Aircraft motor life prediction method and system considering multi-factor coupling

Abstract

The present invention provides an aviation motor life prediction method and system considering the coupling effect of multiple factors, which relates to the technical field of aviation equipment reliability assessment, and includes: conducting a motor life test based on a uniform test to obtain several factors affecting the life of the aviation motor; allowing each factor to take different values ​​and repeating the experiment to obtain several groups of test data, the test data being the corresponding relationship between the numerical value of each factor and the actual value of the life of the aviation motor; preprocessing each test data to obtain a training data set, a test data set and a verification data set; iteratively training, testing and verifying the motor life prediction model based on the training data set, the test data set and the verification data set to obtain a trained motor life prediction model; obtaining the life prediction value of the motor to be tested based on the actual data and the trained motor life prediction model. The present invention can improve the motor life prediction effect and enhance the safety and reliability of the aircraft system.

Description

Technical Field

[0001] The present invention relates to the technical field of aviation equipment reliability assessment, and in particular to an aviation motor life prediction method and system taking into account the coupling effect of multiple factors. Background Art

[0002] Brushed DC motor is a common type of motor with advantages such as relatively simple design, low maintenance cost, stable and reliable operation. In the field of aerospace, brushed DC motors are widely used. Brushed DC motors on aircraft are mostly used to provide precise speed and position control, drive loads, etc. For example, the attitude and direction control of aircraft elevators and ailerons, the retraction and extension of landing gear, and the drive of avionics equipment. Therefore, the reliable and stable operation of brushed DC motors is the key to the reliable operation of aircraft wing control systems, landing gear systems, avionics equipment systems, etc., and is of great significance to ensure the safe and reliable flight of aircraft.

[0003] The life prediction and reliability assessment of aviation motors are not only of great significance to the improvement of their own functional performance and the formulation of maintenance strategies, but also crucial to the improvement of the durability indicators of their application systems. DC brush motors mainly generate mechanical rotational motion through the interaction between internal brushes and electromagnetic fields. Common failure modes during operation include brush wear, insulation aging, bearing wear, excitation loss, corrosion and rust. For aircraft serving in coastal areas and high-temperature areas, the DC brush motors in various systems are susceptible to environmental conditions such as high humidity, high temperature, and high salt spray corrosion. In addition, the motors are frequently started and stopped, and the load is variable. Long-term operation will cause environmental damage (corrosion, aging, etc.) and mechanical damage (wear, fatigue, etc.), and these damages have significant coupling effects and cumulative effects, which accelerate the occurrence of motor failures.

[0004] The life prediction methods of brushed DC motors include physical model method and data-driven method. The physical model method starts from studying the failure mode and mechanism of the motor. However, this method needs to conduct reliability and life assessment under the condition of clear understanding of the internal working principle and failure mechanism of the motor. It is based on the modeling of specific parts of the motor. For complex systems or multi-factor coupling, this method is even more difficult to establish a complete, accurate and comprehensive failure degradation model. Therefore, it is impossible to accurately describe the motor degradation law under multi-factor conditions. The life prediction methods based on data-driven mainly include mathematical statistics methods and machine learning. The mathematical statistics method is mainly based on the historical moment information of the system state change collected, and by establishing different degradation models, to predict the change of the system state at the current moment. This method has a large dependence on the model, but it is not widely used in nonlinear aspects, and its life prediction ability in complex variable conditions has certain limitations. The machine learning method uses a large number of simple neuron sensors to deal with complex nonlinear problems. It has learning, memory and association functions, can establish implicit mapping relationships between data, and can establish mapping relationships between input information and output information through learning between network layers. It has advantages in nonlinear modeling. However, the current motor life prediction method based on machine learning estimates the life of the motor under a single fault mode by collecting degradation data of certain performance parameters of the motor. It does not adequately consider the combined impact of the motor's environmental factors and load conditions. Therefore, it cannot well characterize the life of the motor under multiple fault modes. Summary of the invention

[0005] The purpose of the present invention is to provide an aviation motor life prediction method and system taking into account the coupling of multiple factors, which can improve the motor life prediction effect, better characterize the motor life under multiple fault modes, and enhance the safety and reliability of the aircraft system.

