Aviation converter fault feature extraction and diagnosis method based on multi-source information

Abstract

The invention relates to the technical field of aeronautical equipment fault detection, and discloses an aeronautical converter fault feature extraction and diagnosis method based on multi-source information, which is characterized in that time sequence data and spatial distribution data are respectively collected and preprocessed, and the accuracy and the stability of the data are improved. Time feature extraction, spatial feature extraction, frequency spectrum acquisition and feature extraction are respectively carried out on the preprocessed data to realize fusion of multi-source information; information from different types of sensors and different feature domains is integrated, and multi-dimensional information of the equipment operation state is fully excavated, so that the accuracy and reliability of fault detection are improved. And finally, performing envelope detection on the signal to obtain an envelope line of signal amplitude change, and analyzing characteristic parameters of the envelope to detect fault information or pulse type fault characteristics in the modulation signal. And based on a deep convolutional neural network, learning a complex relationship between the envelope features and the fault, and realizing the diagnosis of the aviation airborne converter fault.

Description

Technical Field

[0001] The present invention relates to the technical field of aviation equipment fault detection, and more specifically, to an aviation inverter fault feature extraction and diagnosis method based on multi-source information. Background Art

[0002] Traditional fault detection methods for aircraft airborne converters mainly rely on a single signal analysis method, such as only performing time domain analysis on electrical signals such as voltage and current, or only focusing on the temperature changes of the equipment. These methods cannot fully and accurately extract fault feature quantities. On the one hand, single signal analysis is easily interfered by noise, resulting in the masking of fault features; on the other hand, it is unable to fully utilize the multi-dimensional information during the operation of the equipment, and the detection capabilities for some complex faults and early faults are insufficient. For example, when a slight local overheating fault occurs inside the converter, it is difficult to detect it in time only through time domain analysis of electrical signals, and the single-point measurement of the temperature sensor cannot accurately reflect the spatial distribution of the fault.

[0003] Therefore, the present invention provides an aviation inverter fault feature extraction and diagnosis method based on multi-source information, which improves the above technical problems. Summary of the invention

[0004] The disclosed embodiments aim to address the deficiencies of the prior art and provide an aviation inverter fault feature extraction and diagnosis method based on multi-source information. The present invention adopts multi-source information fusion technology to integrate information from different types of sensors and different feature domains, fully mining the multi-dimensional information of the equipment operating status, thereby improving the accuracy and reliability of fault detection.

[0005] The above technical objectives of the present invention are achieved through the following technical solutions: A method for extracting and diagnosing fault features of an aviation inverter based on multi-source information comprises the following steps:

[0006] S1, collect and preprocess time series data and spatial distribution data respectively;

[0007] S2. Perform time feature extraction, spatial feature extraction, spectrum acquisition and feature extraction on the preprocessed data to achieve the fusion of multi-source information.

[0008] As a preferred technical solution of the present invention, the time series data collection process is: through sensors installed at key parts of the aircraft airborne converter, the data series of electrical parameters and environmental parameters of the converter during operation are continuously collected to form a data set of time dimension;

[0009] The spatially distributed data acquisition process is as follows: multiple sensors are arranged at different physical positions of the converter to obtain information data from different spatial regions inside the converter; these spatially distributed sensor data are organized and numbered according to the physical structure and circuit topology of the converter, and a corresponding relationship between the data and the spatial position is established to form a data set of spatial dimension.

[0010] As a preferred technical solution of the present invention, the steps of preprocessing the time series data include: wavelet decomposition, noise standard deviation estimation, threshold determination, threshold processing, and wavelet reconstruction;

[0011] The steps of spatial distribution data preprocessing include: information entropy calculation, anomaly determination, and spatial interpolation.

[0012] As a preferred technical solution of the present invention, the process of wavelet decomposition is: ) for discrete wavelet transform, using Coiflet wavelet Coif3 with good symmetry and regularity for processing: Approximate coefficients of layer decomposition The calculation formula is: ,in, , are the low-pass filter coefficients corresponding to the selected wavelet basis, for its conjugation; Layer decomposition detail factor The calculation formula is: ,in , are the high-pass filter coefficients corresponding to the selected wavelet basis, for its conjugation;

[0013] The process of estimating the noise standard deviation is as follows: Calculate the detail coefficient The noise standard deviation estimate of ; Using the median-based estimation method: ;

[0014] The process of determining the threshold is as follows: initial threshold The calculation formula is: ;

[0015] Dynamic Adjustment Factor ,in is the weight coefficient, is the temperature change, is the maximum allowable temperature change; the final threshold for ;

[0016] The process of threshold processing is as follows: For mixed parameters , the processed detail coefficient For: When hour, ;

[0017] Among them, the soft threshold part Can smooth noise, hard threshold part Can preserve the sharp features of the signal;

[0018] The process of wavelet reconstruction is: using the processed detail coefficients and approximate coefficients Perform inverse wavelet transform to reconstruct the denoised signal The formula is: ,in is the maximum number of decomposition levels.

[0019] As a preferred technical solution of the present invention, the process of calculating information entropy is as follows: The data collected by a sensor at a certain moment ; First, calculate the mean of the data in the neighborhood ; Then, calculate the information entropy , the formula is ,in ;

[0020] The process of abnormal judgment is: set a normal information entropy threshold range for each sensor neighborhood ; Collect all sensor neighborhood information entropy data when the converter operates normally for a period of time and calculate its mean and standard deviation ,set up , ,in and is an empirical coefficient and can be adjusted according to actual conditions, such as ; When the information entropy of a sensor neighborhood When this threshold is exceeded, or , then it is determined that there may be anomalies in the spatial area where the neighborhood is located;

[0021] The process of spatial interpolation is as follows: for any point in the converter space , its physical quantity Estimated value of Calculated by the following formula: ,in is the known sensor position The measured value of the physical quantity at is the number of sensors involved in interpolation, is the weight coefficient, obtained by solving the following system of equations: ,here is the semivariogram function, which is used to describe two points in space and Variability of physical quantities between, Gaussian semivariogram function ,in is the distance between two points, It's the nugget effect. is the sill value, is the range of variation, and these parameters are determined by analyzing the variogram of the sensor data.

