hot starts, impacting less on the rotor rupture than cold starts.
4.4 LCC-calculations on maintenance costs
For the equations using a time-based maintenance (TBM) policy, the inspection
interval, E , was calculated for each scenario and resulted in intervals ranging
TBM
from 16 883 to 21 279 hours. In the equations on an opportunity-based maintenance
(OBM) policy, there was instead a t-variable present. With the variable defined as
the interval of an opportunity window, it was defined as t = 8000 for the cases with
relatively low MTFF (8000 cycles) and t = 16000 for cases with higher MTTF. This
reflects a risen opportunity once per year or once every second year.
The results on the LCC-computations are presented in Table 4.5. The table presents
each scenario by expressing the maintenance cost in SEK/MWh. As the method-
ology evaluates four maintenance policies and three labour services, there are 72
possible combinations of the costs. It must further be pointed out, that if the cal-
culated scenario exceeded the threshold of Z = 35%, the cost of spare parts was
F
ten-folded as the threshold limit was considered as a failure. This meant that such
a scenario would increase the cost of spare parts significantly. Table 4.5 is followed
by visualization of the given numbers in 3D-plotted graphs in Figure 4.5, 4.6 and
4.7.
41 |
Chalmers University of Technology | 4. Results
Table 4.5: LCCresultsformaintenanceexpressedinSEK/MWhforinternal(INT),
external (EXT) and contract maintenance (FWC), applied to the four policies:
failure- (FBM), time- (TBM), condition- (CBM) and opportunity-based mainte-
nance (OBM).
Maintenance costs [SEK/MWh]
Labour Policy
S1 S2 S3 S4 S5 S6
INT FBM 42.71 42.68 58.21 25.92 27.03 27.62
TBM 13.12 13.66 28.78 25.31 22.72 21.29
CBM 14.32 14.32 34.96 12.77 12.88 12.93
OBM 25.55 25.56 55.56 30.40 27.90 26.78
EXT FBM 41.04 41.01 56.55 22.88 24.08 24.71
TBM 8.78 9.34 24.50 21.54 18.84 17.34
CBM 12.44 12.43 33.08 10.89 11.00 11.05
OBM 21.69 21.71 51.70 26.85 24.24 23.07
FWC FBM 25.78 25.77 40.33 21.98 22.24 22.37
TBM 16.63 17.16 31.38 28.82 26.24 24.82
CBM 16.02 16.02 36.66 14.56 14.65 14.71
OBM 28.83 28.85 56.97 33.88 31.38 30.26
When analyzing the values in Table 4.5, starting with S1 to S4, it is of interest
to remember the analysis and conclusions derived from Table 4.4. It must also be
remembered that S1 is the base line scenario and that S3 is exceeding the Z = 35%
threshold. Comparing S1 to S2 for all labour and policy types, there is no significant
trend noted as the proportion of starts remain the same after the 40% increase and
both scenarios make sure not to trespass life usage. For S3, the spare part cost
is ten-folded for all labour and policy cases and it can be seen that the cost per
MWh has radically increased. In terms of S4, there are interesting changes taking
place and when comparing S4 against S3 as well as S1, the difference in the chosen
numbers of start-up types becomes present. With a higher number of hot starts in
S4 than the other cases and rather low numbers on cold and warm stars, an FBM-
policy becomes the cheapest in an S4-case if it would be of interest to be using.
When comparing the TBM-policy among first four scenarios, it is seen to be more
expensive for S4, which could be explained by the higher reliability from Table 4.4
and interventions will become unnecessary and lead to extra costs even for a long
time after the rotor investment.
Considering the two scenarios S5 and S6, both of the scenarios are making the Z-
limit and there is actually not a big difference among the cost for the policies relative
to each other for the two scenarios. From Table 4.5 in combination with Table 4.4,
one can determine that a scenario with more cold starts, S5, would cost almost the
same as a scenario with half as many cold starts but twice as many hot starts, S6.
It is therefore possible to double the number of starts and almost keeping the same
maintenance cost by just adjusting the proportion of the number of start types.
This is in line with severity notation by Venkatesh et al. stating that 1 cold start
42 |
Chalmers University of Technology | 4. Results
equals 100 hot starts [14].
To conclude from Table 4.5, an FBM-policy would generally be the most expensive,
with a contribution by high collateral damage costs. It can, however, be seen that
when having a higher portion of hot starts and less cold, the FBM-policy decreases
in expensiveness and becomes more interesting. One must though remember the
possible consequences that could occur if letting the machinery just run until it fails
leading to unforeseen production shutdowns, eventual collateral damages, long spare
part deliveries etc. It must also be pointed out that the presented costs are only
overhead maintenance costs and do not vary with the changed number of starts. Ac-
cording to Schill et al. an increased number of starts by 81% is expected to lead to
increased start costs of 119% [51]. An interesting case would be to investigate that
correlation to maintenance. Further, as theory presented, preventive maintenance
is known to be cheaper than reactive [7] and would actually be able to increase the
numbers on the reliability after an intervention has been performed. In the current
methodology, this is discarded as the reliability is untouched after an intervention,
leading to shorter lifetimes than it actually should. This could most certainly be
impacting on the maintenance costs.
Internal maintenance
FBM TBM CBM OBM
60
50
40
30
20
10
0
S6
10000 S3
S5
8000
S4 300
S2
6000 250
S1
4000 200
150
2000
100
Total starts [-] 0 50 Cold starts [-]
Figure 4.5: A visual comparison of the internal maintenance costs for FBM-,
TBM-, CBM- and OBM-policies. The maintenance cost is expressed in SEK/MWh
on the z-axis, the total number of starts in the scenario on the y-axis and lastly the
number of cold starts in the scenario on the x-axis. The scenarios are indicated by
the vertical lines and pointed out using the prefix S.
Figure 4.5 supports the numbers presented in Table 4.5 and shows how the internal
43
]hWM/KES[
tsoc
ecnanetniaM |
Chalmers University of Technology | 4. Results
Maintenance with framework contract
FBM TBM CBM OBM
60
50
40
30
20
10
0
S6
10000 S3
S5
8000
S4 300
S2
6000 250
S1
4000 200
150
2000
100
Total starts [-] 0 50 Cold starts [-]
Figure 4.7: A visual comparison of maintenance costs using a framework contract
for FBM-, TBM-, CBM- and OBM-policies.
Looking at Figure 4.6, it can be seen that the costs for external maintenance policies
are close to the internal, making it hard to distinguish any new trends based on the
graphics. When having the supporting numbers on hand from Table 4.5, the small
difference between internal and external maintenance service can be explained by
analyzing the equations in section 2.8. It can then be concluded that the only dif-
ference between internal and external maintenance is the labour costs and the time
at the site.
For maintenance service based on a framework contract, the costs are changing com-
pared to the other cases, as the equations are defined slightly different. This might
be, since Figure 4.7 shows that the OBM-policy is the most expensive policy due to
the fact that maintenance is carried out no matter if it is necessary or not. With a
contract, this creates forced high costs. A CBM-policy is among the cheapest poli-
cies to use as the contracted company is responsible for monitoring and replacing
parts whenever it is needed. This also impacts on the previously expensive FBM-
policy with the other services, since the cost of a failed machinery is taken on by
the contracted company, becomes the contracted company’s cost.
On the bottom line, any conclusions will not be drawn in terms of suggested mainte-
nance based on the obtained results. The developed methodology and tool aims to
support the discussions on maintenance where many possible discussion points have
been highlighted. Even though it could be possible to conclude which scenario, pol-
icy and labour service should be recommended to be used, there are many external
45
]hWM/KES[
tsoc
ecnanetniaM |
Chalmers University of Technology | 5
Discussion
The developed methodology shows how an operating parameter like the number of
starts may be used to estimate the lifetime expenditure of the rotor, and how it
relates to the maintenance costs. The method could, thus, be used for guidance in
design of new operational modes for thermal power plants for more volatile future
energy market scenarios. Further more, the view on maintenance of production sys-
temshasbeenseenasadevelopmentovertime,previouslybeinganecessaryevil,but
now a strategic question [7], the discussions on how and when maintenance should
be performed is a key driver in a plant strategy [8]. Not only will maintenance help
to increase the reliability of the plant resources, such as the steam turbine rotor, but
could also help to gain economic advantages by decreasing the chances of unwanted
downtime.
A relevant notation in the discussion of this thesis is whether or not an expected
change in the number of starts is relevant for all markets. The increased share of
renewable energy sources will probably only affect the operational pattern at ther-
mal power plants in markets where electricity is the primary product. As for an
instance, most plants in Sweden are focusing on the delivery of heat as primary
product [32], this is what drives the plants revenue and not the demand on elec-
tricity. The generated electricity at these plants is more of a bi-product and the
plants are therefore not interested to schedule their operations accordingly in the
sameextent, especiallyinashorttermperspective. Thisthesisworkwouldtherefore
be more suitable for markets like Denmark, Germany and the Netherlands where
electricity is what drives the revenue at many thermal power plants and where the
operational pattern of the plants will be significantly affected by the increasing share
of renewables.
5.1 Methodology and sensitivity
The results shows how a scenario with a different numbers of start-ups connects to
rupture of the rotor in the steam turbine. With an increasing need of start and
stops and possible load ramps due to varying net loads, there will be changes in
the maintenance costs for power plant owners. Section 4.4 elaborated on how the
costs in an LCC-perspective would change based on the scenarios, but as there are
many input parameters to consider, one must also be aware of the possibly varying
results from case to case. Therefore, note that the aim of the thesis has never been
to suggest how future maintenance should be performed at power plants, but to act
as a supporting tool in discussions on how to evaluate different perspectives.
47 |
Chalmers University of Technology | 5. Discussion
All the defined input data originates in some form from assumptions as described in
the methodology chapter. Some assumptions are more reasonable than others since
some are highly challenging to estimate without insight. A comparable example
of this is how internal salaries for internal maintenance labour is easier to assume
rather than costs for collateral damages or spare parts in case of rotor failures on
a steam turbine in the considered size. The costs can range from a couple of thou-
sands SEK to millions depending on the severity. Furthermore the turbine might be
custom made by the OEM and spare part can take a year to design, produce and
deliver in a worst case scenario [32]. For the results, this could mean miss-calculated
costsindicatingthewrongcorrelationbetweenoperationalpatternandmaintenance.
