A method for monitoring sudden changes in gas path components under concurrent failures of an aircraft engine control system

Abstract

The present invention discloses a method for diagnosing the failure of abnormal gas path performance under concurrent faults of an aero-engine control system, and uses a virtual sensor reconstruction signal technology to replace the fault sensor signal for gas path fault diagnosis. The method includes: isolating and locating the fault sensor and the actuator through a hybrid Kalman filter matrix; reconstructing the virtual sensor signal based on the mechanism model and the non-faulty sensor signal according to the located faulty sensor; and combining the EKF gas path selection component fault diagnosis. The present invention solves the problem that the engine gas path fault is difficult to diagnose in the case of sensor and actuator failures, and has a high positioning accuracy for the sudden faults of rotating components at different working points within the flight envelope of the aero-engine, providing reliable protection for engine health management and reducing maintenance costs.

Description

Technical Field

[0001] The invention belongs to the technical field of air path fault diagnosis of aircraft engines, and in particular relates to a method for monitoring sudden changes in air path components under concurrent faults of an aircraft engine control system. Background Art

[0002] As the heart of the aircraft, the aircraft engine has a complex structure and a harsh working environment. Engine fault diagnosis technology is an important means to ensure engine performance and reliability and reduce maintenance costs. Failure of gas path components affects the performance and reliability of aircraft engines, and the accuracy of sensor signals directly affects the fault diagnosis results of gas path components. It is necessary to isolate and locate gas path components, sensors, and actuators when they fail.

[0003] The Kalman filter algorithm is a recursive form of linear minimum variance estimation. As long as there are enough sensor measurements available, the Kalman filter method has a high accuracy in the diagnosis of aircraft engine gas path faults. However, due to the harsh working environment of the sensor, it is easy to fail, which easily causes performance monitoring to fail to accurately diagnose engine faults. The present invention is based on the extended Kalman filter method. On the basis of aviation thermodynamics theory, a hot end virtual sensor is established, and a method for isolating sensor and actuator faults and reconstructing fault signals is proposed. Combined with the EKF (Extended Kalman Filter) algorithm, a method for monitoring sudden abnormal changes in engine components is realized, which has high accuracy at different working points within the flight envelope. Summary of the invention

[0004] In view of the above technical problems, the present invention provides a method for monitoring sudden changes in gas path components under concurrent faults of an aircraft engine control system, which uses a hybrid Kalman filter matrix (MHKF) to locate fault sensors and actuator faults, replaces fault sensors with virtual sensors, and combines state monitoring to track sudden degradation of engine components. In view of different sudden failure modes of aircraft engines within the envelope, gas path fault diagnosis of sudden abnormal performance changes of engine components with sensor actuator fault isolation capabilities at different operating points within the flight envelope is achieved.

[0005] Technical solution: To achieve the above purpose, the technical solution adopted by the present invention is:

[0006] A method for monitoring sudden changes in gas path components under concurrent faults of an aircraft engine control system comprises the following steps:

[0007] Step 1) isolating and locating sensor and actuator faults based on a hybrid Kalman filter matrix;

[0008] Step 2) Establish a virtual sensor to reconstruct the fault sensor signal, and combine EKF to monitor sudden abnormal changes in component gas path performance.

[0009] Preferably, the specific steps of isolating and locating sensor and actuator faults based on the hybrid Kalman filter matrix in step 1) are as follows:

[0010] Step 1.1) Calculate the A, C, L, M matrices of the linearized model of the engine component level model; use the buffer to reduce the noise of the input health parameter h and the output pressure ratio π of the linearized model of the engine component level model, and update the two OBEM (OnBoard Engine Model) outputs y OBEM , where the control quantity of OBEM1 is Wfreal, A8model; the control quantity of OBEM2 is Wfmodel, A8real;

[0011] Step 1.2) Calculate the weighted sum of squares residual matrix (WSSR) using the hybrid Kalman filter matrix, compare the WSSR with the threshold λ, and isolate and locate the actuator and sensor faults according to the judgment logic.