[0022] As a preferred technical solution of the present invention, the process of time feature extraction is: find the time series All local maximum and minimum points of are interpolated by cubic spline to obtain the upper envelope and lower envelope , calculate the mean of the upper and lower envelopes , subtract the mean from the original sequence to get ;like If it satisfies the IMF condition, it is the first IMF component; otherwise, Repeat the above process as a new sequence until a satisfying condition is obtained. ; Then, subtract from the original sequence , and get the remaining sequence ,right Repeat the above decomposition steps to obtain and the residual ,Right now ;

[0023] In order to more accurately reflect the intrinsic characteristics of the multivariate time series of aircraft airborne converters, a weight is assigned to each variable according to the importance of each converter parameter on the fault. ; Coarse-grained sequence Calculated as ;

[0024] For the coarse-grained sequence , calculate its sample entropy; given the embedding dimension and tolerance , define the vector ;make Represents two vectors and The Chebyshev distance of ;definition To satisfy of The number and Ratio ;but ; Similarly, calculate dimensional vector ; Sample entropy for ;

[0025] For time series , its autocorrelation function ,in Is a sequence The mean of

[0026] For time series , when calculating the autocorrelation, taking into account different delays subsequence under and ; The DTW algorithm works by finding a path ,in , so that and Minimize the distance between paths The constraints are , ,and ; The distance metric uses Euclidean distance ; Define DTW distance ; Improved autocorrelation function ;

[0027] For two time series and , the traditional cross-correlation function ; The DTW algorithm is also introduced for improvement; for different delays subsequence under and , find the optimal path through the DTW algorithm , calculate the DTW distance ; Improved cross-correlation function .

[0028] As a preferred technical solution of the present invention, the process of spatial feature extraction is: constructing a deep convolutional neural network to extract the fused spatial features; the network structure includes multiple convolutional layers, pooling layers and fully connected layers;

[0029] The convolution layer: using convolution kernels of different sizes Perform convolution operation on the input feature map; for the input feature map , the convolutional layer outputs feature maps Elements Calculated as ,in is the convolution kernel element, is the bias, is the input feature map element;

[0030] The pooling layer: uses maximum pooling, for the input feature map , in the pooling window Inside, output feature map Elements , reducing data dimensionality while retaining key features;

[0031] The fully connected layer: flattens the pooled feature map and passes the weight matrix and bias Perform linear transformation, and then use the activation function ReLU to get the final output; the output layer uses the Softmax function to map the result to the probability distribution of the fault category , is the output of the fully connected layer, is the number of fault categories; the network is trained by a large amount of spatial feature data with fault labels, so that the network can learn the complex relationship between spatial features and faults and achieve accurate fault location;

[0032] The method based on multi-physics field coupling and deep learning can deeply extract the spatial characteristics of aviation airborne converters and improve the accuracy and reliability of fault location.

[0033] As a preferred technical solution of the present invention, the process of spectrum acquisition is: define each modal function Hilbert transform , and Composition of analytical signals , through the index term Demodulate to baseband and get ; Find the first-order derivative of The square of the norm, that is , which is used as the bandwidth measure of each modal function; the constrained variation problem is constructed as , the constraints are ,in is the preset number of decomposition modes;

[0034] By introducing the quadratic penalty factor and Lagrange multipliers , transforming the constrained variational problem into an unconstrained augmented Lagrangian function ; The augmented Lagrangian function is solved iteratively using the alternating direction multiplier method, and the update and , until the convergence condition is met;

[0035] After VMD decomposition, the original signal is decomposed into IMF components , each component has different frequency characteristics; VMD adaptively separates different frequency components, avoiding problems such as spectrum aliasing, and providing purer signal components for subsequent spectrum feature extraction;

[0036] The process of spectrum feature extraction is as follows: time-frequency atomic decomposition is used to decompose the signal into a linear combination of a series of atomic functions with specific time-frequency characteristics; Gauss-Hermitian time-frequency atomic function is selected ,in is the normalization constant, yes Hermitian polynomial of order, Indicates the time center, represents the frequency center, Control the width of the time-frequency window;

[0037] For a given signal , the matching pursuit algorithm is used to find the best time-frequency atom to approximate the signal; the specific process is as follows:

[0038] Initialize residual ;

[0039] For each iteration Calculate residuals With all possible time-frequency atoms The inner product of ,turn up The largest time-frequency atom ;

[0040] Update residual ;

[0041] go through After iterations, the signal It can be approximately expressed as ;

[0042] Based on the time-frequency atomic decomposition results, the frequency center of gravity and bandwidth of the spectrum are recalculated; the traditional frequency center of gravity formula is , in the framework of time-frequency atomic decomposition, consider the contribution of each time-frequency atom to the frequency center of gravity; suppose , then the improved frequency center of gravity for ;

[0043] Introduce a bandwidth calculation method based on time-frequency atoms; first calculate the effective bandwidth of each time-frequency atom , can be determined by the frequency support range or the relevant energy distribution of the time-frequency atom; then, the bandwidth Calculated as .

[0044] The fault diagnosis method of aviation inverter based on multi-source information includes the following steps: empirical mode decomposition, Hilbert transform envelope, envelope characteristic parameter calculation, deep convolution neural network construction, model training and optimization.