In addition, material properties of the rotor could vary over its axis due to different
geometrical dimensions, certainly impacting on the results. In the rotor model, the
rotor was defined as a cylinder of infinite length for simplifications, but would prefer-
ably be modelled in a FEM-software with its real manufactured dimensions. That
would increase the accuracy of the results even more as it would consider a more
advanced mathematical relationship. This could also help to evaluate both tensile
and compressing stresses in geometric complex locations. By changing some of the
material properties, one could expect other results in temperature differences in the
rotormodel,impactingthestressesandeventuallytheresultsonmaintenancencosts.
When considering how the methodology is computing the input data, only fatigue
due to start-ups was considered. In a more extensive model with more detailed sce-
narios, there would be reasonable to add support for the impact of creep impacted
by the temperature difference as well as support for the impact of load changes
on the turbine due to the more volatile operational patterns. This could be done
today by changing the start-up times, where the temperature differences would be
even bigger resulting in even more severe thermal stresses. This would then lead to
an even shorter lifetime with sustained numbers on start-ups. This is a case likely
case in the future when a sudden drop could appear in the non-dispatchable power
generation.
The fact that preventive interventions are performed on the steam turbine rotor was
alsodiscardedinthemethodology. Inarealcase, aprecautionaryactionwouldlikely
contribute to an extended lifetime and hence an increase in reliability by elevating
the reliability curve at the current time. Also, as Schill et al. [51] present in their
work on how renewable energy sources are expected to impact on increased start-up
costs, questions on increased maintenance costs would also be of interest. Today,
the methodology is not considering an increase in the assumed overhead costs. This
is becoming clear when assessing S1 with S2 and one can see that the cost per is
almost the same. Hence, questions are raised that varying costs could be applied to
the maintenance cost data.
In a management perspective, it is also of interest to ask oneself what is wanted. In
the methodology today the cost of maintenance is calculated in a life cycle costing
48 |
Chalmers University of Technology | 5. Discussion
perspective where the goal is to consider all the costs affiliated with maintenance
during the rotor lifetime. How the results are then used is a management ques-
tion. Do one want to minimize the costs and use the cheapest policy? Or does
one want to minimize the number of interventions by just monitor the health on a
computer screen? Or is it of interest to discard the costs and just focus on maximum
availability by taking every possible chance to increase or maintain the reliability
of the rotor? It could even be the case that reliability is the key, if the plant is
working close to industrial production, for example the oil industry, and losing one
day of production would mean the cost of several rotors. These are questions that
the methodology is not considering as of today and could be possible work for the
future, preparing for how to plan, schedule and optimize the maintenance. Another
maintenance concept, rather than LCC would also be of interest to use, for the eval-
uation of maintenance and is often sometimes based on a company level. Perhaps
the change towards an even more dynamic electricity market would have more use of
a life cycle profit approach, where one is estimating how the equipment can instead
contribute to profit [27]. The importance of the maintenance strategy would hence
depend on the role of the power plant in the energy system.
However, the methodology successfully manages to provide results based on the
available input data and is indicating the general knowledge in the industry, that
preventive maintenance is often cheaper than reactive maintenance [7]. It is how-
everhardtotellwhatchangingoperationalpatternswouldmeanintermsofpossible
changesinmaintenance. Today, mostplantownersareworkingactivelywithpreven-
tive maintenance on a daily, weekly, monthly and yearly basis, often with directions
and schedules from the OEMs [32], [33]. It is often common that the plant own-
ers are using frameworks with contract maintenance where expertise is brought in
on revision periods for performing and leading maintenance on set intervals. This
shows strong applications of TBM-policy usage in the industry, and at the same
time having CBM-policies for which hundreds of sensors are monitoring the turbines
in parameters such as temperatures, pressures and most importantly vibrations [32].
The way electricity is consumed as a product is also constraining how maintenance
of the power producing equipment can be performed. Any other type of typically
produced product, is most likely able to be put on delay and let the customer wait
for delivery if the producing equipment breaks down or has to be maintained. But
as electricity cannot be stored at a reasonable cost today, the production must con-
stantly meet the demand and the customer or consumer cannot be put on hold in
the same way. This is creating a need of regarding maintenance of power plants
connected on the same grid in a holistic perspective, as all plants cannot shut down
their power production at the same time. There must be detailed planning and
scheduling behind the decisions, needing a time variable behind the planning. Cer-
tainly, the machinery can also be condition monitored, but if all power plants are
running FBM-policies and break down at the same time, there would not only be
consequences for each plant but for substantial parts of our society since it is relying
heavily on the power grid. This strongly highlights the need for usage of the de-
veloped tool with clearly defined market conditions along with its production from
49 |
Chalmers University of Technology | 5. Discussion
different energy sources.
5.2 Experience for future work
The rotor model could eventually have been developed by the help of a dynamic
simulation model in a software like Dymola. In this project, it was investigated to
doitwithoutincludingtheentiresteamcycle, butitwasdiscoveredthatforaproper
Dymola simulation, the turbine model required good implementation of boundary
conditions, or having a close-cycle model with control structures and bypass valves,
which was too challenging without close plant collaboration or knowledge of the
exact process configuration. The work on a dynamic model would however be rec-
ommended to be done in future work by following previous research approaches
[53]–[55], in order to define more complex scenarios containing more parameters and
get a more extensive picture of the start-up results.
For this thesis, if was concluded that the usage of a more simple start-up rotor
model would be enough to create a transient simulation based on work by Can
Gülen and Kim [39]. The developed rotor model considered temperatures, pressures
and times, which could possibly be changed due to different scenarios by the user
with the goal to match other types of load ramps and start-ups than those defined
for the specifically considered turbine. This was however not included in the de-
fined scenarios for the thesis. Further, the current transient model did not vary the
Nu-, Re- or Pr-numbers which would be the case in a more mathematical complex
model. This type of model could then have been for instance made in Dymola and
in the combination of an FEM tool. But the simplifications that were made in the
thesis, are still of such character that they are describing the rotor start-ups and the
relationships taking place, and are certainly good enough for the scope of this thesis.
There was also an initial thought on calculating reliability using a more complicated
stress model with an inverse power law relationship using an underlying Weibull
distribution. This was tried out by the usage of ReliaSoft ALTA Pro, but was
eventually discarded as there was not time to find a solution creating a seamless
integration from the transient simulations and later over to the cost model. Hence,
it was decided to develop the complete methodology in Matlab even though spe-
cialized software could have given more accurate results for the different models. An
even more dynamic rotor model could have also considered dynamic life limits of the
turbinebygivingconditionmonitoredparametersasinputtoreliabilitycalculations.
Another challenge in the project was the acquisition of data. Most of the data was
collected from other research projects where collaborations had been conducted with
power plant owners. From the interviews, it was learned that before the 2000s, an
international organization used to gather data on maintenance performance from
power plant members all across the world and later printed publications for all
organization members. This type of data was not found in the literature review
performed in this project, but could have helped to gain more interesting insights.
50 |
Chalmers University of Technology | 5. Discussion
The interviews also concluded that more and more maintenance is being handed
over to the OEMs [32], which most likely don’t want to disclose data to the public
nor competitors, as it can be important information for their research and develop-
ment. No conclusions have been drawn based on this, but would be in line with the
evolving view of shifting maintenance issues into a strategic question [7]. Another
interesting insight from the interviews, was the view on external maintenance in the
industry. The current trend is that maintenance is carried out either internally or
by the OEM, especially when it comes to critical parts as the steam turbine [32],
[33]. So even if the LCC-results in the thesis pointed towards external maintenance,
this might not be preferred in the reality.
When Bokrantz et al. studied how maintenance is expected to change in the manu-
facturing industry until 2030, using Delphi-based scenarios, it was found with high
probability that the evolution of digitalized manufacturing will introduce a high
presence of data collection and the usage of data analytics and big data manage-
ment [56]. It was expected that maintenance organizations will use the data to
identify patterns, root causes and make decisions, eventually using support decision
systems. An example of this is the developed digital twin by DecisionLab on a
Siemens Gas Turbine, named the Agent-based Turbine Operations & Maintenance
model. The digital model monitors maintenance repair and overhaul of Siemens’
aero-derivative gas turbine division using live data available in their supply chain
[57] and helps to do better data-driven decision makings and to predict KPIs as well
as overwatch maintenance operations.
With this amount of data gathering and pattern identification, it would also be
possible to connect events and scenarios to variable maintenance costs and see how
the costs follows the scenarios and turbine usage more dynamically and accurately
rather than using overhead costs. Though this requires a close collaboration with
industry.
51 |
Chalmers University of Technology | 6
Conclusion
This master’s thesis has developed a method to relate the thermal stresses caused
by frequent starts and stops to maintenance costs of a steam turbine rotor by con-
sidering temperature differences in the rotor and relating them to lifetime and costs.
The development of the method is based on previous work experience from industry.
This thesis shows that the cost of maintenance can be estimated depending on the
operational pattern and maintenance policies, constrained by scenario prerequisites.
The developed methodology is considered possible to be used for initial testings in
real cases but must be reviewed and given real input data for accurate validation.
The application can be made for plants on those markets where electricity is driving
the revenue and not heat. In an even further extent, the purpose of the methodology
is also believed to eventually be achieved by enabling cost efficiency in the power
industry.
Furthermore, from the application of the proposed method to a set of scenarios three
main conclusions may be drawn:
1. A higher number of starts impact the steam turbine rotor maintenance. The
extent of the impact depends on what type of starts that are initiated on the
turbine. A cold start causes more wear on the rotor and will contribute to a
faster rupture of the rotor, which results in higher maintenance costs - 1 cold
start equals almost 100 hot starts. The number of cold starts is, therefore,
important to consider if one wants to aim for a minimized maintenance cost.