[0012] Preferably: in step 1.1), the A, C, L, M matrices are calculated, the health parameter h and the pressure ratio π are denoised using the buffer, and the OBEM output y is updated. OBEM The specific steps are as follows:

[0013] Step 1.1.1) Use the small perturbation method and fitting method to calculate the A, C, L, M matrices corresponding to the linearized model of the engine component level model, and use the M-dimensional buffer buffer1 k and N-dimensional buffer buffer2 k , respectively denoise h and π; for the health parameter h at the kth moment k With pressure ratio π k , store it in the cache and update h k buffer and π k buffer ;

[0014] buffer1 k =[h k-i ,…,h k-1 ,h k ,h k+1 ,…,h k+M-1-i ]

[0015] buffer2 k =[π k-i ,…,π k-1 ,π k ,π k+1 ,…,πk+N-1-i ]

[0016] h k buffer =sum(buffer1 k ) / M

[0017] π k buffer =sum(buffer2 k ) / N

[0018] According to the calculated h k buffer , update the output of the two OBEM

[0019] Preferably: in step 1.2), the WSSR is calculated by using the hybrid Kalman filter matrix, the WSSR is compared with the threshold λ, and the specific steps of isolating and locating the actuator and sensor faults according to the judgment logic are as follows:

[0020] Step 1.2.1) Using the updated Combined with the current measurement signal y And the hybrid Kalman filter HKF corresponding to the i-th actuator and the j-th sensor i,j The structure is:

[0021]

[0022]

[0023]

[0024]

[0025] Where a represents the number of actuators, b represents the number of sensors, and u i m The ith control quantity is the real actuator signal value, and the remaining control quantities are the actuator model calculation values. e is the matrix composed of height and Mach number, and x OBEM is the state quantity of OBEM, The value calculated by the real sensor signal of the jth sensor and the j-1 sensor model is

[0026] Step 1.2.2) HKF i,j Calculated WSSR i,j HKF With WSSR i,j OBEM Get the weight to get WSSR i,j :

[0027]

[0028]

[0029]

[0030] By comparing the relationship between WSSR and threshold λ, if no fault occurs, all WSSRs are lower than λ; if the mth actuator fails, the WSSR of the mth row corresponding to the faulty actuator will be lower than λ, and the WSSR of the other m-1 rows will be higher than λ; if the nth sensor fails, the WSSR of the nth column corresponding to the faulty sensor will be lower than λ, and the WSSR of the other n-1 columns will be higher than λ; if the mth actuator fails and the nth sensor fails, the corresponding WSSR of the mth row and nth column will be lower than λ, and the other WSSR will be higher than λ.

[0031] Preferably, the steps of establishing a virtual sensor in step 2) to reconstruct the fault sensor signal and combining EKF to monitor the sudden abnormal change of the gas path performance of the component are as follows:

[0032] Step 2.1) Determine the signal reconstruction method based on the logic selection, and establish a virtual sensor to replace the located faulty sensor;

[0033] Step 2.2) Combine the reconstructed signal of the virtual sensor and use EKF to monitor component performance and isolate sudden abnormal changes in gas path components.

[0034] Preferably, the specific steps of diagnosing and locating sensor and actuator faults based on the hybrid Kalman filter matrix in step 2.1) are as follows:

[0035] Step 2.1.1) Due to the small amount of sensor step faults during simulation and the influence of noise, not all WSSRs are below the threshold, and it is necessary to make a misdiagnosis judgment on the WSSR. If the sensor position located at the current moment is consistent with the confirmed fault sensor position, it is not a misdiagnosis. If they are inconsistent, it is a misdiagnosis;

[0036] Step 2.1.2) If there is no misdiagnosis, based on the faulty sensor located in step 1, the engine component model is used to reduce the noise of h buffer and π buffer , calculate the corresponding fault sensor signal value and output the reconstructed signal y re ; If the fan outlet sensor fails, the air intake duct components and fan components are used to calculate the fan outlet sensor signal based on the fan pressure ratio and health parameters after noise reduction. The temperature or pressure sensor signal corresponding to the located sensor output;

[0037] Step 2.1.3) If a misdiagnosis occurs, use the reconstructed signal of the virtual sensor at the previous moment as the virtual sensor output y at the current moment re .