[0006] A method for predicting the life of an aviation motor considering the coupling of multiple factors includes:

[0007] S1, conduct motor life test based on uniform test and obtain several factors affecting the life of aviation motor;

[0008] S2, making each of the factors take different values ​​and repeating the experiment to obtain several groups of test data, wherein the test data is the corresponding relationship between the value of each factor and the actual value of the life of the aircraft motor;

[0009] S3, preprocessing each of the test data to obtain a training data set, a test data set, and a verification data set;

[0010] S4, iteratively training, testing and verifying the motor life prediction model based on the training data set, the test data set and the verification data set to obtain the trained motor life prediction model;

[0011] S5, obtaining a life prediction value of the motor to be tested based on actual data and the trained motor life prediction model;

[0012] The motor life prediction model includes an input layer, a hidden layer and an output layer;

[0013] The training process of the motor life prediction model includes two stages: forward propagation and back propagation;

[0014] In the forward propagation stage, the input sample is transmitted to the hidden layer through the input layer, and then transmitted to the output layer through the hidden layer. Each layer of neurons performs weighted summation on the input and performs normalization processing through the activation function before outputting.

[0015] The output value of the output layer is as follows:

[0016] y=NN(x)=σ(W k ...σ(W2σ(W1x+b1)+b2)+…+b k );

[0017] Where: y is the output value of the output layer, x is the input sample of the input layer, σ is the activation function, W k is the weight vector of the kth layer, b k is the bias of the kth layer;

[0018] In the back propagation stage, the weight vectors of neurons in each layer are updated based on the LM method according to the actual value and the output value of the output layer to reduce the error between the actual value and the output value of the output layer:

[0019] The weight update formula is as follows:

[0020] (J T J+I)(W k+1 -W k )=J T (Y-NN(x));

[0021] Where: J is the Jacobian function of the motor life prediction model, Y is the actual value, W k is the weight vector of the back-propagation k-th layer, W k+1 is the weight vector for back-propagation of the k+1th layer, I is the identity matrix, and T is the transpose.

[0022] Optionally, the motor life prediction model is a back propagation model based on a neural network, namely, a BP-ANN model.

[0023] Optionally, the preprocessing includes data cleaning, data conversion, data segmentation and data normalization.

[0024] Optionally, the S4 is specifically:

[0025] S41, let i = 1; i = [1, 2, ..., M], M is an iterative training parameter;

[0026] S42, iteratively training the motor life prediction model based on the training data set;

[0027] S43, testing the motor life prediction model based on the test data set to obtain a test prediction error of the motor life prediction model;

[0028] S44, judging i, if i is less than M, setting i=i+1 and returning to S42, if i is greater than or equal to M, selecting the motor life prediction model corresponding to the minimum value of the test prediction error as the initially trained motor life prediction model and executing S45;

[0029] S45, verifying the initially trained motor life prediction model based on the verification data set to obtain a verification prediction error;

[0030] S46, judging the verification prediction error, if the verification prediction error is less than the error setting value, using the initially trained motor life prediction model as the trained motor life prediction model, if the verification prediction error is greater than or equal to the error setting value, returning to S41.

[0031] Optionally, the test prediction error calculation formula is as follows:

[0032]

[0033] Where: n is the number of test samples in the test data set, y i is the predicted value of the i-th test sample, y is the average of the predicted values ​​of n test samples, Q is the test prediction error, and Y i is the actual value of the i-th test sample, and Y is the average of the actual values ​​of n test samples.

