Claims:1. A method (600) comprising:
receiving (602) machine data (104) of a machine among a fleet of machines, wherein the machine data (104) comprise operational data obtained from a plurality of sensors coupled to the machine, virtual sensor data of a sub-assembly of the machine, and fleet data of the fleet of machines;
generating (604) at least one data processing model (130) of the machine based on the machine data (104);
generating (606) simulation data (132) based on the operational data, the virtual sensor data, and the at least one data processing model;
predicting (608) an operating condition (240) of the sub-assembly of the machine at a future time instant based on the simulation data (132); and
controlling (610) the machine by modifying at least one operating parameter of the machine based on the predicted operating condition (240).
2. The method (600) of claim 1, further comprising providing (612) a maintenance recommendation (244) for the sub-assembly of the machine based on the predicted operating condition (240).
3. The method (600) of claim 1, wherein the fleet data comprise non-operational data comprising at least one of manufacturing data, inspection data, usage data, environment data, geographic data, customer data, and expert opinion data.
4. The method (600) of claim 1, further comprising generating feedback data based on at least one of the machine data (104) and the at least one data processing model.
5. The method (600) of claim 1, wherein the at least one data processing model (130) comprises at least one of a physics based model, a data driven model, a hybrid model, a surrogate model, a data model, a machine learning model, a fault detector model, a work-scope prediction model, and a damage model.
6. The method (600) of claim 1, wherein the machine comprises a locomotive engine.
7. The method (600) of claim 6, wherein the at least one data processing model (130) comprises a mean value engine model of the locomotive engine.
8. The method (600) of claim 7, wherein the virtual sensor data comprises a peak combustion pressure value of a cylinder of the locomotive engine.
9. The method (600) of claim 7, further comprising generating the virtual sensor data by determining a virtual distortion sensor and a virtual stiffness senor based on the piston dynamics model.
10. The method (600) of claim 6, wherein the at least one data processing model (130) comprises a piston dynamics model of a power assembly of the locomotive engine.
11. The method (600) of claim 6, wherein the virtual sensor data comprises a cavitation safety factor value representative of a liner pitting of a cylinder of the locomotive engine, obtained using a finite element analysis model.
12. A system (118) comprising:
a data acquisition unit (120) configured to receive machine data (104) of a machine among a fleet of machines, wherein the machine data (104) comprise operational data obtained from a plurality of sensors coupled to the machine, virtual sensor data of a sub-assembly of the machine, and fleet data of the fleet of machines:
a model generator unit (122) communicatively coupled to the data acquisition unit (120) and configured to generate at least one data processing model (130) of the machine based on the machine data (104);
a simulator unit (124) communicatively coupled to the model generator unit (122) and the data acquisition unit (120) and configured to generate simulation data (132) based on the operational data, the virtual sensor data, and the at least one of the data processing model;
a machine management unit (126) communicatively coupled to the simulator unit (124) and configured to:
predict an operating condition (240) of the sub-assembly of the machine at a future time instant based on the simulation data (132); and
control the machine by modifying at least one operating parameter of the machine based on the predicted operating condition (240).
13. The system (118) of claim 12, wherein the machine management unit (126) is further configured to provide a maintenance recommendation (244) for the sub-assembly of the machine based on the predicted operating condition (240).
14. The system (118) of claim 12, wherein the data acquisition unit (120) is configured to receive a plurality of databases comprising a manufacturing database, an inspection database, a usage database, an environment database, a customer database, or combinations thereof.
15. The system (118) of claim 12, wherein the simulator unit (124) is configured to generate at a feedback data based on at least one of the machine data (104) and the at least one data processing model.
16. The system (118) of claim 12, wherein the at least one data processing model (130) comprises at least one of a physics based model, a data driven model, a hybrid model, a surrogate model, a data model, a machine learning model, a fault detector model, a work-scope prediction model, and a damage model.
17. The system (118) of claim 12, wherein the data acquisition unit (120) is configured to receive the machine data (104) from a locomotive engine among a fleet of locomotives.
18. The system (118) of claim 17, wherein the at least one data processing model (130) comprises a mean value engine model of the locomotive engine.
19. The system (118) of claim 18, wherein the simulator unit (124) is configured to simulate a virtual pressure sensor and a virtual temperature sensor to generate the virtual sensor data comprising a peak combustion pressure value and a peak combustion temperature value respectively of a cylinder of the locomotive engine based on the mean value engine model.
20. The system (118) of claim 17, wherein the at least one data processing model (130) comprises a piston dynamics model of a power assembly of the locomotive engine.
21. The system (118) of claim 20, wherein the simulator unit (124) is configured to simulate a virtual distortion sensor and a virtual stiffness senor based on the piston dynamics model.
22. The system (118) of claim 17, wherein the model generator unit (122) is configured to generate a finite element analysis model of a power assembly of the locomotive engine.
23. The system (118) of claim 22, wherein the simulator unit (124) is configured to simulate a virtual pitting sensor configured to estimate a cavitation safety factor value representative of a liner pitting of a cylinder of the locomotive engine, using the finite element analysis model.
