An electric drive system fault detection and state prediction method, device and product

Abstract

The application discloses a kind of electric drive system fault detection and state prediction method, equipment and product, it is related to fault detection technical field, the method includes obtaining current signal data collected by current sensor, wherein current sensor is located on electric drive system;Feature extraction is carried out on current signal data using short-time Fourier transform, and the final two-dimensional time-frequency spectrum is obtained;Final two-dimensional time-frequency spectrum includes the amplitude value of multiple time points;The final two-dimensional time-frequency spectrum and the latest time-frequency data are input into Informer network model, and the frequency information and fault detection result of a period of time in the future are obtained;The latest time-frequency data is the time-frequency data of final time in the final two-dimensional time-frequency spectrum, and Informer network model includes embedding layer, encoder, decoder and output layer connected in turn.The application can realize the identification of fault mode and the accurate prediction of future state.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fault detection, in particular to an electric drive system fault detection and state prediction method, device and product. BACKGROUND

[0002] Electric drive system fault detection technology has a wide range of applications in modern industry, aerospace, rail transportation, intelligent manufacturing and other fields. With the increase in complexity of industrial equipment and the improvement of intelligent demand, traditional fault detection methods gradually show their limitations, especially in dealing with nonlinear problems, the superposition of multiple fault modes and complex working condition changes, which puts forward more stringent requirements on the accuracy, real-time performance and prediction ability of the fault detection system. SUMMARY

[0003] The purpose of the present application is to provide an electric drive system fault detection and state prediction method, device and product, which can realize the identification of fault modes and accurate prediction of future states.

[0004] To achieve the above purpose, the present application provides the following solutions:

[0005] In a first aspect, the present application provides an electric drive system fault detection and state prediction method, comprising:

[0006] Obtaining current signal data collected by a current sensor; the current sensor is located on an electric drive system;

[0007] Performing feature extraction on the current signal data using short-time Fourier transform to obtain a final two-dimensional time-frequency spectrum; the final two-dimensional time-frequency spectrum includes amplitude values at multiple time points;

[0008] Inputting the final two-dimensional time-frequency spectrum and the latest time-frequency data into an Informer network model to obtain frequency information and fault detection results for a future period of time; the latest time-frequency data is the time-frequency data at the final time point in the final two-dimensional time-frequency spectrum, and the Informer network model includes an embedding layer, an encoder, a decoder and an output layer connected in turn.

[0009] In a second aspect, the present application provides a computer device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the electric drive system fault detection and state prediction method of any one of the above.

[0010] In a third aspect, the present application provides a computer program product comprising a computer program, which, when executed by a processor, implements the electric drive system fault detection and state prediction method of any one of the above.

[0011] According to the specific embodiments provided in the present application, the present application discloses the following technical effects:

[0012] The present application provides an electric drive system fault detection and state prediction method, device and product, current signal data collected by a current sensor is obtained; the current sensor is located on the electric drive system; the current signal data is extracted by using short-time Fourier transform to obtain the final two-dimensional time-frequency spectrum; the final two-dimensional time-frequency spectrum includes amplitude values at multiple time points; the final two-dimensional time-frequency spectrum and the latest time-frequency data are input into an Informer network model to obtain frequency information and fault detection results in a future period of time; the latest time-frequency data is the time-frequency data at the final time point in the final two-dimensional time-frequency spectrum, and the Informer network model includes an embedding layer, an encoder, a decoder and an output layer connected in turn. The present application detects and predicts the current signal of the electric drive control system by combining short-time Fourier transform (SFTF) and Informer network. The short-time Fourier transform is used to extract the time-frequency characteristics of the current signal, and the Informer network is used for deep analysis of the characteristics and fault prediction. The two-dimensional time-frequency spectrum generated by the short-time Fourier transform is used as the input feature of the deep learning model, and the long sequence modeling capability and efficient attention mechanism of the Informer network are used to realize high-precision fault detection and prediction, solving the problems of low efficiency and poor generalization of traditional fault detection technology. BRIEF DESCRIPTION OF DRAWINGS

