Model-based iterative estimation of gas turbine engine component qualities

• US 20070118271A1
• Filed: 11/18/2005
• Published: 05/24/2007
• Est. Priority Date: 11/18/2005
• Status: Active Grant

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Ashby, Malcolm, Wiseman, Matthew
• Granted Patent: US 7,505,844 B2
• Time in Patent Office: Days
• US Class Current: 701/100
• CPC Class Codes: G05B 23/0254 based on a quantitative model, e.g. mathematical relationships between inputs and outputs; functions: observer, Kalman filter, residual calculation, Neural Networks
• CPC Class Codes: G05B 23/0283 Predictive maintenance, e.g. involving the monitoring of a system and, based on the monitoring results, taking decisions on the maintenance schedule of the monitored system; Estimating remaining useful life [RUL] preventive maintenance, i.e. planning maintenance according to the available resources without monitoring the system G06Q10/06

Abstract

A method of estimating quality parameters for a plurality of engine components of a gas turbine engine is provided. The engine components have at least one sensor responsive to the engine component operation. The method includes providing an engine model having virtual sensor values and quality parameters corresponding to the plurality of sensors of the engine components; comparing the virtual sensor values to actual sensor values of the plurality of sensors of the engine components to determine the difference between the actual and virtual sensor values; amplifying the difference by a predetermined gain; generating a plurality of quality parameter deltas in response to the sensed difference; iteratively updating the embedded engine model by inputting a predetermined portion of the generated quality parameter deltas into the embedded engine model; adjusting the embedded engine model for engine operating conditions; and recalculating the virtual sensor values.

Claims

1. A method of estimating quality parameters for a plurality of engine components of a gas turbine engine, the engine components having a plurality of sensors responsive to the engine component operation, the method comprising:

generating virtual sensor values and quality parameters with an engine model, the virtual sensor values corresponding to values from a plurality of sensors of the plurality of engine components;

comparing the virtual sensor values to actual sensor values provided by the plurality of sensors to calculate a difference between the actual and virtual sensor values;

amplifying the difference by a predetermined gain;

generating a plurality of quality parameter deltas in response to the difference;

iteratively updating the engine model by inputting a predetermined portion of the generated plurality of quality parameter deltas into the engine model;

adjusting the embedded engine model for engine operating conditions; and

repeating the steps of generating virtual sensor values, comparing the virtual sensor values, amplifying the difference, generating a plurality of quality parameter deltas and updating the embedded model.

2. The method of claim 1, also including generating a set of quality adjustment parameters for the plurality of engine components.

3. The method of claim 1, wherein the engine model is one of a physics-based model, regression model or neural net model.

4. The method of claim 1, wherein the engine model includes models of a fan, at least one compressor, a high-pressure turbine and a low pressure-turbine.

5. The method of claim 4, wherein the engine model also includes models of an air inlet, a burner, a core nozzle, a bypass duct, a bypass nozzle and an exhaust.

6. The method of claim 5, wherein each model of a component of the engine model is individually modeled and assembled into a component level model.

7. The method of claim 1, wherein the quality parameter deltas include flow modifier and efficiency modifier.

8. The method of claim 1, wherein the predetermined gain is computed based on a number of virtual sensor values and quality parameters, and the virtual sensor values are a function of the quality parameters.

9. The method of claim 8, wherein generating of the quality parameters deltas includes computing the quality parameter deltas as a function of a design constant multiplied by an error vector and an algorithm, wherein the algorithm is a Jacobian function defined as:

$J_Y=\left(\begin{array}{ccc}\frac{\partial <br><br>y_1}{\partial <br><br>q_1}& \mathrm{\cdots <br><br>}& \frac{\partial <br><br>y_1}{\partial <br><br>q_m}\\ \mathrm{⋮<br><br>}& \mathrm{⋰<br><br>}& \mathrm{⋮<br><br>}\\ \frac{\partial <br><br>y_n}{\partial <br><br>q_1}& \mathrm{\cdots <br><br>}& \frac{\partial <br><br>y_n}{\partial <br><br>q_m}\end{array}\right)$

where in n represents the number of virtual sensor values;

m represent the number of quality parameters;

y_i corresponds to the i^th computed sensor value; and

q_i represents the i^th component quality parameter.

