Fault diagnosis method, device and system for driving motor of electric vehicle

Abstract

The invention discloses a fault diagnosis method, device and system for an electric vehicle driving motor, relates to the technical field of motor diagnosis, and mainly aims to reduce consumption of vehicle end computing resources and improve the efficiency of motor fault diagnosis. According to the main technical scheme, the method comprises the steps of receiving stator current data corresponding to a permanent magnet synchronous motor sent by a vehicle end; the stator current data are input into a trained multi-classification fault diagnosis model, a diagnosis result corresponding to the permanent magnet synchronous motor is obtained, the multi-classification fault diagnosis model is a 1D convolutional neural network model, and the diagnosis result comprises the occurrence probability corresponding to each fault category; determining a fault result corresponding to the permanent magnet synchronous motor according to the occurrence probability corresponding to each fault category; and synchronizing the fault result to the vehicle end, so that the vehicle end triggers a corresponding early warning operation. The method is used for fault diagnosis of the permanent magnet synchronous motor.

Description

Technical Field

[0001] The present application relates to the technical field of motor diagnosis, and in particular to a fault diagnosis method, device and system for a drive motor of an electric vehicle. Background Art

[0002] Permanent magnet synchronous motor is one of the most commonly used motor types in electric vehicle drive systems. It has the advantages of high efficiency, high power density and high torque, and can provide better power performance and driving range. However, various faults will inevitably occur during long-term continuous operation of the motor. These faults will seriously affect the reliability and safety of the motor operation. Therefore, the fault detection and diagnosis technology of permanent magnet synchronous motor is particularly important.

[0003] At present, traditional motor fault diagnosis is usually deployed on the vehicle side, and generally the corresponding fault diagnosis model is deployed for a specific fault. In other words, if you want to conduct a comprehensive diagnosis of various motor faults, you need to integrate multiple fault diagnosis models on the vehicle side, which will not only consume a large amount of computing resources on the vehicle side, but also lead to low efficiency of fault diagnosis as computing resources are continuously consumed, affecting the timeliness and accuracy of motor fault diagnosis. Summary of the invention

[0004] In view of the above problems, the present application provides a method, device and system for fault diagnosis of an electric vehicle drive motor, the main purpose of which is to reduce the consumption of vehicle-side computing resources and improve the efficiency of motor fault diagnosis.

[0005] In order to solve the above technical problems, this application proposes the following solutions:

[0006] In a first aspect, the present application provides a fault diagnosis method for an electric vehicle drive motor, which is applied to the cloud, and the method comprises:

[0007] Receive the stator current data of the permanent magnet synchronous motor sent by the vehicle end;

[0008] Inputting the stator current data into a trained multi-classification fault diagnosis model to obtain a diagnosis result corresponding to the permanent magnet synchronous motor, wherein the diagnosis result includes an occurrence probability corresponding to each fault category;

[0009] Determine a fault result corresponding to the permanent magnet synchronous motor according to the occurrence probability corresponding to each of the fault categories;

[0010] The fault result is synchronized to the vehicle end so that the vehicle end can trigger a corresponding early warning operation.

[0011] In a second aspect, the present application provides a fault diagnosis method for a drive motor of an electric vehicle, which is applied to a vehicle side, and the method comprises:

[0012] Collect stator current data corresponding to the permanent magnet synchronous motor;

[0013] Sending the stator current data to a cloud so that the cloud can perform multi-classification fault diagnosis on the permanent magnet synchronous motor according to the stator current data;

[0014] Determining whether the permanent magnet synchronous motor has a fault tendency based on the fault result corresponding to the permanent magnet synchronous motor received from the cloud feedback;

[0015] If so, a warning operation corresponding to the fault result is triggered.