, Description:BACKGROUND
[0001] Embodiments of the present invention relate generally to predicting condition of a machine, and
more particularly to systems and methods for predicting condition based faults in sub-assemblies and
components of a machine.
[0002] Accurate and cost-effective maintenance planning of a machine, for example, a locomotive
engine and an aircraft engine, requires accurate prediction of condition of engine components based on
acute degradations and failures modes. In many scenarios, the machine is a part of the fleet of machines
operating under similar or varied operating conditions. Conventional maintenance practices are based on
collection of operational data from various sensors and inspection of the machines during outages or
maintenance. The operational data is compared with data patterns associated with known lifespan durations
determined at a design stage for making maintenance related decisions. However, the progression of the
failure modes as determined at the design stage may not match accurately with the actual progression of
failure modes during actual field operations. Such a mismatch may result in increased costs to the machine
manufacturer, reduced availability of the machines, and reduced revenue to the machine owner.
[0003] Currently available solutions for determining the condition of the machine components entail use
of diagnostics, prognostics, or health monitoring techniques. Diagnostics, prognostics and/or health
monitoring rely primarily on data collected from on-board sensors and fusion algorithms to combine the
sensor data. Poor sensor performance may yield an inaccurate prediction of the state of the machine
component. Further, the sensor data may not be available for some critical operational parameters due to
legacy issues or physical limitations associated with accessing the data. Limited sensor data may fail to
accurately convey the underlying physics of any degradation of the machine components. Moreover, other
parameters which are not captured by on-board sensors, play an active role for component damage. Such
parameters include geographical parameters (like location in coastal regions v/s location in deserts),
environmental factors (like dust particle size, salinity content, sulphate content, fuel quality), manufacturing
factors (like manufacturing vendor). Hence, complex physics based analytical models are required to assess
such factors.
[0004] Certain lifespan-prediction techniques combine design models, remote monitoring and diagnostic
(RMD) data, and inspection data into a single, probabilistic, total lifespan predictive model to calibrate and
improve the predictions of the failure modes. However, inspection data of the machine may not be available
immediately after installation. Although the quantum of data available increases with respect to time,
determining an effective model for a failure mode of a component may not be sufficient for determining
condition of sub-assemblies of the machine. Further, validation of the failure models may be difficult in
absence of actual failure data.
BRIEF DESCRIPTION
[0005] In accordance with one aspect of the present invention, a method is disclosed. The method
includes receiving machine data of a machine among a fleet of machines. The machine data include
operational data obtained from a plurality of sensors coupled to the machine, virtual sensor data of a subassembly
of the machine, and fleet data of the fleet of machines. The method further includes generating
at least one data processing model of the machine based on the machine data. The method also includes
generating simulation data based on the operational data, the virtual sensor data, and the at least one data
processing model. The method further includes predicting an operating condition of the sub-assembly of
the machine at a future time instant based on the simulation data. The method also includes controlling the
machine by modifying at least one operating parameter of the machine based on the predicted operating
condition.
[0006] In accordance with another aspect of the present invention, a system is disclosed. The system
includes a data acquisition unit configured to receive machine data of a machine among a fleet of machines.
The machine data include operational data obtained from a plurality of sensors coupled to the machine,
virtual sensor data of a sub-assembly of the machine, and fleet data of the fleet of machines. The system
further includes a model generator unit communicatively coupled to the data acquisition unit and configured
to generate at least one data processing model of the machine based on the machine data. The method also
includes a simulator unit communicatively coupled to the model generator unit and the data acquisition unit
and configured to generate simulation data based on the operational data, the virtual sensor data, and the at
least one of the data processing model. The system also includes a machine management unit
communicatively coupled to the simulator unit and configured to predict an operating condition of the subassembly
of the machine at a future time instant based on the simulation data. The machine management
unit is further configured to control the machine by modifying at least one operating parameter of the
machine based on the predicted operating condition.
DRAWINGS
[0007] These and other features and aspects of embodiments of the present invention will become better
understood when the following detailed description is read with reference to the accompanying drawings
in which like characters represent like parts throughout the drawings, wherein:
[0008] FIG. 1 is a block diagram representation of a system for predicting condition of a machine in
accordance with an exemplary embodiment;
[0009] FIG. 2 is schematic block diagram illustrating signal flow in the system of FIG. 1 in accordance
with an exemplary embodiment;
[0010] FIG. 3 is block diagram illustrating use of virtual sensors for a power sub-assembly of a
locomotive in accordance with an exemplary embodiment;
[0011] FIG. 4 is a block diagram representative of power sub-assembly of a locomotive for predicting
pitting liner fault in accordance with an exemplary embodiment;
[0012] FIG. 5 is a graphical representation of a surrogate model in accordance with an exemplary
embodiment; and
[0013] FIG. 6 is a flow chart illustrating a method for predicting condition of a machine in accordance
with an exemplary embodiment.
DETAILED DESCRIPTION
[0014] As will be described in detail hereinafter, systems and methods for predicting condition of a
machine are disclosed. In particular, systems and methods for predicting condition based faults of subassemblies
and components of a locomotive engine.