10. The method of claim 9, wherein the Jacobian function is a derivative of a function Y(Q), wherein the function Y(Q) is defined as:

$Y<br><br>\left(Q\right)=\left(\begin{array}{c}{y}_1<br><br>\left(q_1,\mathrm{\dots <br><br>}<br><br>\text{\hspace{1em}<br><br>},q_m\right)\\ y_2<br><br>\left(q_1,\mathrm{\dots <br><br>}<br><br>\text{\hspace{1em}<br><br>},q_m\right)\\ \mathrm{⋮<br><br>}\\ y_n<br><br>\left(q_1,\mathrm{\dots <br><br>}<br><br>\text{\hspace{1em}<br><br>},q_m\right)\end{array}\right)$

11. The method of claim 10, wherein generating a plurality of quality parameter deltas includes computing the quality parameter deltas as the product of the predetermined gain having a gain value h, the difference having an error vector E, and the matrix J_Y, the error vector E being defined as:

E=(e₁ . . . e_n)=(s₁−

y₁ . . . s_n−

y_n), s_i representing the i^th actual engine sensor value.

12. The method of claim 11, wherein at each iteration of the algorithm, the quality parameter deltas Δ

Q are computed based on the current values of E, and the quality parameters updated according to the following equation;

[Q]_k+1=[Q]_k+Δ

Q

where [Q]_k is the quality parameter vector at an iteration step k.

13. A system for estimating quality parameters for a plurality of engine components of a gas turbine engine comprising:

an engine model configured to compute virtual sensor values and quality parameters, the virtual sensor values corresponding to sensor values from a plurality of sensors of the plurality of engine components;

a comparator to compare the virtual sensor values to actual sensor values from the plurality of sensors to calculate the difference between the actual and virtual sensor values;

an amplifier having a predetermined gain;

a processor unit to calculate a plurality of quality parameter deltas in response to the calculated difference between actual and virtual sensor values;

automatic switching means for iteratively updating the engine model by switching a predetermined portion of the calculated quality parameter deltas into the engine model; and

the engine model having at least one input to receive engine operating conditions, the engine operating conditions being used to generate the virtual sensor values provided to the comparator.

14. The system of claim 13, wherein the engine model includes an output configured to provide a set of quality adjustment parameters for the plurality of engine components.

15. The system of claim 13, wherein the engine model is a physics-based, model, regression model or neural net model.

16. The system of claim 13, wherein the engine model includes models of a fan, at least one compressor, a high-pressure turbine and a low pressure-turbine.

17. The system of claim 16, wherein the engine model also includes models of an air inlet, a burner, a core nozzle, a bypass duct, a bypass nozzle and an exhaust.

18. The system of claim 17, wherein each model of a component of the engine model is individually modeled and assembled into a component level model.

19. The system of claim 13, wherein the quality parameter deltas include flow modifier and efficiency modifier.

20. The system of claim 13, wherein the predetermined gain is computed based on a number of virtual sensor values and quality parameters, and the virtual sensor values are a function of the quality parameters.

21. The system of claim 20, wherein generating of the quality parameters deltas includes computing the quality parameter deltas as a function of a design constant multiplied by an error vector and an algorithm, wherein the algorithm is a Jacobian function defined as:

$J_Y=\left(\begin{array}{ccc}\frac{\partial <br><br>y_1}{\partial <br><br>q_1}& \mathrm{\cdots <br><br>}& \frac{\partial <br><br>y_1}{\partial <br><br>q_m}\\ \mathrm{⋮<br><br>}& \mathrm{⋰<br><br>}& \mathrm{⋮<br><br>}\\ \frac{\partial <br><br>y_n}{\partial <br><br>q_1}& \mathrm{\cdots <br><br>}& \frac{\partial <br><br>y_n}{\partial <br><br>q_m}\end{array}\right)$

wherein n represents the number of virtual sensor values;

m represent the number of quality parameters;

y_i corresponds to the i^th computed sensor value; and

q_i represents the i^th component quality parameter.

22. The system of claim 21, wherein generating a plurality of quality parameter deltas includes computing the quality parameter deltas as the product of the predetermined gain having a gain value h, the difference having an error vector E, and the matrix J_Y, the error vector E being defined as:

E=(e₁ . . . e_n)=(s₁−

y₁ . . . s_n−

y_n), s_i representing the i^th actual engine sensor value; and

wherein at each iteration of the algorithm, the quality parameter deltas Δ

Q are computed based on the current values of E, and the quality parameters updated according to the following equation;

[Q]_k+1=[Q]_k+Δ

Q

where [Q]_k is the quality parameter vector at an iteration step k.