[0013] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0014] Figure 1 An application environment diagram of an electric drive system fault detection and state prediction method in an embodiment of the present application;

[0015] Figure 2 A flowchart of an electric drive system fault detection and state prediction method provided in an embodiment of the present application;

[0016] Figure 3 An Informer network model structure diagram provided in an embodiment of the present application;

[0017] Figure 4 An encoder structure diagram in an Informer network model provided in an embodiment of the present application;

[0018] Figure 5A decoder structure diagram in an Informer network model provided by an embodiment of the present application is shown in the figure;

[0019] Figure 6 A self / cross attention mechanism structure diagram in an Informer network model provided by an embodiment of the present application is shown in the figure;

[0020] Figure 7 A structure diagram of a computer device provided by an embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0022] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0023] The electric drive system fault detection and state prediction method provided by the embodiments of the present application can be applied in an application environment as shown in the figure. Figure 1 The terminal 102 communicates with the server 104 through a network. The data storage system can store data required to be processed by the server 104. The data storage system can be separately arranged, or integrated on the server 104, or placed on the cloud or other servers. The terminal 102 can send the current signal data to the server 104. After receiving the current signal data, the server 104 extracts features from the current signal data by using short-time Fourier transform to obtain a final two-dimensional time-frequency spectrum. The final two-dimensional time-frequency spectrum and the latest time-frequency data are input into an Informer network model to obtain frequency information and fault detection results in a future period of time. The server 104 can feed back the obtained frequency information and fault detection results in the future period of time to the terminal 102. In addition, in some embodiments, the electric drive system fault detection and state prediction method can also be realized by the server 104 or the terminal 102 alone, for example, the terminal 102 can directly process the current signal data, or the server 104 can obtain the current signal data from the data storage system and process it.

[0024] The terminal 102 can be, but is not limited to, various desktop computers, notebook computers and Internet of Things devices, etc. The server 104 can be realized by an independent server or a server cluster composed of multiple servers, and can also be a cloud server.

[0025] In one exemplary embodiment, as shown in Figure 2 , there is provided an electric drive system fault detection and state prediction method, which is executed by a computer device, specifically, can be executed by a terminal or a server alone, or by a terminal and a server together. In the embodiments of the present application, the method is applied to the server 104 in Figure 1 , and the following steps 201 to 203 are included. Among them:

[0026] Step 201, obtaining current signal data collected by a current sensor x(t) , the current sensor is located on the electric drive system.

[0027] Step 202, using short-time Fourier transform to extract features from the current signal data x(t) , to obtain a final two-dimensional time-frequency spectrum, the final two-dimensional time-frequency spectrum includes amplitude values at multiple time points.

[0028] Step 203, inputting the final two-dimensional time-frequency spectrum and the latest time-frequency data X_near(t,f) into an Informer network model to obtain frequency information and fault detection results for a future period of time; the latest time-frequency data is the time-frequency data at the final time point in the final two-dimensional time-frequency spectrum, as shown in Figure 3 , the Informer network model includes an embedding layer, an encoder, a decoder and an output layer connected in turn.

[0029] Further, in step 202, the short-time Fourier transform is used to extract features from the current signal data x(t) , to obtain a final two-dimensional time-frequency spectrum, which specifically includes the following contents.

[0030] Step 2021, preprocessing the current signal data x(t) to obtain preprocessed current signal data x_new (t) .

[0031] Step 2022, using short-time Fourier transform to extract features from the preprocessed current signal data, to obtain a final two-dimensional time-frequency spectrum.

[0032] Further, in step 2021, the current signal data x(t) is preprocessed to obtain preprocessed current signal data x_new(t) , wherein the preprocessing is denoising processing, and an adaptive filtering algorithm is used to remove high-frequency noise, which specifically includes:

[0033] obtaining an expected signal collected in the early stage which is not affected by noise d(t) , and simultaneously collecting an interference signal with noise as an input signalx1(t), Thus the order of the filter is determined N .

