Internal combustion engine fault diagnosis method and device based on time domain and frequency domain feature fusion

Abstract

The invention discloses an internal combustion engine fault diagnosis method and device based on time domain and frequency domain feature fusion, and relates to the technical field of fault detection, and the method comprises the steps: obtaining a vibration acceleration signal of an internal combustion engine during operation; based on a preset center frequency, a variational mode decomposition method is adopted to decompose the vibration acceleration signal to obtain a plurality of signal components, the sum of the broadband of the plurality of signal components is minimum, and the vibration acceleration signal can be obtained by reconstructing signal energy; and splicing the signal components, and inputting the obtained spliced signal and the vibration acceleration signal into a pre-trained fault prediction model to obtain a fault diagnosis result output by the fault prediction model. The method and the device are used for solving the problem of low prediction accuracy during fault diagnosis of the internal combustion engine in the prior art, and the accuracy of fault diagnosis is improved.

Description

Technical Field

[0001] The present application relates to the technical field of fault detection, and in particular to a method and device for internal combustion engine fault diagnosis based on the fusion of time domain and frequency domain features. Background Art

[0002] As one of the core power sources of mechanical equipment, internal combustion engines are widely used in transportation, agriculture, engineering machinery, generator sets, general aviation and other fields due to their high thermal efficiency and easy mobility. As internal combustion engines are the core power source of mechanical equipment, if they fail, accidents may occur.

[0003] Existing technologies improve the safety of equipment by predicting faults and minimize the impact of faults. For example, machine learning and deep learning methods are used to diagnose internal combustion engine faults. Machine learning methods such as support vector machines and random forests are widely used, but these methods have poor feature extraction capabilities, resulting in low accuracy of the final fault diagnosis results. Deep learning methods are mainly based on convolutional neural networks, which are suitable for image signals and have poor capabilities for mining time domain signal features, resulting in low accuracy of the final fault diagnosis results. Summary of the invention

[0004] In response to the above-mentioned problems and technical needs, the applicant has proposed an internal combustion engine fault diagnosis method and equipment based on the fusion of time domain and frequency domain features, so as to solve the problem of low prediction accuracy in internal combustion engine fault diagnosis in the prior art and improve the accuracy of fault diagnosis.

[0005] The embodiment of the present application provides an internal combustion engine fault diagnosis method based on time domain and frequency domain feature fusion, the method comprising:

[0006] Obtaining vibration acceleration signals of the internal combustion engine during operation;

[0007] Based on a preset center frequency, the vibration acceleration signal is decomposed by a variational mode decomposition method to obtain a plurality of signal components, wherein the sum of the broadbands of the plurality of signal components is the smallest, and the vibration acceleration signal can be obtained by reconstructing the signal energy;

[0008] Splicing the signal components, inputting the obtained spliced ​​signal and the vibration acceleration signal into a pre-trained fault prediction model, and obtaining a fault diagnosis result output by the fault prediction model;

[0009] The obtained spliced ​​signal is input into an attention network to obtain a target attention matrix corresponding to the vibration acceleration signal output by the attention network; the target attention matrix is ​​input into a time domain feature extraction network to obtain the time domain features of the vibration acceleration signal output by the time domain feature extraction network; the vibration acceleration signal is input into a frequency domain feature extraction network to obtain the frequency domain features of the vibration acceleration signal output by the frequency domain feature extraction network; the time domain features and the frequency domain features are fused to obtain fused features, and the fault diagnosis of the internal combustion engine is performed based on the fused features;

[0010] Wherein, the fault prediction model includes: the attention network, the time domain feature extraction network and the frequency domain feature extraction network;

[0011] The fault prediction model is trained based on spliced ​​signal samples, vibration acceleration signal samples and fault category samples.

[0012] According to an internal combustion engine fault diagnosis method based on time domain and frequency domain feature fusion according to an embodiment of the present application, the obtained spliced ​​signal is input into an attention network to obtain a target attention matrix corresponding to the vibration acceleration signal output by the attention network, including:

[0013] The spliced ​​signal is input into the attention network, the weight of each signal component is calculated through the attention network, and the attention matrix corresponding to the spliced ​​signal is obtained based on the weight; the product of the attention matrix and the spliced ​​signal is calculated to obtain the attention-weighted signal matrix, and the attention-weighted signal matrix is ​​determined as the target attention matrix.

[0014] According to an internal combustion engine fault diagnosis method based on time domain and frequency domain feature fusion according to an embodiment of the present application, the weight of each signal component is calculated by the attention network, and the attention matrix corresponding to the spliced ​​signal is obtained based on the weight, including:

[0015] Extracting initial features of each of the signal components;

[0016] Assigning a corresponding shallow neural network to each of the initial features, and determining the probability that the initial feature is selected by the shallow neural network for classification;

[0017] The attention matrix is ​​obtained by combining the probabilities for classification corresponding to each of the initial features.

[0018] According to an internal combustion engine fault diagnosis method based on time domain and frequency domain feature fusion according to an embodiment of the present application, the initial feature of each signal component is extracted, including:

[0019] For each of the signal components the following initial feature extraction process is performed:

[0020] Inputting the signal component into a preset feature extraction formula to obtain an initial feature output by the feature extraction formula;

[0021] Wherein, the feature extraction formula includes:

[0022]

[0023] Among them, E k represents the initial feature corresponding to the kth signal component, Tanh represents the activation function, W represents the coefficient, b is a constant, represents the kth signal component, and ω represents the center frequency.