[0015] FIG. 1 is a block diagram representation of a system 100 for predicting condition of a machine
102a among a plurality of machines 102 in accordance with an exemplary embodiment. The plurality of
machines 102 having similar constructional and operational characteristics is also referred herein as a fleet
of machines. Each machine among the fleet of machines 102 includes a plurality of sub-assemblies, parts
and/or components. In the illustrated embodiment, the fleet of machines 102 is a fleet of locomotives.
Although locomotives are discussed herein, other types of machines are also envisioned. The machine 102a
includes a plurality of sensors 114 for sensing a plurality of parameters representative of operating
conditions. For example, the plurality of sensors 114 may include a temperature sensor, a pressure sensor,
a current sensor, and a voltage sensor. Other sensors may be used to measure but not limited to an amount
of remaining fuel, brake hours, idle hours, and the like.
[0016] During operation, the system 100 generates machine data 104 from the fleet of machines 102. In
one embodiment, the machine data 104 includes operational data 108 of the machine 102a measured by the
plurality of sensors 114. The operational data 108 of the machine 102a includes one or more operating
parameters measured from one or more corresponding sensors. In another embodiment, the machine data
104 includes fleet data 112 representative of non-operational data of the fleet of machines 102. The fleet
data 112 may include historical data of the fleet of machines 102. The machine data 104 include nonoperational
data of the machine 102a and operational data of the fleet of machines 102. Further, in some
embodiments, the machine data 104 include virtual sensor data 110 of an operating parameter of a subassembly
of the machine 102a obtained from a virtual sensor 116. In certain embodiments, the virtual
sensor data 110 may include a plurality of operating parameters of one or more sub-assemblies of the
machine 102a. In one embodiment, when the operating parameters are either independent or when the
operating parameters are required at different time scales, a plurality of virtual sensors 116 may be used to
obtain the virtual sensor data 110. In one embodiment, two virtual sensors 116 may be used when a first
operating parameter is required in real-time and a second operating parameter is required once in a day. In
one embodiment, the virtual sensor data 110 is an estimate of one or more operating parameters of the
machine 102a. The working of the virtual sensor 116 and the technique of obtaining the virtual sensor data
110 are explained in subsequent paragraphs with reference to subsequent figures.
[0017] The system 100 is communicatively coupled to a condition prediction system 118 used for
predicting condition based faults in sub-assemblies or parts of the machine 102a. Specifically, the condition
prediction system 118 is communicatively coupled to the the fleet of machines 102 and configured to
receive the machine data 104. The condition prediction system 118 is further configured to control the
operation of the machine 102a based on the predicted faults. In one embodiment, the condition prediction
system 118 is configured to generate the virtual sensor 116 and then generate the virtual sensor data 110.
[0018] The condition prediction system 118 includes a data acquisition unit 120, a model generator unit
122, a simulator unit 124, a machine management unit 126, a memory unit 134, and a processor unit 136
communicatively coupled to each other via a communications bus 140.
[0019] The data acquisition unit 120 is communicatively coupled to the fleet of machines 102 and
configured to receive the fleet data 112. Further, specifically, the data acquisition unit 120 is
communicatively coupled to the machine 102a and configured to receive the machine data 104. In another
embodiment, the fleet data 112 generated by the fleet of machines 102 may be retrieved from a plurality of
databases and stored in the memory unit 134. The data acquisition unit 120 is configured to retrieve the
fleet data 112 from the memory unit 134. The plurality of databases include, but not limited to, a database
for operational data of the fleet of machines 102, a database for non-operational data of the fleet of machines
102, and a database for the historical data of the fleet of machines 102.
[0020] The model generator unit 122 is communicatively coupled to the data acquisition unit 120 and
configured to generate at least one data processing model 130 of a sub-assembly of the machine 102a based
on the machine data 104. The model generator unit 122 is configured to generate at least one of a physics
based model, a data driven model, a hybrid model, a surrogate model, a data model, a machine learning
model, a fault detector model, a work-scope prediction model, and a damage model. In one embodiment,
the model generator unit 122 is configured to generate a mean value engine model of a locomotive engine
246, for example. In another embodiment, the model generator unit 122 is configured to generate a piston
dynamics model corresponding to a power assembly 248 of the locomotive engine 246. In yet another
embodiment, the model generator unit 122 is configured to gnerate a finite element analysis model of the
power assembly 248 of the locomotive engine 246. In some embodiments, the model generator unit 122 is
configured to select a type of model based on at least one of the type of machine, the type of fault condition
to be predicted, and a parameter to be estimated. In one particular embodiment, the type of model or the
inputs required for selecting the type of the model may be provided by an operator of the fleet of machines
102.