[0034] The desired signal and the interference signal are input into the initialized filter to determine the filter parameters. The filter weight is initialized first, and the weight W =[ w 0 ,w 1 ,w 2 ,…,w N-1 is set to a zero vector, and an appropriate step size parameter μ is taken, and then the filter output y(t) = Σ N-1 i=0 wi(t)x(t-i) is calculated, and the calculation error e(t)=d(t)-y(t) is obtained, and then the weight is updated according to the error signal: w i (t+1)=w i (t)+2μe(t)x(t-i) Finally, iteration is performed to gradually converge the weight, and the filter parameter is obtained as the parameter for removing high-frequency noise in this environment.

[0035] The initialized filter is updated based on the filter parameter to obtain an updated filter.

[0036] The current signal data x(t) is denoised using the updated filter to obtain preprocessed current signal data x_new(t) .

[0037] Further, in step 2022, the preprocessed current signal data is feature extracted using short-time Fourier transform to obtain a final two-dimensional time-frequency spectrogram. The preprocessed current signal is segmented into fixed-length time windows, and the short-time Fourier transform is performed on the signal in each time window to calculate the joint distribution of the signal in the frequency domain and the time domain, and a two-dimensional time-frequency spectrogram is obtained, and finally the two-dimensional time-frequency spectrogram is normalized to reduce the influence of amplitude difference on model training. Specifically, the following contents are included.

[0038] (1) Determine the time window parameter

[0039] In order to segment the preprocessed current signal data into fixed-length time windows, the length of the time window, the type of the window, and the overlap rate need to be set. The length of the time window l t is set to 256 ms, and the sampling rate f s is combined to calculate the number of sample points corresponding to each time window:

[0040] N samples = l t × f s 。

[0041] Select a suitable window function (after experiment, Hamming window is selected here) to reduce spectral leakage. Hamming window:

[0042] 。

[0043] To improve the time-frequency resolution, increase the partial overlap when sliding the time window, and set the overlap rate to 60%.

[0044] (2) Perform short-time Fourier transform

[0045] At this time, the signal is x_new(t) , which is divided into several windows with a length of N samples x i (t) The STFT (short-time Fourier transform) calculation formula of each time window is:

[0046] 。

[0047] X(t,f) Indicates the complex value at time t and frequency f . For each X(t,f) , take its amplitude X(t,f) to generate a two-dimensional time-frequency spectrum.

[0048] (3) Normalize the two-dimensional time-frequency spectrum

[0049] In order to eliminate the influence of amplitude difference on model training and make the feature distribution more uniform, the two-dimensional time-frequency spectrum is normalized by using zero-mean unit variance normalization to obtain the final two-dimensional time-frequency spectrum:

[0050] 。

[0051] Further, in step 203, the final two-dimensional time-frequency spectrum and the latest time-frequency data X_near(t,f) ), (i.e. the last single-point data of the two-dimensional time-frequency spectrum) are input into the Informer network model to obtain the frequency information and fault detection result of a future period of time, which specifically includes the following contents.

[0052] Step 2031, input the final two-dimensional time-frequency spectrum and the latest time-frequency data into the embedding layer to obtain first feature data.

[0053] ​Step 2032, input the first feature data into the encoder to obtain the second feature data.

[0054] Step 2033, input the first feature data and the second feature data into the decoder to obtain the third feature data.

[0055] Step 2034, input the third feature data into the output layer to obtain the frequency information and the fault detection result in the future period of time.

[0056] Further, the embedding layer is designed as a first fully connected layer and a position encoding layer connected in sequence; the first fully connected layer plays a role in changing the dimension of data, and the position encoding layer adds position information to each time step, and generates position encoding by using a sine function.

[0057] In step 2031, the final two-dimensional time-frequency spectrum and the latest time-frequency data are input into the embedding layer to obtain the first feature data, which specifically includes the following contents.

[0058] The final two-dimensional time-frequency spectrum and the latest time-frequency data are input into the first fully connected layer for latitude conversion processing, so as to realize latitude conversion and obtain output data of the first fully connected layer.

