A method for estimating a state of a degradation factor of an aeroengine component

Abstract

The application discloses an aero-engine component degradation factor state estimation method, constructs an engine component degradation factor state transition equation; the engine gas path process is modeled, and a "degradation factor-gas path measurement parameter" observation model is obtained; real flight data is evaluated by using the observation model, measurement noise is obtained according to the obtained evaluation value and the real flight data, and adaptive Gaussian mapping is carried out on the measurement noise; the process noise and the measurement noise of the engine component degradation factor state transition equation are iterated cooperatively, and the process noise variance and the measurement noise variance are determined; kalman filtering is used to continuously predict and update the degradation factor, and the change curve of the engine component degradation factor with time is obtained. The application can comprehensively consider the influence of modeling errors and sensor noise on the degradation factor estimation, more accurately evaluate the degradation state of the engine, and improve the accuracy of the aero-engine monitoring process.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of aero-engine performance monitoring, and particularly relates to a method for estimating the state of aero-engine component degradation factors. BACKGROUND

[0002] Since the engine degradation factors cannot be directly measured or directly calculated from the measured parameters, the degradation assessment of the aero-engine is essentially an estimation problem of unmeasurable parameters.

[0003] According to the difference in the estimation method of the gas path degradation factors, the degradation assessment of the aero-engine can be divided into the following categories:

[0004] (1) Linear and nonlinear regression method

[0005] In the 1970s, limited by the computing capacity, it was extremely difficult to perform the optimization regression of the multiple degradation factors in the nonlinear equation set in the thermodynamic model. The linear regression method converts the nonlinear relationship between the component degradation factor vector and the measurable parameter vector into a linear relationship by introducing a linear approximation at a given working point (such as the maximum or cruising state), and then the component degradation factor can be estimated by the matrix inversion process. The linear method inevitably introduces modeling errors in the linearization process, which reduces the accuracy of the degradation assessment.

[0006] With the development of computing capacity and optimization algorithms, more and more nonlinear regression methods have been introduced into the solution of the original nonlinear equation set, and the Newton-Raphson method is used to directly solve the nonlinear equation set by iteration, which takes into account the nonlinear thermodynamic process of the engine, and the evaluation accuracy is effectively improved.

[0007] (2) Genetic algorithm

[0008] In order to more accurately solve the nonlinear model, the genetic algorithm is introduced into the aero-engine degradation factor assessment. The genetic algorithm first defines a fitness function to measure the difference between the estimated value of the measured parameters calculated by the engine model and the actual measured parameters, and iterates the degradation factors through the "cross-mutation-selection" process to gradually reduce the difference. Through the fine and professional adjustment of the genetic algorithm model parameters, the degradation factor solution that makes the fitness function globally optimal can be obtained.

[0009] (3) Data-driven method

[0010] Data-driven methods use machine learning algorithms such as expert systems, support vector regression, neural networks, etc. to describe the working state of the aero-engine as a "black box" model. Through learning of the training data, a mapping relationship between the measurement parameters and the degradation factor is directly established. In the training phase, a large number of "measurement parameter-degradation factor" mapping samples are used to train the "black box" model, so that the model can learn a certain mapping relationship between the measurement parameters and the degradation factor. In the test phase, the measurement value of the cross-section parameter is input into the "black box" model, and the degradation factor estimate value can be calculated through the mapping relationship obtained in the training phase.

[0011] (4) Kalman filtering method

[0012] In order to consider the influence of modeling errors and measurement noise on the degradation factor estimation, Kalman filtering is introduced into the field of aero-engine gas path analysis.

[0013] Kalman filtering introduces the degradation prior knowledge of the engine through the state transition equation, and introduces the engine modeling error and sensor noise in the form of measurement equation. The evaluation process of the degradation factor is divided into two stages of prediction and update: in the prediction stage, the prior estimate value of the degradation factor is calculated through the state transition equation; in the update stage, the degradation factor prior value and the degradation factor observation value reflected by the measurement parameter are fused to obtain the statistically optimal degradation factor posterior estimate value.