[0034] The present invention also provides an aviation motor life prediction system considering the coupling effect of multiple factors, which includes:

[0035] The influencing factor module is used to conduct motor life tests based on uniform tests to obtain several factors that affect the life of aviation motors;

[0036] A data acquisition module is used to make each of the factors take different values ​​and conduct repeated experiments to obtain several groups of test data; the test data is the corresponding relationship between the value of each factor and the actual value of the life of the aircraft motor;

[0037] A data processing module is used to pre-process the test data to obtain a training data set, a test data set and a verification data set;

[0038] A model training module, used for iteratively training, testing and verifying the motor life prediction model based on the training data set, the test data set and the verification data set to obtain the trained motor life prediction model;

[0039] A life prediction module, used to obtain a life prediction value of the motor to be tested based on actual data and the trained motor life prediction model;

[0040] The motor life prediction model includes an input layer, a hidden layer and an output layer;

[0041] The training process of the motor life prediction model includes two stages: forward propagation and back propagation;

[0042] In the forward propagation stage, the input sample is transmitted to the hidden layer through the input layer, and then transmitted to the output layer through the hidden layer. Each layer of neurons performs weighted summation on the input and performs normalization processing through the activation function before outputting.

[0043] The output value of the output layer is as follows:

[0044] y=NN(x)=σ(W k ...σ(W2σ(W1x+b1)+b2)+…+b k );

[0045] Where: y is the output value of the output layer, x is the input sample of the input layer, σ is the activation function, W k is the weight vector of the kth layer, b k is the bias of the kth layer;

[0046] In the back propagation stage, the weight vectors of neurons in each layer are updated based on the LM method according to the actual value and the output value of the output layer to reduce the error between the actual value and the output value of the output layer:

[0047] The weight update formula is as follows:

[0048] (J T J+I)(W k+1 -W k )=J T (Y-NN(x));

[0049] Where: J is the Jacobian function of the motor life prediction model, Y is the actual value, W k is the weight vector of the back-propagation k-th layer, W k+1is the weight vector for back-propagation of the k+1th layer, I is the identity matrix, and T is the transpose.

[0050] Optionally, the motor life prediction model is a back propagation model based on a neural network, namely, a BP-ANN model.

[0051] Optionally, the preprocessing includes data cleaning, data conversion, data segmentation and data normalization.

[0052] Optionally, the model training module includes:

[0053] An instruction unit is used to set i=1; i=[1,2,…,M], where M is an iterative training parameter;

[0054] A training unit, configured to iteratively train the motor life prediction model based on the training data set;

[0055] A testing unit, used for testing the motor life prediction model based on the test data set to obtain a test prediction error of the motor life prediction model;

[0056] A first judgment unit is used to judge i. If i is less than M, i=i+1 is set and the training unit is returned. If i is greater than or equal to M, the motor life prediction model corresponding to the minimum value of the test prediction error is selected as the initially trained motor life prediction model and the verification unit is executed;

[0057] A verification unit, used to verify the initially trained motor life prediction model based on the verification data set to obtain a verification prediction error;

[0058] The second judgment unit is used to judge the verification prediction error. If the verification prediction error is less than the error setting value, the initially trained motor life prediction model is used as the trained motor life prediction model. If the verification prediction error is greater than or equal to the error setting value, it is returned to the instruction unit.

[0059] Optionally, the test prediction error calculation formula is as follows:

[0060]

[0061] Where: n is the number of test samples in the test data set, y i is the predicted value of the i-th test sample, is the average of the predicted values ​​of n test samples, Q is the test prediction error, and Y i is the actual value of the i-th test sample, is the average of the actual values ​​of n test samples.

[0062] The effects of the present invention are as follows:

[0063] The invention provides an aviation motor life prediction method taking into account the coupling effect of multiple factors, adopts a uniform test design method, conducts motor life tests under a comprehensive environment, obtains motor life test data under the comprehensive environment, and solves the problem of lack of life test data of aviation brushed DC motors under the influence of multiple factors.

[0064] The invention provides an aviation motor life prediction method that takes into account the coupling of multiple factors, improves the motor life prediction accuracy of aviation motors in high-temperature coastal areas under the influence of complex environmental factors and the motor's own load, and enhances the safety and reliability of aircraft systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 It is a flow chart of the aviation motor life prediction method considering the coupling effect of multiple factors of the present invention. DETAILED DESCRIPTION

[0066] Hereinafter, embodiments of the present invention will be described with reference to the drawings.