[0024] According to an internal combustion engine fault diagnosis method based on time domain and frequency domain feature fusion according to an embodiment of the present application, a corresponding shallow neural network is assigned to each of the initial features, and the probability of the initial feature being selected by the shallow neural network for classification is determined, including:

[0025] The following probability calculation process is performed for each of the initial features:

[0026] Input the initial feature into the corresponding shallow neural network, obtain the selection probability through the selection formula in the shallow neural network, and obtain the non-selection probability through the non-selection formula in the shallow neural network; input the selection probability and the non-selection probability into the classification formula to obtain the probability that the initial feature is selected for classification;

[0027] Wherein, the selection formula includes:

[0028]

[0029] Among them, p k represents the probability of selection, represents the network parameters of the kth shallow neural network when the initial features are selected, represents the network constant of the k-th shallow neural network when the initial features are selected;

[0030] Wherein, the unselected formula includes:

[0031]

[0032] Among them, n k represents the probability of not being selected, represents the network parameters of the kth shallow neural network when the initial features are not selected, represents the network constant of the kth shallow neural network when the initial feature is not selected;

[0033] Among them, the classification formula includes:

[0034]

[0035] Among them, a k Represents the probability used to make the classification.

[0036] According to an internal combustion engine fault diagnosis method based on time domain and frequency domain feature fusion according to an embodiment of the present application, based on a preset center frequency, a variational mode decomposition method is used to decompose the vibration acceleration signal to obtain multiple signal components, including:

[0037] Decomposing the vibration acceleration signal by using a variational mode decomposition method to obtain a plurality of signal components;

[0038] Converting the signal components into analytical signals based on Hilbert transform and Dirac function;

[0039] multiplying the analytical signal with an exponential term tuned to a corresponding center frequency, shifting the spectrum of the signal component to baseband, and estimating the broadband of the signal component by Gaussian smoothing;

[0040] When it is determined that the sum of the broadbands of the plurality of signal components is the smallest and the vibration acceleration signal can be obtained by reconstructing the signal components, it is determined that the decomposition is completed.

[0041] According to an internal combustion engine fault diagnosis method based on time domain and frequency domain feature fusion according to an embodiment of the present application, the training process of the fault prediction model includes:

[0042] Acquire a model training sample, wherein the model training sample includes: the splicing signal sample, the vibration acceleration signal sample and the fault category sample;

[0043] Input the spliced ​​signal sample into the attention network of the fault prediction model to obtain the target attention matrix sample corresponding to the vibration acceleration signal sample output by the attention network; input the target attention matrix sample into the time domain feature extraction network of the fault prediction model to obtain the predicted time domain feature output by the time domain feature extraction network; input the vibration acceleration signal sample into the frequency domain feature extraction network of the fault prediction model to obtain the predicted frequency domain feature output by the frequency domain feature extraction network; fuse the time domain feature and the frequency domain feature to obtain the fused feature, and obtain the predicted fault diagnosis result based on the fused feature; determine the consistency between the fault category sample and the fault diagnosis result; adjust the objective function of the attention network, the network parameters of the time domain feature extraction network and the network parameters of the frequency domain feature extraction network based on the consistency until the consistency reaches the preset consistency.

[0044] According to an internal combustion engine fault diagnosis method based on time domain and frequency domain feature fusion according to an embodiment of the present application, the objective function includes:

[0045]

[0046] Where A represents the attention matrix, θ=<θ l ,θ α >,θ l and θ α All represent the network parameters of the attention network, R(θ) represents the L2 norm, λ represents the strength of regularization, Denote the network parameter as θ l The corresponding network layer, represents the predicted attention matrix, and Y is the actual attention matrix.

[0047] According to an internal combustion engine fault diagnosis method based on time domain and frequency domain feature fusion according to an embodiment of the present application, the time domain feature extraction network includes a residual neural network structure;

[0048] The training process of the fault prediction model also includes:

[0049] During the training process, when the number of iterations is greater than a preset number and the consistency does not reach the preset consistency, a residual layer of the residual neural network structure is added, and training is performed based on the changed fault prediction model.

[0050] An embodiment of the present application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the internal combustion engine fault diagnosis method based on the fusion of time domain and frequency domain features as described in any one of the above items are implemented.

[0051] The internal combustion engine fault diagnosis method and device based on the fusion of time domain and frequency domain features provided in the embodiment of the present application obtains the vibration acceleration signal of the internal combustion engine when it is in operation; determines the bandwidth corresponding to the vibration acceleration signal, and based on the determined bandwidth and the preset center frequency, uses the variational mode decomposition method to decompose the vibration acceleration signal to obtain multiple signal components. The present application uses variational mode decomposition to effectively separate the data of the vibration acceleration signal in the time domain line, providing an effective data basis for the subsequent extraction of time domain features; splices the signal components, and inputs the obtained spliced ​​signal and the vibration acceleration signal into a pre-trained fault prediction model to obtain the fault diagnosis result output by the fault prediction model.