[0045] As a preferred technical solution of the present invention, the process of empirical mode decomposition is: find the signal All local maximum and minimum points of are interpolated by cubic spline to obtain the upper envelope and lower envelope , calculate the mean of the upper and lower envelopes , subtract the mean value from the original signal to get ;like satisfy condition, then it is the first otherwise, Repeat the above process as a new sequence until a satisfying condition is obtained. ; Then, subtract from the original signal , and get the remaining sequence ,right Repeat the above decomposition steps to obtain and the residual ,Right now ;

[0046] The process of finding the envelope by Hilbert transform is: for each Performing Hilbert transform, we get ; Construct the analytical signal from this , its amplitude That is The instantaneous amplitude of , that is, the envelope;

[0047] The process of calculating the envelope characteristic parameters is: envelope mean ,in is the time length of the signal; the envelope mean reflects the average level of the envelope amplitude and can be used to determine the overall deviation of the signal amplitude;

[0048] Envelope Variance , the envelope variance measures the dispersion of the envelope amplitude around the mean, and can reflect the fluctuation of the envelope amplitude;

[0049] Envelope Crest Factor , the envelope crest factor is used to evaluate the relative size of the peak value and the mean value in the envelope, and is very sensitive to detecting pulse-type faults. When a pulse-type fault occurs, the crest factor will increase significantly;

[0050] The process of building a deep convolutional neural network is as follows: the network structure includes multiple convolutional layers, pooling layers and fully connected layers;

[0051] Convolutional layer: In the convolutional layer, convolution kernels of different sizes are used Perform convolution operation on the input feature matrix; for the input feature matrix , the convolutional layer outputs feature maps Elements The calculation formula is ,in is the convolution kernel element, is the bias, It is the input feature matrix element; the convolution layer can automatically extract the local feature pattern in the feature matrix. Combination relationship between envelope characteristic parameters;

[0052] Pooling layer: Use the maximum pooling method to reduce the dimension of the feature map output by the convolution layer; The index of the feature map is output within the pooling window; here, the pooling layer reduces the amount of data and the computational complexity without losing key features, while enhancing the model's translation invariance to the input;

[0053] Fully connected layer: Flatten the feature map processed by the pooling layer and connect it to the fully connected layer; the output of the fully connected layer is passed through the weight matrix and bias Perform linear transformation, and then pass the Softmax function to obtain the final fault diagnosis result; for the input vector , the output of the fully connected layer for ,in The function maps the output value to interval, and the sum of all output values ​​is 1, and each output value represents the probability of the corresponding fault category;

[0054] The process of model training and optimization is as follows: DCNN is trained using the envelope feature data of aircraft airborne converter signals with fault labels; during the training process, the cross entropy loss function is used to To update the network weight parameters, is the sample size, is the number of fault categories, It is a sample Belongs to category The real label, is the model prediction sample Belongs to category probability; at the same time, stochastic gradient descent is used to accelerate model convergence and improve training efficiency.

[0055] In summary, the present invention has the following beneficial effects:

[0056] First, sensors are used to collect time series data and spatial distribution data respectively to form time dimension data sets and space dimension data sets. The two-dimensional data sets are preprocessed separately to improve the accuracy and stability of the data.

[0057] Second, through time feature extraction, the characteristics of the time series of aircraft airborne converters can be extracted more deeply and accurately. Through spatial feature extraction, the spatial characteristics of aircraft airborne converters can be deeply extracted to improve the accuracy and reliability of fault location. Through spectrum acquisition and feature extraction, the frequency components and energy distribution of the signal can be determined. Based on the above-mentioned time-space-spectrum feature extraction, the fusion of multi-source information is realized, which can integrate information from different types of sensors and different feature domains, and fully explore the multi-dimensional information of the equipment operation status, thereby improving the accuracy and reliability of fault detection.

[0058] Third, first perform envelope detection on the signal to obtain the envelope line of the signal amplitude change, and analyze the characteristic parameters of the envelope to detect fault information or pulse-type fault characteristics in the modulated signal; then use a deep convolutional neural network to learn the complex relationship between envelope characteristics and faults to realize the diagnosis of aircraft airborne inverter faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 A flow chart of extracting fault features of an aviation inverter based on multi-source information provided by an embodiment of the present invention.

[0060] Figure 2 A flowchart of an aviation inverter fault diagnosis method based on multi-source information provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0061] The present application is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present application, but do not limit the present application in any form. It should be noted that, for those of ordinary skill in the art, several variations and improvements can also be made without departing from the concept of the present application. These all belong to the protection scope of the present application.

[0062] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0063] It should be noted that, if there is no conflict, the various features in the embodiments of the present application can be combined with each other, all within the scope of protection of the present application. In addition, although the functional module division is performed in the device schematic diagram and the logical order is shown in the flow chart, in some cases, the steps shown or described can be performed in a sequence different from the module division in the device or the flow chart. In addition, the words "first", "second", "third", etc. used herein do not limit the data and execution order, but only distinguish the same items or similar items with basically the same functions and effects.

[0064] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used in this specification and in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application. The term "and / or" used in this specification includes any and all combinations of one or more of the related listed items.

[0065] In addition, the technical features involved in each embodiment of the present application described below can be combined with each other as long as they do not conflict with each other.

[0066] The disclosed embodiments aim to solve the problem that the traditional aviation airborne inverter fault detection method relies on a single signal analysis method and cannot fully and accurately extract fault feature quantities. In view of this, the disclosed embodiments propose an aviation inverter fault feature extraction and diagnosis method based on multi-source information. Through multi-source information fusion, it is possible to integrate information from different types of sensors and different feature domains, fully explore the multi-dimensional information of the equipment operating status, and thus improve the accuracy and reliability of fault detection.

[0067] Please refer to Figure 1 , Figure 1 The flowchart of the method for extracting and diagnosing fault features of an aviation inverter based on multi-source information according to an embodiment of the present disclosure is shown. The overall process mainly includes the following five steps:

[0068] Step 1: Time series data collection and preprocessing.

[0069] Specifically, by installing sensors at key locations of aircraft-borne converters, data sequences of electrical parameters and environmental parameters of the converters during operation are continuously collected to form a data set in the time dimension.

[0070] The collected time series data is denoised and filtering algorithms are used to remove high-frequency noise and measurement errors to improve the accuracy and stability of the data.

[0071] Embodiment 1: The steps of time series data collection and preprocessing include: S1.1 wavelet decomposition, S1.2 noise standard deviation estimation, S1.3 threshold determination, S1.4 threshold processing, S1.5 wavelet reconstruction.