[0021] The simulator unit 124 is communicatively coupled to the model generator unit 122 and the data
acquisition unit 120 and configured to generate simulation data 132 based on the operational data 108, the
virtual sensor data 110, and the at least one of the data processing model 130. In one embodiment, the
simulator unit 124 is configured to generate the virtual sensor 116 based on the machine data 104 and the
at least one data processing model 130. In one embodiment, the simulator unit 124 is configured to generate
a feedback data based on at least one of the machine data 104 and the at least one data processing model
130. Specifically, the virtual sensor data 110 may be part of the feedback data and is determined as an
estimate of one or more operating parameters of the machine 102a obtained based on the operational data
108 and the at least one data processing model 130. The virtual sensor 116 refers to a mechanism of the
simulator unit 124 used for generating the feedback data. The virtual sensor data 110 refers to the feedback
data available to the units 120, 122 from the simulator unit 124. In one embodiment, simulator unit 124 is
configured to simulate a virtual pressure sensor configured to generate a peak combustion pressure (PCP)
value of a cylinder of a locomotive engine and a virtual temperature sensor configured to generate a peak
combustion temperature (PCT) value of the cylinder of the locomotive engine based on the mean value
engine model. In one embodiment, the simulator unit 124 is configured to simulate a virtual distortion
sensor and a virtual stiffness senor based on the piston dynamics model. In one embodiment, the simulator
unit 124 is configured to simulate an virtual impairment sensor configured to generate a measure of wear
in a cylinder of the power assembly 248 and a virtual pitting sensor configured to generate a measure of
cavition of a liner (also referred to as “liner pitting”) in the cylinder of the power assembly 248.
[0022] The machine management unit 126 is communicatively coupled to the simulator unit 124 and
configured to predict an operating condition of the sub-assembly of the machine 102a at a future time instant
based on the simulation data 132. In one embodiment, the future time instant may be a time instant of a
scheduled maintenance of the machine 102. In another embodiment, the future time is an estimated time
instant of an emergency maintenance of the machine 102. The machine management unit 126 is further
configured to control the machine 102a by modifying one or more operating parameters of the machine
102a based on the predicted operating condition. The machine management unit 126 is also configured to
provide a maintenance recommendation (also referred as work-scope recommendation) for the subassembly
of the machine 102a in real-time to an operator of the fleet of machines 102. Further, the machine
management unit 126 is also configured to generate a control signal for modifying the operation of the
machine 102a to perform a scheduled maintenance. The machine management unit 126 generates an output
138 representative of the control signal, the predicted operating condition, and the maintenance
recommendation.
[0023] The processor unit 136 is communicatively coupled to units 120, 122, 124, 126 and may include
one or more of an arithmetic logic unit, a microprocessor, a general purpose controller, and a processor
array to perform desired computations. In one embodiment, the processor unit 136 may be configured to
perform tasks associated with at least one of the data acquisition unit 120, the model generator unit 122, the
simulator unit 124, and the machine management unit 126. In another embodiment, the processor unit 136
may retrieve one or more non-operational databases and historical databases for the fleet data 112 from the
memory unit 134. While the processor unit 136 is shown as a single unit, in certain embodiments, the
processor unit 136 may include more than one processor co-located or distributed at different locations.
[0024] The memory unit 134 is communicatively coupled to the processor unit 136. In certain
embodiments, the memory unit 134 may include a plurality of memory subunits. The memory unit 134
may be a non-transitory storage medium. For example, the memory unit 134 may be a dynamic random
access memory (DRAM) device, a static random access memory (SRAM) device, flash memory or other
memory devices. In one embodiment, the memory unit 134 may include a non-volatile memory or similar
permanent storage device, media such as a hard disk drive, a floppy disk drive, a compact disc read only
memory (CD-ROM) device, a digital versatile disc read only memory (DVD-ROM) device, a digital
versatile disc random access memory (DVD-RAM) device, a digital versatile disc rewritable (DVD-RW)
device, a flash memory device, or other non-volatile storage devices. In one embodiment, a non-transitory
computer readable medium may be encoded with a program having a sequence of instructions to instruct
the processor unit 136 to perform functions of one or more of the data acquisition unit 120, the model
generator unit 122, the simulator unit 124, and the machine management unit 126. In one embodiment,
the memory unit 134 may store one or more of the non-operational databases and historical databases for
the fleet data 112.
[0025] FIG. 2 is a detailed schematic diagram of the system 100 in accordance with an exemplary
embodiment of FIG. 1. The machine data 104 of the fleet of machines 102 include the operational data
108, the virtual sensor data 110, and the fleet data 112. The operational data 108 is representative of a
plurality of parameters acquired from the fleet of machines 102 continuously via a plurality of sensors. In
one example, the operational data 108 include but not limited to at least one of a current parameter, a voltage
parameter, mileage parameter, a fuel consumption parameter, and a speed parameter.
[0026] The fleet data 112 include data acquired from a plurality of databases 202. The plurality of
databases 202 include, but not limited to an inspection database 204, an environment database 206, a
manufacturing database 208, a customer database 210, an expert database 212, an usage database 214, and
a geographical database 216. In some embodiments, the plurality of databases 202 include internal and
proprietary databases related to repair shop reports, design reviews, life data management, diagnostics data,
and the like.