[0052] Specifically, the attention network of the fault prediction model inputs the spliced ​​signal and outputs a target attention matrix corresponding to the vibration acceleration signal; the target attention matrix is ​​input into the time domain feature extraction network to obtain the time domain features of the vibration acceleration signal output by the time domain feature extraction network. The present application utilizes the attention mechanism to quickly obtain the target attention matrix, and then quickly enables the time domain feature extraction network to obtain the time domain features; at the same time, the vibration acceleration signal is input into the frequency domain feature extraction network to obtain the frequency domain features of the vibration acceleration signal output by the frequency domain feature extraction network. The present application processes the time domain features and the frequency domain features separately, and can accurately mine the signal features of the vibration acceleration signal, providing an accurate data basis for subsequent fault diagnosis; finally, the time domain features and the frequency domain features are fused to obtain the fused features, and the internal combustion engine fault diagnosis is performed based on the fused features. The whole process improves the accuracy, timeliness and comprehensiveness of the signal feature extraction, and improves the accuracy of the fault diagnosis results as a whole. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0054] Figure 1 It is one of the flow charts of the internal combustion engine fault diagnosis method based on the fusion of time domain and frequency domain features provided in the embodiment of the present application;

[0055] Figure 2 This is the second flow chart of the internal combustion engine fault diagnosis method based on the fusion of time domain and frequency domain features provided in the embodiment of the present application;

[0056] Figure 3 It is a structural schematic diagram of a fault diagnosis device for an internal combustion engine based on the fusion of time domain and frequency domain features provided in an embodiment of the present application;

[0057] Figure 4 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] The embodiment of the present application provides a method for diagnosing internal combustion engine faults based on the fusion of time domain and frequency domain features. The method can be applied to a smart terminal or a server. Some other descriptions in the embodiment of the present application are for illustrative purposes only and are not intended to limit the scope of protection of the present application. They will not be described one by one in the following. The specific implementation of the method is as follows Figure 1 As shown:

[0060] Step 101, obtaining a vibration acceleration signal of the internal combustion engine during operation.

[0061] The vibration acceleration signal may be collected by an acceleration signal collector installed at the internal combustion engine.

[0062] Step 102 : determining a bandwidth corresponding to the vibration acceleration signal, and based on the determined bandwidth and a preset center frequency, using a variational mode decomposition method to decompose the vibration acceleration signal to obtain a plurality of signal components.

[0063] Step 103 , concatenating the signal components, inputting the obtained concatenated signal and the vibration acceleration signal into a pre-trained fault prediction model, and obtaining a fault diagnosis result output by the fault prediction model.

[0064] Among them, the obtained spliced ​​signal attention network is used to obtain the target attention matrix corresponding to the vibration acceleration signal output by the attention network; the target attention matrix is ​​input into the time domain feature extraction network to obtain the time domain features of the vibration acceleration signal output by the time domain feature extraction network; the vibration acceleration signal is input into the frequency domain feature extraction network to obtain the frequency domain features of the vibration acceleration signal output by the frequency domain feature extraction network; the time domain features and the frequency domain features are fused to obtain the fused features, and the internal combustion engine fault diagnosis is performed based on the fused features.

[0065] Among them, the fault prediction model includes: attention network, time domain feature extraction network and frequency domain feature extraction network.

[0066] Among them, the fault prediction model is trained based on splicing signal samples, vibration acceleration signal samples and fault category samples.

[0067] The internal combustion engine fault diagnosis method based on the fusion of time domain and frequency domain features provided in the embodiment of the present application obtains the vibration acceleration signal of the internal combustion engine when it is in operation; determines the bandwidth corresponding to the vibration acceleration signal, and based on the determined bandwidth and the preset center frequency, uses the variational mode decomposition method to decompose the vibration acceleration signal to obtain multiple signal components. The present application uses variational mode decomposition to effectively separate the data of the vibration acceleration signal in the time domain line, thereby providing an effective data basis for the subsequent extraction of time domain features; splices the signal components, and inputs the obtained spliced ​​signal and the vibration acceleration signal into a pre-trained fault prediction model to obtain the fault diagnosis result output by the fault prediction model.

[0068] Specifically, the attention network of the fault prediction model inputs the spliced ​​signal and outputs a target attention matrix corresponding to the vibration acceleration signal; the target attention matrix is ​​input into the time domain feature extraction network to obtain the time domain features of the vibration acceleration signal output by the time domain feature extraction network. The present application utilizes the attention mechanism to quickly obtain the target attention matrix, and then quickly enables the time domain feature extraction network to obtain the time domain features; at the same time, the vibration acceleration signal is input into the frequency domain feature extraction network to obtain the frequency domain features of the vibration acceleration signal output by the frequency domain feature extraction network. The present application processes the time domain features and the frequency domain features separately, and can accurately mine the signal features of the vibration acceleration signal, providing an accurate data basis for subsequent fault diagnosis; finally, the time domain features and the frequency domain features are fused to obtain the fused features, and the internal combustion engine fault diagnosis is performed based on the fused features. The whole process improves the accuracy, timeliness and comprehensiveness of the signal feature extraction, and improves the accuracy of the fault diagnosis results as a whole.