[0072] S1.1. Collected discrete time series data , where N is the data length, discrete wavelet transform is performed, and Coiflet wavelet Coif3 with good symmetry and regularity is used for processing.

[0073] No. Approximate coefficients of layer decomposition The calculation formula is: ,in, , are the low-pass filter coefficients corresponding to the selected wavelet basis, This step is to filter the signal, extract the low-frequency components of the signal, and reflect the general trend of the signal. In the current data of the aviation inverter, the low-frequency components usually represent the basic current level of the inverter during normal operation.

[0074] No. Layer decomposition detail factor The calculation formula is: ,in , are the high-pass filter coefficients corresponding to the selected wavelet basis, This step is to extract the high-frequency components of the signal. High-frequency noise is concentrated in these detail coefficients. For example, high-frequency noise such as electromagnetic interference to aviation inverters is reflected in the detail coefficients.

[0075] S1.2. Calculation of detail coefficients The noise standard deviation estimate of . Using the median-based estimation method: In aviation environments, noise is usually non-Gaussian and bursty. This median-based estimation method is more robust than the traditional mean-based estimation method and can estimate the noise intensity more accurately.

[0076] S1.3. Considering the dynamic changes of the operating state of the aircraft-borne converter, combined with the converter temperature change rate The threshold is adjusted based on factors such as the following:

[0077] Initial Threshold The calculation formula is: .

[0078] Dynamic Adjustment Factor ,in, is the weight coefficient, is the temperature change, is the maximum allowable temperature change.

[0079] Final Threshold for Through the above steps, the threshold It can adaptively adjust according to the actual operating status of the converter and remove noise more effectively when the noise intensity changes or the converter operating conditions change.

[0080] S1.4, adopt an improved soft and hard threshold combination method. For mixed parameters , the processed detail coefficient for:

[0081] when hour, .

[0082] Here the soft threshold part Can smooth noise, hard threshold part Can preserve the sharp features of the signal.

[0083] when hour, .

[0084] S1.5. Using the processed detail coefficient and approximate coefficients Perform inverse wavelet transform to reconstruct the denoised signal The formula is: ,in is the maximum number of decomposition layers. Through inverse wavelet transform, the denoised high-frequency and low-frequency components are recombined to obtain the denoised signal, which improves the accuracy and stability of the data and provides a reliable data basis for subsequent fault detection.

[0085] Step 2: Spatial distribution data collection and preprocessing.

[0086] Specifically, multiple sensors are arranged at different physical locations of the converter to obtain information from different spatial regions inside the converter. Temperature sensors are set at power modules, heat sinks, control circuit boards, etc. to monitor the spatial distribution of temperature; voltage and current sensors are set at input and output terminals and key circuit nodes to obtain spatial distribution information of electrical parameters. These spatially distributed sensor data are organized and numbered according to the physical structure and circuit topology of the converter, and a corresponding relationship between data and spatial position is established to form a data set of spatial dimension.

[0087] Embodiment 2: The steps of spatial distribution data collection and preprocessing include: S2.1 information entropy calculation, S2.2 anomaly determination, and S2.3 spatial interpolation.

[0088] S2.1. For the data collected by each sensor, consider the information entropy of its spatial neighborhood. Assume that in the converter space, with a certain sensor as the center, its neighborhood is defined as the set of sensors that include the sensor and its immediate surroundings. Suppose that there are Sensor, for the The data collected by a sensor at a certain moment .

[0089] First, calculate the mean of the data in the neighborhood Then, calculate the information entropy , the formula is ,in Information entropy reflects the uncertainty or confusion of the data in the neighborhood. When the data in the neighborhood differ greatly, the information entropy is high; when the data is relatively consistent, the information entropy is low. In a certain area of ​​the heat sink, if the temperature sensor measurement values ​​differ greatly, it means that the heat dissipation in this area may be abnormal, and the information entropy will be relatively high.

[0090] S2.2. Set a normal information entropy threshold range for each sensor neighborhood This threshold range is obtained by statistically analyzing a large amount of spatial neighborhood information entropy data when the converter is in normal operation. Collect all sensor neighborhood information entropy data when the converter is in normal operation for a period of time and calculate its mean value. and standard deviation ,set up , ,in and is an empirical coefficient and can be adjusted according to actual conditions, such as .

[0091] When the information entropy of a sensor neighborhood When this threshold is exceeded, or , it is determined that there may be an abnormality in the spatial area where the neighborhood is located. If the information entropy of the temperature sensor neighborhood near a power module suddenly increases and exceeds the normal range, it may mean that the power module has local overheating or uneven heat dissipation and other abnormal conditions.

[0092] S2.3. Since the distribution of sensors in the converter space is discrete, in order to more comprehensively understand the changes of physical quantities in space, the Kriging interpolation method is used.

[0093] For any point in the converter space , its physical quantity Estimated value of Calculated by the following formula: ,in is the known sensor position The measured value of the physical quantity at is the number of sensors involved in interpolation, is the weight coefficient,

[0094] By solving the following system of equations: ,here is the semivariogram function, which is used to describe two points in space and Variability of physical quantities between, Gaussian semivariogram function ,in is the distance between two points, It's the nugget effect. is the sill value, These parameters are determined by analyzing the variation function of the sensor data. Through spatial interpolation, the continuous distribution of physical quantities in the converter space is obtained, which can more accurately capture the changing trend of physical quantities.

[0095] Step 3: Temporal feature extraction.

[0096] The statistical characteristics of time series data can reflect the changing trend, fluctuation degree and stability of the signal over time. Extracting the trend characteristics of the time series can determine the long-term trend of the signal and determine whether there is an abnormal upward or downward trend, which helps to discover the slow changes in the converter performance.