[0027] The inspection database 204 includes data generated during inspection and/or scheduled
servicing of the fleet of machines 102. The environment database 206 includes parameters related to the
environmental conditions of the fleet of machines 102. The manufacturing database 208 includes design
parameters and tolerances, machine specifications, vendor/component historical reliability of the fleet of
machines 102. The customer database 210 includes information regarding customers such as fuel quality
checks, operating procedures, and historical reliability of the machines. The expert database 212 includes
patterns observed in the other databases and decisions related to faults of the subassemblies of the machine
102a. The usage database 214 includes data related to operating regimes of the machine 102a. The
geographical database 216 includes data related to a geographical region where the the fleet of machines
102 is deployed during operation.
[0028] It may be noted herein that the fleet data 112 further include historical data generated by storing
the operational data 108 and the virtual sensor data 110 in the memory unit 134. The historical data may
include data related to a component, a part or a sub-assembly of the machine 102a or the fleet of machines
102. The historical data would be used in combination with operational data 108 as inputs for the at least
one data processing model 130 .
[0029] The at least one data processing model 130 includes, but not limited to, a physics based model
218, a data driven model 220, a hybrid model 222, a data model 224, a machine learning model 226, a
damage model 228, a work-scope prediction model 230, a fault detector model 232, and a surrogate model
234.
[0030] The physics based model 218 for a sub-assembly of the machine 102a refers to mathematical or
simulation models based on physical principles of operation of the sub-assembly. The physics based model
218 may be determined based on experiments and validated by real-life operational data generated from
the sub-assembly. The data driven model 220 for the sub-assembly of the machine 102a refers to a
mathematical model generated by exploratory analysis and historical operational data generated from the
sub-assembly. The hybrid model 222 refers to a combined model generated from the physics based model
218 and the data driven model 220. In one embodiment, the data driven model 220 is configured to predict
an operating parameter of the sub-assembly based on the values of the operating parameter obtained in the
past.
[0031] The machine learning model 226 is determined based on the fleet data 112 obtained from the
inspection database 204 corresponding to a fault of the sub-assembly of the machine 102a. The machine
learning model 226 may be obtained using one or more machine learning algorithms. Inone specific
embodiment, a supervised machine learning technique is employed to generate the machine learning model
226. Use of supervised learning technique entails the generation of the machine learning model 226 based
on a plurality of input datasets and corresponding output datasets obtained from one of the plurality of
databases 202. In certain other embodiments, an unsupervised learning technique may be employed to
generate the machine learning model 226. It may be noted that use of the unsupervised learning technique
to generate the machine learning model 226 calls involves use of statistical techniques including, but not
limited to, clustering, method of moments, and matrix factorization techniques.
[0032] The damage model 228 refers to a model used for estimating damage of a part, a component, or
a sub-assembly of the machine 102a. The damage may refer to the damage condition at present instant of
time or a predicted damage at a future instant of time. In one embodiment, the damage model 228 is used
to determine a physical state of a part or a component. In another emmbodiment, the damage model 228 is
used to determine an operational state of a sub-assembly or a machine. A physics based mathematical
model or a simulation model may be used to determine the physical state of the part/component or the
operational state of the sub-assembly/machine. The work-scope prediction model 230 provides a
recommendation or an initial plan of work-scope for performing maintenance. The work-scope prediction
model 230 may use the damage model 228, hybrid models 222, or other models for providing the
recommendation.
[0033] The fault detector model 232 is used to determine a fault or an impending fault based on the
damage model 228. The fault detector model 232 is determined based on the machine data 104. In one
embodiment, the fault detector model 232 is calibrated based on the contents of the inspection database
204. The calibration of the fault detector model 232 involves determining differences between an outcome
of an inspection and a corresponding output of the fault detector model 232. The calibration further involves
modifying the fault detector model 232 when the difference between the outcome of the inspection and the
corresponding output of the fault detector model 232 are not negligible. A life duration of the machine
102a is determined based on the fault detector model 232. Furthermore, the life duration of the machine
102a is used to determine the output 138 from the machine management unit 126.
[0034] The surrogate model 234 refers to a model for one parameter determined from a model of another
parameter. In some embodiments, the virtual sensor 116 is used as a surrogate model. In such
embodiments, the virtual sensor data are the data obtained from the surrogate model 234.
[0035] Further, as previously noted with reference to FIG. 1, the simulator unit 124 is configured to
generate the simulation data 132 based on the at least one data processing model 130 and the machine data
104. In one embodiment, the simulation data 132 are obtained by at least one of a software simulator, a
hardware simulator using the at least one data processing model 130. The simulation data 132 include the
virtual sensor 238 used for generating virtual sensor data or surrogate data used for generating the surrogate
model 234. In one embodiment, the virtual sensor 238 is a scheme for monitoring an operating parameter
of the machine 102a which is not measurable by the plurality of sensors 114. The virtual sensor 238 may
be in the form of a software routine or a model for monitoring the operating parameter. The virtual sensor
data 110 are not accessible by the plurality of sensors 114 either because of incompatibility of legacy
systems or issues related to physically accessing the interiors of the machine 102a for performing the
measurements.