[0038] Preferably, in step 2.2), the reconstructed signal of the virtual sensor is combined, and the EKF is used to monitor the performance of the component to isolate the sudden abnormality of the gas path component. The specific steps are as follows:

[0039] Step 2.2.1) Create the extended matrix x E , A E , C E , use EKF to calculate the health parameters of each aircraft engine component and realize abnormal monitoring of component performance mutation;

[0040]

[0041]

[0042] Among them, y K =[y 1 …y n-1 ,y re ,y n+1 …y b ] is the measurement signal combined with the virtual sensor signal when the nth sensor fails.

[0043] Beneficial effects: The method for monitoring sudden changes in gas path components under concurrent faults of an aircraft engine control system designed by the present invention solves the problem of decreased accuracy in monitoring the performance of gas path components and even misdiagnosis when sensors and actuators fail. Under different working conditions within the envelope, accurate positioning of faulty sensors and actuators is achieved, and virtual sensor reconstruction signals corresponding to faulty sensor signals are established, which can well identify and monitor sudden changes in the performance of aircraft engines in the flight envelope. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 Fault sensor and actuator isolation and positioning system based on MHKF.

[0045] Figure 2 Fault location logic diagram based on MHKF.

[0046] Figure 3 Flowchart of sudden engine performance anomalies with the ability to isolate sensor and actuator failures.

[0047] Figure 4 When the gas path suddenly changes abnormally, the fault sensor and actuator diagnosis results based on MHKF.

[0048] Figure 5When there is a sudden abnormal change in the gas path or a failure in the sensor actuator, the signal result is reconstructed.

[0049] Figure 6 Histogram of component performance mutation results based on three different signals. DETAILED DESCRIPTION

[0050] The specific implementation of the present invention will be further described below in conjunction with the accompanying drawings.

[0051] The present invention discloses a method for monitoring sudden changes in gas path components under concurrent failures of an aircraft engine control system, which specifically comprises the following steps:

[0052] Step 1) isolating and locating sensor and actuator faults based on a hybrid Kalman filter matrix;

[0053] Step 1.1) Calculate the A, C, L, M matrices, use the buffer to denoise the health parameter h and the pressure ratio π, and update the OBEM output y OBEM .

[0054] Step 1.1.1) Use the small perturbation method and fitting method to calculate the A, C, L, and M matrices corresponding to the model, and use two 10-dimensional buffers to reduce the noise of h and π respectively. For the health parameter h at the kth moment k With pressure ratio π k , store it in the cache and update h k buffer and π k buffer .

[0055] buffer1 k =[h k-i ,…,h k-1 ,h k ,h k-9 ,…,h k+9-i ]

[0056] buffer2 k =[π k-i ,…,π k-1 ,π k ,π k-9 ,…,π k+9-i ]

[0057] h k buffer =sum(buffer1 k ) / 10

[0058] π k buffer =sum(buffer2 k ) / 10

[0059] According to the calculated h k buffer , update the output of the two OBEM

[0060] Step 1.2) Calculate the weighted sum of squares residual matrix (WSSR) using the hybrid Kalman filter matrix, compare the WSSR with the threshold λ, and isolate and locate the actuator and sensor faults according to the judgment logic.

[0061] Step 1.2.1) Using the updated Combined with the current measurement signal y And the hybrid Kalman filter HKF corresponding to the i-th actuator and the j-th sensor i,j The structure is:

[0062]

[0063]

[0064]

[0065]

[0066] Among them, u i m The ith control quantity is the real actuator signal value, and the other control quantities are the actuator model calculation values. e is the flight conditions such as altitude and Mach number, and x is OBEM is the state quantity of the model, The value calculated by the real sensor signal of the jth sensor and the j-1 sensor model is

[0067] Step 1.2.2) There are 2 actuators and 10 sensors in the text, so there are 20 corresponding HKFs in MHKF, generating 20 WSSRs HKF As attached Figure 1 shown.