[0059] The output data of the first fully connected layer (the final two-dimensional time-frequency spectrum and the latest time-frequency data processed by the fully connected layer) are respectively input into the position encoding layer to obtain the first feature data; the first feature data is data with position encoding. Taking the former as an example, first, a blank tensor with the same size is constructed, and position encoding is generated for each position of the tensor by using a sine function, and the generated position encoding is superimposed with the input of this step to complete the processing process of the position encoding layer.

[0060] Further, as shown in Figure 4 , the encoder includes a first self-attention mechanism layer, a first normalization layer, a first feed-forward layer and a second normalization layer connected in sequence; the first self-attention mechanism layer adopts a sparse attention mechanism, and the normalization layer is used to adjust the size range of the output data.

[0061] In step 2032, the first feature data is input into the encoder to obtain the second feature data, which specifically includes:

[0062] The first feature data is input into the first self-attention mechanism layer for preliminary feature extraction to obtain output data of the first self-attention mechanism layer; wherein the self-attention mechanism in the encoder adopts a sparse attention mechanism, as shown in Figure 5 , the correlation between different time steps in the input sequence is calculated through the sparse attention mechanism. First, the input data is processed through three different fully connected layers to obtain three groups of Q, K and V data, then L2 norm is calculated for each query point of Q, and the highest proportion of kUsing these query points as key query points, a new sparse mask is constructed. A zero matrix of the same size as the Q matrix is ​​created, with 1s placed only at the key query point positions. The newly constructed matrix is ​​then processed using the following formula to obtain the preliminary features:

[0063] Preliminary features = softmax(Q·K) T )·V / (d K )^0.5.

[0064] Where, d K The dimension of the key K is used to scale the dot product to avoid excessively large values; softmax normalizes the values.

[0065] The output data of the first self-attention mechanism layer is input into the first normalization layer for normalization processing to obtain the output data of the first normalization layer.

[0066] The output data of the first normalization layer is input into the first feedback layer for further feature extraction, and the output data of the first feedback layer is obtained. The first feedback layer is composed of a fully connected layer, an activation layer, and a fully connected layer.

[0067] The output data of the first feedback layer is input into the second normalization layer for further normalization, resulting in the output data of the second normalization layer.

[0068] Determine whether the first preset number of iterations is met.

[0069] When the preset number of iterations is met, the output data of the second normalized layer is output.

[0070] If the preset number of iterations is not met, the output data of the second normalization layer replaces the first feature data, and the process returns to the step "Input the first feature data into the first self-attention mechanism layer for feature extraction to obtain the output data of the first self-attention mechanism layer".

[0071] In this embodiment, the above process is repeated twice before outputting the result to obtain more accurate information.

[0072] Furthermore, such as Figure 5 As shown, the decoder includes a third self-attention mechanism layer, a third normalization layer, a first cross-attention mechanism layer, a fourth normalization layer, a second pre-feedback layer, and a fifth normalization layer connected in sequence. The third self-attention mechanism layer and the first cross-attention mechanism layer are identical to the first self-attention mechanism layer, both employing a sparse attention mechanism, such as... Figure 6As shown, the difference lies in the different sources of K and V matrices, the source of the self-attention mechanism is the data of the encoder, and the source of the cross-attention mechanism is the feature data of the latest time-frequency data processed by the embedding layer. The second front feedback layer is the same as the first front feedback layer of the encoder, which is composed of a fully connected layer, an activation layer, and a fully connected layer.

[0073] In step 2033, the first feature data and the second feature data are input into the decoder to obtain third feature data, which specifically includes the following contents.

[0074] The second feature data is input into the third self-attention mechanism layer for feature extraction to obtain output data of the third self-attention mechanism layer, which is consistent with the processing mode of the first self-attention mechanism layer in the encoder and processes the time-frequency data features processed by the encoder.

[0075] The output data of the third self-attention mechanism layer is input into the third normalization layer for normalization processing to obtain output data of the third normalization layer.