[0097] Embodiment 3, the specific algorithm is as follows:

[0098] Empirical mode decomposition (EMD) is an adaptive decomposition method for nonlinear and non-stationary time series. Decomposed into several intrinsic mode functions (IMFs) and a residual term Each IMF must satisfy two conditions: first, in the entire data segment, the number of extreme points (maxima and minima) must be equal to or differ by at most one in the number of zero-crossing points; second, at any time, the mean of the upper and lower envelopes formed by the local maximum points and the local minimum points is zero.

[0099] First, find the time series All local maximum and minimum points of are interpolated by cubic spline to obtain the upper envelope and lower envelope , calculate the mean of the upper and lower envelopes , subtract the mean from the original sequence to get .like If it satisfies the IMF condition, it is the first IMF component; otherwise, Repeat the above process as a new sequence until a satisfying condition is obtained. . Then, subtract from the original sequence , and get the remaining sequence ,right Repeat the above decomposition steps to obtain and the residual ,Right now .

[0100] In the time series data of aircraft onboard converters, EMD can separate fluctuations of different time scales. The high-frequency IMF components may reflect short-term interference or noise during converter operation, while the low-frequency IMF components and residual terms contain more trend information related to progressive faults.

[0101] By analyzing the autocorrelation function and cross-correlation function of the time series, the correlation of the signal under different time delays is determined, and the relationship between periodic fault characteristics and signals is detected. The specific algorithm is as follows:

[0102] In order to more accurately reflect the intrinsic characteristics of the multivariate time series of aircraft airborne converters, a weight is assigned to each variable according to the importance of each converter parameter on the fault. . Coarse-grained sequence Calculated as .

[0103] For the coarse-grained sequence , calculate its sample entropy. Given the embedding dimension and tolerance , define the vector .make Represents two vectors and The Chebyshev distance of .definition To satisfy of The number and Ratio .but Similarly, calculate dimensional vector Sample entropy for Through this method, the complexity of the converter multivariable time series at different scales can be more accurately characterized, providing more effective features for fault diagnosis.

[0104] For time series , its autocorrelation function ,in Is a sequence The mean of .

[0105] However, the time series of aircraft-borne converters may have local deformation or phase differences, which affects the accuracy of autocorrelation analysis.

[0106] To this end, the autocorrelation improvement of dynamic time warping (DTW) is introduced, and the dynamic time warping (DTW) algorithm is used to solve this problem. , when calculating the autocorrelation, taking into account different delays subsequence under and The DTW algorithm works by finding a path ,in , so that and Minimize the distance between the paths. The constraints are , ,and The distance metric is Euclidean distance . Define DTW distance . Improved autocorrelation function This more accurately captures the similarity of time series under different delays, and can effectively analyze even if there are local deformations, which helps to discover periodic fault characteristics in the converter time series.

[0107] For two time series and , the traditional cross-correlation function . The DTW algorithm is also introduced for improvement. For different delays subsequence under and , find the optimal path through the DTW algorithm , calculate the DTW distance . Improved cross-correlation function This improved cross-correlation analysis can more accurately determine the relationship between different signals and provide stronger support for fault diagnosis of aircraft airborne converters based on multivariate time series analysis.

[0108] Through the above improved multivariate multiscale sample entropy combined with the autocorrelation and cross-correlation analysis method of DTW, the characteristics of the time series of aviation airborne converters can be extracted more deeply and accurately.

[0109] Step 4: Spatial feature extraction.

[0110] A deep convolutional neural network (DCNN) is constructed to extract the fused spatial features.

[0111] Embodiment 4: The network structure includes: multiple convolutional layers, pooling layers and fully connected layers.

[0112] Convolutional layer: using convolution kernels of different sizes Perform convolution operation on the input feature map. , the convolutional layer outputs feature maps Elements Calculated as ,in is the convolution kernel element, is the bias, It is the input feature map element. The convolution operation extracts local spatial features, such as local temperature change patterns, thermal stress concentration areas, etc.

[0113] Pooling layer: using maximum pooling, for the input feature map , in the pooling window Inside, output feature map Elements , reducing the data dimension while retaining key features.

[0114] Fully connected layer: Flatten the pooled feature map and pass it through the weight matrix and bias Perform linear transformation and then pass the activation function ReLU to get the final output. The output layer uses the Softmax function to map the result to the probability distribution of the fault category. , is the output of the fully connected layer, is the number of fault categories. The network is trained by a large amount of spatial feature data with fault labels, so that the network can learn the complex relationship between spatial features and faults and achieve accurate fault location.

[0115] Through this method based on multi-physics field coupling and deep learning, the spatial characteristics of aviation airborne inverters can be deeply extracted, improving the accuracy and reliability of fault location.

[0116] Step 5: Spectrum acquisition and feature extraction.

[0117] Perform short-time Fourier transform (STFT) on the collected time series data (especially current and voltage signals), convert the time domain signal into a frequency domain signal, obtain the signal's spectrum data and characteristics, calculate the amplitude and phase of each harmonic, and determine the signal's frequency component and energy distribution.

[0118] Embodiment 5, the specific algorithm is as follows:

[0119] Variational mode decomposition (VMD) is an adaptive signal decomposition method that aims to decompose complex signals into a series of intrinsic mode functions (IMFs) with different center frequencies. For a given input signal , the goal of VMD is to determine a series of mode functions by solving the variational problem and center frequency .

[0120] Define each modal function Hilbert transform , and Composition of analytical signals , through the index term Demodulate to baseband and get . Find the first-order derivative of The square of the norm, that is , which is used as the bandwidth measure of each mode function. The constrained variational problem is constructed as , the constraints are ,in is the preset number of decomposition modes.

[0121] By introducing the quadratic penalty factor and Lagrange multipliers , transforming the constrained variational problem into an unconstrained augmented Lagrangian function . The alternating direction method of multipliers (ADMM) is used to iteratively solve the augmented Lagrangian function and update and , until the convergence condition is met.

[0122] After VMD decomposition, the original signal is decomposed into IMF components , each component has different frequency characteristics. VMD adaptively separates different frequency components, avoiding problems such as spectrum aliasing, and providing purer signal components for subsequent spectrum feature extraction.

