Imported: 17 Feb '17 | Published: 10 Jan '12

USPTO - Utility Patents

A method for multivariate analysis using a mathematical model for generating model data and actual data for more than one variable includes for each variable, determining a difference between the model data and the actual data. The model data is substantially representative of more than one variable. The method also includes for each variable, determining a fractional impact on performance. The method further includes for each variable, determining a weighted deviation based on the determined difference and the determined fractional impact. The method also includes transmitting the weighted deviation to an output device.

The field of this invention relates generally to engines, and more particularly to a method and system for use in monitoring engines via decomposition of data using a graphical multivariate analysis.

Generally, known engines include a plurality of known components that facilitate a plurality of known operational processes. During operation of an engine, a user typically monitors a selected number of engine processes by collecting data associated with parameters that are at least partially representative of the existing conditions of the components. Specifically, a user may monitor a plurality of operational parameters through raw data supplied directly to the user, or by processing at least a portion of the data collected and providing the results of such processing to the user. Some known processed data is generated from one parameter, while other processed data is derived from a plurality of parameters. Moreover, such raw and processed data may enable trending activities to be generated, such as a trend chart of the raw and processed data. Typically, such trend charts illustrate such data, whether raw or processed, versus time. Moreover, some known trend charts include a plurality of trend plots, wherein each trend plot is associated with one parameter.

At least one known method of monitoring the operation of an engine includes generating a plurality of such trend charts, and/or generating a plurality of data tables, and having a user attempt to determine the relative proportion of how each parameter is impacting the process. If more than one parameter or issue is adversely affecting any of the processes, assessing the extent of how each parameter impact the process may be time-consuming, difficult, and/or inaccurate. For example, it is possible that an anomaly embedded within the data may be identified that ultimately is not adversely impacting the parameters and processes being evaluated. In addition, for those trend charts that indicate a plurality of parameters, such parameters may be scaled such that an apparent disproportionate relationship between parameters that do not share a common unit of measurement may be insinuated.

In one aspect, a method for multivariate analysis is provided. The method uses a mathematical model for generating model data and actual data for more than one variable. The model data is substantially representative of more than one variable. The method includes for each variable, determining a difference between the model data and the actual data. The method also includes for each variable, determining a fractional impact on performance. The method further includes for each variable, determining a weighted deviation based on the determined difference and the determined fractional impact. The method also includes transmitting the weighted deviation to an output device.

In another aspect, a monitoring system for use with a turbine engine is provided. The monitoring system includes at least one processor coupled in data communication with a plurality of measuring instruments. Each of the measuring instruments is associated with at least one of a plurality of variables and collects actual engine data. The system also includes a mathematical model at least partially resident within the at least one processor. The model is substantially representative of at least a portion of an engine process and is at least partially derived from at least one of the variables. The mathematical model stores model data. The system further includes at least one algorithm programmed within the at least one processor. The at least one algorithm is programmed to determine a fractional impact on engine performance for each variable. The at least one algorithm is also programmed to determine a difference between the model and the actual data. The at least one algorithm is further programmed to determine a weighted deviation based on the determined difference and the determined fractional impact. The system also includes at least one output device coupled in data communication with the at least one processor.

In a further aspect, a gas turbine engine is provided. The gas turbine engine includes at least one compressor and at least one turbine rotatably coupled to the at least one compressor. The engine also includes a monitoring system coupled in data communication with at least one of the at least one compressor and the at least one turbine. The monitoring system includes at least one processor coupled in data communication with a plurality of measuring instruments. Each of the measuring instruments is associated with at least one of a plurality of variables and collects actual engine data. The system also includes a mathematical model at least partially resident within the at least one processor. The model is substantially representative of at least a portion of an engine process and is at least partially derived from at least one of the variables. The mathematical model stores model data. The system further includes at least one algorithm programmed within the at least one processor. The at least one algorithm is programmed to determine a fractional impact on engine performance for each variable. The at least one algorithm is also programmed to determine a difference between the model and the actual data. The at least one algorithm is further programmed to determine a weighted deviation based on the determined difference and the determined fractional impact. The system also includes at least one output device coupled in data communication with the at least one processor.

Utilizing graphical multivariate analysis techniques as described herein facilitates identifying associated parameters that may be driving certain process behaviors while also facilitating elimination of those parameters that are less likely to be driving such process behaviors. In an active operating environment, such short-term, or substantially real-time, determinations may facilitate early isolation of the initiating conditions and subsequent restoration of process parameters to desired ranges. Moreover, early elimination of apparently unaffected parameters facilitates focusing operation and maintenance resources toward higher probability causes. Furthermore, earlier determination of apparent causes of the transient facilitates actions by operating personnel to mitigate deleterious effects on the process and/or equipment by taking preemptive and/or corrective actions within a sort time frame after the transient initiates. Also, erroneous values of operational signals due to drift or failures of instrumentation may be manifested within a graphical display by an independent shift of one of the signals with no associated changes in other portion of the graphical display.