[0076] The first feature data and the output data of the third normalization layer are input into the first cross-attention mechanism layer for feature extraction to obtain output data of the first cross-attention mechanism layer; wherein the sources of K and V matrices come from the feature data of the latest time-frequency data processed by the embedding layer, the source of Q matrix comes from the output data of the third normalization layer, and the rest is consistent with the processing mode of the self-attention mechanism, to obtain the final two-dimensional time-frequency spectrum and the comprehensive features of the processed latest time-frequency data.

[0077] The output data of the first cross-attention mechanism layer is input into the fourth normalization layer for normalization processing to obtain output data of the fourth normalization layer.

[0078] The output data of the fourth normalization layer is input into the second front feedback layer for further feature extraction to obtain output data of the second front feedback layer.

[0079] The output data of the second front feedback layer is input into the fifth normalization layer for normalization processing to obtain output data of the fifth normalization layer.

[0080] It is judged whether the second preset number of cycles is met.

[0081] When the preset number of cycles is met, the output data of the fifth normalization layer is output.

[0082] When the preset number of cycles is not met, the output data of the fifth normalization layer is replaced with the second feature data, and the step of inputting the second feature data into the third self-attention mechanism layer for feature extraction to obtain output data of the third self-attention mechanism layer is returned.

[0083] In this embodiment, the above process is cycled twice and then output to obtain deeper feature information.

[0084] Further, the output layer is two different fully connected layers. The output of the decoder is passed through two different fully connected layers respectively to obtain the frequency information and fault detection results of the future period of time respectively.

[0085] When training the Informer network model, the data used comes from the current signal data generated by the electric drive control system, which is collected in real time by a current sensor, and these data are labeled according to different fault states (IGBT / MOSFET switch fault, rectifier diode damage, DC bus fault). The data contains a training data set of 12000 samples, each sample includes 20 groups of current data, 10 groups of current prediction data, fault label (a classification label representing the current fault type), among which IGBT / MOSFET switch fault, rectifier diode damage, DC bus fault and no fault label data account for one fourth of the total data.

[0086] The electric drive system fault detection and state prediction method of the embodiment is different from the traditional detection method. The application does not rely on single time domain or frequency domain analysis, but combines the multi-level fusion strategy of short-time Fourier transform (SFTF) and Informer network, and realizes the recognition of fault mode and accurate prediction of future state through deep learning. The main innovations include:

[0087] 1) Fusion of SFTF and Informer network: Use SFTF to extract the time-frequency features of the current signal, combine the sparse attention mechanism of the Informer network, effectively capture the long-time dependence relationship and key mode in the signal, and significantly improve the analysis ability of complex sequence fault characteristics;

[0088] 2) Get rid of the limitations of traditional models: The application does not rely on complex physical modeling or prior assumptions of fault behavior, avoiding the risk of over-simplification and misjudgment of fault mechanism, and still maintaining high reliability and adaptability under various complex working conditions;

[0089] 3) Realize active early warning and accurate intervention: By combining the time-frequency features extracted by SFTF with the prediction output of the Informer network, not only can the fault type be detected in real time, but also the evolution trend and potential occurrence time of the fault can be predicted, providing active early warning capability for the system and supporting early intervention and optimized maintenance of faults.

[0090] Through the above technical breakthroughs, the embodiment significantly improves the diagnostic accuracy, efficiency and practicability in electric drive system fault detection, providing reliable support for intelligent industrial equipment health management.

[0091] The application further provides an application scenario of the electric drive system fault detection and state prediction method. Specifically, the electric drive system fault detection and state prediction method provided in the embodiment can be applied in an electric drive system fault detection and state prediction scenario. The electric drive system fault detection and state prediction scenario includes a fault detection and state prediction link and a fault detection result and future frequency information display link. The fault detection and state prediction link is configured to obtain the fault detection result and the future frequency information according to the current signal data. The fault detection result and future frequency information display link is configured to display the fault detection result and the future frequency information. The electric drive system fault detection and state prediction method provided in the embodiment belongs to the fault detection and state prediction link and the fault detection result and future frequency information display link.

