Airborne radio frequency system filter circuit fault diagnosis method based on confidence fusion

Abstract

The invention discloses an airborne radio frequency system filter circuit fault diagnosis method based on confidence fusion, and belongs to the technical field of analog circuit fault diagnosis. The method comprises the following steps: firstly, acquiring an output end voltage data set in a circuit health state and each fault state, converting time domain data into frequency domain data by using FFT (Fast Fourier Transform), then carrying out logarithmic compression, and normalizing the time domain data; constructing a 1D-CNN model, and performing fault diagnosis by using the time domain data to obtain a time domain probability vector; constructing a Transform model, and performing fault diagnosis by using the frequency domain data to obtain a frequency domain probability vector; a fusion probability vector is obtained through weighted fusion, and a fault category corresponding to the maximum probability in the fusion probability vector is selected as a diagnosis result; training by using the training set, and verifying validity by using the test set. According to the method, time-frequency double-domain feature extraction is adopted to capture multi-dimensional representation of a circuit state; in combination with 1D-CNN local feature extraction and Transform global association, computing resource consumption is reduced, and the accuracy and robustness of a diagnosis result are greatly improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of analog circuit fault diagnosis, and particularly relates to a filter circuit fault diagnosis method based on confidence fusion for an airborne radio frequency system. BACKGROUND

[0002] The airborne radio frequency system is an important system for realizing communication, radar, navigation and other functions of an airplane, and is a key system for realizing information exchange between airplanes and between an airplane and a ground station. The filter circuit plays an important role in the airborne radio frequency system. The filter circuit plays a role of frequency band selection in the wireless communication process, can eliminate noise interference, filter out harmonics and suppress clutter, and is crucial for realizing the functions of the airplane. However, due to the complexity of the airplane operating environment and the natural aging of components, the circuit is prone to failure. In order to ensure flight safety and reduce the occurrence of dangerous conditions, it is very important to study the filter circuit fault diagnosis method for the airborne radio frequency system.

[0003] Deep learning is widely used in the field of fault diagnosis due to its efficient feature extraction capability and powerful classification capability, but it requires a large number of training samples and high computing resources, while the actual circuit has fewer fault samples. At the same time, a single deep learning model often cannot balance local features and global features, which easily leads to misdiagnosis of the diagnosis result. Therefore, a fault diagnosis method that balances local features and global features is needed to ensure the normal operation of the filter circuit of the airborne radio frequency system. SUMMARY

[0004] In order to overcome the above problems, the application provides a filter circuit fault diagnosis method based on confidence fusion for an airborne radio frequency system, which belongs to the technical field of analog circuit fault diagnosis. First, the output voltage data set of the circuit in the healthy state and each fault state is collected, the time domain data is converted into frequency domain data by using FFT, and then logarithmic compression is performed, and the time domain data is normalized; a 1D-CNN model is constructed, and the time domain data is used for fault diagnosis to obtain a time domain probability vector; a Transformer model is constructed, and the frequency domain data is used for fault diagnosis to obtain a frequency domain probability vector; a fusion probability vector is obtained by weighted fusion, and the fault category corresponding to the maximum probability in the fusion probability vector is selected as the diagnosis result; the training set is used for training, and the test set is used for verifying the effectiveness. The application uses time-frequency dual-domain feature extraction to capture the multi-dimensional representation of the circuit state; combines 1D-CNN local feature extraction with Transformer global correlation, reduces the consumption of computing resources, and greatly improves the accuracy and robustness of the diagnosis result.

[0005] A filter circuit fault diagnosis method based on confidence fusion for an airborne radio frequency system, as shown in Figure 1 , comprises the following steps:

[0006] Step 1, collect the output voltage data set under the health state and each fault state of the circuit;

[0007] Step 2, data processing is performed on the collected voltage data set;

[0008] The data processing process is as follows:

[0009] Step 2.1, normalize the time domain data in the voltage data set to obtain a voltage time domain data set;

[0010] Step 2.2, pre-process the time domain data in the voltage data set in the frequency domain to obtain a voltage frequency domain data set;

[0011] Step 2.3, divide the voltage time domain data set and the voltage frequency domain data set into a training set and a test set according to a 7:3 ratio, and add labels of the circuit fault state to the training set samples;

[0012] Step 3, a 1D-CNN model (one-dimensional convolutional neural network model) is constructed, and the normalized voltage time domain data is used to perform fault diagnosis through the 1D-CNN model to obtain a time domain probability vector;

[0013] Step 4, a Transformer model is constructed, and the frequency domain data is used to perform fault diagnosis through the Transformer model to obtain a frequency domain probability vector;

[0014] Step 5, a weighted fusion method is used to perform decision-level fusion on the time domain probability vector and the frequency domain probability vector to obtain a fusion probability vector; the fault category corresponding to the maximum probability in the fusion probability vector is selected as the final diagnosis result;

[0015] Step 6, the 1D-CNN model, the Transformer model, and the decision-level fusion constitute a double-branch confidence fusion fault diagnosis model; the double-branch confidence fusion fault diagnosis model is trained using the training set samples in the voltage data set;

[0016] Step 7, the effectiveness of the double-branch confidence fusion fault diagnosis model is verified using the test set samples.