[0123] The purpose of using time-frequency atomic decomposition is to decompose the signal into a linear combination of a series of atomic functions with specific time-frequency characteristics. ,in is the normalization constant, yes Hermitian polynomial of order, Indicates the time center, represents the frequency center, Controls the width of the time-frequency window.

[0124] For a given signal , using the matching pursuit algorithm to find the best time-frequency atom to approximate the signal. The specific process is as follows:

[0125] Initialize residual .

[0126] For each iteration Calculate residuals With all possible time-frequency atoms The inner product of ,turn up The largest time-frequency atom .

[0127] Update residual .

[0128] go through After iterations, the signal It can be approximately expressed as .

[0129] Based on the time-frequency atomic decomposition results, the frequency center of gravity and bandwidth of the spectrum are recalculated. The traditional frequency center of gravity formula is , in the framework of time-frequency atomic decomposition, the contribution of each time-frequency atom to the frequency center of gravity is considered. , then the improved frequency center of gravity for This calculation method can better reflect the frequency contribution of the signal in different time-frequency regions. For converter signals, it can more accurately capture the frequency center of gravity shift caused by component failure.

[0130] The bandwidth calculation method based on time-frequency atoms is introduced. First, the effective bandwidth of each time-frequency atom is calculated. , can be determined by the frequency support range or the related energy distribution of the time-frequency atom. Then, the bandwidth Calculated as This formula comprehensively considers the frequency position, amplitude and relationship with the frequency center of gravity of each time-frequency atom, and can more accurately measure the discrete degree of signal frequency distribution, and is more sensitive to detecting bandwidth changes caused by converter faults.

[0131] Please refer to Figure 2 , Figure 2 A flow chart of an aviation inverter fault diagnosis method based on multi-source information according to an embodiment of the present disclosure is shown.

[0132] By performing envelope detection on the signal, the envelope of the signal amplitude change is obtained, and the characteristic parameters of the envelope are analyzed to detect fault information or pulse-type fault characteristics in the modulated signal.

[0133] Embodiment 6: The overall process mainly includes the following five steps: empirical mode decomposition (EMD), Hilbert transform to obtain envelope, envelope feature parameter calculation, construction of deep convolutional neural network (DCNN), model training and optimization.

[0134] S1. Use Empirical Mode Decomposition (EMD) to transform complex non-stationary signals Decomposition into multiple intrinsic mode functions (IMFs) and a residual term Each Two conditions must be met: first, in the entire data segment, the number of extreme points (maxima and minima) and the number of zero-crossing points must be equal or differ by at most one; second, at any time, the mean of the upper and lower envelopes formed by the local maximum points and local minimum points is zero. The specific decomposition process is as follows:

[0135] Find the signal All local maximum and minimum points of are interpolated by cubic spline to obtain the upper envelope and lower envelope , calculate the mean of the upper and lower envelopes , subtract the mean value from the original signal to get .like satisfy condition, then it is the first otherwise, Repeat the above process as a new sequence until a satisfying condition is obtained. . Then, subtract from the original signal , and get the remaining sequence ,right Repeat the above decomposition steps to obtain and the residual ,Right now .

[0136] S2. For each Performing Hilbert transform, we get . This constructs the analytical signal , its amplitude That is In this way, the envelope characteristics of the signal at different time scales can be accurately extracted. These envelope characteristics reflect the change of the signal amplitude and are of great significance for detecting fault information or pulse fault characteristics in the modulated signal. If there is a pulse fault in the current signal of the aircraft airborne converter, its corresponding The envelope will show specific mutations or abnormal fluctuations.

[0137] S3. In order to more comprehensively describe the envelope characteristics, the following characteristic parameters are calculated:

[0138] Envelope mean ,in is the time length of the signal. The envelope mean reflects the average level of the envelope amplitude and can be used to determine the overall deviation of the signal amplitude.

[0139] Envelope Variance The envelope variance measures the degree of dispersion of the envelope amplitude around the mean, and can reflect the fluctuation of the envelope amplitude.

[0140] Envelope Crest Factor The envelope crest factor is used to evaluate the relative size of the peak value to the mean value in the envelope. It is very sensitive to detecting pulse-type faults. When a pulse-type fault occurs, the crest factor will increase significantly.

[0141] S4. The extracted envelope feature parameters (envelope mean, envelope variance, envelope peak factor, etc.) are combined into a feature matrix as the input of the deep convolutional neural network. The network structure includes multiple convolutional layers, pooling layers and fully connected layers.

[0142] Convolutional layer: In the convolutional layer, convolution kernels of different sizes are used Perform convolution operation on the input feature matrix. , the convolutional layer outputs feature maps Elements The calculation formula is ,in is the convolution kernel element, is the bias, is the input feature matrix element. The convolution layer can automatically extract the local feature patterns in the feature matrix. The combined relationship between envelope feature parameters.

[0143] Pooling layer: Use the maximum pooling method to reduce the dimension of the feature map output by the convolution layer. The index of the feature map is output within the pooling window. Here, the pooling layer reduces the amount of data and the computational complexity without losing key features, while enhancing the model's translation invariance to the input.

[0144] Fully connected layer: Flatten the feature map processed by the pooling layer and connect it to the fully connected layer. The output of the fully connected layer is passed through the weight matrix and bias Perform linear transformation, and then pass the Softmax function to obtain the final fault diagnosis result. , the output of the fully connected layer for ,in The function maps the output value to The sum of all output values ​​is 1, and each output value represents the probability of the corresponding fault category.

[0145] S5. Use the envelope feature data of the aircraft airborne converter signal with fault labels to train the DCNN. During the training process, the cross entropy loss function To update the network weight parameters, is the sample size, is the number of fault categories, It is a sample Belongs to category The real label, is the model prediction sample Belongs to category At the same time, stochastic gradient descent (SGD) is used to accelerate model convergence and improve training efficiency.

[0146] In this way, the deep convolutional neural network can learn the complex relationship between envelope features and faults, and realize the diagnosis of aviation onboard inverter faults.