2. Itisgenerallyknownthatpreventivemaintenanceispreferredtoreactivemain-
tenance, where collateral damages could be serious and extremely costly when
considering a steam turbine. Also, in case of a framework contract service, the
opportunity-based maintenance policy is most expensive. This could be ex-
plained by the original equipment manufacturer often wants to perform more
maintenance than the plant owner is expecting.
3. The method is supporting the industry standard of today. It is common that
the original equipment manufacturer is involved in the maintenance service,
providing a vast experience on how to run, care and repair the machines. This
is typically done on time-based intervals and decision from condition-based
parameters. As the turbine machinery is a rather complex and expensive
system with demands on high availability and reliability, the cost for lost
production in high peak seasons is devastating for plant owners.
53 |
Chalmers University of Technology | 7
Future work
The first step in future work is suggested to be the questioning of what is really
wanted to be obtained from the methodology, asking questions like; What should
the application of the methodology be able to indicate? How does one want to
use the results from the application of the methodology? Is there something in the
methodology today that is not in line with future desires? Secondly, the authors
are emphasizing the need for accessing data for proper input. If possible, it is also
recommended to validate the results from the rotor and cost model in terms of the
degree of accuracy and relate those results to the previously discussed areas of im-
provements. Then, prioritization can be done on what must first be adjusted before
continuing the development of the methodology.
With the expected volatile and dynamic operational patterns, the authors are also
suggesting the development of the methodology to become more dynamic and to al-
lowmorechangesofparametersovertime. Itshouldthenbepossibletocomputethe
results cumulatively and to use more extensive boundary conditions for the whole
process. Assuming a severe scenario is given as input, resulting in a break down
earlier than the expected rotor lifetime. At the moment, the developed methodology
will not consider those cases, other than increasing the spare part costs. In reality,
this would likely impact more critically and it must therefore be scenarios in the
methodology on how to handle such cases. An example would be if a new rotor is
purchased after a breakdown and the reliability must then be reset.
As the developed methodology also focuses on the start-ups of a steam turbine, a
more extensive and dynamic simulation model could help to investigate other crit-
ical components of the production process that are of interest. This could help to
identify the approach on maintenance strategies not only for the steam turbine rotor
but also the whole plant.
The methodology is also suggested to be implemented with simple planning and
scheduling algorithms, that can predict suitable maintenance windows based on
conditions such as forecasted demand, prices and production from renewable energy
sources, as well as planned stops in other power plant, fuel prices and full company
economic insights. This could be done by using economic dispatch models where
wind speed, solar radiation and power demand could be used as random input
variables. Whenandifinvolvingthistypeofamoreextensivedynamicandeconomic
support, thecostofmaintenancewouldbepossibletorelatetotheproducedproduct
inamoreprecisewayandcalculatedregardinghowitvarieswithchangedconditions.
This would also create a good support for the identification of KPI:s and how they
55 |
Chalmers University of Technology | Data management
The data gathered from the interview will be used as a background to the master’s thesis work
at Chalmers University of Technology. The interview will be stored as a personal file and will
only be published if the respondent gives his/her consent.
No personal data will be stored if the respondent does not give his/her consent. Personal data
will not be shared.
Ethics
This information is to ensure the respondent about his/her rights during the interview.
The respondent has no obligation to complete the interview and have the possibility to leave at
any time. He or she have the possibility to retrieve all answers from the interview, if he or she
wish so. There are no mandatory questions during the interview, which means that respondent
choses which questions he or she answers. If the interview exceeds the stated time, the
interviewer will ask the respondent to continue the interview. If the answer is no, the interview
will end.
General information
The aim of the project is to develop a first assessment method on how the maintenance cost
would be varying based on changed start and stop of a steam turbine. This will be done by
combining theory and methodologies from the fields of energy technology and production
engineering, that will help to simulate, estimate and understanding the relation between
maintenance and operational patterns in combined heat and power plants with a narrow focus
on the steam turbine exposed to thermal stresses. The relation is assessed in terms of life
cycle cost (LCC) for different maintenance policies and services related to different energy
market scenarios.
The purpose of the development of the method is sought to be a beginning of a supporting tool
that can help initial discussion of different maintenance strategies at power plants.
The interview is expected to take approximately 30 minutes. |
Chalmers University of Technology | Interview questions.
The following questions are to be answered by the respondent over phone and answers are
noted by the interviewer on a separate paper.
1. Background questions
a. Name
b. Current work tasks
c. For how long have you been working at the plant/in the industry?
2. General plant questions (if applicable, otherwise give from experience)
a. Please provide a short description of your plant process.
i. Size (MWth and MWel)
ii. Fuel type
iii. Main product
iv. Process configuration (GT, ST, HRSG etc.)
b. How is the plant operationally running over a year?
i. Load types?
ii. Revenue optimized for heat production or electricity?
c. What is the yearly average operating hours of your plant?
d. How many starts and stops of the plant do you perform on a yearly basis?
e. How many starts and stops do you perform in the categorization of cold, warm and
hot starts on a yearly basis?
i. How are they defined?
3. General maintenance questions (if applicable, otherwise give from experience)
a. How are you working with maintenance of the plant today in general?
b. How are you scheduling and planning preventive maintenance tasks?
i. How are you working with maintenance during operational periods?
Especially during the winter.
4. Maintenance of the steam turbine questions (if applicable, otherwise give from
experience)
a. Which are the major crucial components in the steam turbine to monitor?
b. How are starts and stops of the steam turbine correlated to lifetime expenditure of the
steam turbine and the previously mentioned components?
c. How are you working with maintenance of the steam turbine?
d. How are you scheduling and planning maintenance activities?
e. Does a change in the number of start and stops impact the planning of maintenance?
f. Is the concept of Equivalent Operating Hours (EOH) familiar when considering
maintenance of the steam turbine? If yes, how are you using it?
g. With societal expectations on more volatile electricity prices in the future causing
more cyclic operations, how do you expect this to impact the mechanical parts in the
steam turbine?
i. Have you already noticed changes due to changing operational patterns? |
Colorado School of Mines | ABSTRACT
Face ignitions at the longwall are a serious hazard in underground coal operations and can
lead to a major mine explosion. Despite having methane monitoring systems mounted on the
shearer and at various locations on the longwall face, undetected methane accumulations can still
occur and result in face ignitions. With the use of Computational Fluid Dynamics (CFD), the
interaction between the air flow at the longwall face and factors that contribute to accumulations
around the face can be modeled and visualized in great detail. The results confirm that the tailgate
corner of the longwall face is a critical area prone to face ignitions and thus needs to be properly
monitored. Roof falls at the tailgate entry inby the face and/or poor caving conditions behind the
shields can both pose a safety risk at any longwall operation. Poor gob caving can lead to
insufficient face air quantity with which to dilute methane at the tailgate corner, while a blocking
of the tailgate by a roof fall can carry methane-contaminated air from behind the shields back into
the face near the tailgate corner and pull the explosive gas zones (EGZs) inside the gob and closer
to the face. Additional monitoring locations are deemed necessary to provide early indicators for
such events.
iii |
Colorado School of Mines | ACKNOWLEDGEMENTS
I would like to thank my advisor, Dr. Jürgen Brune for the opportunity to be part of the
NIOSH-funded project that he led. I am very grateful for his knowledge sharing, guidance, and
continuous support of my efforts to complete my thesis. I would also like to thank Dr. Gregory
Bogin for all the input and discussions to help me better understand the computational fluid
dynamics part of this study. Special thanks to Dr. John Grubb, who initiated the first phase of this
research project and sharing his valuable knowledge and experience in longwall mining. This
research would not have been possible without the financial support of the National Institute for
Occupational Safety and Health (NIOSH) under contract number 211-2014-60050.
I would also like to thank all my predecessors in this project, Dan Worrall, Elizabeth Wachel,
Jon Marts, Richard Gilmore and Saqib Saki, all of whom had make invaluable contribution in
completing the early CFD models in this research area. Thanks also for my fellow graduate
students, Samuel Lolon, Claire Strebinger and Matt Fig for the inputs and research discussions.
Thanks also to the rest of the CSM Mining Engineering Department Faculty and staff for their
support. Finally, I wish to thank my family and friends for their continuous support during my
study.
xii |
Colorado School of Mines | 1 CHAPTER 1
INTRODUCTION
1.1 Introduction to longwall mining
Longwall mining is the most commonly used underground coal mining method in the United
States; in 2013, it contributed 59% of the nation’s total underground coal output (U.S. Energy
Information Administration, 2015). Longwall mining is best suited for relatively flat seams with
uniform thickness and few discontinuities. Compared to the conventional room and pillar mining
method, longwall mining provides higher productivity and recovery rates. This method utilizes a
high degree of mechanization at the requirement of high capital cost and less geological flexibility.
The mining process begins by developing two entries of access into the target coal seam.
These accesses can be in the form of drift for shallower coal seams or a shaft in the cases of deeper,
thicker seams. After this, a series of entries called main headings are developed using one or more
continuous miners (CMs) and connected by series of crosscuts, leaving behind chains of pillars
that are used for roof support. These main headings serve as major transport arteries for equipment,
workers and mined coal while also ventilating the mine. In a exhaust ventilation system, a
ventilation shaft will be sunk early in the development to carry return air from the mine. To support
that, ventilation controls are installed on several crosscuts to separate the headings: intake entries
to deliver fresh air from surface, and return entries to deliver contaminated air out of the mine.
From the main headings, a set of gateroads are developed perpendicular to the mains to form a
large coal panel. Figure 1.1 shows the schematic of longwall mining.
The coal panel is extracted in increments of 3 to 4 feet using a longwall shearer; the drums of
the shearer move from the panel’s rear to the front towards the main headings, also called retreat
mining. Because of that, access to the rear of the longwall panel must be obtained prior to mining,
and this is achieved by constructing a set of gateroads at both sides of the panel mined with the
continuous miner. These gateroads are connected by a series of crosscuts and leaving pillars of
coal called chain pillars to support these openings. The gateroad adjacent to the previously mined
longwall panel is called the tailgate, the other the headgate; thus, the tailgate of the current panel
was the headgate of the previous panel.