FIG. 1 is a cross-sectional schematic view of a gas turbine engine assembly **10** that includes a core engine **12**. Core engine **12** includes a compressor **14**, a combustor **16**, and a power turbine **18**. Compressor **14** is coupled in flow communication with combustor **16**, which in turn, is coupled in flow communication with power turbine **18**. Compressor **14** includes a plurality of rotating blades **20** and stationary blades **22**, wherein rotating blades **20** are coupled to an engine shaft **24**. Combustor **16** is also coupled in flow communication with a source of fuel (not shown) via a fuel supply manifold **26**. Turbine **18** includes a plurality of rotating blades **28** and stationary blades **30**, wherein rotating blades **28** are coupled to engine shaft **24**. Blades **20** and **28** extend radially outward from shaft **24** towards casing **32**. Core engine **12** also includes a casing **32** that extends about compressor **14**, combustor **16**, and power turbine **18**. Blades **20** and **28** extend radially outward from shaft **24** towards casing **32**. Blades **22** and **30** are coupled to casing **32** and extend radially inward toward shaft **24**. Core engine **12** further includes an air intake side **34** and a combustion exhaust side **36**. Core engine **12** is substantially symmetrical about a centerline **38**.

Engine assembly **10** also includes a power transmission device **40** that is coupled to power transmission drive shaft **42**, wherein shaft **42** is coupled to engine shaft **24** via a coupling **44**. In the exemplary embodiment, engine assembly **10** is a T64 model turboshaft engine commercially available from GE Aviation, Cincinnati, Ohio. Alternatively, engine assembly **10** is any engine including, but limited to, a gasoline engine, a diesel engine, a steam turbine engine, and any combustion turbine engine. Also, in the exemplary embodiment, device **40** is a main transmission for a propeller shaft (not shown) and is configurable to receive drive power from a plurality of shafts **42**. Alternatively, device **40** is any device that facilitates operation of engine assembly **10** as described herein.

During operation, air is drawn into core engine **12** at air intake side **34** and is channeled through compressor **14** axially along centerline **38**. The air is compressed and the compressed air is channeled to combustor **16**, wherein the air is combusted with fuel channeled to combustor **16** via manifold **26**, thereby generating combustion exhaust gases. The gases are channeled through turbine **18**, wherein the gases transmit rotational power to shaft **24** and are subsequently channeled through and exhausted from core engine **12** via exhaust side **36**. Shaft **24** drives compressor **14**. Shaft **24** also drives shaft **42** via coupling **44**, wherein shaft **42** drives transmission device **40**.

FIG. 2 is a schematic illustration of at least a portion of an exemplary monitoring system **100** that may be used to monitor a plurality of processes within gas turbine engine assembly **10** (shown in FIG. 1). In the exemplary embodiment, monitoring system **100** includes a plurality of measuring instruments **102** that are coupled to gas turbine engine assembly **10** that measure a plurality of predetermined operational parameters associated with engine assembly **10**. Moreover, in the exemplary embodiment, such predetermined operational parameters include, but are not limited to, temperature, torque, humidity, fuel heating values, fuel flow, and turbine speed. Alternatively, system **100** includes instrumentation that enables measurements of any parameters that facilitate operation of engine assembly **10** as described herein.

Specifically, in the exemplary embodiment, instruments **102** include a temperature measuring instrument **104**, a torque measuring instrument **106**, a humidity measuring instrument **108**, a lower heating value measuring instrument **110**, a fuel flow measuring instrument **112**, and a turbine speed measuring instrument **114**. Moreover, each measuring instrument **102** generates at least one electronic or electrical signal that is representative of the operational process being measured. Also specifically, in the exemplary embodiment, each of instruments **104**, **106**, **108**, **110**, **112**, and **114** generates a respective temperature, or T_{Actual }signal **118**, a torque, or τ_{Actual }signal **120**, a humidity, or H_{Actual }signal **122**, a lower heating value, or LHV_{Actual }signal **124**, a fuel flow, or F_{Actual }signal **126**, and a turbine speed, or S_{Actual }signal **128**. Signals **118** through **128** are collectively referred to herein as operating signals **130** wherein T is temperature, τ is torque, H is humidity, LHV is lower heating value, F is fuel flow, and S is turbine speed.

In the exemplary embodiment, operating signals **130** are generated during operation of engine assembly **10** and are measured using a deterministic control and scanning scheme. For example, in one embodiment, the scan time of the scanning scheme is approximately one millisecond and operating signals **130** are time-stamped. In an alternative embodiment, operating signals **130** are measured using a non-deterministic control and scanning scheme. In another embodiment, the scan time of the scanning scheme may be any suitable elapsed time increment that enables monitoring system **100** to function as described herein.

System **100** also includes at least one processor **200**. Each instrument **102** is coupled in data communication with processor **200**, wherein processor **200** receives and uses signals **130** as described herein. Processor **200** processes signals **130** received from measuring instruments **102**. As used herein, the term “processor” is not limited to just those integrated circuits referred to in the art as a processor, but broadly refers to at least one microcontroller, microcomputer, programmable logic controller (PLC), application-specific integrated circuit, and other programmable circuits, and these terms are used interchangeably herein. In the exemplary embodiment, processor **200** includes an electronic memory (not shown) that includes, but is not limited to, a computer-readable medium such as a random access member (RAM). Alternatively, a floppy disk, a compact disc-read only memory (CD-ROM), a magneto-optical disk (MOD), and/or a digital versatile disc (DVD) may also be used.