[0147] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A method for extracting fault features of aviation inverters based on multi-source information, characterized in that: The method comprises the following steps: S1, collect and preprocess time series data and spatial distribution data respectively; S2. Perform time feature extraction, spatial feature extraction, spectrum acquisition and feature extraction on the preprocessed data to achieve the fusion of multi-source information.

2. The method for extracting fault features of an aviation inverter based on multi-source information according to claim 1, characterized in that: The time series data collection process is as follows: through sensors installed at key locations of aircraft-borne converters, the data series of electrical parameters and environmental parameters of the converters during operation are continuously collected to form a data set in the time dimension; The spatial distribution data acquisition process is as follows: multiple sensors are arranged at different physical locations of the converter to obtain information data of different spatial regions inside the converter; These spatially distributed sensor data are organized and numbered according to the physical structure and circuit topology of the converter, and a corresponding relationship between the data and the spatial position is established to form a data set of spatial dimension.

3. The method for extracting fault features of an aviation inverter based on multi-source information according to claim 1, characterized in that: The steps of time series data preprocessing include: wavelet decomposition, noise standard deviation estimation, threshold determination, threshold processing, and wavelet reconstruction; The steps of spatial distribution data preprocessing include: information entropy calculation, anomaly determination, and spatial interpolation.

4. The method for extracting fault features of an aviation inverter based on multi-source information according to claim 3, characterized in that: The process of wavelet decomposition is: the collected discrete time series data ) for discrete wavelet transform, using Coiflet wavelet Coif3 with good symmetry and regularity for processing: Approximate coefficients of layer decomposition The calculation formula is: ,in, , are the low-pass filter coefficients corresponding to the selected wavelet basis, for its conjugation; Layer decomposition detail factor The calculation formula is: ,in , are the high-pass filter coefficients corresponding to the selected wavelet basis, for its conjugation; The process of estimating the noise standard deviation is as follows: Calculate the detail coefficient The noise standard deviation estimate of ; Using the median-based estimation method: ; The process of determining the threshold is as follows: initial threshold The calculation formula is: ; Dynamic Adjustment Factor ,in is the weight coefficient, is the temperature change, is the maximum allowable temperature change; the final threshold for ; The process of threshold processing is as follows: For mixed parameters , the processed detail coefficient For: When hour, ; Among them, the soft threshold part Can smooth out noise, hard threshold part Can preserve the sharp features of the signal; The process of wavelet reconstruction is: using the processed detail coefficients and approximate coefficients Perform inverse wavelet transform to reconstruct the denoised signal The formula is: ,in is the maximum number of decomposition levels.

5. The method for extracting fault features of an aviation inverter based on multi-source information according to claim 3, characterized in that: The process of calculating information entropy is as follows: The data collected by a sensor at a certain moment ; First, calculate the mean of the data in the neighborhood ; Then, calculate the information entropy , the formula is ,in ; The process of abnormal judgment is: set a normal information entropy threshold range for each sensor neighborhood ; Collect all sensor neighborhood information entropy data when the converter operates normally for a period of time and calculate its mean value and standard deviation ,set up , ,in and is an empirical coefficient and can be adjusted according to actual conditions, such as ; When the information entropy of a sensor neighborhood When this threshold is exceeded, or , then it is determined that there may be anomalies in the spatial area where the neighborhood is located; The process of spatial interpolation is as follows: for any point in the converter space , its physical quantity Estimated value of Calculated by the following formula: ,in is the known sensor position The measured value of the physical quantity at is the number of sensors involved in interpolation, is the weight coefficient, obtained by solving the following system of equations: ,here is the semivariogram function, which is used to describe two points in space and Variability of physical quantities between, Gaussian semivariogram function ,in is the distance between two points, It's the nugget effect. is the sill value, is the range of variation, and these parameters are determined by analyzing the variogram of the sensor data.

6. The method for extracting fault features of aviation inverters based on multi-source information according to claim 1, characterized in that: The process of time feature extraction is: find the time series All local maximum and minimum points of are interpolated by cubic spline to obtain the upper envelope and lower envelope , calculate the mean of the upper and lower envelopes , subtract the mean from the original sequence to get ;like If it satisfies the IMF condition, it is the first IMF component; otherwise, Repeat the above process as a new sequence until a satisfying condition is obtained. ; Then, subtract from the original sequence , and get the remaining sequence ,right Repeat the above decomposition steps to obtain and the residual ,Right now ; In order to more accurately reflect the intrinsic characteristics of the multivariate time series of aircraft airborne converters, a weight is assigned to each variable according to the importance of each converter parameter on the fault. ; Coarse-grained sequence Calculated as ; For the coarse-grained sequence , calculate its sample entropy; given the embedding dimension and tolerance , define the vector ;make Represents two vectors and The Chebyshev distance of ;definition To satisfy of The number and Ratio ;but ; Similarly, calculate dimensional vector ; Sample entropy for ; For time series , its autocorrelation function ,in Is a sequence The mean of For time series , when calculating the autocorrelation, taking into account different delays subsequence under and ; The DTW algorithm works by finding a path ,in , so that and Minimize the distance between paths The constraints are , ,and ; The distance metric is Euclidean distance ; Define DTW distance ; Improved autocorrelation function ; For two time series and , the traditional cross-correlation function ; The DTW algorithm is also introduced for improvement; for different delays subsequence under and , find the optimal path through the DTW algorithm , calculate the DTW distance ; Improved cross-correlation function .