[0017] Further, in step 1, the fault states include resistance value drift, capacitor failure, and operational amplifier saturation, each of which is a single component failure; resistance value drift refers to a single resistance component value rising or falling by more than 10%, capacitor failure refers to a single capacitor component short circuit, and operational amplifier saturation refers to a phenomenon that the output of the operational amplifier cannot continue to follow the input change after the input signal amplitude is too large, the operational amplifier driving capacity is insufficient, or the operational amplifier supply voltage drops.

[0018] Further, in step 2, the normalization equation is as follows:

[0019] (1);

[0020] wherein, is the time domain data signal in the voltage data set; is the minimum value in the voltage data set; is the maximum value in the voltage data set; is the normalized voltage time domain data signal, the value range is between 0 and 1;

[0021] The frequency domain preprocessing process is as follows:

[0022] Step 2.2.1, using fast Fourier transform (FFT) to convert time domain data into frequency domain data;

[0023] The window added by the fast Fourier transform is the Hanning window, and the complex frequency domain data is obtained by performing fast Fourier transform on the windowed signal; the real part and the imaginary part of the frequency domain data are squared and summed to obtain the voltage frequency domain data set amplitude spectrum;

[0024] Step 2.2.2, logarithmic compression and standardization are performed on the voltage frequency domain data set amplitude spectrum;

[0025] First, logarithmic compression is performed to enhance weak signals and obtain compressed data L[k]:

[0026] (2);

[0027] wherein, M[k] represents the amplitude spectrum, ; , for preventing zero value;

[0028] Then, the compressed data L[k] is standardized to obtain the voltage frequency domain data F[k]:

[0029] (3);

[0030] wherein, is the mean value of the compressed data L[k]; is the standard deviation of the compressed data; the frequency domain data F[k] is used as the input of the Transformer model.

[0031] Further, in step 3, the 1D-CNN model includes an input layer and a classification output layer; for example, Figure 2As shown, the input layer sequentially includes a first convolutional layer, a first max-pooling layer, a second convolutional layer, a second max-pooling layer, a third convolutional layer, and a third max-pooling layer; the convolutional kernel size of the three convolutional layers is 5, and the pooling window size of each max-pooling layer is 2; the third max-pooling layer accesses a classification output layer; the classification output layer includes two fully connected layers, the first fully connected layer is used to receive fused feature information, the second fully connected layer uses a softmax function for fault classification, and a dropout layer with a dropout rate of 0.5 is added between the two fully connected layers to prevent model overfitting.

[0032] Further, in step 4, the Transformer model is a lightweight model, such as Figure 3 As shown, the Transformer model sequentially includes an input layer, an embedding layer, position encoding, an added CLS classification token, a Transformer encoder layer, and a linear classification output layer; the Transformer encoder layer is three layers of encoders, sequentially including a first layer of encoders, a second layer of encoders, and a third layer of encoders; the classification output layer is two layers of fully connected layers, sequentially including a first fully connected layer and a second fully connected layer; each layer of Transformer encoders has a four-headed self-attention mechanism, a 256-dimensional feedforward network, and a GELU activation function; the Transformer encoder layer is stabilized by residual connection and layer normalization for training, extracts the state vector of the added CLS token, and inputs the layer-normalized state vector to the linear classification output layer for fault classification.

[0033] Further, in step 5, the process of decision-level fusion is as follows:

[0034] Step 5.1, calculate the confidence of the time domain probability vector output by the 1D-CNN model

[0035] (4);

[0036] wherein, represents the time domain probability vector output by the 1D-CNN model; represents the information entropy of the time domain probability vector output by the 1D-CNN model, and the information entropy calculation formula is as follows:

[0037] (5);

[0038] wherein, n represents the total number of elements in the time domain probability vector; represents an element in the time domain probability vector,

[0039] Similarly, the confidence of the frequency domain probability vector is calculated using the method of step 5.1 ;​​

[0040] Step 5.2, respectively calculate the confidence of the time domain probability vector and the confidence of the frequency domain probability vector The fusion weight of the time domain probability vector and the weight of the frequency domain probability vector ;

[0041] The fusion weight is calculated as follows:

[0042] (6);

[0043] (7);

[0044] Step 5.3, weight fusion is performed on the weight of the time domain probability vector and the weight of the frequency domain probability vector to obtain the fusion probability vector.