FIG. 3 is a block diagram of an exemplary logic module **300** that may be used with monitoring system **100** (shown in FIG. 2). Monitoring system **100** includes at least one logic module **300**, and in the exemplary embodiment, logic module **300** is implemented within processor **200**. Logic module **300** includes a data storage block **301** that is coupled in data communication with instruments **102**. Data storage block **301** is receives and stores operating signals **130** transmitted from instruments **102** for later use, that is, other than immediate use, within monitoring system **100** as described herein.

Logic module **300** also includes a mathematical model function block **302**. Mathematical model function block **302** includes at least one algorithm (not shown) that at least partially represents a mathematical model of a process. Such an algorithm, or algorithms, is/are herein referred to as a mathematical model (not shown). In the exemplary embodiment, the mathematical model is substantially representative of the real-world performance and behavior of at least a portion of at least one process associated with engine assembly **10** (shown in FIGS. 1 and 2), and is a function of a plurality of measured parameters. For example, in the exemplary embodiment, specific fuel consumption (SFC) is a function of a plurality of variables, or measured parameters including temperature, torque, humidity, lower heating value, fuel flow, and turbine speed. Alternatively, any process with any number of variables, or functional parameters is monitored as described herein.

As is known, SFC for a gas turbine engine with a mechanical shaft output, for example, a turboshaft engine, is substantially a measure of efficiency. Moreover, SFC measures a mass of fuel needed to provide a specific power output for a specific period of time. SFC is typically measured in units of kilograms per kilowatt-hours (kg/kW-h) (pounds-mass per horsepower-hour (lb/HP-h)), that is, kilograms (pounds) of fuel consumed for every kilowatt (horsepower) generated during one hour of operations. As engine efficiency increases, SFC decreases. Conversely, as engine efficiency decreases, SFC increases.

More specifically, the following expression, or algorithm is representative of specific fuel consumption:

SFC(*T,τ,H*,LHV,*F,S*) (1)

wherein T is temperature, τ is torque, H is humidity, LHV is lower heating value, F is fuel flow, and S is turbine speed. The variables associated with algorithm (1) are substantially similar to the respective parameters measured by instruments **102** and represented by signals **130**. That is, each of the six measured parameters T, τ, H, LHV, F, and S has a known proportionate effect on specific fuel consumption of engine assembly **10**.

In the exemplary embodiment, the mathematical model implemented within function block **302** is static. Alternatively, the mathematical model implemented within function block **302** is dynamic. Moreover, in the exemplary embodiment, the mathematical model implemented within function block **302** is derived by backfitting data collected via instruments **102** over a predetermined period of time through a predetermined plurality of parameter transients within predetermined ranges. Backfitting as used herein is defined as generating mathematical expressions that substantially predict future results using collected data substantially representative of past results. Methods of backfitting include, but are not limited to, rigorous statistical analysis, standard regression analyses, and “trial and error” algorithm determinations.

Alternatively, the mathematical model implemented within function block **302** is derived by methods that include, but are not limited to, implementing the model within a neural network. Function block **302** generates and transmits a mathematical model signal **304** for use within logic module **300**. Mathematical model signal **304** includes model data, or more specifically, values for each of temperature (T_{Model}), torque (τ_{Model}), humidity (H_{Model}), lower heating value (LHV_{Model}), fuel flow (F_{Model}), and turbine speed (S_{Model}).

Logic module **300** also includes at least one comparator function block **306**. Function block **306** is coupled in data communication with data storage block **301**, function block **302** and instruments **102**. Function block **306** receives at least one mathematical model signal **304** from function block **302** and operating signals **130** from instruments **102** and/or data storage block **301**. Function block **306** uses at least one comparator algorithm (not shown) to facilitate comparing mathematical model signal **304** with each of operating signals **130**. As such, function block **306** generates a plurality of residual value signals **308**, that is, one signal for each of the six measured parameters per the following algorithms:

*T*_{R}*=T*_{Model}*−T*_{Actual} (2)

τ_{R}=τ_{Model}−τ_{Actual} (3)

*H*_{R}*=H*_{Model}*−H*_{Actual} (4)

LHV_{R}=LHV_{Model}−LHV_{Actual} (5)

*F*_{R}*=F*_{Model}*−F*_{Actual} (6)

*S*_{R}*=S*_{Model}*−S*_{Actual} (7)

wherein each residual value signal generated by algorithms (2) through (7) above are substantially representative of a difference between the appropriate portions of mathematical model signal **304** and each of respective signals **118** through **128**. Therefore, such differences as generated within signals **308** are representative of deviations between the substantially real world model of the process and actual measurements.

Function block **306** also sorts residual value signals **308** chronologically, that is, in time order of receipt. Each of residual value signals **308** is transmitted by block **306** for use within logic module **300** as described herein.