[0014] Through the above analysis, the problems and defects of the prior art are:

[0015] First, the extremely limited measurable gas path parameters of military engines make the non-linear function of the degradation factor to the measurement parameter itself unable to meet the regression conditions. In order to solve it, more condition constraints need to be added, and these constraints are difficult to fit the complex degradation process of the engine, resulting in inaccurate estimation of the degradation factor;

[0016] Second, the above methods do not consider the influence of modeling errors and sensor noise. The complex working state of the military turbofan engine makes the modeling error of the engine thermodynamic model large; at the same time, the complex flow field distortion makes the sensor measurement value at a certain point difficult to represent the measurement parameter distribution of the entire cross-section, resulting in large sensor noise. The above factors greatly reduce the accuracy of the degradation factor estimation.

[0017] The data-driven method directly establishes the mapping relationship between the measurement parameters and the degradation factor, which can to some extent alleviate the influence of the above factors, but it requires a large number of "measurement parameter-degradation factor" training samples reflecting the real engine state for learning, which is difficult to meet in actual work.

[0018] In view of the above problems, the Kalman filtering method is an effective military turbofan engine gas path analysis idea, which can consider the modeling error and sensor noise at the same time, and can add the engine degradation priori in the form of state equation to the data fusion method of degradation evaluation process.

[0019] The difficulty of solving the above problems and defects is:

[0020] However, the Kalman filter applied to military turbofan engines still has the following problems:

[0021] The Kalman filter requires that the state transition equation of the degradation factor be as accurate as possible, and the measurement noise should meet the Gaussian distribution and be able to give a relatively accurate noise variance estimate. However, in the actual engine gas path analysis process, the above three requirements are difficult to meet. This leads to the fact that the Kalman filter-based gas path analysis method performs well on the simulation data set with known degradation process, noise distribution and variance, but performs poorly on the actual flight data with unknown degradation process, noise distribution and variance.

[0022] The significance of solving the above problems and defects is:

[0023] The solution to the above problems has theoretical significance and application value.

[0024] In terms of theoretical significance, the solution to the above problems can provide a solution to the non-convergence problem of iterative solution of nonlinear equations and the noise matching problem of system state evaluation, and thus provide a feasible method for related research directions such as health assessment of mechanical parts, tracking guidance of missiles and state monitoring of aircraft.

[0025] In terms of application value, the solution to the above problems can make up for the missing short board of the military turbofan engine gas path analysis method. On the basis of comprehensively evaluating the degradation state of each engine in the formation, the specific task is reasonably allocated to the individual fighter / engines, the task completion rate and flight safety are improved, and the purpose of optimizing the formation task deployment and enhancing the combat effectiveness of the aviation troops is achieved. SUMMARY

[0026] To overcome the problems in the related art, the embodiments of the present application provide an aero-engine component degradation factor state estimation method.

[0027] The technical solution is as follows: an aero-engine component degradation factor state estimation method, and the specific method steps are as follows:

[0028] S1, constructing an engine component degradation factor state transition equation according to flight parameter data and endurance test data;

[0029] S2, based on the degradation factor state transition equation in S1, modeling the engine gas path process to obtain a "degradation factor-gas path measurement parameter" observation model;

[0030] S3, using the observation model in S2 to evaluate the real flight data, obtaining the measurement noise according to the evaluation value and the real flight data, and performing adaptive Gaussian mapping on the measurement noise;

[0031] S4, iteratively determining the process noise variance and the measurement noise variance of the engine component degradation factor state transition equation and the measurement noise in S3;

[0032] S5, based on the process noise variance and the measurement noise variance obtained in S4, using Kalman filtering to continuously predict and update the degradation factor, and obtaining the change curve of the engine component degradation factor over time.

[0033] As a preferred embodiment, step S1 specifically comprises:

[0034] S11, using the endurance test data of multiple engines to respectively calculate multiple degradation curves of flow rate, efficiency, etc. over time for each component;

[0035] S12, selecting multiple degradation function forms to fit the distribution of the multiple degradation curves in S11 to obtain a form expression of the degradation function:

[0036] g i = g i (t) + ∈ i

[0037] Where g i is the i-th degradation envelope, g i (t) is the function fitting form of the i-th degradation curve, t is the service time or cycle number of the engine, and ∈ i is the error tolerance of the i-th degradation envelope;

[0038] S13, selecting flight data of multiple engines of the same type to the end of life, and inputting the flight time sequence into the degradation function to obtain different degradation factor time sequence;

[0039] S14, inputting the flight data and the degradation factor time sequence in S13 into the thermodynamic model to obtain multiple measurement parameter estimated value time sequence of different degradation processes;

[0040] S15, matching the estimated value time sequence obtained in S14 with the actual value in the flight data, selecting the degradation function with the highest matching degree within the error tolerance as the degradation process curve of the component, discretizing the degradation curve to obtain a degradation factor state transition equation for Kalman filtering:

[0041] x t+1 = x t + ε t

[0042] where x t is the degradation factor of each component of the engine at time t, and ε t is the process noise at time t.