1 |
Colorado School of Mines | Figure 1.1: Schematic of longwall mining (Consol Energy, 2016)
All headgate and tailgates are connected to the main headings, where fresh air is distributed
and exhaust air leaves the mine. Fresh air is supplied to the face, where mine personnel operate
the shearer and retreat the shields. Depending on the ventilation design, the contaminated air (often
called bad air) leaving the longwall face is exhausted to the surface either via a main exhaust shaft
in the main headings or through a bleeder fan at the back of the panel. At the rear of the longwall
panel, a start drift called the startup room must also be excavated; the start drive acts as the working
face and provide space for the shearer, shields and other equipment to be installed. At the other
end of the panel, a barrier pillar is left behind, shown as a dashed line in Figure 1.2. This pillar will
not be extracted and will serve to support the main headings.
A longwall panel typically ranges from 10,000 feet to 20,000 ft in length and from 800 ft to
1,500 ft in width. From a productivity and recovery perspective, the panel should be as wide and
as long as possible. However, panel length and width are restricted by ground stability, geological
disturbances such as major faults, or the potential of roof cave. The height of a typical longwall-
mined coal seam ranges from 9 ft to 11 ft in the western United States, and 5 ft to 8 ft in the eastern
United States (Gilmore, 2015). Factors such as coal quality and the size of mining equipment can
influence height. Figure 1.2 shows the overview of a longwall face.
2 |
Colorado School of Mines | After being mined, the coal is then transferred to the crusher located at the face’s headgate
side via the AFC. From here, the conveyer belt transports the crushed coal to the surface. As coal
is extracted and the longwall face advances, the immediate roof behind the shields collapses and
forms a caved zone known as the gob. The gob consists of various sizes of rock ranging from large
boulders to finely crushed gravel, all from the failed overlying roof material that fills the void of
the mined-out coal seam. Figure 1.4 shows the formation of the gob as the longwall retreats.
Figure 1.4: Representation of an active longwall panel and the formation of the gob (Karacan,
2008)
According to a study by Esterhuizen and Karacan (2005), gob height typically ranges from
three to six times the mining height and is characterized by relatively high permeability and void
ratios ranging from 30%-45%. Directly above the gob, there is a disturbed zone, the fracture zone,
which is characterized by having near-vertical fractures and bedding plane shearing. This zone can
extend up to 60 times the mining height. The region above the fracture zone, the bending zone,
receives minimal disturbance that results in rock that is not fractured but is deflected over the edges
of the extracted panel. The extent of each zone is site-specific and highly dependent on surrounding
geology. Figure 1.5 shows a cross section view of a caved longwall panel.
In addition to the surface disturbance, panel extraction also causes rock mass disturbances in
the mine floor. Although the extent of floor disturbance is significantly less compared to surface
disturbance, in some cases it can result in floor heave producing floor gas emissions or even a gas
outburst. The gob, because it is filled with rubble and has an unstable roof, is considered an
4 |
Colorado School of Mines | inaccessible area. As a result, taking direct measurements or monitoring inside the gob is nearly
impossible.
Figure 1.5: Long section view of a caved longwall panel (Esterhuizen and Karacan, 2005)
1.2 Motivation for Longwall Mining Ventilation Research
Longwall face ignitions from accumulated methane gas are known to be among the most
common causes of methane explosions at underground coal operations. Some of these ignitions
can lead to major mine explosions, such as the Upper Big Branch mine disaster in 2010. This
incident highlights the importance of understanding airflow and methane distribution, especially
at the tailgate of the longwall face, and the impact that roof falls in the tailgate entry inby the face
can have on tailgate ventilation and methane accumulations.
The objective of this research is to utilize Computational Fluid Dynamics (CFD) modeling to
identify critical areas near the longwall tailgate where ventilation must be closely monitored, and
from those, develop general best practices for longwall face ventilation and methane monitoring.
This was achieved by developing a detailed CFD model to analyze airflow in and around the
longwall face, analyzing the methane (CH ) distribution and explosibility near the tailgate for a
4
bleeder ventilation system, and predicting the impact of tailgate roof falls on ventilation and
methane distribution at the face and inside the gob.
5 |
Colorado School of Mines | 2 CHAPTER 2
BACKGROUND OF THE STUDY
2.1 Regulatory requirements for ventilation in U.S. longwall mining
In the United States, ventilation requirements for underground coal operations are regulated
under the Code of Federal Regulations (CFR) and enforced by the Mine Safety and Health
Administration (MSHA). For longwall operations, 30 CFR Part 75.325 requires that at least 30,000
cubic feet per minute (cfm) of fresh air is supplied to the longwall face, unless the mine operator
can demonstrate that a lesser air quantity would be sufficient to maintain methane and respirable
dust below prescribed limits. The ventilation plan, which must be approved by MSHA, must
include air velocity measurement locations; the specified locations must be at least 50 ft but not
more than 100 ft from the headgate and tailgate.
30 CFR Part 75.321 requires the mine operator to maintain a minimum of 19.5 percent oxygen
and not more than 0.5 percent carbon dioxide in areas where persons work or travel. The supplied
air must also be sufficient to dilute, render harmless and carry away flammable, explosive, noxious
and harmful gases, dusts, smoke and fumes. For the maximum allowable methane concentrations,
30 CFR (§75.323) requires operators to make the necessary ventilation adjustments to reduce
methane concentrations to below the prescribed limit when:
1.0 percent or more methane is present in a working area or intake air course, including a
belt conveyor air course or in an area where mechanized mining equipment is being
installed or removed;
1.0 percent or more methane is present in a return air split between the last working area
on an active section and where that split of air meets another split of air, or the location at
which the split is used to ventilate seals or worked-out areas;
1.5 percent or more methane is present in a return air split between a point in the return
opposite the section loading point and where that split of air meets another split of air, or
where the split of air is used to ventilate seals or worked-out areas; and
2.0 percent or more methane is present in a bleeder split of air immediately before the air
in the split joins another split of air, or in a return air course other than as described above.
6 |
Colorado School of Mines | The regulation also requires all longwall operators to use bleeder systems unless it can be
demonstrated that the mined coal is prone to spontaneous combustion. In that case, the operator is
allowed to use a progressive sealed method known as U-type ventilation. In the U.S., almost all
longwall operations utilize bleeder systems and only few mines are known to use the progressive
sealed system. Figure 2.1 shows a comparison of ventilation layouts for bleeder systems and U-
type systems.
Figure 2.1: Longwall ventilation types: Bleeder (left) and U-type (right) (Grubb, 2008)
In a bleeder design, fresh air is provided from the headgate entries and splits towards the belt
entry, longwall face and inby the face. Some of this air is expected to pass through the gob and
dilute the methane inside before merging again at the bleeder entries at the back of the panel and
exhaust to the surface through the bleeder fan.
In U-type ventilation, the gob area is sealed progressively by constructing seals in the
crosscuts along the headgate entries as the longwall face advances; this method also utilizes
nitrogen injection to inertize gob and gob ventilation boreholes (GVB) to reduce gob methane.
Although rarely used in United States, U-type ventilation is commonly used in other mining-rich
countries such as Australia and South Africa.
2.2 Source of methane inflow at an active longwall panel
In a longwall operation, methane may come from three main locations: the seam being mined,
lower coal seams, and the upper seam (the rider seam). Lidin (1961), Thakur (1981), Winter
(1975), Gunther and Belin (1967) have all conducted studies to determine the extent of gas
emissions from surrounding gas sources to the working seam, as shown in Figure 2.2.
7 |
Colorado School of Mines | Figure 2.3: Explosibility of methane-air mixtures (Coward and Jones, 1952)
According to the graph, the mixtures become explosive when 12-20% oxygen and 5.3-15%
methane are present; this is known as the explosive range. In this case, 15% CH is considered the
4
upper explosive limit (UEL) and 5.3% CH is considered the lower explosive limit (LEL). Outside
4
these limits, the mixture is either fuel-rich (for mixtures containing above 15% CH ), where the
4
addition of oxygen can shift the mixture back into the explosive range, or in the fuel-lean state (for
mixtures containing below 5% CH ), where it can no longer form an explosive mixture if more
4
oxygen is added. This explosibility range is used as a reference by MSHA in 30 CFR.
In addition to methane, other flammable gases are commonly found in mine gas mixtures,
such as ethane, hydrogen and carbon monoxide. The addition of other flammable gases to the air
can change the LEL of the methane-air mixtures. The explosive limits of these flammable gas
mixtures can be calculated using Le Chatelier’s Law shown below:
Where P1 + P2 + PX = 100. The number #1, #2, #X represent gas mixtures of gas #1, gas
#2, and up through gas #X. L is the lower explosive limit of the mixture, P is the proportion of
⋅⋅⋅⋅
9 |
Colorado School of Mines | each gas in the mixture, and L1, L2, and LX are the lower explosive limits for each combustible
gas (Jones, 1929).
According to this law, if one gas mixture at its lower explosive limit is added to another gas
mixture, also at its lower explosive limit, then the combination of the mixtures will be at the lower
explosive limit of the combination. However, this equation can only be used if each flammable
gas concentration is known.
Other factors such as pressure and temperature can also change the explosibility limits of
methane-air mixtures. A study done by Zabetakis (1965) shows that temperature changes only
have slight impacts on the explosibility limits of methane. For example, at room temperature,
methane-air mixtures have a LEL and UEL of around 5% and 15%, respectively. The LEL will
change to 5.6% methane at -100°C (-148°F) and to 4.8% methane at 100°C (212°F), while the
UEL of methane-air mixtures will change to 16.3% methane at 100°C. Conversely, a study done
by Kuchta (1985) shows that a change in pressure has a more significant impact on explosibility
limits of methane. Figure 2.4 shows the variations in methane LEL and UEL with increases in
pressure.
Figure 2.4: Effect of elevated pressure on methane explosibility limits (Kuchta, 1985)
The result shows that methane-air mixture explosibility limits vary slightly with reductions in
pressure, except at very low pressures. However, at higher pressures, the UEL increases greatly
while the LEL decreases slightly.