[0016] In a third aspect, the present application provides a fault diagnosis device for an electric vehicle drive motor, which is applied to the cloud, and the device includes:

[0017] A receiving unit, used for receiving stator current data corresponding to the permanent magnet synchronous motor sent by the vehicle end;

[0018] A diagnosis unit, used for inputting the stator current data obtained by the receiving unit into a trained multi-classification fault diagnosis model to obtain a diagnosis result corresponding to the permanent magnet synchronous motor, wherein the diagnosis result includes an occurrence probability corresponding to each fault category;

[0019] a determination unit, configured to determine a fault result corresponding to the permanent magnet synchronous motor according to the occurrence probability corresponding to each of the fault categories obtained by the diagnosis unit;

[0020] The sending unit is used to synchronize the fault result obtained by the determination unit to the vehicle end, so that the vehicle end can trigger a corresponding early warning operation.

[0021] In a fourth aspect, the present application provides a fault diagnosis device for a drive motor of an electric vehicle, which is applied to a vehicle end, and the device comprises:

[0022] A collection unit, used for collecting stator current data corresponding to the permanent magnet synchronous motor;

[0023] A sending unit, used to send the stator current data obtained by the acquisition unit to a cloud, so that the cloud can perform multi-classification fault diagnosis on the permanent magnet synchronous motor according to the stator current data;

[0024] a processing unit, configured to determine whether the permanent magnet synchronous motor has a fault tendency based on a fault result corresponding to the permanent magnet synchronous motor received from the cloud feedback;

[0025] The early warning unit is used to trigger an early warning operation corresponding to the fault result if the processing unit determines that the permanent magnet synchronous motor has a fault tendency.

[0026] In a fifth aspect, the present application provides a fault diagnosis system for an electric vehicle drive motor, the system comprising: a cloud end and a vehicle end,

[0027] When the cloud performs fault diagnosis of the electric vehicle drive motor, the electric vehicle drive motor fault diagnosis method of the first aspect is executed;

[0028] When the vehicle end performs fault diagnosis of the electric vehicle drive motor, the electric vehicle drive motor fault diagnosis method of the second aspect is executed.

[0029] In order to achieve the above-mentioned purpose, according to the sixth aspect of the present application, a storage medium is provided, wherein the storage medium includes a stored program, wherein when the program is running, the device where the storage medium is located is controlled to execute the fault diagnosis method of the electric vehicle drive motor of the first aspect or the second aspect mentioned above.

[0030] In order to achieve the above-mentioned purpose, according to the fourth aspect of the present application, a processor is provided, and the processor is used to run a program, wherein when the program is running, the fault diagnosis method of the electric vehicle drive motor of the above-mentioned first aspect or second aspect is executed.

[0031] By means of the above technical scheme, the present application provides a method, device and system for fault diagnosis of an electric vehicle drive motor. When it is necessary to perform fault diagnosis on a permanent magnet synchronous motor, after the cloud receives the stator current data corresponding to the permanent magnet synchronous motor sent by the vehicle end, the stator current data is input into the trained multi-classification fault diagnosis model to obtain the probability of occurrence of each fault category of the permanent magnet synchronous motor in the multi-classification fault, and then the fault result corresponding to the permanent magnet synchronous motor is determined according to the probability of occurrence corresponding to each fault category, and the fault result is synchronized to the vehicle end, so that the vehicle end triggers the corresponding early warning operation. Through the technical scheme provided by the present application, the motor fault diagnosis can be effectively transferred from the vehicle end to the cloud, thereby reducing the consumption of computing resources on the vehicle end, and the cloud end does not need to deploy the corresponding fault diagnosis model corresponding to a specific fault, but performs fault diagnosis on the permanent magnet synchronous motor through the multi-classification fault diagnosis model, which can identify multiple potential fault types, not just limited to a single fault, effectively reducing the number of model deployments, not only ensuring the efficiency of motor fault diagnosis, but also improving the timeliness and accuracy of fault diagnosis, and the diagnosis result also includes the probability of occurrence of each fault category, making the vehicle end early warning operation more intelligent and accurate.