[0043] As a preferred embodiment, a data method is used in S2 to model the engine gas path process, the engine inlet parameters and operating parameters in the flight data are selected as inputs, the cross-section parameters in the flight data are selected as outputs, a data-driven method is used to train the data model, and a mapping relationship from the inlet parameters and operating parameters to the cross-section parameters is obtained.

[0044] As a preferred embodiment, a thermodynamic model is used in S2 to model the engine gas path process, a component-level thermodynamic model of the engine is established, and the model is corrected using flight data.

[0045] As a preferred embodiment, S3 specifically includes:

[0046] S31, using the observation model obtained in S2 to evaluate the cross-section measurement parameters z t of the real flight data to obtain h(x t ), and by comparing the evaluation value with the actual value of the flight data, the measurement noise δ t = z t - h(x t ) is obtained.

[0047] S32, Gaussian regression is performed on the measurement noise to obtain a Gaussian distribution function;

[0048] S33, the Gaussian distribution function is sampled to obtain the same number of regression noise σ t as the evaluation error of the observation model, and the regression noise and the measurement noise are one-to-one corresponding according to the noise size, and the noise difference δ t - σ t is calculated.

[0049] S34, taking the observation model estimation value h(x t ) as input and the noise difference δ t - σ t as output, a noise mapping network δ t - σ t = w(h(x t )) is trained, and after conversion by the noise mapping network, the observation equation is obtained:

[0050] z t = h(x t ) + w(h(xt ))+σ t .

[0051] As a preferred embodiment, S4 specifically comprises:

[0052] S41, initializing process noise and measurement noise covariance matrix Q0, R0;

[0053] S42, inputting the covariance matrix of process noise and measurement noise into Kalman filter estimation to obtain the posteriori estimation value of the degradation factor;

[0054] S43, bringing the posteriori estimation value into the measurement equation to obtain the posteriori estimation value of the cross-section parameter;

[0055] S44, comparing the posteriori estimation value of the cross-section parameter with the true value of the flight data to calculate the estimation deviation;

[0056] S45, judging the ratio of the estimation deviation and the pre-set threshold value, if the estimation deviation is greater than the pre-set threshold value, executing step S46, otherwise executing step S47;

[0057] S46, calculating the derivative of the estimation deviation to the covariance matrix respectively, and updating the covariance matrix of process noise and measurement noise using the derivative information and repeating steps S42-S45;

[0058] S47, if the estimation deviation is less than the pre-set threshold value, considering that the covariance matrix Q i +1,R i+1 of the last iteration can meet the estimation accuracy of the cross-section parameter value of the flight data, and considering that the noise variance can reflect the process noise and measurement noise distribution of the degradation evaluation model, thereby determining the process noise variance Q = Q i+1 and the measurement noise variance R = R i+1 .

[0059] As a preferred embodiment, S5 specifically comprises:

[0060] S51, initializing the degradation factor: optionally initializing the degradation factor of each component of a newly entered aero-engine to 1, obtaining the process noise variance Q and the measurement noise variance R of the degradation factor from S4;

[0061] S52, predicting using Kalman filter: using the Kalman filter method, predicting the mean and variance of the degradation factor using the state transition equation of the degradation factor obtained in S15 to obtain the estimation prior value of the mean and variance of the degradation factor;

[0062] S53, updating using Kalman filtering: using the Kalman filtering method, updating the mean and variance of the degradation factor using the observation equation obtained in S34, and obtaining a posteriori estimates of the mean and variance of the degradation factor;

[0063] S54. Tracking using Kalman filtering: As the engine usage time increases, the degradation factor is continuously predicted and updated using Kalman filtering, ultimately obtaining a curve showing how the degradation factors of various engine components change over time.