7. The method for extracting fault features of an aviation inverter based on multi-source information according to claim 1, characterized in that: The process of spatial feature extraction is: constructing a deep convolutional neural network to extract the fused spatial features; the network structure includes multiple convolutional layers, pooling layers and fully connected layers; The convolution layer: using convolution kernels of different sizes Perform convolution operation on the input feature map; for the input feature map , the convolutional layer outputs the feature map Elements Calculated as ,in is the convolution kernel element, is the bias, is the input feature map element; The pooling layer: uses maximum pooling, for the input feature map , in the pooling window Inside, output feature map Elements , reducing data dimensionality while retaining key features; The fully connected layer: flattens the pooled feature map and passes the weight matrix and bias Perform linear transformation, and then use the activation function ReLU to get the final output; the output layer uses the Softmax function to map the result to the probability distribution of the fault category , is the output of the fully connected layer, is the number of fault categories; the network is trained by a large amount of spatial feature data with fault labels, so that the network can learn the complex relationship between spatial features and faults and achieve accurate fault location; The method based on multi-physics field coupling and deep learning can deeply extract the spatial characteristics of aviation airborne converters and improve the accuracy and reliability of fault location.

8. The method for extracting fault features of an aviation inverter based on multi-source information according to claim 1, characterized in that: The process of spectrum acquisition is: define each mode function Hilbert transform , and Composition of analytical signals , through the index term Demodulate to baseband and get ; Find the first-order derivative of The square of the norm, that is , which is used as the bandwidth measure of each modal function; the constrained variation problem is constructed as , the constraints are ,in is the preset number of decomposition modes; By introducing a quadratic penalty factor and Lagrange multipliers , transforming the constrained variational problem into an unconstrained augmented Lagrangian function ； The augmented Lagrangian function is solved iteratively using the alternating direction multiplier method, and the update and , until the convergence condition is met; After VMD decomposition, the original signal is decomposed into IMF Components , each component has different frequency characteristics; VMD adaptively separates different frequency components, avoiding problems such as spectrum aliasing, and providing purer signal components for subsequent spectrum feature extraction; The process of spectrum feature extraction is as follows: time-frequency atomic decomposition is used to decompose the signal into a linear combination of a series of atomic functions with specific time-frequency characteristics; Gauss-Hermitian time-frequency atomic function is selected ,in is the normalization constant, yes Hermitian polynomial of order, Indicates the time center, represents the frequency center, Control the width of the time-frequency window; For a given signal , the matching pursuit algorithm is used to find the best time-frequency atom to approximate the signal; the specific process is as follows: Initialize residual ; For each iteration Calculate residuals With all possible time-frequency atoms The inner product of ,turn up The largest time-frequency atom ; Update residual ; go through After iterations, the signal It can be approximately expressed as ; Based on the time-frequency atomic decomposition results, recalculate the frequency center of gravity and bandwidth and other characteristics of the spectrum; The traditional frequency center of gravity formula is , in the framework of time-frequency atomic decomposition, consider the contribution of each time-frequency atom to the frequency center of gravity; suppose , then the improved frequency center of gravity for ; Introduce a bandwidth calculation method based on time-frequency atoms; first calculate the effective bandwidth of each time-frequency atom , can be determined by the frequency support range or the relevant energy distribution of the time-frequency atom; then, the bandwidth Calculated as .

9. The method for diagnosing faults of aviation inverters based on multi-source information is characterized in that: The method uses the aviation inverter fault feature extraction method based on multi-source information described in claims 1-8 for diagnosis, and includes the following steps: empirical mode decomposition, Hilbert transform envelope, envelope feature parameter calculation, construction of deep convolutional neural network, model training and optimization.

10. The aviation inverter fault diagnosis method based on multi-source information according to claim 9, characterized in that: The process of empirical mode decomposition is to find the signal All local maximum and minimum points of are interpolated by cubic spline to obtain the upper envelope and lower envelope , calculate the mean of the upper and lower envelopes , subtract the mean value from the original signal to get ;like satisfy condition, then it is the first otherwise, Repeat the above process as a new sequence until a satisfying condition is obtained. ; Then, subtract from the original signal , and get the remaining sequence ,right Repeat the above decomposition steps to obtain and the residual ,Right now ; The process of finding the envelope by Hilbert transform is: for each Performing Hilbert transform, we get ; Construct the analytical signal from this , its amplitude That is The instantaneous amplitude of , that is, the envelope; The process of calculating the envelope characteristic parameters is: envelope mean ,in is the time length of the signal; the envelope mean reflects the average level of the envelope amplitude and can be used to determine the overall deviation of the signal amplitude; Envelope Variance , the envelope variance measures the dispersion of the envelope amplitude around the mean, and can reflect the fluctuation of the envelope amplitude; Envelope Crest Factor , the envelope crest factor is used to evaluate the relative size of the peak value and the mean value in the envelope, and is very sensitive to detecting pulse-type faults. When a pulse-type fault occurs, the crest factor will increase significantly; The process of building a deep convolutional neural network is as follows: the network structure includes multiple convolutional layers, pooling layers and fully connected layers; Convolutional layer: In the convolutional layer, convolution kernels of different sizes are used Perform convolution operation on the input feature matrix; for the input feature matrix , the convolutional layer outputs the feature map Elements The calculation formula is ,in is the convolution kernel element, is the bias, It is the input feature matrix element; the convolution layer can automatically extract the local feature pattern in the feature matrix. Combination relationship between envelope characteristic parameters; Pooling layer: Use the maximum pooling method to reduce the dimension of the feature map output by the convolution layer; The index of the feature map is output within the pooling window; here, the pooling layer reduces the amount of data and the computational complexity without losing key features, while enhancing the model's translation invariance to the input; Fully connected layer: Flatten the feature map processed by the pooling layer and connect it to the fully connected layer; the output of the fully connected layer is passed through the weight matrix and bias Perform linear transformation, and then pass the Softmax function to obtain the final fault diagnosis result; for the input vector , the output of the fully connected layer for ,in The function maps the output value to interval, and the sum of all output values ​​is 1, and each output value represents the probability of the corresponding fault category; The process of model training and optimization is as follows: DCNN is trained using the envelope feature data of aircraft airborne converter signals with fault labels; during the training process, the cross entropy loss function is used to To update the network weight parameters, is the sample size, is the number of fault categories, It is a sample Belongs to category The real label, is the model prediction sample Belongs to category probability; at the same time, stochastic gradient descent is used to accelerate model convergence and improve training efficiency.

Images