[0045] The technical effects of the present application are as follows:

[0046] The present application adopts time-frequency dual-domain feature extraction to capture the multi-dimensional representation of circuit state, and uses confidence fusion to realize on-board radio frequency system filter circuit fault diagnosis, and the dual-branch structure realizes the complementation of the ability of 1D-CNN to extract local features and the ability of Transformer to globally associate, overcomes the problem that deep learning method cannot consider local features and global features, and the problem of large consumption of computing resources, greatly improves the accuracy and robustness of the diagnosis result. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 is the overall flowchart of the present application;

[0048] Figure 2 is a schematic diagram of the time domain branch 1D-CNN model in the present application;

[0049] Figure 3 is a schematic diagram of the frequency domain branch Transformer model in the present application. DETAILED DESCRIPTION

[0050] The present application will be further described in detail below in combination with the drawings and examples:

[0051] A method for diagnosing filter circuit faults of an on-board radio frequency system based on confidence fusion, as shown in Figure 1 , comprising the following steps:

[0052] Step 1, collect the output voltage data set under the healthy state and each fault state of the circuit;

[0053] The fault states include resistance value drift, capacitor failure and operational amplifier saturation, all of which are single component failures; the resistance value drift refers to a single resistance component value rising or falling by more than 10%, the capacitor failure refers to a single capacitor component short circuit, and the operational amplifier saturation refers to a phenomenon that the output of the operational amplifier cannot continue to follow the input change after exceeding a certain value due to an excessively large input signal amplitude, insufficient operational amplifier driving capacity or a decrease in operational amplifier supply voltage.

[0054] Step 2, data processing is performed on the collected voltage data set;

[0055] The data processing process is as follows:

[0056] Step 2.1, the time domain data in the voltage data set is normalized to obtain a voltage time domain data set;

[0057] The normalization equation is as follows:

[0058] (1);

[0059] wherein, is the time domain data signal in the voltage data set; is the minimum value in the voltage data set; is the maximum value in the voltage data set; is the normalized voltage time domain data signal, the value range is between 0 and 1;

[0060] Step 2.2, the time domain data in the voltage data set is preprocessed in the frequency domain to obtain a voltage frequency domain data set;

[0061] The frequency domain preprocessing process is as follows:

[0062] Step 2.2.1, the time domain data is converted into frequency domain data using fast Fourier transform (FFT);

[0063] The window added by the fast Fourier transform is the Hanning window, and the complex frequency domain data is obtained by performing fast Fourier transform on the windowed signal; the real part and the imaginary part of the frequency domain data are squared and summed to obtain the voltage frequency domain data set amplitude spectrum;

[0064] Step 2.2.2, logarithmic compression and standardization are performed on the voltage frequency domain data set amplitude spectrum;

[0065] First, logarithmic compression is performed to enhance weak signals and obtain compressed data L[k]:

[0066] (2);

[0067] wherein M[k] represents an amplitude spectrum, ; , for preventing zero values;

[0068] Then, the compressed data L[k] is normalized to obtain voltage frequency domain data F[k]:

[0069] (3);

[0070] wherein, is the mean of the compressed data L[k]; is the standard deviation of the compressed data; the frequency domain data F[k] is used as the input of the Transformer model.

[0071] Step 2.3, the voltage time domain data set and the voltage frequency domain data set are divided into training set and test set according to the ratio of 7:3, and the training set sample is labeled with the circuit fault state;

[0072] Step 3, a 1D-CNN model (one-dimensional convolutional neural network model) is constructed, and the normalized voltage time domain data is used to perform fault diagnosis through the 1D-CNN model to obtain a time domain probability vector;

[0073] The 1D-CNN model includes an input layer and a classification output layer; as shown in Figure 2 the input layer includes a first convolutional layer, a first max pooling layer, a second convolutional layer, a second max pooling layer, a third convolutional layer and a third max pooling layer in sequence; the convolution kernel size of the three convolutional layers is 5, and the pooling window size of each max pooling layer is 2; the third max pooling layer is connected to the classification output layer; the classification output layer includes two fully connected layers, the first fully connected layer is used to receive fused feature information, the second fully connected layer uses a softmax function for fault classification, and a dropout layer with a dropout rate of 0.5 is added between the two fully connected layers to prevent model overfitting.