Logic module **300** also includes at least one differentiator function block **310** that is coupled in data communication with function block **302** and receives mathematical model signal **304** transmitted from function block **302**. Function block **310** also uses at least one partial differentiation algorithm (shown below) to facilitate performing at least one partial differentiation, or decomposition operation on at least a portion of signal **304** and to generate a plurality of partially differentiated, or decomposed signals **312**. Therefore, function block **310** effectively decomposes the mathematical model into its constituent variables, and signals **312** substantially represent decomposed portions of the mathematical model. Moreover, function block **310** transmits signals **312** for use within logic module **300** as described herein.

For example, in the exemplary embodiment, the following partial differentiation, or decomposition algorithms are used wherein SFC, T, τ, H, LHV, F, and S are as defined above:

*A*_{1}=(δSFC/δ*T*) (8)

*A*_{2}=(δSFC/δτ) (9)

*A*_{3}=(δSFC/δ*H*) (10)

*A*_{4}=(δSFC/δLHV) (11)

*A*_{5}=(δSFC/δ*F*) (12)

*A*_{6}=(δSFC/δ*S*) (13)

wherein A_{1}, A_{2}, A_{3}, A_{4}, A_{5}, and A_{6 }are each short-hand expressions that substantially represent predetermined weighted values for each of the six variables relied upon in determining SFC as defined in algorithm (1). That is, algorithms (8) through (13) are mathematical representations of a fractional impact of how much each variable affects the overall process. Therefore, as used herein, decomposition is defined as determining the fractional impact of how much each variable affects the overall process.

Specifically, A_{1 }is substantially representative of a weighted effect that the T variable has on SFC. Also, specifically, A_{2 }is substantially representative of a weighted effect that the τ variable has on SFC. Further, specifically, A_{3 }is substantially representative of a weighted effect that the H variable has on SFC. Moreover, specifically, A_{4 }is substantially representative of a weighted effect that the LHV variable has on SFC. Also, specifically, A_{5 }is substantially representative of a weighted effect that the F variable has on SFC. Further, specifically, A_{6 }is substantially representative of a weighted effect that the S variable has on SFC. As such, algorithms (8) through (13) implemented within function block **310** generate partially differentiated signals **312** for each measured parameter. In the exemplary embodiment, algorithms (8) through (13) are statically resident within function block **310**. Alternatively, algorithms (8) through (13) are dynamically resident within function block **310**.

Logic module **300** also includes a change determination function block **314**. Function block **314** is coupled in data communication with function blocks **306** and **310** and receives signals **308** and signals **312** from function blocks **306** and **310**, respectively. Function block **314** uses at least one change determination algorithm (below) to facilitate generating a set of signals proportional to both the associated residual values and their relative effect on SFC. Specifically, function block **314** uses at least one multiplication algorithm (shown below) to determine a plurality of values representative of the associated weighted effects of the deviations of actual measurements from the model and generate a plurality of weighted deviation signals **316**. Moreover, function block **314** transmits signals **316** for use within logic module **300** as described herein. At least a portion of the decomposition of the mathematical model performed within function block **310** is transmitted within signals **316**.

For example, in the exemplary embodiment, the following multiplication algorithms are used wherein T_{R}, τ_{R}, H_{R}, LHV_{R}, F_{R}, S_{R}, and A_{1 }through A_{6 }are as defined above:

*B*_{1}*=A*_{1}**T*_{R} (14)

*B*_{2}*=A*_{2}**τ*_{R} (15)

*B*_{3}*=A*_{3}**H*_{R} (16)

*B*_{4}*=A*_{4}**LHV*_{R} (17)

*B*_{5}*=A*_{5}**F*_{R} (18)

*B*_{6}*=A*_{6}**S*_{R} (19)

wherein B_{1 }through B_{6 }substantially represent the plurality of values representative of the associated weighted effects of the deviations of actual measurements from the model. Therefore, weighted effect signals **316** are substantially representative of SFC effectively decomposed into the six associated parameters as described herein.

Specifically, B_{1 }substantially represents the proportional, or weighted, effect a change in temperature would have, or is having, upon SFC. Also, specifically, B_{2 }substantially represents the proportional, or weighted, effect a change in torque would have, or is having, upon SFC. Further, specifically, B_{3 }substantially represents the proportional, or weighted, effect a change in humidity would have, or is having, upon SFC. Also, specifically, B_{4 }substantially represents the proportional, or weighted, effect a change in lower heating value would have, or is having, upon SFC. Further, specifically, B_{5 }substantially represents the proportional, or weighted, effect a change in fuel flow would have, or is having, upon SFC. Also, specifically, B_{6 }substantially represents the proportional, or weighted, effect a change in turbine speed would have, or is having, upon SFC.