[0092] In an exemplary embodiment, a computer device, which can be a server or a terminal, is provided, and an internal structure diagram of the computer device can be as shown in Figure 7 The computer device includes a processor, a memory, an input / output interface (I / O), and a communication interface. The processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The database of the computer device is configured to store processing data. The input / output interface of the computer device is configured to exchange information between the processor and external devices. The communication interface of the computer device is configured to communicate with external terminals through a network connection. The computer program is executed by the processor to implement an electric drive system fault detection and state prediction method.

[0093] Those skilled in the art can understand that Figure 7 The structure shown in the above embodiment is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. Specifically, the computer device can include more or fewer components than those shown in the diagram, or combine certain components, or have a different arrangement of components.

[0094] In an exemplary embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps in the above method embodiments.

[0095] In an exemplary embodiment, a computer readable storage medium is provided, storing a computer program which, when executed by a processor, implements the steps of any of the above method embodiments.

[0096] In an exemplary embodiment, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the steps of any of the above method embodiments.

[0097] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. Any reference to memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0098] The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a blockchain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.

[0099] Any technical features in the above embodiments can be combined, and for the sake of brevity, not all possible combinations are described above, however, it should be understood that the application encompasses all possible combinations of the technical features unless such a combination is not technically possible.

[0100] The principles and implementation manners of the present application are described herein by using specific examples, and the above embodiments are only used to help understand the method of the present application and its core idea; meanwhile, according to the idea of the present application, a person skilled in the art can make changes in specific implementation manners and application scopes. In conclusion, the content of the present specification should not be understood as a limitation of the present application.

Claims

1. A method for fault detection and state prediction of an electric drive system, characterized in that, The electric drive system fault detection and status prediction method includes: Acquire current signal data collected by a current sensor; the current sensor is located on the electric drive system; The current signal data is subjected to short-time Fourier transform for feature extraction to obtain the final two-dimensional time spectrum; the final two-dimensional time spectrum includes amplitude values ​​at multiple time points. The final two-dimensional time-frequency spectrum and the latest time-frequency data are input into the Informer network model to obtain frequency information and fault detection results for a future period of time. The latest time-frequency data is the time-frequency data at the final moment in the final two-dimensional time-frequency spectrum. The Informer network model includes an embedding layer, an encoder, a decoder, and an output layer connected in sequence.

2. The method for fault detection and status prediction of an electric drive system according to claim 1, characterized in that, Feature extraction is performed on the current signal data using short-time Fourier transform to obtain the final two-dimensional time-frequency spectrum, specifically including: The current signal data is preprocessed to obtain preprocessed current signal data; The preprocessed current signal data is subjected to feature extraction using short-time Fourier transform to obtain the final two-dimensional time spectrum.

3. The method for fault detection and status prediction of an electric drive system according to claim 2, characterized in that, The current signal data is preprocessed to obtain preprocessed current signal data, specifically including: To acquire the desired signal unaffected by noise and the interference signal with noise; The desired signal and the interference signal are input into the initialized filter to determine the filtering parameters; The initialized filter is updated based on the filtering parameters to obtain the updated filter; The updated filter is used to denoise the current signal data to obtain preprocessed current signal data.

4. The method for fault detection and status prediction of an electric drive system according to claim 2, characterized in that, Feature extraction is performed on the preprocessed current signal data using short-time Fourier transform to obtain the final two-dimensional time-frequency spectrum, specifically including: Based on the preset time window length, window type and overlap rate, the preprocessed current signal data is segmented to obtain current signal data in multiple time windows. Perform a short-time Fourier transform on the current signal data within each time window, calculate the joint distribution of the current signal data in the frequency domain and time domain, and obtain an initial two-dimensional time spectrum. The initial two-dimensional time spectrum is normalized using the zero-mean unit variance normalization method to obtain the final two-dimensional time spectrum.