[0067] Figure 1 The flowchart of the method for predicting the life of an aviation motor in the present invention taking into account the coupling effect of multiple factors is shown in FIG. Figure 1 As shown, the present invention provides a method for predicting the life of an aviation motor taking into account the coupling effect of multiple factors, which includes:

[0068] S1, based on the uniform test, the motor life test is carried out to obtain several factors affecting the life of aviation motors.

[0069] S2, let each factor take different values ​​and repeat the experiment to obtain several groups of test data. The test data is the corresponding relationship between the value of each factor and the actual value of the life of the aircraft motor.

[0070] S3, preprocessing each test data to obtain a training data set, a test data set and a validation data set.

[0071] Preprocessing includes data cleaning, data transformation, data segmentation, and data normalization.

[0072] Data cleaning removes duplicates and missing items from the original data. Data transformation converts the cleaned data into a format that is easy for machine learning models to read. Data normalization normalizes the transformed data to fall between [-1, 1] or [0, 1]. Data segmentation divides the normalized data into several groups and constructs training data sets, test data sets, and validation data sets.

[0073] S4, iteratively training, testing and verifying the motor life prediction model based on the training data set, the test data set and the validation data set to obtain a trained motor life prediction model.

[0074] Specifically, S4 includes:

[0075] S41, let i=1; i=[1,2,…,M], where M is an iterative training parameter.

[0076] S42, iteratively training the motor life prediction model based on the training data set.

[0077] S43, testing the motor life prediction model based on the test data set to obtain a test prediction error of the motor life prediction model.

[0078] The test prediction error calculation formula is as follows:

[0079]

[0080] Where: n is the number of test samples in the test data set, y i is the predicted value of the i-th test sample, is the average of the predicted values ​​of n test samples, Q is the test prediction error, and Y i is the actual value of the i-th test sample, is the average of the actual values ​​of n test samples.

[0081] S44, judge i, if i is less than M, set i=i+1 and return to S42, if i is greater than or equal to M, select the motor life prediction model corresponding to the minimum value of the test prediction error as the initial trained motor life prediction model and execute S45.

[0082] S45, verifying the initially trained motor life prediction model based on the verification data set to obtain a verification prediction error.

[0083] S46, judging the verification prediction error, if the verification prediction error is less than the error setting value, the initially trained motor life prediction model is used as the trained motor life prediction model, if the verification prediction error is greater than or equal to the error setting value, returning to S41.

[0084] S5, obtaining a life prediction value of the motor to be tested based on actual data and the trained motor life prediction model.

[0085] The motor life prediction model is a back propagation model based on neural network, namely BP-ANN model. The motor life prediction model includes input layer, hidden layer and output layer.

[0086] The training process of the motor life prediction model includes two stages: forward propagation and back propagation.

[0087] In the forward propagation stage, the input sample is passed from the input layer to the hidden layer, and then to the output layer through the hidden layer. Each layer of neurons performs weighted summation on the input and normalizes it through the activation function before outputting it.

[0088] The output value of the output layer is as follows:

[0089] y=NN(x)=σ(W k ...σ(W2σ(W1x+b1)+b2)+…+b k );

[0090] Where: y is the output value of the output layer, x is the input sample of the input layer, σ is the activation function, W k is the weight vector of the kth layer, b k is the bias of the kth layer.

[0091] In the back propagation stage, the weight vectors of neurons in each layer are updated based on the LM method according to the actual value and the output value of the output layer to reduce the error between the actual value and the output value of the output layer.

[0092] The weight update formula is as follows:

[0093] (J T J+I)(W k+1 -W k )=J T (Y-NN(x));

[0094] Where: J is the Jacobian function of the motor life prediction model, Y is the actual value, W k is the weight vector of the back-propagation k-th layer, W k+1 is the weight vector for back-propagation of the k+1th layer, I is the identity matrix, and T is the transpose.

