Aerospace equipment digital-real fusion test small sample data analysis method

Abstract

The application discloses a kind of space equipment number real fusion test small sample data analysis methods, comprising: design space equipment non-destructive data acquisition module, the module is obtained by health management system actual operation data, for extracting the health data of space equipment;Design space equipment twin prediction module, the module will actual operation data as input, train twin learning model to process data, output the prediction result of health state;Design online transfer learning module, the module is constantly updated and optimized twin learning model according to new operation data.This method can guarantee under the premise of only obtaining a small amount of space equipment information, predict its running health state data, and improve the prediction accuracy through continuous learning mechanism, provide strong support for the maintenance and fault diagnosis of space equipment.

Description

Technical Field

[0001] This invention belongs to the fields of electronic engineering and computer science, and specifically relates to a method for analyzing small sample data from the fusion of data and reality in aerospace equipment testing. Background Technology

[0002] Aerospace equipment plays a crucial role in the aerospace industry, including aircraft engines, navigation systems, and flight control systems. Unexpected failures in any of these systems can severely impact safety and economic performance. Therefore, real-time monitoring and fault diagnosis, along with highly real-time and accurate health status predictions, are essential. Due to the high cost and small volume of data generated during the operation of aerospace equipment, a major challenge in data-real-time fusion testing of aerospace equipment is how to quickly extract useful information from this limited data and maintain high accuracy and real-time performance in analysis with small sample sizes. Summary of the Invention

[0003] The technical problem this invention aims to solve is to provide a method for analyzing small-sample data from data-real fusion tests of aerospace equipment. This method encompasses the design of a non-destructive data acquisition module, an aerospace equipment twin prediction module, and an online transfer learning module. It can improve the real-time performance and accuracy of aerospace equipment health status prediction to a certain extent, thereby enhancing the safety and economy of aerospace equipment. Therefore, this invention discloses a method for analyzing small-sample data from data-real fusion tests of aerospace equipment. Through the design of a non-destructive data acquisition module, an aerospace equipment twin prediction module, and an online transfer learning module, and with the help of digital twin technology, it improves the real-time performance and accuracy of aerospace equipment health status prediction to a certain extent, further ensuring the safety and economy of aerospace equipment.

[0004] The technical problem solved by this invention is achieved through the following technical solution: a method for analyzing small sample data from aerospace equipment data-real fusion tests, comprising:

[0005] Step S1: Design a non-destructive data acquisition module for aerospace equipment. This module obtains raw equipment data from the aerospace equipment and performs data cleaning.

[0006] Step S2: Aerospace equipment twin prediction module. This module first constructs a neural network based on physical information, then trains the model using measured health status data of historical aerospace equipment under normal operating conditions, and finally receives some measured health status data of aerospace equipment under normal operating conditions from step S1 to make real-time predictions of health status.

[0007] Step S3: Online transfer learning module, which is used to predict the health status based on small sample data of real-time detection of health status of aerospace equipment under normal operating conditions.

[0008] The advantages of this invention compared to existing technologies are as follows: by using digital twin technology and online physical information networks to obtain historical health data of aerospace equipment, network parameters are initialized. Under this premise, by obtaining real-time health data of aerospace equipment, the health status can be quickly predicted with small sample data. At the same time, the digital twin model is iterated based on the loss function output, which can improve the real-time performance and accuracy of aerospace equipment health status prediction to a certain extent, and enhance the safety and economy of aerospace equipment. Attached Figure Description

[0009] Figure 1 This is a flowchart of a method for analyzing small sample data in aerospace equipment data fusion testing according to the present invention. Detailed Implementation

[0010] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other. To achieve the above objectives, this invention adopts the following technical solution.

[0011] This invention relates to a method for analyzing small sample data from aerospace equipment data-real fusion tests, including the design of a non-destructive data acquisition module, an aerospace equipment twin prediction module, and an online transfer learning module. This method can improve the real-time performance and accuracy of aerospace equipment health status prediction to a certain extent, thereby enhancing the safety and economy of aerospace equipment.

[0012] The flowchart of the present invention is as follows Figure 1 As shown, taking the engine in aerospace equipment as an example, the specific implementation method is as follows:

[0013] Step S1: Design a non-destructive data acquisition module for the engine. This module obtains raw equipment data from the engine and performs data cleaning. The specific implementation is as follows:

[0014] To address the discrepancy between engine health data under normal operating conditions and non-operational conditions, a health management system is designed. This system, through sensor networks, data acquisition cards, and communication interface facilities, ensures the continuous collection of real-time health data at every stage of engine operation.

[0015] To address the complex operating conditions and high noise levels of engines, a data cleaning module was designed. Utilizing an outlier detection algorithm, the quartiles were first calculated. and interquartile range : =25th percentile =75th percentile Then determine the outlier limits: lower bound = Upper bound = Finally, outlier detection is performed on all data: if lower bound or The upper realm, then Outliers are removed, achieving deep purification of the collected data and ensuring the reliability of health status data used in subsequent analysis.