[0032] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present application. Also, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0034] Figure 1 A flow chart of a fault diagnosis method for an electric vehicle drive motor applied in the cloud provided in an embodiment of the present application is shown;

[0035] Figure 2 A flowchart of another method for diagnosing a fault of an electric vehicle drive motor applied to a vehicle side provided by an embodiment of the present application is shown;

[0036] Figure 3 A block diagram of a fault diagnosis device for an electric vehicle drive motor applied in the cloud provided in an embodiment of the present application is shown;

[0037] Figure 4 A block diagram of another fault diagnosis device for an electric vehicle drive motor applied in the cloud provided in an embodiment of the present application is shown;

[0038] Figure 5 A block diagram of a fault diagnosis device for an electric vehicle drive motor applied to a vehicle side provided by an embodiment of the present application is shown;

[0039] Figure 6 A block diagram of another fault diagnosis device for an electric vehicle drive motor applied to a vehicle side provided by an embodiment of the present application is shown;

[0040] Figure 7 A block diagram of a fault diagnosis system for an electric vehicle drive motor provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0041] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0042] Traditional motor fault diagnosis is usually deployed on the vehicle side, and generally the corresponding fault diagnosis model is deployed for a specific fault. In other words, if you want to conduct a comprehensive diagnosis of various motor faults, you need to integrate multiple fault diagnosis models on the vehicle side, which will not only consume a large amount of computing resources on the vehicle side, but also lead to low efficiency of fault diagnosis as computing resources are continuously consumed, affecting the timeliness and accuracy of motor fault diagnosis.

[0043] After research, the inventor found that since the computing resources on the vehicle side are limited, and the cloud can provide almost unlimited computing and storage capabilities, the motor fault diagnosis can be transferred to the cloud. That is, the stator current data of the permanent magnet synchronous motor is collected through the vehicle side, and the stator current data is transmitted to the cloud. The cloud deploys a multi-classification fault model that can identify each fault category and give the corresponding probability of occurrence through the stator current data, and the final fault result of the permanent magnet synchronous motor determined based on the diagnosis result of the model is synchronized to the vehicle side, so that the vehicle side can perform early warning according to the fault result. In this way, the motor fault diagnosis can be effectively transferred from the vehicle side to the cloud, thereby reducing the consumption of computing resources on the vehicle side, and the cloud does not need to deploy the corresponding fault diagnosis model for a specific fault, which effectively reduces the number of model deployments, further reduces the consumption of cloud computing resources, and ensures the efficiency of motor fault diagnosis, that is, the timeliness and accuracy of motor fault diagnosis are improved through vehicle-cloud collaboration.

[0044] Based on the above considerations, the embodiment of the present application provides a fault diagnosis method for an electric vehicle drive motor, which can reduce the consumption of vehicle-side computing resources and improve the efficiency of motor fault diagnosis. The specific execution steps are as follows: Figure 1 As shown, it is applied to the cloud, including:

[0045] 101. Receive the stator current data corresponding to the permanent magnet synchronous motor sent by the vehicle end.

[0046] In this step, the stator current data is collected by a current sensor connected or fixed to the stator of the vehicle's permanent magnet synchronous motor and sent to the cloud through the vehicle end. The current sensor is a device for measuring current, which can convert the current signal into a digital signal output, and the stator current data refers to a current signal sequence, which is composed of current signals continuously sampled at a preset sampling frequency within a preset total sampling time, arranged from left to right in the order of sampling time. Through the current signal sequence, the change of the stator current over time can be captured, which is helpful for subsequent fault detection and diagnosis.

[0047] The collection of the stator current data can be performed according to the set time measurement or mileage measurement, that is, the data is collected once every certain period of time or every certain mileage. The time measurement or mileage measurement can be customized by the equipment manufacturer or the user to set an initial time measurement or initial mileage measurement; or a corresponding prediction model can be set in the cloud to predict a corrected time measurement or corrected mileage measurement, and fed back to the vehicle side for execution. This embodiment does not limit this.