[0095] The present invention also provides an aviation motor life prediction system considering the coupling effect of multiple factors, which includes:

[0096] The influencing factor module is used to conduct motor life tests based on uniform tests to obtain several factors that affect the life of aviation motors.

[0097] The data acquisition module is used to make each factor take different values ​​and conduct repeated experiments to obtain several groups of test data; the test data is the corresponding relationship between the value of each factor and the actual value of the life of the aviation motor.

[0098] The data processing module is used to preprocess the test data to obtain the training data set, the test data set and the verification data set.

[0099] The model training module is used to iteratively train, test and verify the motor life prediction model based on the training data set, the test data set and the verification data set to obtain a trained motor life prediction model.

[0100] The life prediction module is used to obtain the life prediction value of the motor to be tested based on actual data and the trained motor life prediction model.

[0101] The motor life prediction model includes an input layer, a hidden layer and an output layer.

[0102] The training process of the motor life prediction model includes two stages: forward propagation and back propagation.

[0103] In the forward propagation stage, the input sample is passed from the input layer to the hidden layer, and then to the output layer through the hidden layer. Each layer of neurons performs weighted summation on the input and normalizes it through the activation function before outputting it.

[0104] The output value of the output layer is as follows:

[0105] y=NN(x)=σ(W k ...σ(W2σ(W1x+b1)+b2)+…+b k );

[0106] Where: y is the output value of the output layer, x is the input sample of the input layer, σ is the activation function, W k is the weight vector of the kth layer, b k is the bias of the kth layer.

[0107] In the back propagation stage, the weight vectors of neurons in each layer are updated based on the LM method according to the actual value and the output value of the output layer to reduce the error between the actual value and the output value of the output layer:

[0108] The weight update formula is as follows:

[0109] (J T J+I)(W k+1 -W k )=J T (Y-NN(x));

[0110] Where: J is the Jacobian function of the motor life prediction model, Y is the actual value, W k is the weight vector of the back-propagation k-th layer, W k+1 is the weight vector for back-propagation of the k+1th layer, I is the identity matrix, and T is the transpose.

[0111] Optionally, the motor life prediction model is a back propagation model based on a neural network, namely, a BP-ANN model.

[0112] Optionally, preprocessing includes data cleaning, data transformation, data segmentation, and data normalization.

[0113] Optionally, the model training module includes:

[0114] The instruction unit is used to set i=1; i=[1,2,…,M], where M is an iterative training parameter.

[0115] The training unit is used to iteratively train the motor life prediction model based on the training data set.

[0116] The testing unit is used to test the motor life prediction model based on a test data set to obtain a test prediction error of the motor life prediction model.

[0117] The first judgment unit is used to judge i. If i is less than M, i=i+1 is set and returned to the training unit. If i is greater than or equal to M, the motor life prediction model corresponding to the minimum value of the test prediction error is selected as the initial trained motor life prediction model and the verification unit is executed.

[0118] The verification unit is used to verify the initially trained motor life prediction model based on the verification data set to obtain a verification prediction error.

[0119] The second judgment unit is used to judge the verification prediction error. If the verification prediction error is less than the error setting value, the initially trained motor life prediction model is used as the trained motor life prediction model. If the verification prediction error is greater than or equal to the error setting value, it is returned to the instruction unit.

[0120] Optionally, the test prediction error calculation formula is as follows:

[0121]

[0122] Where: n is the number of test samples in the test data set, y i is the predicted value of the i-th test sample, is the average of the predicted values ​​of n test samples, Q is the test prediction error, and Y i is the actual value of the i-th test sample, is the average of the actual values ​​of n test samples.

[0123] Specifically, based on the design method of uniform test, the motor life test under comprehensive environment is designed. The three factors affecting the life of aviation motors are determined to be temperature, humidity and load, where the influence level of each factor is 7 levels, the temperature is 25℃, 50℃, 75℃, 100℃, 125℃, 150℃ and 175℃; the humidity is 30%, 40%, 50%, 60%, 70%, 80% and 90%; the load is 10mN·m, 20mN·m, 30mN·m, 40mN·m, 50mN·m, 60mN·m and 70mN·m.