[0074] Step 4, a Transformer model is constructed, and the frequency domain data is used to perform fault diagnosis through the Transformer model to obtain a frequency domain probability vector;

[0075] The Transformer model is a lightweight model, as shown in Figure 3As shown, the Transformer model sequentially includes an input layer, an embedding layer, position encoding, an added CLS classification token, a Transformer encoder layer, and a linear classification output layer; the Transformer encoder layer is a three-layer encoder, sequentially including a first-layer encoder, a second-layer encoder, and a third-layer encoder; the classification output layer is a two-layer fully connected layer, sequentially including a first fully connected layer and a second fully connected layer; each layer of the Transformer encoder has a four-head self-attention mechanism, a 256-dimensional feedforward network, and a GELU activation function; the Transformer encoder layer is stabilized in training through residual connection and layer normalization, extracts a state vector of the added CLS token, and inputs the state vector to the linear classification output layer after layer normalization for fault classification.

[0076] Step 5, using a weighted fusion method to perform decision-level fusion on the time domain probability vector and the frequency domain probability vector to obtain a fused probability vector; selecting a fault category corresponding to a maximum probability in the fused probability vector as a final diagnosis result;

[0077] The process of the decision-level fusion is as follows:

[0078] Step 5.1, calculating a confidence of the time domain probability vector output by the 1D-CNN model :

[0079] (4);

[0080] wherein, represents the time domain probability vector output by the 1D-CNN model; calculating an information entropy of the time domain probability vector output by the 1D-CNN model, and the information entropy calculation formula is as follows:

[0081] (5);

[0082] wherein, n represents a total number of elements in the time domain probability vector; represents an element in the time domain probability vector, ;

[0083] Similarly, the confidence of the frequency domain probability vector is calculated by the method of step 5.1 ;

[0084] Step 5.2, respectively calculating a fusion weight of the confidence of the time domain probability vector and a fusion weight of the confidence of the frequency domain probability vector to obtain a weight of the time domain probability vector and a weight of the frequency domain probability vector;

[0085] The fusion weight is calculated as follows:

[0086] (6);

[0087] (7);

[0088] weighting the time domain probability vector and the weight of the frequency domain probability vector to obtain a fused probability vector.

[0089] Step 6, the 1D-CNN model, the Transformer model and the decision level fusion constitute a double-branch confidence fusion fault diagnosis model; the training set samples in the voltage data set are used to train the double-branch confidence fusion fault diagnosis model;

[0090] Step 7, the effectiveness of the double-branch confidence fusion fault diagnosis model is verified by using the test set samples.

Claims

1. A method for fault diagnosis of filter circuits in airborne radio frequency systems based on confidence fusion, characterized in that: The filter circuit fault diagnosis method comprises the following steps: Step 1: Collect output voltage data sets under circuit health status and various fault conditions; Step 2: Processing the collected voltage data set; The data processing process is as follows: Step 2.1, normalize the time domain data in the voltage data set to obtain a voltage time domain data set; Step 2.2, preprocess the time domain data in the voltage data set in the frequency domain to obtain a voltage frequency domain data set; Step 2.3: Divide the voltage time domain dataset and the voltage frequency domain dataset into a training set and a test set in a ratio of 7:3, and add circuit fault status labels to the training set samples; Step 3: Build a 1D-CNN model and use the normalized voltage time domain data to perform fault diagnosis through the 1D-CNN model to obtain a time domain probability vector; Step 4: Build a Transformer model and use the frequency domain data to perform fault diagnosis through the Transformer model to obtain a frequency domain probability vector; Step 5: Use weighted fusion to perform decision-level fusion on the time domain probability vector and the frequency domain probability vector to obtain a fused probability vector; select the fault category corresponding to the maximum probability in the fused probability vector as the final diagnosis result; Step 6: The 1D-CNN model and the Transformer model are fused at the decision level to form a dual-branch confidence fusion fault diagnosis model. The dual-branch confidence fusion fault diagnosis model is trained using the training set samples in the voltage dataset. Step 7: Use the test set samples to verify the effectiveness of the dual-branch confidence fusion fault diagnosis model.

2. The method for fault diagnosis of filter circuit of airborne radio frequency system based on confidence fusion according to claim 1, characterized in that: In step 1, the fault conditions include resistance drift, capacitor failure and operational amplifier saturation. Resistance drift, capacitor failure and operational amplifier saturation are all failures of a single component; resistance drift refers to an increase or decrease in the value of a single resistor component by more than 10%, capacitor failure refers to a short circuit of a single capacitor element, and operational amplifier saturation refers to the phenomenon that the input signal amplitude is too large, the operational amplifier driving capability is insufficient, or the operational amplifier power supply voltage drops, causing the output of the operational amplifier to be unable to continue to follow the input changes after exceeding a predetermined value.

