System and method for estimating turbine engine deterioration rate with noisy data

• US 20070078585A1
• Filed: 09/30/2005
• Published: 04/05/2007
• Est. Priority Date: 09/30/2005
• Status: Active Grant

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Aragones, James, Doganaksoy, Deniz, Pomeroy, Bruce
• Granted Patent: US 7,286,923 B2
• Time in Patent Office: Days
• US Class Current: 701/100
• CPC Class Codes: G01C 23/00 Combined instruments indicating more than one navigational value, e.g. for aircraft; Combined measuring devices for measuring two or more variables of movement, e.g. distance, speed or acceleration

Abstract

A method for monitoring engine performance includes sampling exhaust gas temperature associated with a turbine engine over an interval of operational time of the turbine engine. The method further includes applying a first test to identify statistical outliers on the sampled exhaust gas temperature data and removing identified statistical outliers from the sampled exhaust gas temperature data. Subsequently, the method includes applying a second test to identify step changes in slope of the exhaust gas temperature data and dividing the interval of operational time into one or more segments based upon the identified step changes. Finally, the method includes determining a slope for each segment and combining the segments to obtain a rate of performance deterioration of the turbine engine.

Claims

1. A method for monitoring engine performance, comprising:

sampling exhaust gas temperature data associated with a turbine engine over an interval of operational time of the turbine engine;

applying a first test to identify statistical outliers on the sampled exhaust gas temperature data;

removing identified statistical outliers from the sampled exhaust gas temperature data;

applying a second test to identify step changes in slope of the exhaust gas temperature data;

dividing the interval of operational time into one or more segments based upon the identified step changes;

determining a slope for each segment; and

combining the segments to obtain a rate of performance deterioration of the turbine engine.

2. The method of claim 1, comprising applying the first and second tests to exhaust gas temperature data obtained from a plurality of engines in a fleet and determining an average rate of performance deterioration for the fleet.

3. The method of claim 2, comprising applying a weighted regression fit to exhaust gas temperature data from the plurality of engines at each sampled point of operational time, wherein applying the weighted regression fit further comprises:

determining a weight for each sampled point, wherein the weight at each sampled point is computed as an inverse of standard error for each sampled point, and wherein the standard error is in a slope fit at that point; and

regression fitting the slope fit along the sampled points of operational time incorporating a linear model utilizing said weights.

4. The method of claim 1, further comprising applying a third test to identify one or more X-range gaps in the exhaust gas temperature data, fitting the identified one or more X-range gaps for estimating deterioration rate and deterioration error, and estimating an overall slope with gaps.

5. The method of claim 4, wherein applying a third test to identify one or more X-range gaps further comprises:

utilizing a Loess regression fit to obtain a trend line of slope of the sampled exhaust gas temperature data and determining a standard error at each sampled point of operational time;

computing an upper limit on gap size based upon a pre-selected step size and computed values of averaged standard error of the trend line and an average slope; and

computing X-range differences at each sample point, and identifying a gap when the X-range difference at a point is greater than the upper limit on the gap size.

6. The method of claim 5, wherein the Loess regression fit comprises a low-damping Loess regression fit comprising a polynomial of degree 1 and having a span of less than or equal to about 5%.

7. The method of claim 5, wherein the preselected step size is computed as one standard deviation of the sampled exhaust gas temperature data relative to the trend line.

8. The method of claim 5, further comprising utilizing a high damping linear Loess regression fit to estimate slope and standard error at each point in the range of sampled points and using the Loess regression fit to predict slope and standard error at additional points in an identified X-range gap.

9. The method of claim 1, wherein applying the first test to identify statistical outliers further comprises:

applying a Loess regression fit to the sampled exhaust gas temperature data to obtain a trend line of exhaust gas temperature data;

computing residuals around the trend line, wherein residuals each comprise a difference between an actual data point of the sampled exhaust gas temperature data and a corresponding point on the trend line;

estimating a trimmed standard deviation of the residuals;

standardizing each of the residuals as a fraction of the trimmed standard deviation; and

identifying as statistical outliers all points in which the modulus of the standard residual is greater than a selected value.

10. The method of claim 9, wherein the Loess regression fit comprises a low-damping Loess regression fit comprising a polynomial of degree 1 and having a span of less than or equal to about 5%.

11. The method of claim 9, wherein the trimmed standard deviation comprises standard deviation of the lowest 90% of the residuals.