[0064] Combining all the above technical solutions, the advantages and positive effects of the present invention are as follows:

[0065] (1) The present invention fully combines test data to establish a degradation state transfer equation that can reflect the actual degradation trend of engine components.

[0066] (2) The established observation equation of “compressor efficiency degradation factor-turbine after-total temperature” is corrected by using the flight data of the on-wing engine, which can better reflect the actual working status of the engine.

[0067] (3) The measurement noise adaptive Gaussian mapping method constructed by the present invention fundamentally solves the problem that the measurement noise does not conform to the Gaussian distribution when Kalman filtering is used for aircraft engine degradation assessment, thereby improving the convergence and accuracy of degradation assessment.

[0068] (4) The constructed noise collaborative iterative estimation algorithm can fully evaluate the noise distribution variance of the state equation and measurement equation on the flight data, so that the filtering process matches the actual operation and degradation process of the engine, and improves the accuracy of the Kalman filter degradation assessment based on real flight data.

[0069] (5) In addition, experimental results and analysis show that compared with traditional degradation assessment methods, the model and algorithm of the present invention can reasonably estimate the distribution of measurement noise and the prior degradation trend, more reasonably evaluate the degradation of the engine, and improve the accuracy of the aircraft engine monitoring process.

[0070] It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0072] Figure 1 This is a flow chart of a method for estimating the degradation factor state of an aircraft engine component provided by an embodiment of the present invention.

[0073] Figure 2 is a construction flowchart of the degradation state transition equation of step S1 of at least one embodiment provided by the present application.

[0074] Figure 3 is a flowchart of the adaptive Gaussian mapping method of the measurement noise of step S3 of at least one embodiment provided by the present application.

[0075] Figure 4 is a flowchart of the cooperative iteration method of the process noise and the measurement noise of step S4 of at least one embodiment provided by the present application.

[0076] Figure 5 is a degradation factor tracking result map of step S53 provided by the present application. DETAILED DESCRIPTION

[0077] In order to make the above objectives, characteristics and advantages of the present application more apparent and easy to understand, the specific embodiments of the present application are described in detail below with reference to the accompanying drawings. In the following description, a large number of specific details are set forth in order to facilitate a full understanding of the present application. However, the present application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without departing from the scope of the present application, so the present application is not limited to the specific embodiments disclosed below.

[0078] As shown in Figure 1 , the aero-engine component degradation factor state estimation method provided by the present application is as follows:

[0079] S1, construct an engine component degradation factor state transition equation, the engine component selected in the present embodiment is a compressor, that is, a compressor efficiency degradation state transition equation is constructed.

[0080] Step S1 includes the following specific steps, and the specific flow is as shown in Figure 2 :

[0081] S11, use the bench endurance test data of multiple engines to respectively calculate the degradation curves of the compressor efficiency and the like varying with time.

[0082] S12, select multiple degradation function forms to fit the degradation curves, and obtain the form expression of the degradation function corresponding to the degradation curves in Figure 2 :

[0083] g i (t) + ∈ i i

[0084] g i is the i-th degradation envelope, and g i ​(t) is the function fitting form of the ith degradation curve, t is the service time or cycle number of the engine, ∈ i is the error tolerance of the ith degradation envelope.

[0085] S13, flight data of multiple engines of the same model to the end of life are selected, the flight time sequence is input into the degradation function, and different degradation factor time sequence is obtained:

[0086] g i (t1), g i (t2), …, g i (t i ) n )

[0087] S14, the flight data and the above degradation sequence are input into the thermodynamic model, and a plurality of turbine rear temperature estimated value time sequence of different degradation processes is obtained:

[0088]

[0089] wherein, T (i) is the estimated value of the degradation factor g i (t1) at flight time t1, the turbine rear temperature estimated value calculated by the thermodynamic model f(x1, g i (t1)) when the flight data is x1.

[0090] S15, the estimated value time sequence is matched with the actual value of the turbine rear temperature in the flight data, and the degradation function with the highest matching degree is obtained:

[0091]

[0092] wherein, g fit is the degradation sequence with the smallest estimation error in all degradation sequences; T j is the actual value of the turbine rear temperature, and f(u1, g1(t j )) is the turbine rear temperature estimated value calculated by the thermodynamic model when the degradation sequence is g1(t j ) under the current flight parameter input u1 factor.