[0016] For situations where engine operating data is scarce and complex, the Pearson correlation coefficient of the data is calculated: ,in and There are two characteristics. Sample points In features The value below, Sample points In features The value below, and These are the samples in terms of features and The mean of the following, It is a feature and characteristics The Pearson correlation coefficient between them The number of samples is used to compare the correlation between features and health status by comparing the magnitude of coefficient values, and feature selection is made by considering the physical meaning of the parameters. For each feature's peak value, mean, standard deviation, kurtosis, curve entropy, and curve slope, principal component analysis is used: first, the covariance matrix of the data is calculated. Then solve for the covariance matrix. The eigenvalues ​​and corresponding eigenvectors are used to extract and enhance features that are highly correlated with health status, thereby improving the accuracy and efficiency of data analysis. The eigenvalues ​​are arranged in descending order and the eigenvectors corresponding to the three largest eigenvalues ​​are selected as principal components.

[0017] The formula for calculating kurtosis is: ,in It is the first in the dataset Observations of each sample point It is the mean of the entire sample. It is the number of sample points. It is the variance of the entire sample;

[0018] The formula for calculating curve entropy is: ,in Represents the first in the dataset Normalized values ​​of each data point This represents the total number of data points in the dataset;

[0019] The formula for calculating the slope of a curve is: ,in This represents the observation at the beginning of a certain iteration of the loop. This represents the observation at the termination time of a certain iteration of the loop. This represents the total time taken from the start time to the end time of a certain loop.

[0020] Step S2: Design an engine twin prediction module. This module first constructs a neural network based on physical information, including structural mechanical properties, material properties, and fluid dynamics properties. Then, it trains the model using historical measured health status data of the engine under normal operating conditions. Finally, it receives partial measured health status data of the engine under normal operating conditions from step S1 to perform real-time health status prediction. The specific implementation is as follows:

[0021] Based on the actual physical characteristics of the engine, specifically, a neural network model including a health state decay rate model and a health state decay model was built. The input of the network is the feature data selected in step S1. The health state decay rate is calculated through two layers of MLP and one layer of FC. Then, the predicted health state decay rate and the data selected in S1 are fed into the second part of the model consisting of one layer of MLP and one layer of FC to predict the health state of the aerospace equipment.

[0022] MLP stands for Multilayer Perceptron, a type of feedforward neural network. It consists of three layers: an input layer, a hidden layer, and an output layer. Each node (except the input node) is a neuron with a non-linear activation function. Through full connections between these layers, input samples are fed forward layer by layer, ultimately outputting high-level abstract information extracted from the data.

[0023] FC stands for Fully Connected Layer, where each neuron is connected to all neurons in the previous layer. Each connection corresponds to a learnable weight, which can then integrate information from the MLP layer and output the final prediction result.

[0024] After establishing the above network, the model parameters are initialized. Then, the neural network model is trained using measured health status data under normal engine operating conditions. The specific training method is as follows: First, the input data is fed forward. After each round of feedforward calculation, the Euclidean distance between the predicted health status decay rate and the true value is calculated. ,in It is the network's predicted output. The predicted health state is the true value, and the calculated result is used as the first penalty term. The Euclidean distance between the predicted health state and the true health state is used as the second penalty term. The sum of the two penalty terms is used as the total penalty term, and then the backpropagation algorithm is used. ,in It is the learning rate, Model parameters This is the overall penalty item; the model parameters are adjusted to ensure that the model can accurately simulate changes in the engine's health status.

[0025] Step S3: Design an online transfer learning module. This module is used to predict the health status based on small sample data of real-time detection of engine health status under normal operating conditions. The specific implementation is as follows:

[0026] The small sample data of real-time engine features after outlier processing, feature selection and principal component analysis in step S1 are input into the neural network framework whose parameters have been updated with historical data in step S2, so as to output the engine health status predicted by the network.

[0027] The predicted real-time engine health status is compared with the actual health status data obtained in step S1 to calculate the loss function. Based on the result of the loss function, the parameters of the physical information network are fine-tuned through the backpropagation algorithm to improve the model's prediction performance and generalization ability on small sample datasets.

[0028] In summary, this invention discloses a small sample data analysis method for data-real fusion testing of aerospace equipment, including the design of a non-destructive data acquisition module, a twin prediction module, and an online transfer learning module for aerospace equipment. This method can improve the real-time performance and accuracy of aerospace equipment health status prediction to a certain extent, thereby enhancing the safety and economy of aerospace equipment.

[0029] The contents not described in detail in this specification are existing technologies known to those skilled in the art.