10 |
Colorado School of Mines | 2.4 Hybrid mixtures of methane and coal dust
Normally, methane gas has a lower flammable limit (LFL) of 5% methane in methane-air
mixtures. However, a test done by Cashdollar et al. (1987) shows that the addition of dispersed
coal dust to methane-air mixtures can reduce the mixture’s LFL. Further laboratory tests with
various coal dusts mixed with methane gas confirmed this relationship (Cashdollar, 1996). Figure
2.5 shows changes in methane flammability limits due to the addition of dispersed coal dust.
Figure 2.5: Flammable limits for mixtures of methane and coal dust (Cashdollar, 1987)
The area on the right and above the dashed line represents flammable mixtures. Based on the
graph, without the addition of dispersed coal dust, the LFL of methane is 5%. As more coal dust
is added, this LFL drops on a linear trend until the coal dust concentration reaches 0.1 oz/ft3, where
the coal dust concentration itself is enough to produce ignition without the addition of methane
gas.
2.5 Ignition risks around the longwall face
Mine explosions are commonly caused by poorly designed ventilation systems, insufficient
ventilation and inadequate monitoring in critical areas that are prone to methane accumulations. In
longwall operations, the tailgate corner of the longwall face can be considered one such critical
area.
Verma and Brune (2016) summarized the sources responsible for face ignitions between 1983
and 2014 based on data from the U.S. Mine Safety and Health Administration (MSHA). The
sources include spontaneous combustion, electrical equipment and switchgear, mechanical heat
11 |
Colorado School of Mines | mine disasters began as small face ignitions before transitioning into coal dust explosions, such as
the Upper Big Branch mine explosion in 2010 that killed 29 miners.
Although rare, ignitions due to rock-on-rock friction caused by roof falls caving into the
longwall gob are possible and can also lead to a major mine explosion, such as the Willow Creek
mine explosion in 1998 and 2000 (Elkins et al., 2001; McKinney et al., 2001). This type of face
ignition is difficult to prevent, as they are generally located in inaccessible areas and caused by
ignition sources that can’t be controlled. Regardless of the ignition’s occurrence, each frictional
ignition has the potential to cause a mine explosion.
2.6 Factors that can lead to methane accumulation in the tailgate area
The main purpose of bleeder ventilation is to dilute methane gas and prevent the formation of
explosive gas mixtures in critical areas, including the tailgate corner. However, accounts from real
operations, along with experiments and simulation studies by various researchers, have identified
several factors that can lead to methane accumulation in the tailgate area, and these need to be
properly addressed.
2.6.1 Insufficient face air
Providing proper ventilation to the tailgate side of the longwall face can be a challenge. Thakur
(2006) stated that, in order to be considered adequate, face ventilation should fulfill three criteria:
the ability of fresh air to keep methane concentrations below statutory limits; preventing gas
layering; and the adequacy of air leaking into the gob to push the explosive methane-air mixture
away from the gob area immediately behind the longwall face where the active roof fall occurrs
(typically 30m-45m (100-150ft) behind the shields). Figure 2.8 shows the result of a site study
done in one operating longwall panel with a seam thickness of 1.5m to 1.8m (4.9ft to 5.9ft)
showing the impact of a wider face on the leakage rate across the face and methane emissions at
the tail end of the face.
As the longwall face became wider, the air quantity that managed to reach the tail end of the
face decreased while higher methane emissions were detected at the tail end of the face. This means
that, due to extensive leakage and the amount of methane emissions produced along the face, there
will be a limit when it is no longer possible to supply enough face air to meet the statutory limit.
13 |
Colorado School of Mines | Figure 2.10: Cumulative face methane emissions for three days of monitoring (Schatzel, 2012)
Overall, higher face methane emissions were recorded during the head-to-tail passes, with the
highest methane emissions recorded on day three. Although the results were site-specific, it clearly
shows the relation between longer face lengths and higher cumulative methane emissions at the
tailgate corner.
In addition to face emission increases, caving conditions behind the face should also be a
major concern. Syd Peng (1984) conducted on-site experiments in West Virginia coal mines to
study air velocity distribution on mechanized longwall faces. The experiments were conducted
using smoke tubes at four longwall faces, each with different roof conditions and roof-caving
compactness in the gob. Several surveying stations were set up along the longwall face, starting
from headgate entry T-junction until the third support from the tailgate entry T-junction.
The results show that roof-caving conditions inside the gob have a significant impact on the
extent of air leakage. Figure 2.11 below shows air quantity distribution along the longwall face for
four different panels.
15 |
Colorado School of Mines | Figure 2.11: Air quantity distribution along a longwall face (Peng, 1983).
Panels 1 and 3 represent a case of longwall faces with a relatively stable roof and good caving
conditions behind the shields. In both panels, the immediate roof fell periodically with short delay
after the shields moved forward. As a result, there is no significant air leakage from the face area
into the gob, and most of the fresh air supplied to the longwall face managed to reach the last
survey station near the tailgate corner.
Panels 2 and 4 represent a case of longwall faces with stable roofs and poor caving condition
behind the shields. On Panel 2, the immediate roof in the gob was not tightly compacted and there
was a visible space above the rock pile. The open space allowed the majority of the face air to leak
into the gob and resulted in only 20%-40% of fresh air supplied to the face managing to reach the
tailgate. In the case of Panel 4, the immediate roof in the middle of the panel caved properly right
after support advance; however, large areas on both the headgate and tailgate ends did not cave
completely and resulted in void areas ranging from 13-30m (43ft-100ft) long and 6-10m (20ft-
33ft) deep in the gob. As a result, about 30%-50% of the fresh air supplied to the longwall leaked
into the gob near the headgate corner. Some of this leaked air then returned to the face, in the area
16 |
Colorado School of Mines | where the immediate roof is tightly caved, and continued to gradually leak back into the gob as it
approached the poorly caved area near the tailgate corner. By the time it reached the tailgate corner,
the majority of the air supplied to the face had already leaked into the gob.
Krog et al. (2014) conducted a study on airflow patterns along longwall faces in bleeder
ventilated systems. Three tracer gas tests were conducted at different gas release locations and
bleeder setups. Several sampling locations measuring arrival times and gas concentrations were
set across the longwall face inby the shield line. The tracer gas containing 99.95% sulfur
hexafluoride (SF ) was released at the start of the longwall section in tests 1 and 2. Both tests were
6
done on the same longwall panel that utilizes the t-split bleeder ventilation system. Figure 2.12
shows the schematic of a longwall district and ventilation system experiment setup for tests 1 and
2.
Figure 2.12: Sample tube placement for test 1 (left) and test 2 (right) (Krog et al, 2014)
The result of the first test shows that about half of the airflow measured at the tailgate corner
came from the face, while the remainder came from behind the shields. A similar trend was
observed during the second test, with 60% of total air supplied to the face flowing from behind the
shields. In the third test, the tracer gas was released on the longwall face at shield 19. Unlike the
previous two, the third test was done on a different longwall panel utilizing an internal bleeder
system where most of the tailgate airflow was directed towards the main return and then diverted
towards the bleeder system. Figure 2.13 shows the schematic of the longwall district and
ventilation system experiment setup for test 3.
17 |
Colorado School of Mines | Figure 2.13: Sample tube placement for test 3 (Krog et al, 2014)
Even with a different tailgate setup, the result of the third test also showed a similar airflow
pattern across the longwall face. Half of the airflow moving away from the longwall tailgate corner
and towards the back bleeders did not come from the longwall face, but rather flowed from behind
the shields. These tests clearly showed that significant air leakage across the face is common in
longwall bleeder ventilation systems.
2.6.2 Barometric pressure changes
Changes in barometric pressure, whether it is a gradual daily fluctuation or an abrupt change
such as with a case of stormy weather, can have an impact on methane outflow from the
surrounding coal strata and inside the gob. Previous studies done by various researchers (McIntosh,
1957; Boyer, 1964; Kissell et al., 1973; Fauconnier, 1992; Belle, 2014; Wasilewski, 2014; Lolon
et al., 2015) have identified the relationship between changes in barometric pressure and the
outgassing of methane in underground coal operations. A rise in barometric pressure may push
fresh air into the gob, also known as the gob “breathing in,” while the fall in external barometric
pressures may cause a potential outflow of methane gas from the gob, also known as gob
18 |
Colorado School of Mines | “breathing out.” These phenomena are believed to have contributed to several underground coal
mine explosions cases in the United States.
A computational fluid dynamic (CFD) modeling study done by Lolon et al. (2016) on the
effect of barometric pressure changes on methane distribution inside longwall gob shows that
fluctuations of barometric pressure outside the mine can affect the size and location of explosive
gas zones (EGZs) in bleeder ventilated longwall gobs. An increase of barometric pressure will
result in more ingression of bleeder and face air into the gob, diluting methane and reducing the
EGZ volume inside the gob. Figure 2.14 shows the change of EGZ distribution inside the gob due
to barometric pressure increases.
Figure 2.14: EGZ transformation during rising barometric pressures (Lolon et al., 2016).
An increase in outside pressure induces more ingression of fresh air into the gob from the
surrounding longwall face and bleeder entries. As a result, the EGZ inside the gob is pushed further
toward the center of the gob as the barometric pressure continues to rise. Generally, the rise of
barometric pressure will not pose an issue to the operation’s safety; however, when barometric
pressures drop, the methane inside the gob will expand into the adjacent bleeder entries,
accompanied by more methane inflow strata. This methane outgassing would also push the
explosive gas zones closer to the longwall face, as seen in Figure 2.15.
19 |
Colorado School of Mines | Figure 2.15: EGZ transformation during falling barometric pressures (Lolon et al., 2016)
In addition to changes in barometric pressure, sudden change in mine pressure due to local
events such as roof falls in critical areas, can also affect methane flow in and out of the gob.
2.6.3 Tailgate ventilation setup
Yuan et al. (2012) conducted a study on ventilation flow paths in longwall gobs using CFD.
Three different ventilation systems were compared, which included one-entry and two-entry
bleederless systems and a three-entry bleeder system.