[0069] In a specific embodiment, based on a preset center frequency, the vibration acceleration signal is decomposed by using a variational mode decomposition method to obtain multiple signal components, and the specific implementation includes:

[0070] The vibration acceleration signal is decomposed by using the variational mode decomposition method to obtain multiple signal components; the signal components are converted into analytical signals based on Hilbert transform and Dirac function; the analytical signal is multiplied by an exponential term tuned to the corresponding center frequency, the spectrum of the signal component is moved to the baseband, and the broadband of the signal component is estimated by Gaussian smoothness; when it is determined that the sum of the broadbands of multiple signal components is minimized and the reconstructed signal component can obtain the vibration acceleration signal, the decomposition is determined to be complete.

[0071] Specifically, the signal component (IMF) is a group of amplitude-frequency modulated signals. According to Casson's law, the bandwidth of the IMF is limited, and the IMF is close to the center frequency.

[0072] In order to conveniently and accurately estimate the bandwidth information of each IMF, the Hilbert transform is used to represent the actual signal (signal component) in the form of an analytical signal. For details, see formula (1):

[0073]

[0074] in, represents the analytical signal, ω represents the center frequency, j represents the imaginary unit, and the square of j is equal to -1.

[0075] Among them, one-dimensional Hilbert transform (1DHT for short) is a linear translation invariant operator that can cause a 90° phase shift in a one-dimensional signal after mapping.

[0076] Therefore, the imaginary part of the signal can be constructed through 1DHT to represent the signal component as an analytical signal. For details, see formula (2):

[0077] u k (t) = u k (t)+jH(u k (t)) = A k (t)e jφ(t) =A k (t)(cos(φ(t))+jsin(φ(t)))..(2)

[0078] Among them, u k (t) represents the kth signal component, A k (t) represents the amplitude, φ(t) represents the phase, e jφ(t) Represents the rotation vector of a complex signal.

[0079] Among them, the instantaneous frequency of the complex signal can be obtained by taking the derivative of the phase, see formula (3) for details:

[0080] ω(t)=dφ(t) / dt…………………………………………(3)

[0081] Where ω(t) represents the instantaneous frequency.

[0082] Among them, complex signal refers to a complex signal, because variational mode decomposition is performed in the frequency domain, and there are real and imaginary components, that is, complex signal components.

[0083] In addition, the real part of the analytical signal can be taken to obtain the original signal. Therefore, when performing Hilbert transform on the signal component, the Dirac function is introduced, and the analytical signal obtained is specifically shown in formula (4):

[0084]

[0085] Here, δ(t) represents the Dirac function.

[0086] In a specific embodiment, the obtained spliced ​​signal is input into the attention network, and the specific implementation of obtaining the target attention matrix corresponding to the vibration acceleration signal output by the attention network includes:

[0087] The spliced ​​signal is input into the attention network, and the weight of each signal component is calculated through the attention network. The attention matrix corresponding to the spliced ​​signal is obtained based on the weight. The product of the attention matrix and the spliced ​​signal is calculated to obtain the attention-weighted signal matrix, and the attention-weighted signal matrix is ​​determined as the target attention matrix.

[0088] The splicing signal is represented by X.

[0089] Among them, the attention network calculates the weight of the input data based on the attention mechanism, focusing on the part with a larger weight and focusing on it as important information to improve the data efficiency and accuracy.

[0090] In a specific embodiment, the weight of each signal component is calculated by the attention network, and the specific implementation of obtaining the attention matrix corresponding to the spliced ​​signal based on the weight includes:

[0091] Extract the initial features of each signal component. Assign a corresponding shallow neural network to each initial feature, and determine the probability that the initial feature is selected by the shallow neural network for classification. Combine the probabilities for classification corresponding to each initial feature to obtain the attention matrix.

[0092] In a specific embodiment, the specific implementation of extracting the initial features of each signal component includes:

[0093] The following initial feature extraction process is performed for each signal component:

[0094] The signal components are input into a preset feature extraction formula to obtain the initial features output by the feature extraction formula.

[0095] The feature extraction formula is shown in formula (5):

[0096] E k =Tanh(W·(u k (ω) T +b)…………………………(5)

[0097] Among them, E k represents the initial feature corresponding to the kth signal component, Tanh represents the activation function, W represents the coefficient, b is a constant, and uk represents the kth signal component, and ω represents the center frequency.

[0098] Specifically, the initial features of the signal components are extracted by formula (5), that is, the initial features are compressed into a smaller vector while retaining most of the information (discarding some redundant features and noise). The Tanh function can retain important information during the extraction process because it has both positive and negative values.

[0099] In a specific embodiment, a corresponding shallow neural network is assigned to each initial feature, and the specific implementation of determining the probability that the initial feature is selected by the shallow neural network for classification includes:

[0100] For each initial feature, the following probability calculation process is performed:

[0101] The initial features are input into the corresponding shallow neural network, and the selection probability is obtained through the selection formula in the shallow neural network, and the non-selection probability is obtained through the non-selection formula in the shallow neural network. The selection probability and non-selection probability are input into the classification formula to obtain the probability that the initial features are selected for classification.

[0102] The selection formula is shown in formula (6):

[0103]

[0104] Among them, p k represents the probability of selection, represents the network parameters of the kth shallow neural network when the initial features are selected, Represents the network constant of the kth shallow neural network when the initial features are selected.

[0105] Among them, the unselected formula is shown in formula (7):

[0106]

[0107] Among them, n k represents the probability of not being selected, represents the network parameters of the kth shallow neural network when the initial features are not selected, Represents the network constant of the kth shallow neural network when the initial features are not selected.