[0036] Further, with reference to the machine management unit 126 of FIG. 1, the output 138 from the
machine management unit 126 includes, but not limited to, a predicted operating condition 240, a control
signal 242, and a maintenance recommendation 244. In one embodiment, the predicted operating condition
240 is representative of a parameter value, a state of a sub-assembly or a state of a component of the machine
102a at a future instant of time. The control signal 242 is used to control the machine 102a or subassemblies
of the machine 102a to change the operating condition of the machine 102a during operation.
The maintenance recommendation 244 is representative of recommended repair or replacement of specific
components of the machine 102a, an initial cost estimate to be incurred by an end customer, and also an
extent of work in terms of tear down, inspection, rework and assembling to be performed on specific
components.
[0037] FIG. 3 is block diagram 300 illustrating use of virtual sensors for a sub-assembly of the machine
102a in accordance with an exemplary embodiment of FIG. 1. In the illustrated embodiment, specifically,
the use of virtual sensors for a power sub-assembly of a locomotive is disclosed. The block diagram 300
shows a plurality of inputs 302 required for the working of the power sub-assembly of the locomotive. The
plurality of inputs 302 is obtained from the operational data 108, the virtual sensor data 110, and the fleet
data 112. The plurality of inputs 302 also additionally includes fault data 328 of the power sub-assembly,
statistical data 330 related to the fleet of locomotives, material data 332 representative of properties of
materials used for the components of the power sub-assembly, and teardown data 334 generated during
repair and tearing down operation.
[0038] An engine sub-assembly of the locomotive includes a turbocharger and the power sub-assembly
having an inlet manifold and a plurality cylinders with pistons among other parts. The block diagram 300
further includes a mean value engine model (MVEM) 304, a piston dynamics model 306, a finite element
analysis (FEA) model 308, a computational fluid dynamics (CFD) model 310. The MVEM 304 is
representative of combustion of the engine, the turbocharger, and the engine inlet manifold. The piston
dynamics model 306 is used for simulating the dynamics of a piston having a pin, a connecting rod, a liner,
and rings. The finite element analysis (FEA) model 308 is used for simulating force, displacement, velocity,
and acceleration response of different parts of the piston and the liner . The computational fluid dynamics
(CFD) model 310 is used for determining state variables of fluids of each cylinder of the locomotive engine.
Fluids include but are not restricted to a coolant, a lubricating oil, and an exhaust gas.
[0039] The MVEM 304 is configured to receive a plurality of inputs 312 among the inputs 302 required
for simulating the working of the engine and generate a description of engine dynamics 320. In one
embodiment, the MVEM 304 generates the description of engine dynamics 320 in real-time once in every
ten milliseconds. The MVEM 304 is represented by a set of differential and static equations to predict the
engine parameters such as but not limited to an intake manifold pressure, an intake manifold temperature,
an exhaust manifold pressure, an exhaust manifold temperature, a turbo speed, an indicated efficiency, an
airflow, a peak cylinder pressure, and a peak cylinder temperature. Specifically, aforementioned parameters
associated with manifolds and turbo speed are estimated using the differential equations. The differential
equations may be either linear or nonlinear equations. Further, aforementioned paramaters associated with
the cylinder and flow parameters are estimated using the static equations. The static equations may be
algebraic expressions such as linear or nonlinar polynomials. In one embodiment, the static equation may
also be stored in the form of lookup table. Specifically, equations are represented by:
= ,

(1)
=
, (2)

= , (3)
where x is an engine parameter represented by a differential equation (1), is representative of derivative
of the engine parameter x, u is representative of the plurality of inputs 312, f1 is a static function, yp is a
predicted value of an engine parameter y from the sensor data 110 and y2 is a virtual sensor data 110
representative of an engine parameter represented by a static equation (3). The predicted value yp is
determined by using a static model f2 of equation (2) The function f2 and f3 are static functions that are
generated by the machine data 104.
[0040] The plurality of inputs 312 includes, but not limited to, a fuel value (FV) parameter, an angle of
advance (AA) parameter, and revolutions per minute (RPM) parameter. The description of engine dynamics
includes, among other parameters, a peak combustion pressure (PCP) and a peak combustion temperature
(PCT). In one embodiment, the PCP is used as an input paramter by the piston dynamics model 306. In
another embodiment, the PCT is used as an input parameter by the piston dynamics model 306. PCP and
PCT are examples of the virtual sensor data 110. A part of the MVEM 304 that simulates the PCP is a
virtual PCP sensor 336.
[0041] The piston dynamics model 306 is configured to receive a plurality of inputs 314 among the
inputs 302 and the virtual PCP sensor 336 from MVEM 304 and generate a description of piston dynamics
322. The plurality of inputs 314 includes but not limited to oil temperature parameter and speed. Such
inputs 314 are examples of the virtual sensor data 110. The description of piston dynamics 322 includes,
among other parameters, a bore distortion parameter, a boundary temperature, and a stiffness parameter. In
one embodiment, the bore distortion parameter, the boundary temperature, and the stiffness parameter are
used as inputs by the FEA model 308. The part of the piston dynamics model 306 that generates bore
distortion parameter is a virtual distortion sensor. The bore distortion parameter is representative of changes
in the bore dimensions. Similarly, piston dynamics model 306 simulates a virtual temperature sensor for
generating the boundary temperature and a virtual stiffness sensor for measuring the stiffness parameter
representative of stiffness of the piston. The virtual sensors simulated by the piston dynamics model 306,
viz the virtual bore distortion sensor, the virtual temperature sensor, and the virtual stiffness sensor are
represented by a reference numeral 338.