The modeled longwall panel was 3,300ft long, 1,000ft wide and 164ft high. Porous media
with a permeability of 9x10-8 m2 was used to represent the longwall face shields and the longwall
gob interior was modeled as five zones with permeability ranging from to 7x 0-7 m2 around the
gob edge to 2x10-8 m2 for the gob center. The result in Figure 2.16 shows that, in bleederless
ventilation systems, the flow through the gob was mainly concentrated behind the shields and the
air that leaked through the shields at the headgate side was forced back into the face near the
tailgate side.
20 |
Colorado School of Mines | 2.7 Case study: UBB mine explosion April 5, 2010
The Upper Big Branch mine (UBB) was owned by Massey Energy and operated by
Performance Coal Company. The mine was located in Montcoal, Raleigh County, West Virginia.
It utilized the longwall mining method and produced approximately 41.4 million tons of coal from
1994 to 2010.
On April 5, 2010, at 3:02 p.m., an explosion occurred that killed 29 miners and injured two
others. At the time of the explosion, the active panel was producing coal from the Eagle seam and
had already advanced about 6,000ft from the panel startup room. The seam itself was only 26-40
inches in thickness, at times separated into two benches by a sandstone parting several inches thick.
Due to equipment height limitations, the mine cut 2.5ft of sandstone roof in addition to the 4.5ft
coal seam, resulting in total mining height of 7 feet. The longwall face was 1,000ft wide and was
supported by 176 Joy twin-leg shields. A double drum shearer was used to cut the coal face, where
a bi-directional cutting method was utilized. The average mine entry was 7ft high by 19ft wide.
2.7.1 The UBB ventilation system
The UBB mine was ventilated using three fans installed at the surface, as shown in Figure
2.18. Two of them were blowing fans installed at drift openings of the North and South portals,
while the third fan was an exhaust fan installed atop a 16-foot diameter airshaft located near
Bandytown. Combined, the three fans supplied around 1,014,000 cfm of fresh air to the mine.
Figure 2.18: UBB ventilation layout at the time of the mine explosion in 2010 (Phillips, 2012)
22 |
Colorado School of Mines | The amount of airflow supplied to the face the day of the explosion was significantly lower
than the typical rate. Investigators concluded that the T-split did not provide enough airflow to
safely dilute the methane released prior to the localized methane explosion.
2.7.2 Changes in tailgate ventilation
It is believed that one of the main contributors to the explosion was a change in tailgate
ventilation control due to a tight roof fall blocking the tailgate entry inby the face. Figure 2.21
shows the condition of the tailgate entry inby the longwall face after being blocked by the roof
fall. Although the picture was taken during the mine explosion investigation, several pieces of
evidence were found by investigators; for example, small pieces of freshly fallen white sandstone
and coatings of soot on the fallen rubble both indicate that the intersection had caved prior to the
explosion.
Figure 2.21: Condition of the tailgate entry inby the face after a roof fall occurred (Phillips,
2012)
Investigators determined that inadequate roof control was the cause of the roof fall in the
tailgate entry about 45ft inby the face. The support requirement-based roof control plan involved
installing two rows of posts or two 8ft cable bolts in the tailgate entry as supplemental support.
However, the investigation found that only one row of posts was installed in the tailgate entry. In
addition to inadequate roof control, an analysis done by MSHA on UBB’s coal pillar design
indicated that the pillars did not meet NIOSH recommendations, which specifically outlined a
Pillar Stability Factor (PSF) of 1.13 requiring a 125ft crosscut and entry centers. Instead, the mine
had 100ft crosscuts on 95ft--105ft centers which yielded a Pillar Stability Factor of only 0.82.
24 |
Colorado School of Mines | The mine’s roof fall restricted the face air access to the bleeder entry, thus forcing the operator
to partially open the crosscut stopping outby the face to provide new access to the bleeder entry.
Figure 2.22 below shows an illustration of the ventilation changes in the tailgate T-split area at the
UBB mine.
Figure 2.22: Changes in UBB’s tailgate ventilation due to a roof fall (Phillips, 2012)
2.7.3 Possible ignition source
A post-incident investigation suggested that the explosion started as a methane gas explosion
at the longwall tailgate T-split area before transitioning to a coal dust explosion. The ignition
source was believed to be frictional impact, as the shearer was cutting sandstone roof, or by rock
colliding with steel supports (or other rock) while falling from the sandstone roof behind the
longwall shield; the former was deemed the most likely.
On the day of the explosion, the investigation report indicates that significant methane gas
with flow rate of several hundred cubic feet per minute was released from a series of 81 floor
fractures between shields 160 and 170, when the longwall mined past a fault zone. Due to tailgate
ventilation changes, the gas that was supposed to flow towards the back of the panel was instead
pulled toward the tailgate corner and was believed to have interacted with the shearer’s tailgate
drum.
The shearer used at the UBB mine was equipped with a methane monitor to provide constant
methane readings. Another methane monitor sensor was located under the tail drive motor covers
at the tailgate side of the face. After the explosion, both sensors were examined, but neither sensor
reported any methane reading exceeding regulatory limits prior to the explosion. However, the
report also indicated that the shearer was not equipped with sprays behind the bits, and at least two
25 |
Colorado School of Mines | shearer bits on the tail side cutting drum showed signs of severe wear to both the bit and carbide
tip. A further inspection also found that seven sprays were missing from the shearer tail drum,
creating inadequate dust and explosion control.
During the investigation following the explosion, MSHA conducted smoke tests near the
tailgate of UBB’s longwall to replicate the tailgate condition prior to the explosion and determine
the airflow path around this area that lead to a face ignition by the shearer tailgate drum. Figure
2.23 shows the test setup along with the location of the shearer and methane sensors.
Figure 2.23: Diagram of UBB’s longwall face showing locations of the methane sensors
(Phillips, 2012)
The tests were conducted by releasing smoke from behind shields 160, 170 and 176 separately.
The first test location was behind shield 160, where the smoke traveled downwind behind the
shields until it reached an area where the gob had fallen tight against the shields, near shield 164,
before migrating from behind the shields out into the walkway and continuing to travel over the
shearer, tailgate drive and methane monitor sensors. When the smoke was released from behind
shield 170, it traveled behind the shields until it reached shield 173. From this location, some of
the smoke traveled behind the shields and into the tailgate entry, while the rest of the smoke came
out of the shields into the walkway and traveled toward the shearer tailgate drum. The test
conducted from behind shield 176 produced a similar result. In both tests, the smoke managed to
travel across the tailgate drum of the shearer without passing the two methane sensors.
Although both the longwall face and ventilation conditions at the time of the test were different
than in the moments prior to explosion, this experiment provided a good indication of the existence
of a flow path behind the shields that allowed methane gas from the gob to reach the shearer tailgate
26 |
Colorado School of Mines | drum without passing the two methane sensors. These results are in agreement with several studies
done regarding airflow path across a longwall face in a bleeder ventilated panel that have shown
that flow paths around gobs are concentrated mainly where gob permeability is the highest, which
includes the area behind the shields. This result further suggests that the combination of
insufficient face air and tailgate ventilation changes allow undetected methane to accumulate near
the tailgate corner.
Brune and Sapko (2012) conducted computational fluid dynamics (CFD) simulations using
the Fire Dynamics Simulator (FDS) developed by the National Institute of Standards and
Technology (NIST) to analyze ventilation and potential methane accumulations and mixing
patterns in the longwall tailgate corner area after a roof fall on the tailgate entry inby the face.
Figure 2.24 shows the longwall tailgate corner model used for the simulations.
Figure 2.24: The FDS model for a typical longwall bleeder tailgate (Brune and Sapko, 2012)
The model was set up so that 80,000 cfm of fresh air was flowing from the longwall face while
the tailgate entry was established as an intake, supplying 10,000 cfm of fresh air. The outby
crosscut was partially open to allow access to the bleeder entry after the roof fall occurred inby the
face and restricted the flow. Various tailgate caving conditions were simulated to study methane
outflow patterns from the gob corner. Figure 2.25 represents a case where 1,000 cfm of methane
is outgassing from the tailgate gob corner on a different tailgate opening after a roof fall. It should
be noted that, at a face quantity of 56,000 cfm, 1,000 cfm methane would be just below the 2%
statutory limit for bleeder entries.
27 |
Colorado School of Mines | Figure 2.25: Effect of a roof fall inby the tailgate on explosive mixture formations (Brune and
Sapko, 2012)
After the roof fall reduced the opening area of the tailgate inby the face, the methane that
previously pulled directly toward the back of the panel is now being pulled into the open crosscut
outby the face. This had allowed methane to accumulate around the tailgate corner outby the face
and, depending on the roof fall tightness, the explosive mixtures could reach the shearer tailgate
drum. The results also show that the methane monitor located at the tailgate drive and shearer body
failed to detect this methane accumulation, even after it became explosive and reached the shearer
tailgate drum.
The investigation report also referred to the possibility of falling rock inside the gob as the
ignition source. This theory suggested that the explosion was triggered by a gas ignition located
behind the shields near the tailgate. The ignited methane was believed to burn behind the shield
toward the tailgate, where it dispersed combustible coal dust and initiated the coal dust explosion.
The most likely source of ignition in this case was the friction between the newly caved immediate
roof, which consisted of sandstone rocks, with the rock rubble inside the gob. Previous studies
have demonstrated that collision and rubbing of sandstone rocks can create sparks hot enough to
ignite methane gas.
28 |
Colorado School of Mines | The UBB mine was known to have had problems with floor methane; air quality samples
taken from the longwall face’s immediate return showed 0.06% methane in 61,650 cfm, which
indicates that most of the methane measured at the Bandytown bleeder fan were originated from
the gob floor. Several cases of methane outbursts were previously reported on January 4, 1997;
July 3, 2003; and February 18, 2004. Figure 2.26 shows the locations of the reported methane
outbursts.
Figure 2.26: UBB mine map showing the location of the 2010 explosion and prior gas outburst
events (Phillips, 2012)
The 2003 and 2004 incidents were reported to have had no significant impact on mining
operations. However, the methane outburst in 1997 reportedly caused a gas ignition inside the gob,
and a resulting fire burned behind the shields on the face side of the gob for some time. These
previous three gas outburst events raise the possibility that the 2010 incident was started by a
methane outburst behind the shield near the tailgate corner, before being ignited by a roof fall.