[0108] The classification formula is shown in formula (8):

[0109]

[0110] Among them, a k Represents the probability used for classification.

[0111] Specifically, the initial features are input into the attention network, and a shallow neural network is assigned to each initial feature to determine its probability of being selected. Each shallow neural network generates two values, namely the probability of selection and the probability of non-selection. Then, only the initial features corresponding to the probability of classification greater than the preset probability are paid attention to. Then, the probability of classification corresponding to each initial feature is combined to obtain the attention matrix A.

[0112] in,

[0113] Among them, m represents the number of iterations during attention network training, and d represents the number of initial features whose probability for classification is greater than the preset probability.

[0114] Finally, after obtaining the attention matrix A, calculate the product of A and X to obtain G, that is, G = A*X. At this time, G is the target attention matrix.

[0115] In a specific embodiment, the target attention matrix is ​​input into the time domain feature extraction network, and the specific implementation of obtaining the time domain features of the vibration acceleration signal output by the time domain feature extraction network includes:

[0116] In the time domain line, the target attention matrix is ​​input into the time domain feature extraction network, and the target attention matrix is ​​converted to the input size that adapts to the residual layer. Then, the residual layer of the time domain feature extraction network performs multi-level feature extraction to obtain the intermediate features. Then, a convolutional layer is used to transform the dimension of the intermediate features and further extract features. After that, batch normalization is performed. Then, nonlinear mapping is performed through the activation function, and the time series features are extracted using the self-attention mechanism. Finally, the features of the self-attention mechanism are globally averaged to obtain the time domain features.

[0117] In a specific embodiment, the specific implementation of inputting the vibration acceleration signal into the frequency domain feature extraction network to obtain the frequency domain features of the vibration acceleration signal output by the frequency domain feature extraction network includes:

[0118] In the frequency domain, data is transmitted synchronously with the time domain. The vibration acceleration signal is input into the frequency domain feature extraction network, and the signal is transformed from the time domain to a single-sided spectrum signal in the frequency domain by fast Fourier transform. The frequency domain features are extracted through multiple convolutional layers, residual layers, and pooling layers. The frequency domain signal features have lost the original temporal sequence of the signal, so the use of convolutional neural networks can effectively extract frequency domain features.

[0119] After feature extraction is completed, the feature layer is expanded using a fully connected layer to keep the scale of the frequency domain features consistent with the scale of the time domain features.

[0120] In a specific embodiment, the time domain features and the frequency domain features are fused to obtain fused features, and the specific implementation of the internal combustion engine fault diagnosis based on the fused features includes:

[0121] The time domain features and frequency domain features are concatenated and input into a classifier consisting of two fully connected layers, and the fault diagnosis results are output through the classifier.

[0122] Specifically, almost all the convolutional layers involved in the above network structure have corresponding pooling layers and activation functions, including the ReLU activation function. When this function faces input data less than 1, the forward propagation and back propagation gradient transfers are both 0, which will inactivate some neurons. Therefore, this application uses a smooth Swish function to improve the accuracy and efficiency of fault diagnosis.

[0123] In a specific embodiment, the training process of the fault prediction model includes:

[0124] Obtain model training samples; input the spliced ​​signal samples into the attention network of the fault prediction model to obtain the target attention matrix samples corresponding to the vibration acceleration signal samples output by the attention network; input the target attention matrix samples into the time domain feature extraction network of the fault prediction model to obtain the predicted time domain features output by the time domain feature extraction network; input the vibration acceleration signal samples into the frequency domain feature extraction network of the fault prediction model to obtain the predicted frequency domain features output by the frequency domain feature extraction network; fuse the time domain features and the frequency domain features to obtain the fused features, and obtain the predicted fault diagnosis results based on the fused features; determine the consistency of the fault category samples and the fault diagnosis results; adjust the objective function of the attention network, the network parameters of the time domain feature extraction network, and the network parameters of the frequency domain feature extraction network based on the consistency until the consistency reaches the preset consistency.

[0125] Among them, the model training samples include: splicing signal samples, vibration acceleration signal samples and fault category samples.

[0126] Specifically, the model training samples may be divided according to a preset ratio, such as 4:1, and four-fifths of the data may be used to train the fault prediction model, while one-fifth of the data may be used to test the prediction accuracy of the fault prediction model.

[0127] In addition, the vibration acceleration signal corresponding to the internal combustion engine can be obtained in real time in the form of a time window to add to the model training sample to update the fault prediction model.

[0128] In a specific embodiment, the objective function is shown in formula (9):

[0129]

[0130] Where A represents the attention matrix, θ=<θ l,θ α >,θ l and θ a All represent the network parameters of the attention network, R(θ) represents the L2 norm, λ represents the strength of regularization, Denote the network parameter as θ l The corresponding network layer, represents the predicted attention matrix, and Y is the actual attention matrix.

[0131] In a specific embodiment, the time domain feature extraction network includes a residual neural network structure. The training process of the fault prediction model also includes: during the training process, when the number of iterations is greater than a preset number and the consistency does not reach a preset consistency, a residual layer of the residual neural network structure is added, and training is performed based on the changed fault prediction model.