Logic module **300** also includes a summation function block **332** that is coupled in electronic communication with function block **314** and that receives signals **316** transmitted from function block **314**. Moreover, function block **332** sums each of associated signals **316** representing D_{1 }through D_{6 }using the following algorithm:

ΔSFC=*B*_{1}*+B*_{2}*+B*_{3}*+B*_{4}*+B*_{5}*+B*_{6} (20)

wherein ΔSFC is substantially representative of the overall change in SFC as a result of the sum of the each individual change associated with each of the six parameters. Function block **332** generates and transmits a summation signal **334** that is used within monitoring system **100**.

Monitoring system **100** also includes a calculated SFC determination function block **400** implemented within processor **200**. Function block **400** receives operational signals **130** from instruments **102**, as well as a plurality of historical data signals **402** from data storage block **301**. Signals **402** include historical signals associated with the six variables, or parameters that affect SFC, that is, temperature, torque, humidity, lower heating value, fuel flow, and turbine speed. An algorithm substantially similar to algorithm (1) is implemented within function block **400** to determine a plurality of SFC values as a function of time. A SFC signal **404** that is substantially representative of the calculated SFC values is generated and transmitted. Signal **404** is transmitted to a first output device **406**.

An exemplary method of multivariate analysis using a mathematical model for generating model data and actual data for more than one variable includes for each variable, determining a difference between the model data and the actual data. The model data is substantially representative of more than one variable. The method also includes for each variable, determining a fractional impact on performance. The method further includes for each variable, determining a weighted deviation based on the determined difference and the determined fractional impact. The method also includes transmitting the weighted deviation to output device **406** and/or **408**.

FIG. 4 is a graphical view **500** of an exemplary output generated by monitoring system **100** (shown in FIG. 2). Specifically, graphical view **500** is substantially representative of a signal **404** generated by monitoring system **100**, or more specifically, by calculated SFC determination function block **400** (shown in FIG. 3). Graph **500** includes an ordinate (y-axis) **502** that is substantially representative of SFC as described herein. In the exemplary embodiment, y-axis **502** is graduated in units of kg/kW-h (lb/HP-h). Such graduations include a plurality of substantially identical increments with y_{0 }as a centralized measurement and y_{1 }and y_{2 }as positive increments above y_{0 }and y_{−1 }and y_{−2 }as negative increments below y_{0}. Alternatively, y-axis **502** has any units that facilitates operation of monitoring system **100** as described herein including, not being limited to, unitless graduations and specialized indexed graduations. In the exemplary embodiment, y_{0 }is a predetermined value of SFC representative of a predetermined efficiency of engine assembly **10** (shown in FIGS. 1 and 2).

Graph **500** also includes an abscissa (x-axis) **504** that is substantially representative of time with appropriate temporal units. Graph **500** further includes an SFC curve **506** that substantially represents a behavior of SFC as a function of time coincident with signal **404** generated and transmitted by function block **400** (both shown in FIG. 3).

Curve **506** includes a first portion **508** that substantially represents values of SFC typically expected during operation of engine assembly **10** (shown in FIG. 1). Consistent with the correlation between SFC and engine efficiency as described above, portion **508** indicates engine assembly **10** operating with an SFC above y_{0 }that is also rep efficiency of engine assembly **10** representative of an efficiency below that associated with y_{0}. That is, all components of engine assembly **10** for a given power output of engine assembly **10**, including instruments **102**, are operating substantially within expected operational parameters for such given power output of engine assembly **10** with such operating efficiency.

Curve **506** also includes a second portion **510** that substantially represents values of SFC during a change in at least one of the measured parameters, wherein the change in the at least one measured parameter manifests itself as a change in SFC starting at after a time t_{0 }as indicated by a first distinguishing temporal line **512**. The transient ends at a time t_{1 }as indicated by a second distinguishing temporal line **514**. During this transient, the efficiency of engine assembly **10** improves as SFC decreases.

Curve **506** further includes a third portion **516** that substantially represents values of SFC after the change in at least one of the measured parameters, wherein the change in the at least one measured parameter manifests itself as a semi-permanent decrease in SFC and a semi-permanent increase in the efficiency of engine assembly **10** after time t_{1}.

Referring again to FIG. 3, monitoring system **100** also includes a second output device **408** that receives weighted effect signals **316** and summation signals **334** from function blocks **314** and **332**, respectively. First output device **406** and second output device **408** may be the same device. Device **408** displays a summation graphical image (not shown in FIG. 3) that is substantially representative of summation signal **334**. Device **408** also displays a multivariate graphical image (not shown in FIG. 3) wherein each of the plurality of decomposed signals, or weighted effect signals **316**, is displayed in temporal relation to the summation signal graphical image.

FIG. 5 is a graphical view **600** of an exemplary output generated by logic module **300** (shown in FIG. 3). Specifically, graphical view **600** is substantially representative of signal **334** and a portion of signals **316** generated by logic module **300** (all shown in FIG. 3). More specifically, graphical view **600** is substantially representative of a change in SFC and the associated six parameters described above, that is, T, τ, H, LHV, F, and S, with respect to time.