12. The method of claim 1, wherein applying the second test to identify step changes in slope further comprises:

utilizing a Loess regression fit to obtain a trend line of slope of the sampled exhaust gas temperature data and determining a standard error between the sampled exhaust gas temperature data and the trend line;

computing an average value of the slope of the trend line and determining an average error (dY/dX.avg) of the trend line, wherein the average error is computed as an average of the standard error (dY/dX.se) at each sampled point;

determining a step-detector band having an upper band limit and a lower band limit;

determining step detection limits having an upper confidence limit and a lower confidence limit, wherein the upper confidence limit is computed as a sum of the upper band limit and a multiple of the average error, and the lower confidence limit is computed as a difference of the lower band limit and a multiple of the average error; and

identifying a step change for all points in which the slope does not lie within the step detection limits.

13. The method of claim 12, wherein the Loess regression fit comprises a low-damping Loess regression fit comprising a polynomial of degree 1 and having a span of less than or equal to about 5%.

14. The method of claim 12, wherein the upper band limit is computed as UBL=2*(dY/dX.avg) and the lower band limit is computed as LBL=0.5*(dY/dX.avg).

15. The method of claim 12, further comprising applying a high damping Loess regression fit for each segment and determining performance deterioration rate and an error of performance deterioration rate fit for each segment, wherein the performance deterioration rate at any point is computed as the value of the slope at that point and the error in performance deterioration at any point is computed directly from the standard error at that point.

16. The method of claim 15, wherein applying the high-damping Loess regression fit comprises fitting a polynomial of degree 2 and having a span of less than or equal to about 50%.

17. A method for monitoring engine performance, comprising:

identifying and eliminating any statistical outliers and step changes from exhaust gas temperature data of individual turbine engines;

identifying any X-range gaps in the exhaust gas temperature data of individual turbine engines to segment the exhaust gas temperature data;

obtaining a rate of change of exhaust gas temperature for each turbine engine after identifying and eliminating any statistical outliers and step changes and identifying any X-range gaps; and

combining the rates of change of exhaust gas temperature of individual turbine engines to determine an average rate of change of exhaust gas temperature of a fleet including a plurality of turbine engines.

18. The method of claim 17, comprising applying a weighted regression fit to exhaust gas temperature data from the plurality of turbine engines at each sampled point of operational time, wherein applying the weighted regression fit further comprises:

determining a weight for each sampled point, wherein the weight at each sampled point is computed as weight=1/(dY/dX.se) for each sampled point and wherein dY/dX.se represents a standard error in a slope fit (dY/dX.fit) at that point; and

regression fitting the slope fit (dY/dX.fit) along the sampled points of operational time incorporating a linear model utilizing said weights.

19. A method, comprising:

scheduling downtime for a turbine engine based on a determination of engine performance deterioration, wherein determination of engine performance deterioration comprises identifying and eliminating statistical outliers and step changes from exhaust gas temperature data of the turbine engine and determining the occurrence of an X-range gap in the exhaust gas temperature data of the turbine engine to obtain a rate of change of exhaust gas temperature for the turbine engine therefrom.

20. The method of claim 19, further comprising servicing the turbine engine based upon the determination of a preselected performance deterioration level.

21. A system, comprising:

sensors adapted to sense exhaust gas temperature of one or more turbine engines; and

a performance deterioration rate monitoring system configured for identifying and eliminating statistical outliers and step changes from sensed exhaust gas temperature data of individual turbine engines and determining the occurrence of an X-range gap in said exhaust gas temperature data of individual turbine engines to obtain a rate of change of exhaust gas temperature for each turbine engine therefrom.

22. The system of claim 21, wherein the performance deterioration rate monitoring system is further configured for combining the rates of change of exhaust gas temperature deterioration of individual turbine engines to determine an average rate of change of performance deterioration of the one or more turbine engines in a fleet.

23. A system, comprising:

a tangible media;

computer readable code disposed on the tangible media, wherein the computer readable code comprises;

code adapted to identify and remove any statistical outliers in sampled exhaust gas temperature data of a turbine engine;

code adapted to identify any step changes in slope of the exhaust gas temperature data and divide the interval of operational time into one or more segments based upon the identified step changes; and

code adapted to determine a slope for each segment and to combine the segments to obtain a rate of performance deterioration of the turbine engine therefrom.

24. The system of claim 23, further comprising code adapted to apply a weighted regression fit to exhaust gas temperature data from a plurality of turbine engines in a fleet at each sampled point of operational time to determine an average performance deterioration rate of the fleet.

25. The system of claim 23, further comprising code adapted to identify any X-range gaps in the exhaust gas temperature data and, upon identification of an X-range gap, fitting the identified gap and estimating an overall slope with gaps.