[0124] By making each factor take different values, a uniform design table is obtained, as shown in Table 1. Repeated experiments are carried out based on the uniform design table to obtain several groups of test data.

[0125] Table 1 Uniform design table

[0126] Serial number Temperature grade Humidity level Load level 1 1 2 3 2 2 4 6 3 3 6 2 4 4 1 5 5 5 3 1 6 6 5 4 7 7 7 7

[0127] Each experimental data is preprocessed to obtain a training data set, a test data set, and a validation data set.

[0128] The motor life prediction model is iteratively trained, tested and verified based on the training data set, the test data set and the validation data set to obtain a trained motor life prediction model.

[0129] The life prediction value of the motor to be tested is obtained based on actual data and the trained motor life prediction model.

[0130] The embodiments described above are only descriptions of the preferred implementation modes of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A method for predicting the life of an aviation motor considering the coupling of multiple factors, characterized in that: It includes: S1, conduct motor life test based on uniform test and obtain several factors affecting the life of aviation motor; S2, making each of the factors take different values ​​and repeating the experiment to obtain several groups of test data, wherein the test data is the corresponding relationship between the value of each factor and the actual value of the life of the aircraft motor; S3, preprocessing each of the test data to obtain a training data set, a test data set, and a verification data set; S4, iteratively training, testing and verifying the motor life prediction model based on the training data set, the test data set and the verification data set to obtain the trained motor life prediction model; S5, obtaining a life prediction value of the motor to be tested based on actual data and the trained motor life prediction model; The motor life prediction model includes an input layer, a hidden layer and an output layer; The training process of the motor life prediction model includes two stages: forward propagation and back propagation; In the forward propagation stage, the input sample is transmitted to the hidden layer through the input layer, and then transmitted to the output layer through the hidden layer. Each layer of neurons performs weighted summation on the input and performs normalization processing through the activation function before outputting. The output value of the output layer is as follows: y=NN(x)=σ(W k ...σ(W2σ(W1x+b1)+b2)+…+b k ); Where: y is the output value of the output layer, x is the input sample of the input layer, σ is the activation function, W k is the weight vector of the kth layer, b k is the bias of the kth layer; In the back propagation stage, the weight vectors of neurons in each layer are updated based on the LM method according to the actual value and the output value of the output layer to reduce the error between the actual value and the output value of the output layer: The weight update formula is as follows: (J T J+I)(W k+1 -W k )=J T (Y-NN(x)); Where: J is the Jacobian function of the motor life prediction model, Y is the actual value, W k is the weight vector of the back-propagation k-th layer, W k+1 is the weight vector for back-propagation of the k+1th layer, I is the identity matrix, and T is the transpose.

2. The method for predicting the life of an aviation motor considering the coupling effect of multiple factors according to claim 1 is characterized in that: The motor life prediction model is a back propagation model based on a neural network, namely a BP-ANN model.

3. The method for predicting the life of an aviation motor considering the coupling effect of multiple factors according to claim 1 is characterized in that: The preprocessing includes data cleaning, data conversion, data segmentation and data normalization.

4. The method for predicting the life of an aircraft motor considering the coupling effect of multiple factors according to claim 1 is characterized in that: The S4 is specifically: S41, let i = 1; i = [1, 2, ..., M], M is an iterative training parameter; S42, iteratively training the motor life prediction model based on the training data set; S43, testing the motor life prediction model based on the test data set to obtain a test prediction error of the motor life prediction model; S44, judging i, if i is less than M, setting i=i+1 and returning to S42, if i is greater than or equal to M, selecting the motor life prediction model corresponding to the minimum value of the test prediction error as the initially trained motor life prediction model and executing S45; S45, verifying the initially trained motor life prediction model based on the verification data set to obtain a verification prediction error; S46, judging the verification prediction error, if the verification prediction error is less than the error setting value, using the initially trained motor life prediction model as the trained motor life prediction model, if the verification prediction error is greater than or equal to the error setting value, returning to S41.