Graph **600** includes an ordinate (y-axis) **602** that includes unitless graduations having a substantially identical increments with y′_{0 }as a centralized measurement, y′_{1 }as a positive increment above y_{0}, and y′_{−1 }as a negative increment below y′_{0}. Alternatively, y-axis **602** has any units that facilitate operation of monitoring system **100** as described herein including, not being limited to, specialized indexed graduations as well as units of SFC and the six associated parameters per unit time. In the exemplary embodiment, y′_{0 }is zero, thereby representing a zero rate of change of SFC and the six associated parameters.

In addition, graphical view **600** includes a plurality of curves substantially representative of a plurality of signals that include a rate of change of SFC and two decomposed parameter signals. Specifically, graphical view **600** includes a ΔSFC curve **606** that is substantially representative of summation signal **334**, as generated by function block **332** (both shown in FIG. 3). Also, view **600** includes first and second decomposed parameter curves **608** and **610**, respectively, that substantially represent two of the six weighted deviation signals **316**. In general, graphical view **600** facilitates substantially associating a first decomposed signal with at least a portion of the summation signal graphical image; that is ΔSFC curve **606** and substantially excluding a second decomposed signal from association with at least a portion of ΔSFC curve **606**.

ΔSFC curve **606** includes a first portion **612** that substantially represents a zero value for ΔSFC. Portion **612** is indicative of an overall rate of change, or slope, of portion **508** (shown in FIG. 4), wherein the smaller changes in SFC curve **506** (shown in FIG. 4) are not shown in curve **606**. In the exemplary embodiment, such smaller changes generally are a result of minute changes in one of the six associated parameters and electronic noise within monitoring system **100**. Such minute changes in SFC and system noise are mitigated via filters and/or smoothing functions (neither shown) programmed within processor **200**. FIG. 5 shows portion **612** slightly below the y′_{0 }graduation for illustrative clarity.

ΔSFC curve **606** also includes a second portion **614** that substantially represents a negative value for ΔSFC. Specifically, second portion **614** illustrates a negative slope associated with second portion **510** (shown in FIG. 4) starting at time t_{0 }as indicated by first distinguishing temporal line **512** and ending at time t_{1 }as indicated by second distinguishing temporal line **514**. Portion **614** is smoothed via filtering and noise mitigation in a manner substantially similar to that of portion **612**.

ΔSFC curve **606** further includes a third portion **616** that substantially represents a zero value for ΔSFC after time t_{1 }as indicated by second distinguishing temporal line **514**. Portion **616** is smoothed via filtering and noise mitigation in a manner substantially similar to that of portion **612**. Also, FIG. 5 shows portion **616** offset from the y′_{0 }graduation for illustrative clarity.

First decomposed parameter curve **608** is substantially constant throughout the SFC transient as illustrated via curve **506** initiating at time t_{0}, or line **512**. An apparent lack of a change in the first parameter represented by first parameter curve **608** throughout the transient facilitates eliminating the first parameter from being a significant cause of the transient. FIG. 5 shows curve **608** slightly above the y′_{0 }graduation for illustrative clarity.

Second decomposed parameter curve **610** illustrates a transient of the associated second parameter. Curve **610** indicates a substantial change in behavior of the second parameter initiating at time t_{0}, or line **512**. An apparent coincident change in behavior at line **512** of curves **506**, **606**, and **610** facilitates implicating the second parameter as being a leading driver of the transient.

Curve **610** includes a first portion **618** that substantially represents a zero value for the rate of change, or slope, of the second decomposed parameter. Portion **618** is smoothed via filtering and noise mitigation in a manner substantially similar to that of portion **612**. FIG. 5 shows portion **618** slightly above the y′_{0 }graduation for illustrative clarity.

Curve **610** also includes a second portion **620** that substantially represents a negative value for the second decomposed parameter. Specifically, second portion **620** illustrates a negative slope associated with the second decomposed parameter starting at time t_{0 }as indicated by first distinguishing temporal line **512** and ending at time t_{1 }as indicated by second distinguishing temporal line **514**. Portion **620** is smoothed via filtering and noise mitigation in a manner substantially similar to that of portion **612**.

Curve **610** further includes a third portion **622** that substantially represents a zero value for the second decomposed parameter after time t_{1 }as indicated by second distinguishing temporal line **514**. Portion **622** is smoothed via filtering and noise mitigation in a manner substantially similar to that of portion **612**. Also, FIG. 5 shows portion **622** offset from the y′_{0 }graduation for illustrative clarity.

At least one known method of monitoring the operation of an engine includes generating a plurality of trend charts, and/or generating a plurality of data tables, and having a user attempt to determine the relative proportion of how each parameter is impacting the process. If more than one parameter or issue is adversely affecting any of the processes, assessing the extent of how each parameter impact the process may be time-consuming, difficult, and/or inaccurate. For example, it is possible that resources may be expended on analyzing an anomaly embedded within associated operational data, wherein such anomaly may ultimately be determined to not adversely impact the parameters and processes being evaluated. In addition, for those trend charts that indicate a plurality of parameters, such parameters may be scaled such that an apparent disproportionate relationship between parameters that do not share a common unit of measurement may be insinuated.