5. The method for fault detection and status prediction of an electric drive system according to claim 1, characterized in that, The final two-dimensional time-frequency spectrum and the latest time-frequency data are input into the Informer network model to obtain frequency information and fault detection results for a future period, specifically including: The final two-dimensional time-frequency spectrum and the latest time-frequency data are input into the embedding layer to obtain the first feature data; The first feature data is input into the encoder to obtain the second feature data; The first feature data and the second feature data are input into the decoder to obtain the third feature data; The third feature data is input into the output layer to obtain frequency information and fault detection results for a future period of time.

6. The method for fault detection and status prediction of an electric drive system according to claim 5, characterized in that, The embedding layer includes a first fully connected layer and a position coding layer connected in sequence; the position coding layer includes a sine function, which is used to perform position coding on the output data of the first fully connected layer. The final two-dimensional time-frequency spectrum and the latest time-frequency data are input into the embedding layer to obtain the first feature data, which specifically includes: The final two-dimensional time-frequency spectrum and the latest time-frequency data are input into the first fully connected layer for dimensional transformation processing to obtain the output data of the first fully connected layer; The output data of the first fully connected layer is input into the position encoding layer to obtain the first feature data; the first feature data is data with position encoding.

7. The method for fault detection and status prediction of an electric drive system according to claim 5, characterized in that, The encoder includes a first self-attention mechanism layer, a first normalization layer, a first feedback layer, and a second normalization layer connected in sequence; the first self-attention mechanism layer adopts a sparse attention mechanism. The first feature data is input into the encoder to obtain the second feature data, specifically including: The first feature data is input into the first self-attention mechanism layer for feature extraction, and the output data of the first self-attention mechanism layer is obtained. The output data of the first self-attention mechanism layer is input into the first normalization layer for normalization processing to obtain the output data of the first normalization layer; The output data of the first normalization layer is input into the first front feedback layer for feature extraction, and the output data of the first front feedback layer is obtained. The output data of the first feedback layer is input into the second normalization layer for normalization processing to obtain the output data of the second normalization layer; Determine whether the first preset number of iterations is met; When the preset number of iterations is met, the output data of the second normalization layer is output. If the preset number of iterations is not met, the output data of the second normalization layer replaces the first feature data, and the process returns to the step "input the first feature data into the first self-attention mechanism layer for feature extraction to obtain the output data of the first self-attention mechanism layer".

8. The method for fault detection and status prediction of an electric drive system according to claim 5, characterized in that, The decoder includes a third self-attention mechanism layer, a third normalization layer, a first cross-attention mechanism layer, a fourth normalization layer, a second feedback layer, and a fifth normalization layer connected in sequence; both the third self-attention mechanism layer and the first cross-attention mechanism layer adopt sparse attention mechanism; The first feature data and the second feature data are input into the decoder to obtain the third feature data, which specifically includes: The second feature data is input into the third self-attention mechanism layer for feature extraction to obtain the output data of the third self-attention mechanism layer. The output data of the third self-attention mechanism layer is input into the third normalization layer for normalization processing to obtain the output data of the third normalization layer; The first feature data and the output data of the third normalization layer are input into the first cross-attention mechanism layer for feature extraction, and the output data of the first cross-attention mechanism layer is obtained. The output data of the first cross-attention mechanism layer is input into the fourth normalization layer for normalization processing to obtain the output data of the fourth normalization layer; The output data of the fourth normalization layer is input into the second front feedback layer for feature extraction, and the output data of the second front feedback layer is obtained. The output data of the second feedback layer is input into the fifth normalization layer for normalization processing to obtain the output data of the fifth normalization layer; Determine whether the second preset number of iterations is met; When the preset number of iterations is met, the output data of the fifth normalization layer is output. If the preset number of iterations is not met, the output data of the fifth normalization layer replaces the second feature data, and the process returns to the step "input the second feature data into the third self-attention mechanism layer for feature extraction to obtain the output data of the third self-attention mechanism layer".

9. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the electric drive system fault detection and state prediction method according to any one of claims 1-8.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the electric drive system fault detection and status prediction method as described in any one of claims 1-8.

Images