[0042] The FEA model 308 is configured to receive a plurality of inputs 316 such as geometry variations
of disassembled components obtained from teardown operation and generate an output 324 such as but not
limited to a pitting parameter, cylinder wear parameter, coolant temperature parameter. The FEA model
308 is further configured to receive the description of piston dynamics model 322. In one embodiment, the
FEA model 308 is further configured to perform at least one of a static and a dynamic analysis to estimate
parameters such as but not limited to a force and a velocity response depending on the stiffness parameter
corresponding to piston components such as liner and the cylinder. The output 324 is an example of virtual
sensor data which are provided as input to the CFD model 310. The FEA model 308 simulates a virtual
pitting sensor, a virtual wear sensor, a virtual temperature sensor for measuring the liner pitting, cylinder
wear, and coolant temperature respectively. The virutal sensors simulated by the FEA model 308, viz the
virtual pitting sensor, the virtual wear sensor and the virtual temperature sensor are illustrated by numeral
340. In a further embodiment, the CFD model 310 is configured to receive inputs 318 such as engine
coolant temperature, coolant pressure, and coolant flow and virtual temperature sensors 340 from FEA
model 308 and generate an output 326 such as liner temperature parameter, coolant heat dissipation
parameter, and coolant temperature per cylinder. The CFD model 310 among other applications can
estimate parameters such as but not limited to localised coolant pressure. The output 326 is an example of
virtual sensor data. The CFD model 310 is configured to simulate a virtual temperature sensor for
measuring the liner temperature parameter. The virutal sensors simulated by the CFD model 310, viz the
virtual bore distortion sensor and the virtual temperature sensor are illustrated by numeral 342.
[0043] FIG. 4 is a block diagram 400 illustrating prediction of a fault of a sub-assembly of a machine in
accordance with an exemplary embodiment. The sub-assembly considered in this illustrated embodiment
is a power sub-assembly 248 of FIG.2 having a cylinder 432 with a liner 434. The illustrated embodiment
is specifically representative of a model for prediction of a liner pitting fault in a power sub-assembly of a
locomotive such as 102. The block diagram 400 includes the MVEM 304, the piston dynamics model 306,
and the FEA model 308. In the illustrated embodiment, the MVEM 304 includes a turbocharger model 418
and a manifold model 420. Both the turbocharger model 418 and the manifold model 420 are generated
based on basic mass and energy conservation laws and are configured to estimate a flow, a pressure and a
temperture parameters of the turbocharger. The MVEM is configured to receive the inputs 312 and generate
a manifold air pressure (MAP) parameter, a pre turbine pressure (PTP) parameter, and turbo speed
parameter as outputs 426 and corresponding prediction values 428. The MVEM further includes a cylinder
combustion model 424 configured to generate an initial estimate 412 of the PCT and the PCP parameters.
The cylinder combustion model 424 is further configured to generate a pre turbine temperature value 422
for the manifold model 420. A Kalman adaptation scheme 430 is employed to generate the PCP and PCT
parameters based on the initial estimates 412 and corresponding measured values 414 of the PCP, the PCT,
and the turbo speed parameter. In the illustrated embodiment, the Kalman filter is used to implement a
virtual sensor using the Kalman adaptation scheme 430. The Kalman filter generates an initial prediction,
computes a gain value referred generally as ‘Kalman gain’ and refines the initial prediction using the
Kalman gain value. The Kalman gain is computed based on the measured engine operating parameters and
represented by:
= + - (4)
where xc is corrected engine parameter, x is predicted engine parameter obtained from equation (1), ym is a
measured value of the engine parameter y, yp is predicted engine parameter obtained from equation (2), and
K is Kalman gain. The corrected parameter xc is used in equation (3) to determine the virtual sensor data
corresponding to the parameter y2. In one example, the parameter PCP is obtained as the parameter y2 using
equations (1)-(4). In another example, the parameter PCT is obtained as the parameter y2 using equations
(1)-(4). In the illustrated embodiment, the piston dynamics model 306 receives parameters of the PCP and
the PCT among other input parameters provided by the MVEM 304. The FEA model 308 of a cylinder
432 is generated based on the piston dynamics model 306.
[0044] Liner 434 cavitation condition is detected based on a liner response 410 determined from FEA
model 308 of the cylinder. In one embodiment, the liner response 410 includes a critical coolant
acceleration value and a liner acceleration value. The pitting condition of the liner is determined when the
critical coolant acceleration value is less than the liner acceleration value. The coolant acceleration and the
liner accleration are the forces acting on the cylinder liner causing the cavitation condition. The critical
coolant acceleration (Acr) is determined using following relationship:
= * * !"/$%&&'()*
where pst is static pressure of the coolant, a, b are constants and Tcoolant is coolant temperature, and RPM is
speed parameter. The liner acceleration is determined by a frequency response function given by
+,
+ = -1
/ -0
01

01
- 0
+ 220104
where ? is angular frequency, ?N is natural frequency, A is linear acceleration, F is a piston secondary
force, K is liner stiffness parameter, and ? is a function of the dampness parameter. The piston secondary
force parameter is determined based on the PCP and speed parameter, using piston dynamics model 306.