[0093] The degradation sequence g fit with the smallest estimation error is taken as the degradation process curve of the component, the degradation curve is discretized, and the degradation factor state transition equation for Kalman filtering is obtained:

[0094] x t+1 = x t + g fit (t+1) - g fit (t) + ∈ i .

[0095] S2, based on the degradation factor state transition equation in S1, modeling the engine gas path process to obtain a "degradation factor-gas path measurement parameter" observation model, in this embodiment, a component "compressor efficiency degradation factor-turbine total temperature after the observation model" is constructed, and the specific steps are:

[0096] S21, using data method or thermodynamic model method to model the engine gas path process.

[0097] S22, if the data method is used for modeling, the engine inlet total temperature, total pressure, high pressure rotor speed, ambient static pressure and compressor efficiency degradation factor in the flight data are selected as the input, and the turbine total temperature after the output is selected as the output. A certain data-driven method is used to train the data model to obtain the mapping relationship from the above input to the turbine total temperature after the output:

[0098]

[0099] Among them, is the turbine temperature estimate value obtained by the data-driven method, x t is the degradation factor at time t, u t is the flight parameter input at time t, h net is the data-driven model mapping.

[0100] S23, if the thermodynamic model is used for modeling, the engine component-level thermodynamic model is established and the model is corrected using flight data, and the mapping relationship from the engine inlet total temperature, total pressure, high pressure rotor speed, ambient static pressure and compressor efficiency degradation factor to the turbine total temperature after the output is obtained:

[0101]

[0102] Among them, is the turbine temperature estimate value obtained by the data-driven method, x t is the degradation factor at time t, u t is the flight parameter input at time t, h th is the thermodynamic model mapping.

[0103] S3, using the observation model in S2 to evaluate the real flight data, obtaining the measurement noise according to the evaluation value and the real flight data, and performing adaptive Gaussian mapping on the turbine total temperature after the measurement noise. The specific process of S3 is as shown in Figure 3

[0104] S31, using the mapping relationship established by S2 to evaluate the turbine total temperature T6 of the real flight data to obtain The measurement noise is obtained by comparing the difference between the evaluation value and the actual value of the flight data ​

[0105] S32, Gaussian regression is performed on the measurement noise to obtain a Gaussian distribution function.

[0106] S33, the Gaussian distribution function is sampled to obtain the same number of regression noise σ t as the observation model evaluation error, and the regression noise and the measurement noise are one-to-one corresponding according to the noise size and the noise difference δ t -σ t .

[0107] S34, the observation model estimate h(x t ,u t ) is input, and the noise difference δ t -σ t is output, and a noise mapping network δ t -σ t is trained, that is, w(h(x t ,u t )) is obtained. After conversion by the noise mapping network, the observation equation can be written as:

[0108] T6=h(x t ,u t )+w(h(x t ,u t ))+σ t

[0109] In this way, through Gaussian mapping training of the measurement noise, the measurement noise δ t that does not conform to the Gaussian distribution is converted into the regression noise σ t that conforms to the Gaussian distribution.

[0110] S4, the process noise of the compressor efficiency degradation state transition equation and the turbine total temperature measurement noise are iterated cooperatively to determine the process noise variance and the measurement noise variance. The specific process is shown in Figure 4

[0111] S41, the process noise and the measurement noise covariance matrix Q0, R0 are initialized;

[0112] S42, the covariance matrix of the process noise and the measurement noise is input into the Kalman filter estimation to obtain the posterior estimate of the degradation factor;

[0113] S43, the posterior estimate is brought into the observation equation of step S34 to obtain the posterior estimate of the cross-section parameter;

[0114] S44, the posterior estimate of the cross-section parameter is compared with the true value of the flight data to calculate the estimation deviation;

[0115] ​S45, if the estimation bias is greater than a preset threshold, the derivative of the estimation bias to the covariance matrix is calculated respectively, and the derivative information is used to update the covariance matrix of the process noise and the measurement noise, and steps S42-S45 are repeated;

[0116] S46, if the estimation bias is less than a preset threshold, the covariance matrix Q i+1 ,R i+1 The estimation accuracy of the flight data section parameter value can be met at the same time, and it is considered that the noise variance can reflect the process noise and measurement noise distribution of the degradation evaluation model.