This theory is further supported by the evidence found along the face that did not indicate methane
deflagration occurring there. Figure 2.27 shows the possible explosion location of the 2010 UBB
explosion.
29 |
Colorado School of Mines | Figure 2.27: Summary of flame propagation during the UBB mine explosion (Phillips, 2012)
2.7.4 Lessons learned from the UBB explosion
Maintaining adequate roof support at the T-split area is important in a longwall bleeder
operation. The change in tailgate ventilation due to a roof fall blocking the tailgate entry inby the
face was clearly the main contributor to this event, as it allowed the methane behind the shields to
be carried back into the tailgate corner instead being pulled back toward the bleeder entry.
Several cases of methane ignitions inside the gob in 1997, 2003 and 2004 show that methane
buildup behind the shields can pose a serious risk. Being located in an inaccessible area also makes
performing gas monitoring difficult. One way to prevent ignition inside the gob is to rely on the
fresh air leaking across the longwall face to dilute the methane inside the gob. This would also
mean that sufficient air must be supplied to the longwall face to ventilate both the area behind the
shields and tailgate side of the face.
30 |
Colorado School of Mines | The UBB mine explosion also demonstrated that the current practice of installing methane
monitors on the tailgate drive and shearer body are not effective in detecting possible methane
accumulations behind the shields and tailgate corner. Both monitoring locations failed to detect
the outgassing of methane from the gob. Further study is required to determine additional
monitoring location or locations that can give an indication of change in tailgate ventilation and
possible methane accumulation around the tailgate corner.
2.7 Effective methane monitoring location
30 CFR 27 requires the installation of a methane monitoring system on the shearer and a
warning device that triggers when methane concentrations above 1.0% are detected; additionally,
it must automatically shut off equipment power when the methane concentration reaches 2.0%.
Additional gas readings can be done using portable methane detectors in accessible areas that are
considered poorly ventilated or where the dilution of methane is impaired.
Assessing an explosion hazard based on an instrument reading can be misleading, as it is
highly dependent on the location being measured. MSHA requires that all tests for methane
concentrations must be made at least 12 inches from the roof, face, ribs and floor, primarily
because methane often enters the mine workings as a localized source at a high concentration.
Figure 2.28 shows an example of how methane gas enters mine workings through a crack in the
rock and subsequently diluted by the moving air stream.
Figure 2.28: Illustration of methane being diluted into a moving air stream (Kissell, 2006)
31 |
Colorado School of Mines | This illustration shows how the measuring location can have significant impact on the methane
readings. One can expect to receive high methane readings if it is performed close to the methane
source, especially if there is not enough air to immediately dilute it. Methane, as it is lighter than
air, tends to form a high concentration layer along the mine roof, but if sufficient ventilation air
is present, the incoming methane gas can be progressively diluted until it is below prescribed
limits.
Studies performed on ventilation patterns and methane emission at longwall faces have
suggested that the majority of face methane is produced by coal breakage by the shearer (Cecala
et al., 1985a, 1989; Denk and Wirth, 1991). Two methane monitors are usually installed at the
longwall face, with one monitor mounted at the shearer and the other near the tailgate side of the
face. The methane concentration reading at the shearer is generally higher than the concentration
at the tailgate. An experiment conducted by Kissell and Cecala, et al. (2006) showed that during
the tail-to-head pass, methane concentrations at the shearer would exceeded 1.0% several times,
but no methane concentration above 1.0% was recorded by the tailgate methane monitor. This
indicates that the methane concentration recorded by the tailgate monitor is not a good
representation of face condition.
The shearer is a primary ignition source for longwall operations, which makes the placement
location of the methane monitor on the shearer body critical. Unlike portable handheld detectors,
where a peak emission can be easily missed because of infrequent reading intervals, machine-
mounted monitors are expected to operate continuously and must be able to identify emission
peaks and automatically shut off electrical equipment when methane levels exceed the prescribed
limit. Cecala et al. (1993) conducted a full-scale laboratory study to determine the best methane
monitoring location on the shearer. In it, methane was released at the shearer drums and the
concentrations were measured at four potential monitoring locations atop the shearer body (shown
in Figure 2.29).
It was found that locations A though C gave approximately the same methane concentration
readings. The values measured in these locations were found to be two times higher than that
measured at location D. However, since the location of these three monitors were close to the coal
face, these monitors were prone to being damaged, covered with coal dust and soaked by water
32 |
Colorado School of Mines | the tailgate side. The gob and fracture zone heights are 33ft and 23ft, respectively. Figure 3.2
shows the close-up view of longwall face geometry.
Figure 3.2: Close-up view of longwall face geometry
The longwall face is supported by 176 shields. Each shield is 23ft long, 5.8ft wide and 7ft
high, with the exception of the first and the last three shields that all need to accommodate the
headgate and tailgate drive. On the back of each shield, there is a 3ft2 opening that allows air to
exit and enter the face. On the headgate side, a ventilation curtain extends from the rib of the chain
pillar to shield 6. The shearer was assumed to be cutting the tailgate corner of the coal face. The
shearer body is 3.7ft in height, with cutting drum diameters of 5ft wide and a cutting depth of 3.3ft.
The longwall face model includes the operational components typically found in a longwall
operation, such as a shearer, stage loader, face conveyor, shield supports, face curtain, gob plate
and the headgate and tailgate drives. A detailed view of the longwall face equipment models is
shown in Figure 3.3.
35 |
Colorado School of Mines | Figure 3.4: Ventilation network model
At the front of the panel, 100,000 cfm of fresh air is delivered from the headgate entry; 25,000
cfm assumed to leak through the headgate curtains and 10,000 cfm of air is returned through the
belt entry, resulting in 65,000 cfm of air delivered to the face. Each tailgate entry and three bleeder
entries are set to supply 10,000 cfm of air. The air quantities at the back of the bleeder are
controlled with a series of bleeder regulators. R1 and R2 represent regulators that allow 10,000
cfm and 30,000 cfm of air to pass through, respectively.
Two different ventilation scenarios were considered for this study. The first represents the
normal ventilation conditions for the mine, while the second represents a case of a roof fall in the
tailgate entry 50ft inby the longwall face, forcing a change in tailgate ventilation. Figure 3.5 and
Figure 3.6 each illustrate the two ventilation scenarios from the plan view.
37 |
Colorado School of Mines | In this scenario, it is assumed that a tight roof fall has blocked the tailgate entry inby the face
and is forcing the operator to change tailgate ventilation controls. Since the majority of airflow
coming from the face is no longer able to enter the bleeder entry, the stoppings outby the face must
be opened to allow new ventilation access from the face toward the bleeder entries. As a result,
face airflow is now directed toward the nearest open crosscut outby the face.
3.3 Meshing
The accuracy of the numerical modeling solution is highly dependent on cell quality. ANSYS
identified the major parameters that can be used to assess overall mesh quality, including overall
cell skewness, orthogonal quality and aspect ratio. ANSYS (2014) has provided the following
definition for cell properties:
Skewness is defined as the difference between the shape of the cell and the shape of an
equilateral cell of equivalent volume. Table 3.1 shows guidelines that can be used to assess
the mesh quality based on the maximum cell skewness values.
Table 3.1: Cell Skewness Guidelines (ANSYS, 2014)
Aspect ratio is a measure of cell stretching. The recommended value for this property is below
5 for flow located away from the walls, with an exception for quadrilateral/hexahedral/wedge
cells inside the boundary layer. In general, the maximum aspect ratio should be kept below 35
for the stability of the energy solution.
Orthogonal quality for cells is computed using the vector from the cell centroid to each of its
faces, the corresponding face area vector and the vector from the cell centroid to the centroids
of each of the adjacent cells. Table 3.2 shows guidelines that can be used to assess the mesh
quality based on the maximum cell skewness values.
Table 3.2: Cell Orthogonal Quality Guidelines (ANSYS, 2014)
39 |
Colorado School of Mines | The belt entry was 195ft long and consisted of around 49,000 cells. The geometry was meshed
using a cut-cell method with cell size of 1ft, which resulted in seven cell divisions in the 7ft entry
height and 19 cell divisions in the 20ft entry width. The resulting mesh, shown in Figure 3.9, had
a maximum skewness of 0.64, maximum aspect ratio of 4.9 and a minimum orthogonal quality of
0.55, which meets the standard.
Figure 3.9: Mesh - Belt entry
3.3.2 Mine entries and crosscuts
Both mine entries and crosscuts were meshed as one single geometry consisting of around 2.3
million hexahedral cells. The geometry was meshed using a cut-cell method with a fixed cell size
of 1.3ft, which resulted in six cell divisions in the 7ft entry height and 15 cell divisions in the 20ft
entry width. The resulting mesh, shown in Figure 3.10, had a maximum skewness of 0.78,
maximum aspect ratio of 2.9 and a minimum orthogonal quality of 0.70, which meets the standard.
Figure 3.10: Mesh - Mine entries and crosscuts
41 |
Colorado School of Mines | 3.3.3 Longwall gob
The gob mesh had a total of approximately 5.7 million cells and consisted of the uncaved
immediate roof behind the shields, the gob fringe on the headgate and tailgate side of the gob, the
supported tailgate entry extending 400ft inby the face on tailgate side and the gob itself. These
four components were meshed together using a cut-cell method with cell sizes ranging from 0.92ft,
the smallest size, for the supported tailgate entry, immediate roof and gob fringes, to the largest
size of 3.67ft for the gob center. The resulting overall mesh, shown in Figure 3.11, had a maximum
skewness of 0.77, maximum aspect ratio of 5.7 and a minimum orthogonal quality of 0.40, which
meets the standard.
Figure 3.11: Mesh - Gob, immediate roof, and roof fall area
3.3.4 Fracture zone
The fracture zone was meshed separately from the gob and consisted of around 400,000 cells.
The geometry was meshed using a cut-cell method with cell sizes ranging from the smallest size
42 |
Colorado School of Mines | 3.4 Gob and fracture zone characteristics
The flow inside the gob is highly dependent on gob permeability and porosity distribution.