[0132] In addition, during the model training process, due to multiple iterations of training, the signal component output by the attention network after n iterations is

[0133] Specifically, the initial number of residual layers is determined in advance based on the complexity of the fault and the number of model training samples, and the residual layers are adjusted based on the actual training process.

[0134] In addition, during the model training process, the Dropout layer is used to randomly block some neurons so that the model does not rely too much on certain neurons, preventing the problem of overfitting of the model.

[0135] Next, through Figure 2 The implementation of this application is described in detail:

[0136] Step 201, obtaining a vibration acceleration signal of the internal combustion engine during operation.

[0137] Step 202: Based on a preset center frequency, a variational mode decomposition method is used to decompose the vibration acceleration signal to obtain a plurality of signal components.

[0138] Step 203, splicing the signal components, inputting the obtained spliced ​​signal into a preset attention network, and obtaining a target attention matrix output by the attention network corresponding to the vibration acceleration signal.

[0139] Step 204, input the target attention matrix into a preset time domain feature extraction network to obtain the time domain features of the vibration acceleration signal output by the time domain feature extraction network.

[0140] Step 205: input the vibration acceleration signal into a preset frequency domain feature extraction network to obtain the frequency domain features of the vibration acceleration signal output by the frequency domain feature extraction network.

[0141] Step 206, the time domain features and the frequency domain features are fused to obtain fused features, and the fused features are input into a preset classifier to obtain a fault diagnosis result of the internal combustion engine output by the classifier.

[0142] This application utilizes the advantages of convolutional neural networks, residual neural networks and self-attention mechanisms to improve the model's ability to characterize time-series vibration signals, effectively improving the accuracy of fault diagnosis. In addition, a feature extraction method that uses time domain lines and frequency domain lines in parallel is used to retain the time-series features of the time domain while also obtaining the frequency domain features. Classification is performed after combining the time domain and frequency domain features, effectively improving the training efficiency of the model and improving the accuracy of the model. In addition, the use of the Swich function as an activation function further improves the model's nonlinear fitting ability and improves the model's convergence efficiency. Finally, the incremental learning update strategy is used to update the model, so that the internal combustion engine can continuously adapt to changes in the body during the entire life cycle and maintain a stable diagnostic result.

[0143] The embodiment of the present application also provides an internal combustion engine fault diagnosis device based on the fusion of time domain and frequency domain features. The specific implementation of the device can refer to the description of the internal combustion engine fault diagnosis method based on the fusion of time domain and frequency domain features, and the repeated parts will not be repeated. Figure 3 As shown, the device comprises:

[0144] An acquisition module 301 is used to acquire a vibration acceleration signal of the internal combustion engine during operation;

[0145] The signal decomposition module 302 is used to decompose the vibration acceleration signal based on a preset center frequency by using a variational mode decomposition method to obtain multiple signal components, wherein the sum of the broadbands of the multiple signal components is the smallest and the vibration acceleration signal can be obtained by reconstructing the signal energy;

[0146] Fault prediction 303, used for splicing signal components, inputting the obtained spliced ​​signal and vibration acceleration signal into a pre-trained fault prediction model, and obtaining a fault diagnosis result output by the fault prediction model;

[0147] The obtained spliced ​​signal is input into the attention network to obtain the target attention matrix corresponding to the vibration acceleration signal output by the attention network; the target attention matrix is ​​input into the time domain feature extraction network to obtain the time domain features of the vibration acceleration signal output by the time domain feature extraction network; the vibration acceleration signal is input into the frequency domain feature extraction network to obtain the frequency domain features of the vibration acceleration signal output by the frequency domain feature extraction network; the time domain features and the frequency domain features are fused to obtain the fused features, and the fault diagnosis of the internal combustion engine is performed based on the fused features;

[0148] Among them, the fault prediction model includes: attention network, time domain feature extraction network and frequency domain feature extraction network;

[0149] Among them, the fault prediction model is trained based on splicing signal samples, vibration acceleration signal samples and fault category samples.

[0150] Figure 4 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 4 As shown, the electronic device may include: a processor (processor) 401, a communication interface (Communications Interface) 402, a memory (memory) 403 and a communication bus 404, wherein the processor 401, the communication interface 402, and the memory 403 communicate with each other through the communication bus 404. The processor 401 can call the logic instructions in the memory 403 to execute the internal combustion engine fault diagnosis method based on the fusion of time domain and frequency domain features, the method comprising: obtaining the vibration acceleration signal of the internal combustion engine when it is in operation; based on the preset center frequency, using the variational mode decomposition method to decompose the vibration acceleration signal to obtain multiple signal components, wherein the sum of the broadbands of the multiple signal components is the smallest, and the reconstructed signal energy can obtain the vibration acceleration signal; splicing the signal components, inputting the obtained splicing signal and the vibration acceleration signal into a pre-trained fault prediction model, and obtaining the fault diagnosis result output by the fault prediction model; wherein the obtained splicing signal is input into the attention network to obtain the attention network. The attention network outputs a target attention matrix corresponding to the vibration acceleration signal; the target attention matrix is ​​input into the time domain feature extraction network to obtain the time domain features of the vibration acceleration signal output by the time domain feature extraction network; the vibration acceleration signal is input into the frequency domain feature extraction network to obtain the frequency domain features of the vibration acceleration signal output by the frequency domain feature extraction network; the time domain features and the frequency domain features are fused to obtain the fused features, and the internal combustion engine fault diagnosis is performed based on the fused features; wherein the fault prediction model comprises: an attention network, a time domain feature extraction network and a frequency domain feature extraction network; wherein the fault prediction model is trained based on spliced ​​signal samples, vibration acceleration signal samples and fault category samples.