[0048] 102. Input the stator current data into the trained multi-classification fault diagnosis model to obtain the corresponding diagnosis result of the permanent magnet synchronous motor.

[0049] Among them, the diagnosis result includes the probability of occurrence corresponding to each fault category. In this step, after receiving the stator current data, the cloud will load a pre-trained multi-classification fault diagnosis model. The multi-classification fault diagnosis model can adopt a 1D convolutional neural network model, a long short-term memory network model or a gated recurrent unit and other models with classification and prediction capabilities. It is trained on the basis of a large number of historical fault samples, covering various fault categories that can be identified by the data performance of the stator current, and can predict the probability of occurrence of each fault category, which is used to indicate the possibility of occurrence of the fault category. By inputting the stator current data into the trained multi-classification fault diagnosis model, the multi-classification fault diagnosis model can output the probability of occurrence corresponding to each fault category as the corresponding diagnosis result of the permanent magnet synchronous motor.

[0050] In addition, the multi-classification fault diagnosis model can also be composed of multiple fault diagnosis models for a specific fault, including but not limited to an analytical mathematical model, a magnetic equivalent circuit model (Magnetic Equivalent Circuit Model, MEC Model) and a digital simulation model (Digital Simulation Model), etc. This embodiment does not limit this.

[0051] 103. Determine the fault result corresponding to the permanent magnet synchronous motor according to the occurrence probability corresponding to each fault category.

[0052] Since the diagnosis result includes the probability of occurrence corresponding to each fault category, usually, the fault category with the highest probability of occurrence is the most likely potential fault of the permanent magnet synchronous motor. Therefore, the fault category with the highest probability of occurrence can be regarded as the potential fault category. In some special cases, multiple fault categories with high probability of occurrence may also appear at the same time, that is, the permanent magnet synchronous motor currently has multiple potential faults that may occur, that is, the permanent magnet synchronous motor may currently have a composite fault or a fault state with high uncertainty. In order to enable the vehicle end to know the current fault state of the permanent magnet synchronous motor in real time, the cloud can pre-set a relatively high probability threshold and compare the probability of occurrence of each fault category with it. When there are multiple fault categories higher than the probability threshold, they are all regarded as potential fault categories. The fault results are jointly generated by combining the potential fault category and the corresponding probability of occurrence of the potential fault category, so that the vehicle end can know the potential faults with high probability of the permanent magnet synchronous motor in time, which is conducive to subsequent early warning and maintenance.

[0053] 104. Synchronize the fault result to the vehicle side so that the vehicle side can trigger corresponding warning operations.

[0054] After determining the fault result, the cloud transmits it back to the vehicle, and the transmission can be encrypted to ensure data security. After receiving the fault result, the vehicle triggers the corresponding warning operation according to the preset rules, including but not limited to reminding the driver to take over, automatically pulling over, the dashboard warning light turns on or audio warnings, etc. In addition, the cloud can also synchronize the fault results to the mobile phone APP or user service center, etc. Among them, the mobile phone APP can inform the user through information pop-up prompts and sound prompts. The user service center is a service platform provided to users, aiming to solve problems, provide help and support for users. When the user service platform receives the fault result sent by the cloud, the staff of the user service platform can remind the user of the current fault status of the permanent magnet synchronous motor through telephone contact and other methods, and communicate the inspection / repair time, etc.