[0068] HKF i,j Calculated WSSR i,j HKF With WSSR i,j OBEM The weight is WSSR i,j :

[0069]

[0070]

[0071]

[0072] By comparing the relationship between WSSR and threshold λ, if no fault occurs, all WSSRs are lower than λ. If the mth actuator fails, the WSSR of the mth row corresponding to the faulty actuator will be lower than λ, and the WSSR of the other m-1 rows will be higher than λ. If the nth sensor fails, the WSSR of the nth row corresponding to the faulty actuator will be lower than λ, and the WSSR of the other n-1 rows will be higher than λ. Figure 2 As shown in Figure 1, if the mth actuator fails, the WSSR in the left slash diagram will be lower than λ. If the nth sensor fails, the WSSR in the right slash diagram will be lower than λ. If the mth actuator and the nth sensor fail together, the WSSR in the cross section will be lower than λ.

[0073] Step 2) Establish a virtual sensor to reconstruct the fault sensor signal, and combine EKF to monitor the sudden abnormal change of the gas path performance of the component;

[0074] Step 2.1), according to the logic selection, determine the signal reconstruction method and establish a virtual sensor to replace the located faulty sensor.

[0075] Step 2.1.1) Due to the small amount of sensor step faults during simulation and the influence of noise, not all WSSRs are below the threshold, and a misdiagnosis judgment is required. The current position of the located sensor is consistent with the confirmed fault sensor position.

[0076] Step 2.1.2) If there is no misdiagnosis, based on the faulty sensor located above, use the engine component model to reduce the noise of h buffer and π buffer , calculate the corresponding fault sensor signal value. If the fan outlet sensor fails, the fan outlet sensor signal is calculated using the intake duct component and the fan component according to the fan pressure ratio and health parameters after noise reduction. The corresponding temperature or pressure sensor signal is output according to the located sensor.

[0077] Step 2.1.3) If a misdiagnosis occurs, the reconstructed signal of the virtual sensor at the previous moment is used as the virtual sensor output at the current moment.

[0078] Step 2.2) Combine the reconstructed signal of the virtual sensor and use EKF to monitor component performance and isolate sudden abnormal changes in gas path components.

[0079] Step 2.2.1) Create the extended matrix x E , A E , C E , EKF is used to calculate the health parameters of each aircraft engine component and realize the monitoring of sudden abnormal changes in component performance.

[0080]

[0081]

[0082] Among them, y K =[y 1 …y n-1 ,y re ,y n+1 …y 10 ] is the measurement signal combined with the virtual sensor signal when the nth sensor fails.

[0083] In order to verify the effectiveness of the method for monitoring sudden changes of gas path components under concurrent faults of the aero-engine control system designed in the present invention, a digital simulation of sudden changes of gas path components with sensor actuator fault identification in the envelope was carried out in the MATLAB environment.

[0084] First, in order to identify the sudden abnormal change of the gas path of the aircraft engine within the envelope, 11 fault modes (as shown in Table 1) are selected to verify the accuracy of the virtual sensor and whether the filtering method can well identify the corresponding fault mode. The real sensors and standard deviations selected in the example involved in the present invention are shown in Table 2. By injecting 1% step fault amount into 10 sensors and 2 actuators within the envelope, 11 fault modes (as shown in Table 1) are injected.