[0151] In addition, the logic instructions in the above-mentioned memory 403 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0152] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by the computer, the computer can execute the internal combustion engine fault diagnosis method based on the fusion of time domain and frequency domain features provided by the above methods, and the method includes: obtaining a vibration acceleration signal of the internal combustion engine when it is in operation; based on a preset center frequency, using a variational mode decomposition method to decompose the vibration acceleration signal to obtain multiple signal components, wherein the sum of the broadbands of the multiple signal components is the smallest, and the reconstructed signal energy can obtain a vibration acceleration signal; splicing the signal components, and inputting the obtained spliced ​​signal and the vibration acceleration signal into a pre-trained fault prediction model to obtain a fault prediction signal. The fault diagnosis result output by the test model is obtained; wherein, the obtained spliced ​​signal is input into the attention network to obtain the target attention matrix corresponding to the vibration acceleration signal output by the attention network; the target attention matrix is ​​input into the time domain feature extraction network to obtain the time domain features of the vibration acceleration signal output by the time domain feature extraction network; the vibration acceleration signal is input into the frequency domain feature extraction network to obtain the frequency domain features of the vibration acceleration signal output by the frequency domain feature extraction network; the time domain features and the frequency domain features are fused to obtain the fused features, and the fault diagnosis of the internal combustion engine is performed based on the fused features; wherein, the fault prediction model includes: an attention network, a time domain feature extraction network and a frequency domain feature extraction network; wherein the fault prediction model is trained based on the spliced ​​signal samples, the vibration acceleration signal samples and the fault category samples.

[0153] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the processor executes the internal combustion engine fault diagnosis method based on the fusion of time domain and frequency domain features provided in the above embodiments, the method comprising: obtaining a vibration acceleration signal of the internal combustion engine when it is in operation; based on a preset center frequency, decomposing the vibration acceleration signal using a variational mode decomposition method to obtain a plurality of signal components, wherein the sum of the broadbands of the plurality of signal components is the smallest, and the reconstructed signal energy can obtain a vibration acceleration signal; splicing the signal components, inputting the obtained spliced ​​signal and the vibration acceleration signal into a pre-trained fault prediction model, and obtaining a fault diagnosis result output by the fault prediction model; wherein, The obtained spliced ​​signal is input into the attention network to obtain the target attention matrix corresponding to the vibration acceleration signal output by the attention network; the target attention matrix is ​​input into the time domain feature extraction network to obtain the time domain features of the vibration acceleration signal output by the time domain feature extraction network; the vibration acceleration signal is input into the frequency domain feature extraction network to obtain the frequency domain features of the vibration acceleration signal output by the frequency domain feature extraction network; the time domain features and the frequency domain features are fused to obtain the fused features, and the internal combustion engine fault diagnosis is performed based on the fused features; wherein the fault prediction model includes: an attention network, a time domain feature extraction network and a frequency domain feature extraction network; wherein the fault prediction model is trained based on the spliced ​​signal samples, the vibration acceleration signal samples and the fault category samples.

[0154] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0155] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0156] Finally, it should be noted that the above is only the preferred implementation of the present application, and the present application is not limited to the above embodiments. It is understood that other improvements and changes directly derived or associated by those skilled in the art without departing from the spirit and concept of the present application should be considered to be included in the scope of protection of the present application.

Claims

1. A method for internal combustion engine fault diagnosis based on time domain and frequency domain feature fusion, characterized in that: The method comprises: Obtaining vibration acceleration signals of the internal combustion engine during operation; Based on a preset center frequency, the vibration acceleration signal is decomposed by a variational mode decomposition method to obtain a plurality of signal components, wherein the sum of the broadbands of the plurality of signal components is the smallest, and the vibration acceleration signal can be obtained by reconstructing the signal energy; Splicing the signal components, inputting the obtained spliced ​​signal and the vibration acceleration signal into a pre-trained fault prediction model, and obtaining a fault diagnosis result output by the fault prediction model; The obtained spliced ​​signal is input into an attention network to obtain a target attention matrix corresponding to the vibration acceleration signal output by the attention network; the target attention matrix is ​​input into a time domain feature extraction network to obtain the time domain features of the vibration acceleration signal output by the time domain feature extraction network; the vibration acceleration signal is input into a frequency domain feature extraction network to obtain the frequency domain features of the vibration acceleration signal output by the frequency domain feature extraction network; the time domain features and the frequency domain features are fused to obtain fused features, and the fault diagnosis of the internal combustion engine is performed based on the fused features; Wherein, the fault prediction model includes: the attention network, the time domain feature extraction network and the frequency domain feature extraction network; The fault prediction model is trained based on spliced ​​signal samples, vibration acceleration signal samples and fault category samples.