3. The method for fault diagnosis of filter circuit of airborne radio frequency system based on confidence fusion according to claim 1, characterized in that: In step 2.1, the normalization equation of the normalization process is as follows: （1）； in, is the time domain data signal in the voltage data set; is the minimum value in the voltage data set; is the maximum value in the voltage data set; is the normalized voltage time domain data signal, The value range is between 0 and 1.

4. The method for fault diagnosis of filter circuit of airborne radio frequency system based on confidence fusion according to claim 1, characterized in that: In step 2.2, the frequency domain preprocessing process is as follows: Step 2.2.1, use fast Fourier transform to convert time domain data into frequency domain data; The window added by the fast Fourier transform is a Hanning window. The fast Fourier transform is performed on the windowed signal to obtain the frequency domain data in complex form. The real and imaginary parts of the frequency domain data are squared respectively and then summed and root-summed to obtain the amplitude spectrum of the voltage frequency domain data set. Step 2.2.2: Logarithmically compress and normalize the amplitude spectrum of the voltage frequency domain data set. First, logarithmic compression is performed to enhance weak signals and obtain the compressed data L[k]: （2）； Where M[k] represents the amplitude spectrum, ; , Used to prevent zero value; Then, the compressed data L[k] is normalized to obtain the voltage frequency domain data F[k]: （3）； in, is the mean value of the compressed data L[k]; is the standard deviation of the compressed data; the frequency domain data F[k] serves as the input of the Transformer model.

5. The method for fault diagnosis of filter circuit of airborne radio frequency system based on confidence fusion according to claim 1, characterized in that: In step 3, the 1D-CNN model includes an input layer and a classification output layer; the input layer includes the first convolution layer, the first maximum pooling layer, the second convolution layer, the second maximum pooling layer, the third convolution layer and the third maximum pooling layer in sequence; the convolution kernel size of the three convolution layers is 5, and the pooling window size of each maximum pooling layer is 2; the third maximum pooling layer is connected to the classification output layer; the classification output layer includes two fully connected layers, the first fully connected layer is used to receive the fused feature information, the second fully connected layer uses the softmax function for fault classification, and a dropout layer with a dropout rate of 0.5 is added between the two fully connected layers to prevent model overfitting.

6. The method for fault diagnosis of filter circuit of airborne radio frequency system based on confidence fusion according to claim 1, characterized in that: In step 4, the Transformer model is a lightweight model, which includes an input layer, an embedding layer, a position encoding, an added CLS classification token, a Transformer encoder layer and a linear classification output layer in sequence; the Transformer encoder layer is a three-layer encoder, which includes a first encoder layer, a second encoder layer and a third encoder layer in sequence; the classification output layer is a two-layer fully connected layer, which includes a first fully connected layer and a second fully connected layer in sequence; each layer of the Transformer encoder has a four-head self-attention mechanism, a 256-dimensional feedforward network and a GELU activation function; the Transformer encoder layer is stabilized through residual connection and layer normalization training, extracts the state vector with the added CLS token, and inputs it into the linear classification output layer after layer normalization for fault classification.

7. The method for fault diagnosis of filter circuit of airborne radio frequency system based on confidence fusion according to claim 1, characterized in that: In step 5, the process of decision-level fusion is as follows: Step 5.1, calculate the confidence of the time domain probability vector output by the 1D-CNN model : （4）； in, Represents the time domain probability vector output by the 1D-CNN model; The information entropy representing the time domain probability vector output by the 1D-CNN model is calculated as follows: （5）； Where n represents the total number of elements in the time domain probability vector; represents the elements in the time domain probability vector, ; Similarly, use the method in step 5.1 to calculate the confidence of the frequency domain probability vector ; Step 5.2, calculate the confidence of the time domain probability vector respectively and the confidence of the frequency domain probability vector The fusion weight of the time domain probability vector is obtained and the weight of the frequency domain probability vector ; The fusion weight is calculated as follows: （6）； （7）； Step 5.3, weighting the time domain probability vector and the weight of the frequency domain probability vector Perform weighted fusion to obtain the fusion probability vector.

8. An electronic device, characterized in that: include: one or more processors; Memory; One or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs being configured to perform the method according to any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores program codes, which can be called by a processor to execute the method according to any one of claims 1 to 7.

Images