[0055] Based on the above Figure 1It can be seen from the implementation method that when it is necessary to diagnose the fault of the permanent magnet synchronous motor, after the cloud receives the stator current data corresponding to the permanent magnet synchronous motor sent by the vehicle, the stator current data is input into the trained multi-classification fault diagnosis model to obtain the probability of occurrence of each fault category of the permanent magnet synchronous motor in the multi-classification fault, and then the fault result corresponding to the permanent magnet synchronous motor is determined according to the probability of occurrence of each fault category, and the fault result is synchronized to the vehicle, so that the vehicle triggers the corresponding early warning operation. Through the technical solution provided by the present application, the motor fault diagnosis can be effectively transferred from the vehicle to the cloud, thereby reducing the consumption of computing resources on the vehicle side, and the cloud does not need to deploy the corresponding fault diagnosis model corresponding to a specific fault, but uses the multi-classification fault diagnosis model to diagnose the permanent magnet synchronous motor, which can identify a variety of potential fault types, rather than just a single fault, effectively reducing the number of model deployments, not only ensuring the efficiency of motor fault diagnosis, but also improving the comprehensiveness and accuracy of fault diagnosis, and the diagnosis result also includes the probability of occurrence of each fault category, making the vehicle-side early warning operation more intelligent and accurate.

[0056] Further, in combination with the description of the mileage measurement in step 101, since the failure of the permanent magnet synchronous motor usually occurs during the driving process of the vehicle, in this embodiment, it is preferred to control the collection of the stator current data at the vehicle end according to the mileage measurement. The mileage measurement can be an initial mileage measurement or a corrected mileage measurement. The corrected mileage measurement can undoubtedly better realize the dynamic collection of the stator current data. Therefore, the following steps 101A-101F can be performed for the construction of the prediction model and the determination of the corrected mileage measurement based on the prediction model:

[0057] 101A. Extract historical fault results of a specified number of times and historical mileage measurements corresponding to the historical fault results.

[0058] In this step, since the diagnosis results of each diagnosis of the permanent magnet synchronous motor will be synchronized from the cloud to the vehicle side, and the vehicle side is also controlled according to the mileage measurement when collecting stator current data. Therefore, both the cloud and the vehicle side will record the historical fault results of the permanent magnet synchronous motor and its corresponding historical mileage measurement. The historical fault results of the permanent magnet synchronous motor for a specified number of times and its corresponding historical mileage measurement can be obtained from the storage medium of the vehicle side or the database of the cloud. The specified number of times can be customized according to needs, such as the last 10 times, etc. In order to execute step 101B.

[0059] 101B. Convert each historical fault result and historical mileage measurement into graph sequence data respectively, and train a graph convolution-long short-term memory neural network model based on the graph sequence data of a specified number of times.

[0060] In this step, the historical fault results and the corresponding mileage are converted into a graph data structure. Each node can represent a fault detection event, and the node attributes include the fault category and its probability, mileage, etc. The nodes are connected through time series relationships to form graph sequence data. A combined model of a graph convolutional network and a long short-term memory network, namely a graph convolution-long short-term memory neural network model, is selected for training. The graph convolutional network is used to capture the correlation between nodes in the graph sequence data, and the long short-term memory network is used to memorize long-term dependencies and capture the impact of historical mileage metrics and historical fault results on the current or future time series characteristics. Since the probability of occurrence of each fault category is affected by both historical fault results and the mileage accumulation effect, the model obtained by combining the two can simultaneously process the spatial relationship in the graph structure and the time dependency in the time series, that is, the relationship between the vehicle's fault probability and the mileage metric is obtained. Therefore, the model is trained using the converted graph sequence data, and the model parameters are adjusted to minimize the prediction error, so that the model can accurately predict the future mileage metric.

[0061] 101C. Use the trained graph convolution-long short-term memory neural network model as the mileage prediction model.

[0062] In this step, the trained graph convolution-long short-term memory neural network model is deployed on the cloud as a mileage prediction model to predict the next mileage measurement when the vehicle-side collects the stator current data corresponding to the permanent magnet synchronous motor.

[0063] 101D. Obtain the mileage measurement corresponding to the current time when the vehicle end collects the stator current data corresponding to the permanent magnet synchronous motor.