5. The method for predicting the life of an aircraft motor considering the coupling effect of multiple factors according to claim 4 is characterized in that: The test prediction error calculation formula is as follows: Where: n is the number of test samples in the test data set, y i is the predicted value of the i-th test sample, is the average of the predicted values ​​of n test samples, Q is the test prediction error, and Y i is the actual value of the i-th test sample, is the average of the actual values ​​of n test samples.

6. An aviation motor life prediction system considering the coupling of multiple factors, characterized in that: It includes: The influencing factor module is used to conduct motor life tests based on uniform tests to obtain several factors that affect the life of aviation motors; A data acquisition module is used to make each of the factors take different values ​​and conduct repeated experiments to obtain several groups of test data; the test data is the corresponding relationship between the value of each factor and the actual value of the life of the aircraft motor; A data processing module is used to pre-process the test data to obtain a training data set, a test data set and a verification data set; A model training module, used for iteratively training, testing and verifying the motor life prediction model based on the training data set, the test data set and the verification data set to obtain the trained motor life prediction model; A life prediction module, used to obtain a life prediction value of the motor to be tested based on actual data and the trained motor life prediction model; The motor life prediction model includes an input layer, a hidden layer and an output layer; The training process of the motor life prediction model includes two stages: forward propagation and back propagation; In the forward propagation stage, the input sample is transmitted to the hidden layer through the input layer, and then transmitted to the output layer through the hidden layer. Each layer of neurons performs weighted summation on the input and performs normalization processing through the activation function before outputting. The output value of the output layer is as follows: y=NN(x)=σ(W k ...σ(W2σ(W1x+b1)+b2)+…+b k ); Where: y is the output value of the output layer, x is the input sample of the input layer, σ is the activation function, W k is the weight vector of the kth layer, b k is the bias of the kth layer; In the back propagation stage, the weight vectors of neurons in each layer are updated based on the LM method according to the actual value and the output value of the output layer to reduce the error between the actual value and the output value of the output layer: The weight update formula is as follows: (J T J+I)(W k+1 -W k )=J T (Y-NN(x)); Where: J is the Jacobian function of the motor life prediction model, Y is the actual value, W k is the weight vector of the back-propagation k-th layer, W k+1 is the weight vector for back-propagation of the k+1th layer, I is the identity matrix, and T is the transpose.

7. The aircraft motor life prediction system considering the multi-factor coupling effect according to claim 6 is characterized in that: The motor life prediction model is a back propagation model based on a neural network, namely a BP-ANN model.

8. The aircraft motor life prediction system considering the multi-factor coupling effect according to claim 6 is characterized in that: The preprocessing includes data cleaning, data conversion, data segmentation and data normalization.

9. The aircraft motor life prediction system considering the multi-factor coupling effect according to claim 6 is characterized in that: The model training module includes: An instruction unit is used to set i=1; i=[1,2,…,M], where M is an iterative training parameter; A training unit, configured to iteratively train the motor life prediction model based on the training data set; A testing unit, used for testing the motor life prediction model based on the test data set to obtain a test prediction error of the motor life prediction model; A first judgment unit is used to judge i. If i is less than M, i=i+1 is set and the training unit is returned. If i is greater than or equal to M, the motor life prediction model corresponding to the minimum value of the test prediction error is selected as the initially trained motor life prediction model and the verification unit is executed; A verification unit, used to verify the initially trained motor life prediction model based on the verification data set to obtain a verification prediction error; The second judgment unit is used to judge the verification prediction error. If the verification prediction error is less than the error setting value, the initially trained motor life prediction model is used as the trained motor life prediction model. If the verification prediction error is greater than or equal to the error setting value, it is returned to the instruction unit.

10. The aircraft motor life prediction system considering the coupling of multiple factors according to claim 9, characterized in that: The test prediction error calculation formula is as follows: Where: n is the number of test samples in the test data set, y i is the predicted value of the i-th test sample, is the average of the predicted values ​​of n test samples, Q is the test prediction error, and Y i is the actual value of the i-th test sample, is the average of the actual values ​​of n test samples.

Images