[0030] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for analyzing small sample data from aerospace equipment data-real fusion tests, characterized in that, include: Step S1: Design a non-destructive data acquisition module for aerospace equipment. This module obtains raw equipment data from the aerospace equipment and performs data cleaning. Step S2: Design a twin prediction module for aerospace equipment. This module first constructs a neural network based on physical information, then trains the model using measured health status data of aerospace equipment under normal operating conditions in history, and finally receives some measured health status data of aerospace equipment under normal operating conditions from step S1 to make real-time predictions of health status. Step S3: Design an online transfer learning module, which is used to realize health status prediction based on small sample data of real-time detection of health status under normal operating conditions of aerospace equipment; The non-destructive data acquisition module for aerospace equipment in step S1 is implemented as follows: Step S1.1: To address the discrepancy between the health data status of aerospace equipment under normal operating conditions and non-operational conditions, a health management system is designed. This system ensures the continuous collection of real-time health data at each stage of aerospace equipment operation through sensor networks, data acquisition cards, and communication interface facilities. Step S1.2: Considering the complex operating conditions and high noise interference of aerospace equipment, a data cleaning module is designed. Using an outlier detection algorithm, the quartiles are first calculated. and interquartile range : =25th percentile =75th percentile Then determine the outlier limits: lower bound = Upper bound = Finally, outlier detection is performed on all data: if lower bound or The upper realm, then Outliers are removed, achieving deep cleaning of the collected data; Step S1.3: For situations where the amount of operational data for aerospace equipment is small and complex, calculate the Pearson correlation coefficient of the data: ,in and There are two characteristics. Sample points In features The value to be taken below, Sample points In features The value below, and These are the samples in terms of features and The mean of the following, It is a feature and characteristics The Pearson correlation coefficient between them The number of samples is used to compare the correlation between features and health status by comparing the Pearson correlation coefficient values, and feature selection is made by considering the physical meaning of the parameters. For each feature, the peak value, mean, standard deviation, kurtosis, curve entropy, and curve slope are calculated using principal component analysis, i.e., first calculating the covariance matrix of the data. Then solve for the covariance matrix. The eigenvalues ​​and corresponding eigenvectors are defined, where the eigenvalues ​​reflect the magnitude of the variance along each eigendirection, and the eigenvectors indicate the direction of these variances. Finally, the eigenvalues ​​are arranged in descending order, and the top eigenvalues ​​are selected. The eigenvectors corresponding to the largest eigenvalues ​​are used as principal components to further extract and enhance features highly correlated with health status, thereby improving the accuracy and efficiency of data analysis; among them, The formula for kurtosis is: ,in It is the first in the dataset Observations of each sample point It is the mean of the entire sample. It is the number of sample points. It is the variance of the entire sample; The formula for curve entropy is: ,in Represents the first in the dataset Normalized values ​​of each data point This represents the total number of data points in the dataset; The formula for the slope of a curve is: ,in This represents the observation at the beginning of a certain iteration of the loop. This represents the observation at the termination time of a certain iteration of the loop. This represents the total time taken from the start to the end of a given loop. The specific implementation of the aerospace equipment twin prediction module in step S2 is as follows: Step S2.1: Combining the actual physical characteristics of aerospace equipment, specifically, a neural network model including a health status decay rate model and a health status decay model is built. The input of this network is the feature data selected in step S1. After passing through two layers of MLP and one layer of FC, the health status decay rate is calculated. Then, the predicted health status decay rate and the data selected in S1 are fed into the second part of the model consisting of one layer of MLP and one layer of FC to predict the health status of the aerospace equipment. MLP stands for Multilayer Perceptron, a type of feedforward neural network. It consists of three layers: an input layer, a hidden layer, and an output layer. Each node, except for the input node, is a neuron with a non-linear activation function. Through the full connections between these layers, the input samples are fed forward layer by layer, and the final output is a high-level abstract information extracted from the data. FC stands for fully connected layer, where each neuron is connected to all neurons in the previous layer. Each connection corresponds to a learnable weight, which integrates information from the MLP layer and outputs the final prediction result. After building the above network, the model parameters are initialized. Step S2.2: Train the neural network model using measured health status data of aerospace equipment under normal operating conditions. The specific training method is as follows: First, perform feedforward calculations on the input data. After each round of feedforward calculations, calculate the Euclidean distance between the predicted health status decay rate and the true value. ,in It is the network's predicted output. The predicted health state is the true value, and the calculated result is used as the first penalty term. The Euclidean distance between the predicted health state and the true health state is used as the second penalty term. The sum of the two penalty terms is used as the total penalty term, and then the backpropagation algorithm is used. ,in It is the learning rate, Model parameters This is the overall penalty item; the model parameters are adjusted to ensure that the model can accurately simulate changes in the health status of aerospace equipment.

2. The method for analyzing small sample data from aerospace equipment data fusion tests according to claim 1, characterized in that, The online transfer learning module in step S3 is implemented as follows: Step S3.1: Input the small sample data of real-time features of aerospace equipment obtained in step S1 into the neural network model whose parameters have been updated with historical data in step S2, so as to output the real-time health status of aerospace equipment predicted by the network. Step S3.2: Compare the predicted real-time health status of aerospace equipment with the acquired actual health status data to calculate the loss function, and fine-tune the parameters of the physical information network through the backpropagation algorithm based on the result of the loss function to improve the prediction performance and generalization ability of the model on small sample datasets.

Images