[0117] S5, based on the process noise variance and the measurement noise variance obtained in S4, the Kalman filter is used to continuously predict and update the degradation factor, and the change curve of the engine component degradation factor with time is obtained, that is, the compressor efficiency degradation estimation method considering the compressor efficiency degradation prior and the turbine after total temperature measurement noise distribution is obtained. The specific process is as follows:

[0118] S51, degradation factor initialization process: for a new engine, the degradation factors of all components are initialized to 1, that is, all components are not degraded, and the degradation factor process noise variance Q and the measurement noise variance R can be obtained by steps S41-S46;

[0119] S52, prediction process of Kalman filter: the unscented Kalman filter is used, and the state transition equation of the degradation factor obtained in step S15 is used:

[0120] x t+1 =x t +g fit (t+1)-g fit (t)+∈ i

[0121] The mean and variance of the degradation factor are predicted to obtain the estimation prior value of the mean and variance of the degradation factor:

[0122]

[0123]

[0124] S53, update process of Kalman filter: the unscented Kalman filter is used, and the measurement equation obtained in step S34 is used:

[0125] T6=h(x t ,u t )+w(h(x t ,u t ))+σ t

[0126] Update the mean and variance of the degradation factor to obtain the posterior estimates of the mean and variance of the degradation factor:

[0127]

[0128]

[0129]

[0130] S54, Kalman filter tracking process: As the engine usage time increases, the degradation factor is continuously predicted and updated using Kalman filter, and finally the curve of the degradation factor of each engine component over time is obtained, such as Figure 5 As shown in the figure, it can be seen that the trend estimation of the degradation factors of various engine components is relatively close to the actual value, which proves that the evaluation method proposed in this patent can more effectively track the performance degradation of various components, especially for the flow and efficiency degradation of the low-pressure compressor and the high-pressure compressor, which can be more accurately tracked and evaluated.

[0131] Those skilled in the art will readily appreciate other embodiments disclosed herein after considering the specification and practicing the disclosure disclosed herein. This application is intended to cover any variations, uses, or adaptations disclosed herein that follow the general principles disclosed herein and include common knowledge or customary techniques in the art that are not disclosed herein. The description and examples are intended to be exemplary only, and the true scope and spirit of the present disclosure are indicated by the appended claims. It should be understood that the present disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and that various modifications and changes can be made without departing from the scope thereof. The scope of the present disclosure should be limited by the appended claims.

Claims

1. A method for estimating the degradation factor state of an aircraft engine component, characterized in that: The steps are as follows: S1. Construct the state transition equation of engine component degradation factor based on flight parameter data and endurance test data; S2. Based on the degradation factor state transfer equation in S1, the engine gas path process is modeled to obtain the "degradation factor-gas path measurement parameter" observation model; S3, using the observation model in S2 to evaluate the actual flight data, obtaining the measurement noise based on the obtained evaluation value and the actual flight data, and performing adaptive Gaussian mapping on the measurement noise; S4, co-iterate the process noise of the engine component degradation factor state transfer equation and the measurement noise in S3 to determine the process noise variance and the measurement noise variance; S5. Based on the process noise variance and measurement noise variance obtained in S4, Kalman filtering is used to continuously predict and update the degradation factor to obtain a curve of the engine component degradation factor changing over time; S3 specifically includes: S31, using the observation model obtained in S2 to measure the cross-sectional parameters z of the real flight data t Evaluate and get h(x t ), and the measurement noise δ is obtained by comparing the difference between the evaluation value and the actual value of the flight data t =z t -h(x t ); S32, performing Gaussian regression on the measurement noise to obtain a Gaussian distribution function; S33, sample the Gaussian distribution function to obtain the same number of regression noise σ as the observation model evaluation error t , and according to the noise size, the regression noise and the measurement noise are matched one by one and the noise difference δ is calculated t -σ t ; S34, using the observation model to estimate the value h(x t ) as input, with the noise difference δ t -σ t As the output, the noise mapping network δ is trained t -σ t =w(h(x t )), after being transformed by the noise mapping network, the observation equation is obtained: z t =h(x t )+w(h(x t ))+σ t 。 2. The method for estimating the degradation factor state of an aircraft engine component according to claim 1, characterized in that: Step S1 specifically includes: S11. Using the endurance test data of multiple engines, calculate and obtain multiple flow rate and efficiency degradation curves of each component over time; S12. Select multiple degradation function forms to fit the distribution of multiple degradation curves in S11 and obtain the formal expression of the degradation function: g i =g i (t)+∈ i Among them, g i is the i-th degenerate envelope, g i (t) is the function fitting form of the i-th degradation curve, t is the engine usage time or cycle number, ∈ i is the error tolerance of the i-th degenerate envelope; S13, selecting flight data of multiple engines of the same model at the end of their service life, inputting the flight time series into the degradation function, and obtaining different degradation factor time series; S14, inputting the flight data and the degradation factor time series in S13 into the thermodynamic model to obtain a plurality of time series of estimated values ​​of measurement parameters of different degradation processes; S15. Match the estimated value time series obtained in S14 with the actual value in the flight data, select the degradation function with the highest matching degree within the error tolerance as the degradation process curve of the component, discretize the degradation curve, and obtain the degradation factor state transition equation for Kalman filtering: x t+1 =x t +e t Where x t is the degradation factor of each engine component at time t, ε t is the process noise at time t.