Utilizing graphical multivariate analysis techniques as described herein facilitates identifying associated parameters that may be driving certain process behaviors while also facilitating elimination of those parameters that are less likely to be driving such process behaviors. In an active operating environment, such short-term, or substantially real-time, determinations may facilitate early isolation of the initiating conditions and subsequent restoration of process parameters to desired ranges. Moreover, early elimination of apparently unaffected parameters facilitates focusing operation and maintenance resources toward higher probability causes. Furthermore, earlier determination of apparent causes of the transient facilitates actions by operating personnel to mitigate deleterious effects on the process and/or equipment by taking preemptive and/or corrective actions within a sort time frame after the transient initiates. Also, erroneous values of operational signals **130** due to drift or failures of instrumentation **102** may be manifested within graph **600** by an independent shift of one of signals **316** with no associated change in SFC curve **506**.

The methods and apparatus as described herein facilitate monitoring engines using graphical multivariate analysis. Specifically, the monitoring system as described above facilitates efficient and effective identification of performance deviations associated with predetermined processes and engine components. More specifically, the above-described monitoring system provides a method to evaluate a trend chart by decomposing such trend chart. Therefore, the monitoring system improves the ability of a user to identify an anomaly on the trend chart, and identify whether or not the anomaly is effecting the overall operation of the engine.

Exemplary embodiments of engine monitoring systems are described above in detail. The methods, apparatus and systems are not limited to the specific embodiments described herein nor to the specific illustrated engine monitoring systems.

While the invention has been described in terms of various specific embodiments, those skilled in the art will recognize that the invention can be practiced with modification within the spirit and scope of the claims.

1. A method for multivariate analysis using a mathematical model for generating model data and actual data for more than one variable, wherein the model data is substantially representative of more than one variable, said method comprising:

for each variable, determining a difference between the model data stored in a logic module of a machinery monitoring system and the actual data received from machinery being analyzed;

for each variable, determining by the logic module a factional impact on performance;

for each variable, determining by the logic module a weighted deviation based on the determined difference and the determined factional impact; and

transmitting the weighted deviation to an output device.

for each variable, determining a difference between the model data stored in a logic module of a machinery monitoring system and the actual data received from machinery being analyzed;

for each variable, determining by the logic module a factional impact on performance;

for each variable, determining by the logic module a weighted deviation based on the determined difference and the determined factional impact; and

transmitting the weighted deviation to an output device.

2. A method in accordance with claim 1 wherein determining a difference between the model data and the actual data comprises:

for each variable, collecting operational data that is substantially representative of performance; and

for each variable, determining residuals that are substantially representative of a difference between at least a portion of the operational data and at least a portion of the model data.

for each variable, collecting operational data that is substantially representative of performance; and

for each variable, determining residuals that are substantially representative of a difference between at least a portion of the operational data and at least a portion of the model data.

3. A method in accordance with claim 1 wherein determining a factional impact on performance comprises decomposing at least a portion of the mathematical model.

4. A method in accordance with claim 3 wherein decomposing at least a portion of the mathematical model comprises determining the fractional impact of each variable by executing a partial differential operation on the mathematical model.

5. A method in accordance with claim 1 wherein determining a weighted deviation based on the determined difference and the determined factional impact comprises, for each variable, multiplying the determined difference by the determined fractional impact.

6. A method in accordance with claim 1 wherein transmitting the weighted deviation to an output device comprises:

summing the weighted deviations and displaying a summation graphical image; and

displaying a multivariate graphical image of each of the weighted deviations in temporal relation to the summation graphical image, thereby facilitating graphical multivariate analysis.

summing the weighted deviations and displaying a summation graphical image; and

displaying a multivariate graphical image of each of the weighted deviations in temporal relation to the summation graphical image, thereby facilitating graphical multivariate analysis.

7. A method in accordance with claim 6 wherein displaying a multivariate graphical image comprises at least one of:

associating a first weighted deviation with at least a portion of the summation graphical image; and

substantially excluding a second weighted deviation from association with at least a portion of the summation graphical image.

associating a first weighted deviation with at least a portion of the summation graphical image; and

substantially excluding a second weighted deviation from association with at least a portion of the summation graphical image.

8. A method in accordance with claim 1 further comprising programming the mathematical model comprising one of:

storing a plurality of static algorithms;

storing a plurality of dynamic algorithms that are modified by at least one predetermined operational condition; and

backfitting operational data.

storing a plurality of static algorithms;

storing a plurality of dynamic algorithms that are modified by at least one predetermined operational condition; and

backfitting operational data.