[0085] Table 1 Description of turbofan engine failure modes

[0086]

[0087] Table 2 Sensor selection and standard deviation description

[0088]

[0089] The output h of the buffer buffer , update the onboard engine model (OBEM) corresponding to the two actuator models. Use the hybrid Kalman filter matrix (MHKF) to calculate the WSSR matrix (structure as Figure 1 As shown), according to the judgment logic of WSSR and threshold λ (as shown Figure 2 Using the engine component model and the buffer output π buffer and h buffer Establish virtual sensors, reconstruct signals from virtual sensors and combine them with EKF to monitor engine component performance. Figure 3 The simulation results of the MHKF of fault mode 2 in Table 1 are shown in Figure 1. Figure 4As shown in Figure 2, when the injection sensor fails at 300 flight cycles, the injection actuator fuel pump A1 fails at 600 flight cycles, and the injection component suddenly becomes abnormal at 1200 flight cycles, the 20 WSSRs calculated by MHKF are all below the threshold λ before 300 flight cycles. 22 After the sensor fails, the corresponding numbers A1S4 and A2S4 are below the threshold, and the other WSSRs are above the threshold. According to logical judgment, P No. 4 is located. 22 Sensor failure. When the fuel pump fails, the corresponding A1S4 is below the threshold, and the other WSSRs are above the threshold. According to logical judgment, the No. 4 P 22 The sensor and the fuel pump of actuator No. 1 are faulty. When the performance of a component suddenly fails, a large peak will appear. Figure 3 In the misdiagnosis link, this peak does not affect the positioning accuracy of MHKF. Figure 5 When a sudden failure of a sensor, actuator, or component occurs, P 22 Sensor and T 3 The result of the sensor reconstructing the signal. Figure 4 The misdiagnosis rate of WSSR exceeding the threshold due to noise in MHKF and the error between the reconstructed signal and the true signal are shown in Table 3. 3 When the sensor fails, the 11 failure modes shown in Table 1 are injected. The simulation results based on three different signals are as follows: Figure 6 As shown in the figure, using the real signal y real The corresponding RMSE left slash column is the smallest. fault When monitoring component performance, the estimation results are extremely inaccurate, resulting in misdiagnosis. re The performance monitoring results show that although the accuracy is reduced to a certain extent when compared with the actual value results, it does not affect the diagnosis of component mutation anomalies.

[0090] Table 3 MHKF misdiagnosis rate (MR*10 2 ) and the root mean square error of the reconstructed signal (RMSE*10 2 )

[0091]

[0092] It can be seen from Table 3 that in all actuator and sensor combination faults, the misdiagnosis rate of MHKF is less than 4.39%, and the RMSE of the reconstructed signal is less than 0.8%, which meets the accuracy requirements. Figure 5It can be seen that the proposed method of establishing a virtual sensor to replace the located faulty sensor to monitor the sudden abnormal performance of engine components has greatly improved the gas path diagnosis accuracy calculated using the faulty sensor signal value, and compared with the results based on the real sensor signal, the RMSE of the health parameters is not much different.

[0093] The method for monitoring sudden changes in air path components under concurrent faults of an aircraft engine control system designed in the present invention realizes the isolation and positioning of various fault combinations of sensors and actuators under different sudden changes in the performance of air path components within the flight envelope. The signal reconstructed by the virtual sensor has high accuracy and can well monitor sudden changes in the performance of aircraft engine air path components within the flight envelope.