2. The internal combustion engine fault diagnosis method based on time domain and frequency domain feature fusion according to claim 1 is characterized in that: The obtained spliced ​​signal is input into the attention network to obtain a target attention matrix output by the attention network corresponding to the vibration acceleration signal, including: The spliced ​​signal is input into the attention network, the weight of each signal component is calculated through the attention network, and the attention matrix corresponding to the spliced ​​signal is obtained based on the weight; the product of the attention matrix and the spliced ​​signal is calculated to obtain the attention-weighted signal matrix, and the attention-weighted signal matrix is ​​determined as the target attention matrix.

3. The internal combustion engine fault diagnosis method based on time domain and frequency domain feature fusion according to claim 2 is characterized in that: Calculating the weight of each of the signal components through the attention network, and obtaining the attention matrix corresponding to the spliced ​​signal based on the weight, including: Extracting initial features of each of the signal components; Assigning a corresponding shallow neural network to each of the initial features, and determining the probability that the initial feature is selected by the shallow neural network for classification; The attention matrix is ​​obtained by combining the probabilities for classification corresponding to each of the initial features.

4. The internal combustion engine fault diagnosis method based on time domain and frequency domain feature fusion according to claim 3 is characterized in that: Extracting initial features of each of the signal components, including: For each of the signal components the following initial feature extraction process is performed: Inputting the signal component into a preset feature extraction formula to obtain an initial feature output by the feature extraction formula; Wherein, the feature extraction formula includes: Among them, E k represents the initial feature corresponding to the kth signal component, Tanh represents the activation function, W represents the coefficient, b is a constant, represents the kth signal component, and ω represents the center frequency.

5. The internal combustion engine fault diagnosis method based on time domain and frequency domain feature fusion according to claim 4 is characterized in that: Assigning a corresponding shallow neural network to each of the initial features, and determining the probability that the initial features are selected by the shallow neural network for classification, including: The following probability calculation process is performed for each of the initial features: Input the initial feature into the corresponding shallow neural network, obtain the selection probability through the selection formula in the shallow neural network, and obtain the non-selection probability through the non-selection formula in the shallow neural network; input the selection probability and the non-selection probability into the classification formula to obtain the probability that the initial feature is selected for classification; Wherein, the selection formula includes: Among them, p k represents the probability of selection, represents the network parameters of the kth shallow neural network when the initial features are selected, represents the network constant of the k-th shallow neural network when the initial features are selected; Wherein, the unselected formula includes: Among them, n k represents the probability of not being selected, represents the network parameters of the kth shallow neural network when the initial features are not selected, represents the network constant of the kth shallow neural network when the initial feature is not selected; Among them, the classification formula includes: Among them, a k Represents the probability used to make the classification.

6. The internal combustion engine fault diagnosis method based on time domain and frequency domain feature fusion according to any one of claims 1 to 5, characterized in that: Based on the preset center frequency, the vibration acceleration signal is decomposed by using the variational mode decomposition method to obtain multiple signal components, including: Decomposing the vibration acceleration signal by using a variational mode decomposition method to obtain a plurality of signal components; Converting the signal components into analytical signals based on Hilbert transform and Dirac function; multiplying the analytical signal with an exponential term tuned to a corresponding center frequency, shifting the spectrum of the signal component to baseband, and estimating the broadband of the signal component by Gaussian smoothing; When it is determined that the sum of the broadbands of the plurality of signal components is the smallest and the vibration acceleration signal can be obtained by reconstructing the signal components, it is determined that the decomposition is completed.

7. The internal combustion engine fault diagnosis method based on time domain and frequency domain feature fusion according to any one of claims 1 to 5, characterized in that: The training process of the fault prediction model includes: Acquire a model training sample, wherein the model training sample includes: the splicing signal sample, the vibration acceleration signal sample and the fault category sample; Input the spliced ​​signal sample into the attention network of the fault prediction model to obtain the target attention matrix sample corresponding to the vibration acceleration signal sample output by the attention network; input the target attention matrix sample into the time domain feature extraction network of the fault prediction model to obtain the predicted time domain feature output by the time domain feature extraction network; input the vibration acceleration signal sample into the frequency domain feature extraction network of the fault prediction model to obtain the predicted frequency domain feature output by the frequency domain feature extraction network; fuse the time domain feature and the frequency domain feature to obtain the fused feature, and obtain the predicted fault diagnosis result based on the fused feature; determine the consistency between the fault category sample and the fault diagnosis result; adjust the objective function of the attention network, the network parameters of the time domain feature extraction network and the network parameters of the frequency domain feature extraction network based on the consistency until the consistency reaches the preset consistency.

8. The internal combustion engine fault diagnosis method based on time domain and frequency domain feature fusion according to claim 7 is characterized in that: The objective function includes: Where A represents the attention matrix, θ=<θ l ,θ α >,θ l and θ α All represent the network parameters of the attention network, R(θ) represents the L2 norm, λ represents the strength of regularization, Denote the network parameter as θ l The corresponding network layer, represents the predicted attention matrix, and Y is the actual attention matrix.

9. The internal combustion engine fault diagnosis method based on time domain and frequency domain feature fusion according to claim 7, characterized in that: The time domain feature extraction network includes a residual neural network structure; The training process of the fault prediction model also includes: During the training process, when the number of iterations is greater than a preset number and the consistency does not reach the preset consistency, a residual layer of the residual neural network structure is added, and training is performed based on the changed fault prediction model.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the internal combustion engine fault diagnosis method based on the fusion of time domain and frequency domain features as described in any one of claims 1 to 9 are implemented.

Images