3. The method for estimating the degradation factor state of an aircraft engine component according to claim 2, characterized in that: In S2, a data method is used to model the engine gas path process. The engine inlet parameters and operating parameters in the flight data are selected as input, and the cross-sectional parameters in the flight data are used as output. The data-driven method is used to train the data model to obtain the mapping relationship between the inlet parameters, operating parameters and cross-sectional parameters.

4. The method for estimating the degradation factor state of an aircraft engine component according to claim 2, wherein: In S2, a thermodynamic model is used to model the engine gas path process, an engine component-level thermodynamic model is established, and the model is modified using flight data.

5. The method for estimating the degradation factor state of an aircraft engine component according to claim 1, wherein: S4 specifically includes: S41, initializing the process noise and measurement noise covariance matrix Q0, R0; S42, inputting the covariance matrix of the process noise and the measurement noise into the Kalman filter estimation to obtain a posterior estimate of the degradation factor; S43, substituting the posterior estimated value into the measurement equation to obtain the posterior estimated value of the cross-sectional parameter; S44, comparing the posterior estimated value of the cross-sectional parameter with the true value of the flight data, and calculating the estimation deviation; S45, comparing the estimated deviation with a preset threshold. If the estimated deviation is greater than the preset threshold, proceed to step S46; otherwise, proceed to step S47; S46, respectively calculating the derivatives of the estimated deviation with respect to the covariance matrix, and using the derivative information to update the covariance matrix of the process noise and the measurement noise and repeating steps S42 to S45; S47. If the estimated deviation is less than a preset threshold, the covariance matrix Q of the last iteration is considered to be i+1 ,R i+1 It can simultaneously meet the estimation accuracy of the flight data cross-section parameter value, and then it is believed that the noise variance can reflect the process noise and measurement noise distribution of the degradation assessment model, thereby determining the process noise variance Q = Q i+1 , measurement noise variance R = R i+1 .

6. The method for estimating the degradation factor state of an aircraft engine component according to claim 5, characterized in that: S5 specifically includes: S51. Initialize degradation factors: Select a newly commissioned aircraft engine and initialize the degradation factors of its components to 1. The process noise variance of the degradation factor is obtained from S4 as Q and the measurement noise variance is R. S52. Prediction using Kalman filtering: Using the Kalman filtering method, the degradation factor state transfer equation obtained in S15 is used to predict the mean and variance of the degradation factor, and estimated prior values ​​of the mean and variance of the degradation factor are obtained; S53, updating using Kalman filtering: using the Kalman filtering method, updating the mean and variance of the degradation factor using the observation equation obtained in S34, and obtaining a posteriori estimates of the mean and variance of the degradation factor; S54. Tracking using Kalman filtering: As the engine usage time increases, the degradation factor is continuously predicted and updated using Kalman filtering, ultimately obtaining a curve showing how the degradation factors of various engine components change over time.

Images