The coolant temperature per cylinder is determined from the CFD model 310. Liner acceleration is
calculated using the FEA model 308. It may be noted that liner response parameters and the pitting
condition detection are dependent on a plurality of input parameters measured by the plurality of sensors
and a plurality of input parameters obtained by virtual sensors simulated by the models 304,306,308.
[0045] FIG. 5 is a graph 500 illustrating a surrogate model in accordance with an exemplary
embodiment. The graph 500 includes an x-axis 502 representative of break horse power (BHP) and a yaxis
representative of manifold air pressure (MAP). The graph includes a plurality of dots 506, 508
representative of a correlation between load and the MAP values. Specifically, the shaded dots 508 are
determined based on rotational speed values ranging from 1040 rotations per minute to 1060 roations per
minute, at a throttle notch 8. The graph 500 includes a first curve 510 representative of an upper bound for
the plurality of shaded dots 508 and a second curve 512 representative of a lower bound for the plurality of
shaded dots 508. The first curve 510 and the second curve 512 are used to determine a surrogate model for
the MAP parameter. The surrogate model for the MAP parameter, is used as a virtual MAP sensor for
obtaining a virtual MAP parameter for a particular load.
[0046] FIG. 6 is a flow chart illustrating a method 600 for predicting condition of a machine from a fleet
of machines, for example, a locomotive in accordance with an exemplary embodiment. The method 600
includes receiving machine data of a machine among a fleet of machines in step 602. The machine data
include operational data of the machine obtained from a plurality of sensors and virtual sensor data of a
sub-assembly of the machine. The machine data also include fleet data having non-operational data and
historical data of a fleet of machines. The non-operational data is obtained from a plurality of databases
including at least one of manufacturing data, inspection data, usage data, environment data, geographic
data, customer data, and expert opinion data. The virtual sensor data are generated based on the at least
one of the machine data and the at least one data processing model.
[0047] At step 604, the method further includes generating at least one data processing model of the
machine based on the machine data. The step of generating the at least one data processing model involves
generating at least one of a physics based model, a data driven model, a hybrid model, a surrogate model,
a data model, a machine learning model, a fault detector model, a work-scope prediction model and a
damage model. In one embodiment, the step of generating at least one data processing model includes
generating a mean value engine model of the locomotive engine. In another embodiment, the step of
generating the at least one data processing model involves generating a piston dynamics model of a power
assembly of the locomotive engine.
[0048] The method further includes generating simulation data based on the operational data, the virtual
sensor data, and the at least one data processing model in step 606. In one embodiment, the step of
generating the simulation data includes determining a peak combustion pressure value of a cylinder of a
locomotive engine, using a virtual pressure sensor. In another embodiment, generation of simulation data
includes simulating one or more virtual sensors. In one embodiment, simulating the one or more virtual
sensor includes determining a virtual distortion sensor and a virtual stiffness senor based on the piston
dynamics model. In one exemplary embodiment, the step of generating the virtual sensor data includes
determining a cavitation safety factor value representative of a liner pitting of a cylinder of the locomotive
engine, using a finite element analysis model.
[0049] At step 608, the method includes predicting an operating condition of the sub-assembly of the
machine at a future time instant based on the simulation data. The method further includes controlling the
machine by modifying one or more operating parameters of the machine based on the predicted operating
condition at step 610. Further, at step 612, the method includes generating a maintenance recommendation
for the sub-assembly of the machine in real-time to an operator.
[0050] The systems and methods disclosed hereinabove provide an enhanced prediction of life of a subassembly
of a machine among a fleet of machines. Use of virtual sensor and virtual sensor data facilitate
to enhance a quality of prediction of condition of the machine. Accurate determination of a condition of a
sub-assembly of the machine enables to reduce the frequency of repairs and replacement of parts. Accurate
prediction of the life duration facilitates to optimize the operation of the fleet of machines and thereby
extend the life of components of the machine.
[0051] Not necessarily all such objects or advantages described above may be achieved in accordance
with any particular embodiment. Thus, for example, those skilled in the art will recognize that the systems
and techniques described herein may be embodied or carried out in a manner that achieves or improves one
advantage or group of advantages as taught herein without necessarily achieving other objects or advantages
as may be taught or suggested herein.
[0052] While the technology has been described in detail in connection with only a limited number of
embodiments, it should be readily understood that the specification is not limited to such disclosed
embodiments. Rather, the technology can be modified to incorporate any number of variations, alterations,
substitutions or equivalent arrangements not heretofore described, but which are commensurate with the
spirit and scope of the claims. Additionally, while various embodiments of the technology have been
described, it is to be understood that aspects of the specification may include only some of the described
embodiments. Accordingly, the specification is not to be seen as limited by the foregoing description, but
is only limited by the scope of the appended claims.