9. A monitoring system for use with a turbine engine comprising:

at least one processor coupled in data communication with a plurality of measuring instruments, wherein each of said measuring instruments is associated with at least one of a plurality of variables and collects actual engine data;

a mathematical model at least partially resident within said at least one processor, wherein said model is substantially representative of at least a portion of an engine process and is at least partially derived from at least one of the variables, wherein the mathematical model stores model data;

at least one algorithm programmed within said at least one processor, wherein the at least one algorithm is programmed to:

determine a fractional impact on engine performance for each variable;

determine a difference between said model and the actual data; and

determine a weighted deviation based on the determined difference and the determined fractional impact; and

at least one output device coupled in data communication with said at least one processor.

at least one processor coupled in data communication with a plurality of measuring instruments, wherein each of said measuring instruments is associated with at least one of a plurality of variables and collects actual engine data;

a mathematical model at least partially resident within said at least one processor, wherein said model is substantially representative of at least a portion of an engine process and is at least partially derived from at least one of the variables, wherein the mathematical model stores model data;

at least one algorithm programmed within said at least one processor, wherein the at least one algorithm is programmed to:

determine a fractional impact on engine performance for each variable;

determine a difference between said model and the actual data; and

determine a weighted deviation based on the determined difference and the determined fractional impact; and

at least one output device coupled in data communication with said at least one processor.

determine a fractional impact on engine performance for each variable;

determine a difference between said model and the actual data; and

determine a weighted deviation based on the determined difference and the determined fractional impact; and

10. A monitoring system in accordance with claim 9 wherein the mathematical model is one of

a plurality of algorithms statically stored within said at least one processor; and

a plurality of algorithms stored within said at least one processor programmed to be dynamically modified by at least one predetermined operational condition.

a plurality of algorithms statically stored within said at least one processor; and

a plurality of algorithms stored within said at least one processor programmed to be dynamically modified by at least one predetermined operational condition.

11. A monitoring system in accordance with claim 9 wherein the at least one algorithm is further programmed to decompose at least a portion of said mathematical model with a least one partial differential operation, thereby facilitating determination of a fractional impact on engine performance for each variable.

12. A monitoring system in accordance with claim 11 wherein the at least one algorithm is further programmed to generate a residual for each variable, wherein each residual is substantially representative of a difference between a portion of the model data and the actual data.

13. A monitoring system in accordance with claim 12 wherein the at least one algorithm is further programmed to determine the weighted deviation for each variable by multiplying the fractional impact for each variable by the residual for each variable.

14. A monitoring system in accordance with claim 13 wherein the at least one algorithm is further programmed to sum the weighted deviations.

15. A gas turbine engine comprising:

at least one compressor;

at least one turbine rotatably coupled to said at least one compressor; and

a monitoring system coupled in data communication with at least one of said at least one compressor and said at least one turbine, said monitoring system comprising:

at least one processor coupled in data communication with a plurality of measuring instruments, wherein each of said measuring instruments is associated with at least one of a plurality of variables and collects actual engine data;

a mathematical model at least partially resident within said at least one processor, wherein said model is substantially representative of at least a portion of an engine process and is at least partially derived from at least one of the variables, wherein the mathematical model stores model data;

at least one algorithm programmed within said at least one processor, wherein the at least one algorithm is programmed to:
determine a weighted deviation based on the determined difference and the determined fractional impact; and

determine a fractional impact on engine performance for each variable;

determine a difference between said model and the actual data; and

at least one output device coupled in data communication with said at least one processor.

at least one compressor;

at least one turbine rotatably coupled to said at least one compressor; and

a monitoring system coupled in data communication with at least one of said at least one compressor and said at least one turbine, said monitoring system comprising:
at least one processor coupled in data communication with a plurality of measuring instruments, wherein each of said measuring instruments is associated with at least one of a plurality of variables and collects actual engine data;
a mathematical model at least partially resident within said at least one processor, wherein said model is substantially representative of at least a portion of an engine process and is at least partially derived from at least one of the variables, wherein the mathematical model stores model data;
at least one algorithm programmed within said at least one processor, wherein the at least one algorithm is programmed to:
determine a weighted deviation based on the determined difference and the determined fractional impact; and

determine a fractional impact on engine performance for each variable;

determine a difference between said model and the actual data; and

at least one output device coupled in data communication with said at least one processor.

determine a fractional impact on engine performance for each variable;

determine a difference between said model and the actual data; and

at least one output device coupled in data communication with said at least one processor.

determine a fractional impact on engine performance for each variable;

determine a difference between said model and the actual data; and

16. A gas turbine engine in accordance with claim 15 wherein the mathematical model is one of:

a plurality of algorithms statically stored within said at least one processor; and

a plurality of algorithms stored within said at least one processor programmed to be dynamically modified by at least one predetermined operational condition.

a plurality of algorithms statically stored within said at least one processor; and

17. A gas turbine engine in accordance with claim 15 wherein the at least one algorithm is further programmed to decompose at least a portion of said mathematical model with a least one partial differential operation, thereby facilitating determination of a fractional impact on engine performance for each variable.

18. A gas turbine engine in accordance with claim 17 wherein the at least one algorithm is further programmed to generate a residual for each variable, wherein each residual is substantially representative of a difference between a portion of the model data and the actual data.

19. A gas turbine engine in accordance with claim 18 wherein the at least one algorithm is further programmed to determine the weighted deviation for each variable by multiplying the fractional impact for each variable by the residual for each variable.

20. A gas turbine engine in accordance with claim 19 wherein the at least one algorithm is further programmed to determine the weighted deviation for each variable by multiplying the fractional impact for each variable by the residual for each variable.