Low gob permeability will result in a higher gob resistance and reduction in the ability of fresh air
to pass through the gob. Gob characteristics are site specific and can vary greatly depending on
overburden conditions. Gob characteristic development and validation is not part of this study.
Several studies have been done by other researchers to determine the permeability and
porosity of longwall gob. Some of them, such as Marts (2014), was developed using surface
subsidence data obtained from several mine sites. Other models were developed based on the
predictive approach using fractal scaling in the porous medium with principles of fluid flow
(Karacan, 2010), or caving and block dimensions in relation to the effect of block dimensions and
fall heights on the void ratio (Esterhuizen and Karacan, 2007). A list of gob permeability and
resistance values obtained from literature can be seen in Table 3.4.
Table 3.4: Gob Permeability and Resistance from Literature
Although gob permeability and porosity distribution were site specific, the results of these
studies show similar trends in terms of the characteristics of the gob with the highest permeability
and porosity behind the face shields and around the gob edge; they become significantly less
permeable as it get closer to the center of the gob. Figure 3.13 shows the results of the studies on
permeability distribution inside the gob that has been performed by various researchers.
44 |
Colorado School of Mines | In this study, the fracture zone was simplified and modeled as a 23ft high porous media with
uniform permeability of 1 x 10-13 m2 and 2% porosity. To check the sensitivity of the fracture zone
permeability on the simulation result, three simulations were run using different permeability
values (1 x 10-12, and 1 x 10-14 m2), with the same amount of methane put into each model. The
results, presented in the appendix, show that the change in permeability only affected the pressure
required to supply the same amount of methane to the model, without having any significant impact
on the methane distribution inside the gob.
3.5 Modeling a roof fall blocking the tailgate
In the second ventilation scenario, it was assumed that a tight roof fall was blocking the
tailgate entry 50ft inby the face. Figure 3.15 shows the location of the roof fall in the tailgate entry.
Figure 3.15: Location of roof fall in the model
Prior to the roof fall, the tailgate area inby the face (in yellow) was modeled as an open entry
that represented a supported entry. The mine operator only removed one of every five stoppings,
which required them to maintain at least five tailgate crosscuts inby the face to be fully supported.
Post-fall, this area was changed to a porous media with permeability of 1x10-8 m2 and 5% porosity.
These values are used for the base case to represent a tight roof fall. Different permeability values
were also tested, and the result is presented in Section 4.2.
3.6 Explosive Gas Zone (EGZ) Mixture Analysis
The explosibility of methane-air mixtures is analyzed based on Coward's Triangle. For better
visualization, each cell is colorized based on its methane and oxygen concentration, and further
47 |
Colorado School of Mines | separated into six zones. These zone assignments were added to the simulation results using an
algorithm developed by Worrall (2012), shown in Figure 3.16, that characterizes the mixture and
assigns a value for each cell that corresponds to the appropriate color zone.
Figure 3.16: Explosive gas zone (EGZ) algorithm (Worrall, 2012)
The red color represents explosive gas mixtures, or EGZs, the yellow is fuel-rich inert and
green is fuel-lean inert. Orange represents methane-air mixtures that are close to becoming
explosive. Blue represents inert, oxygen-rich mixtures with less than 4% methane, including fresh
air. This color-coded Coward’s Triangle is used to analyze explosive gas mixtures for the
simulation results presented in this report.
3.7 FLUENT Setup
3.7.1 FLUENT Solver Settings
The FLUENT software settings for the solver are shown in Table 3.6. The pressure-based
solver is used for incompressible flow and the velocity formulation is set to absolute for slow-
moving fluids. The time formulation is set to steady-state and the gravity is set to 9.81 m/s2.
48 |
Colorado School of Mines | 3.7.3 FLUENT Materials Settings
FLUENT material settings for gas species are shown in Table 3.8. The methane-air mixture
species formulation includes five species. The species transport was further simplified by
removing oxygen, carbon dioxide, water vapor and nitrogen from the equation, leaving only two
species, methane and air, to be modeled. The use of these two species reduced the computational
time and model complexity because the solver only required solving the methane species transport
equation instead of multiple species. The remaining parameters are kept at default settings.
Table 3.8: FLUENT Materials Settings
3.7.4 FLUENT Discretization and Solution Settings
The FLUENT solution method settings are given in Table 3.9. Both the default Semi-Implicit
Method for Pressure-Linkage Equations (SIMPLE) and coupled algorithm were tested for the
pressure-velocity coupling method. Although computationally more expensive, it was found that
the coupled method produce a better convergence, as presented in section 3.8 of this report.
The least squares cell based gradient scheme were used to solve spatial discretization
respectively. Standard settings were kept for the pressure discretization scheme, as it was found to
produce the best convergence results compared to other schemes. However, this pressure scheme
is only applied to non-porous media flows. For porous media flows, FLUENT recommended the
use of PRESTO! (PREssure STaggering Option), which computes the face pressure in addition to
the cell pressure. In FLUENT Version 16.0, PRESTO! is intrinsically used by the solver for all
porous media cell zones, which makes the selected pressure schemes only applicable for non-
50 |
Colorado School of Mines | porous media zones. The remaining momentum, turbulent, energy and species transport equations
were set as a second-order scheme for better accuracy.
Table 3.9: FLUENT Solution Method
3.7.5 Boundary Conditions
A parametric study was conducted to determine the most suitable boundary conditions for the
simulation. The convergence results, which will be discussed in Section 3.8, suggested that except
the global continuity residual, velocity boundary condition provides better overall convergence
compared to the pressure boundary, especially for the momentum parameter. However, to achieve
a more realistic pressure field such as in the real mine situation, pressure boundaries have to be
used. Balancing each inlet boundary to achieve the desired airflow quantity using pressure inlets
was found to be difficult. To resolve this problem, the inlets were first set as velocity inlets to
enable the accurate setting for desired flow rate; this approach provides a good estimation of the
pressure required for each inlet boundary to achieve the desired airflow quantity. After required
pressure values were obtained, all boundaries were switched to pressure inlets and the model was
re-run to obtain final results. The ventilation air quantities used for the base case model can be
seen in Table 3.10. This base case represents a normal ventilation set-up for the mine.
51 |
Colorado School of Mines | The inlets of the model are the headgate entry (Point A) supplying air to the face, four tailgate
entries (Point C, D, E, F), and a methane source from the top of the fracture zone (Point G). Each
inlet is first set with a velocity magnitude normal to that boundary, with a turbulent hydraulic
diameter of 3.2 m (10.5 ft) and a turbulent intensity of 3%. The species mole fraction is set to 0%
methane for all fresh air entries. The outlet of the model includes the belt entry (Point B) and the
bleeder entry (Point J), connected to the bleeder fan. Both outlets were set as pressure outlets with
values obtained from the velocity boundary for the belt entry and zero-gauge pressure for the
bleeder fan entry. Both entries were set to have the same turbulent backflow conditions as the
inlets.
In this study, only the methane coming from an upper rider seam located in the fracture zone
is modeled. It is assumed that the rider seam supplies methane from an infinitely large reservoir in
an evenly distributed manner. Figure 3.18 shows the location of the methane source used in the
simulation.
Figure 3.18: Methane source location
The top of the fracture zone has a surface area of around 3,298,000 ft2 on a 3,300ft-long,
1,000ft-wide longwall panel. The species mole fraction is set to 100% methane, while the turbulent
conditions at the methane inlet boundary are set with an intensity of 0.1% and length scale of 2ft.
The methane inlet volume was calibrated to supply a concentration of 1% methane at the bleeder
outlets. Both velocity and pressure boundary conditions were tested and calibrated to achieve the
desired methane inflow. It was found that the use of velocity and pressure boundaries for methane
inlets produced similar results, with slight differences in the methane distribution inside the gob
and the methane volume amount inside the model. The result of this comparison study is presented
53 |
Colorado School of Mines | in Section 3.8. Further consideration of modeling methane source is discussed by Worrall (2012)
and Gilmore (2015).
3.7.6 Wall roughness adjustment
The inclusion of the geometry that represent shields, the armored face conveyor, the tailgate
and headgate drives and the crusher obstructing the headgate entry is necessary to provide better
accuracy of airflow distribution, turbulent flows and the pressure drop across the longwall face.
The pressure drop can significantly affect airflow distribution, especially across the face and gob,
thus affecting the methane distribution inside the gob. However, it was found that the addition of
these geometries was not enough to produce the desired pressure drop, and additional adjustments
on wall roughness for the face and mine entries were required to better reflect the pressure drop
and airway roughnesses typically found in underground coal mines.
Several studies have been done by researchers to determine the friction factor values typically
found in underground coal mines. Table 3.11 and Table 3.12 show the estimated friction factor
values for various coal mine airway conditions.
Table 3.11: Friction Factor Value for Different Airway Types (Kharkar et al., 1974)
*
Table 3.12: Friction Factor Value for Different Airway Types (Kingery, 1960)
54 |
Colorado School of Mines | A more recent study done by Prosser and Wallace (2002) at 14 coal and soft rock mines across
the United States shows a similar result. The surveyed mine airways were divided into four
categories. The intake airway is defined as a clean rectangular entry with roof bolts and limited
mesh lining, while a return airway is defined as a rectangular airway with some irregularities, roof
bolts and limited mesh. Most measurements were taken in straight airways airways and rectangular
surface areas. The results can be seen in Table 3.13.
Table 3.13: Standardized Friction Factors for Coal Mine Airways (Prosser and Wallace, 2002)
The mean friction factor for return entries is generally higher than that of intake entries. This
result is to be expected, considering that intake entries are better maintained than return entries.
The friction factor for the cribbed drift appears to be significantly higher compared to the other
entries, and that seems to vary based on cribbing dimensions and set-up.
A parametric study was done on FLUENT to determine the equivalent wall roughness
parameter that represents the conditions in a real mine. The simulation set-up for this study is
presented in Figure 3.19.
Figure 3.19: Simulation set-up for an equivalent wall roughness study
55 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.