[0094] It should be pointed out that the above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes and substitutions that can be easily thought of by a person skilled in the art within the technical scope of the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for monitoring sudden changes in gas path components under concurrent failures of an aircraft engine control system, characterized in that: The following steps are involved: Step 1) isolating and locating sensor and actuator faults based on a hybrid Kalman filter matrix; Step 2) Establish a virtual sensor to reconstruct the fault sensor signal, and combine EKF to monitor the sudden abnormal change of the gas path performance of the component; Step 2) The specific steps are as follows: Step 2.1) Determine the signal reconstruction method based on the logic selection, and establish a virtual sensor to replace the located faulty sensor; Step 2.2) Combine the reconstructed signal of the virtual sensor and use EKF to monitor component performance and isolate sudden abnormal changes in gas path components; The specific steps of isolating and locating sensor and actuator faults based on the hybrid Kalman filter matrix in step 1) are as follows: Step 1.1) Calculate the A, C, L, M matrices of the linearized model of the engine component level model; use the buffer to reduce the noise of the input health parameter h and the output pressure ratio π of the linearized model of the engine component level model, and update the OBEM output y OBEM ; Step 1.2) Calculate WSSR using the hybrid Kalman filter matrix, compare WSSR with the threshold λ, and isolate and locate actuator and sensor faults according to the judgment logic; In step 1.1), the A, C, L, and M matrices are calculated, and the health parameter h and the pressure ratio π are denoised using the buffer, and the OBEM output y is updated. OBEM The specific steps are as follows: Step 1.1.1) Use the small perturbation method and fitting method to calculate the A, C, L, M matrices corresponding to the linearized model of the engine component level model, and use the M-dimensional buffer buffer1 k and N-dimensional buffer buffer2 k , respectively denoise h and π; for the health parameter h at the kth moment k With pressure ratio π k , and store them in buffer1 k and buffer2 k , and update h k buffer and π k buffer ; bufferl k =[h k-M+1 ,…,h k-1 ,h k ] buffer2 k =[π k-N+1 ,…,p k-1 ,p k ] h k buffer =sum(buffer1 k ) / M π k buffer =sum(buffer2 k ) / N According to the calculated h k buffer and π k buffer , update the output of the two OBEM In step 1.2), the WSSR is calculated using the hybrid Kalman filter matrix, and the WSSR is compared with the threshold λ. The specific steps of isolating and locating the actuator and sensor faults according to the judgment logic are as follows: Step 1.2.1) Using the updated Combined with the current measurement signal y And the hybrid Kalman filter HKF corresponding to the i-th actuator and the j-th sensor i，j The structure is: Where a represents the number of actuators, b represents the number of sensors, and u i m is the ith real actuator signal value of the mth actuator failure, the other control quantities are the values ​​calculated by the actuator model, e is the matrix composed of height and Mach number, x OBEM is the state quantity of OBEM, It is composed of the real sensor signal of the jth sensor and the calculated value of the j-1 sensor model, which is expressed as Step 1.2.2) HKF i，j Calculated WSSR i，j HKF With WSSR i，j OBEM Get the WSSR value i，j : Comparing the relationship between WSSR and threshold λ, if no fault occurs, all WSSRs are lower than λ; if the mth actuator fails, the WSSR of the mth row corresponding to the faulty actuator will be lower than λ, and the WSSR of the other m-1 rows will be higher than λ; if the nth sensor fails, the WSSR of the nth column corresponding to the faulty sensor will be lower than λ, and the WSSR of the other n-1 columns will be higher than λ; if the mth actuator fails and the nth sensor fails, the WSSR of the corresponding mth row and nth column will be lower than λ, and the other WSSR will be higher than λ.

2. The method for monitoring sudden changes of gas path components under concurrent failures of an aircraft engine control system according to claim 1, characterized in that: The specific steps of step 2.1) are as follows: Step 2.1.1) Perform misdiagnosis judgment on WSSR. If the sensor position located at the current moment is consistent with the confirmed fault sensor position, it is not a misdiagnosis. If they are inconsistent, it is a misdiagnosis. Step 2.1.2) If there is no misdiagnosis, based on the faulty sensor located in step 1, the engine component model is used to reduce the noise of h buffer and π buffer , calculate the corresponding fault sensor signal value and output the reconstructed signal y re ; Step 2.1.3) If a misdiagnosis occurs, use the reconstructed signal of the virtual sensor at the previous moment as the virtual sensor output y at the current moment re .

3. The method for monitoring sudden changes of gas path components under concurrent failures of an aircraft engine control system according to claim 1, characterized in that: In step 2.2), the reconstructed signal of the virtual sensor is combined with the EKF to monitor the performance of the components and isolate the sudden abnormality of the gas path components. The specific steps are as follows: Step 2.2.1) Create the extended matrix x E , A E , C E , use EKF to calculate the health parameters of each aircraft engine component and realize abnormal monitoring of component performance mutation; Among them, y K =[y 1 …y n-1 ,y re ,y n+1 …y b ] is the measurement signal combined with the virtual sensor signal when the nth sensor fails.

Images