[0064] In this step, since the vehicle side will simultaneously record the current mileage measurement, such as 10 kilometers, 8 kilometers, etc., when collecting motor data. And the current mileage measurement, whether it is the corrected mileage measurement or the initial mileage measurement, will be recorded in the vehicle side's storage medium or the cloud database, so the current mileage measurement can be directly obtained.

[0065] 101E. Input the occurrence probability corresponding to each fault category and the current mileage measurement into the trained mileage prediction model to obtain the prediction result.

[0066] Among them, the prediction result is used to characterize the next mileage measurement when the vehicle collects the stator current data corresponding to the permanent magnet synchronous motor. During the actual prediction, each fault category corresponding to the current mileage measurement is also obtained from the vehicle's storage medium or the cloud database, and the current mileage measurement and the probability of occurrence of each fault category corresponding to the current mileage measurement are input into the trained mileage prediction model to obtain the next mileage measurement when the vehicle collects the stator current data corresponding to the permanent magnet synchronous motor, so that the next mileage measurement is predicted based on the current fault result, that is, the best time to collect the stator current data corresponding to the permanent magnet synchronous motor next time is determined, which effectively improves the timeliness of stator current data collection to ensure that the subsequent vehicle-side warning is more timely.

[0067] 101F. The next mileage measurement is used as the corrected mileage measurement and fed back to the vehicle end.

[0068] In this step, the corrected mileage measurement may be sent to the vehicle end together with the subsequently obtained fault result, or may be sent to the vehicle end alone. This embodiment does not limit this.

[0069] Further, in combination with the specific description of using the multi-classification fault diagnosis model in step 102, the following steps 102A-102D may be performed for its construction process:

[0070] 102A. Preset the fault category corresponding to the permanent magnet synchronous motor.

[0071] In this step, it should be noted that the fault category is set according to the faults that affect the stator current in different forms. It at least includes no fault, stator phase winding short circuit, full phase open circuit, magnet damage, shaft bending, bolt loosening, bearing fault, air gap eccentricity and demagnetization fault. The fault category can be specifically determined by detecting whether all existing faults of the permanent magnet synchronous motor have an impact on its stator current. After determining the fault category, a unique identifier or label can be assigned to each fault category for use in subsequent steps.

[0072] 102B. Extract target fault samples that match the fault category from historical fault samples of the permanent magnet synchronous motor, and mark a category label for each target fault sample based on the fault category.

[0073] In this step, historical fault samples can be obtained from the operation record of the permanent magnet synchronous motor. These samples include samples of the motor in a normal state and samples of an abnormal state. Among these samples, all fault categories involved in step 102A (including samples without faults, i.e., in a normal state) are extracted, and the extracted samples are marked with the unique identifier or label assigned in step 102, so as to perform step 102C.

[0074] 102C. Select target fault samples with different category labels as training samples according to preset ratios, and train a 1D convolutional neural network model based on the training samples.

[0075] In this step, since it is necessary to train the selected 1D convolutional neural network model based on the training samples, and the training samples usually include a training set and a validation set, they can be selected according to a preset ratio to ensure the representativeness of each fault category in the training samples. After determining the training samples, they need to be normalized, standardized or feature extracted for model processing. A 1D convolutional neural network model is trained using the training set. The 1D convolutional neural network model is suitable for processing time series data, such as a current signal sequence corresponding to stator current data. It can capture local features in the current signal sequence, so that the degree of matching with each fault category can be quickly compared in subsequent use, that is, the probability of occurrence of each fault category. After training with the training set, the hyperparameters of the model, such as learning rate, batch size, convolution kernel size, number of filters, etc., are adjusted through cross-validation with the validation set to optimize the model performance, and the classification effect of the model is evaluated using indicators such as accuracy, recall rate, F1 score, etc., and when the evaluation results meet the requirements, step 102D is executed.

[0076] 102D. Use the trained 1D convolutional neural network model as a multi-classification fault diagnosis model.

[0077] In this step, the trained and verified 1D convolutional neural network model is deployed in the cloud as a multi-classification fault diagnosis model to identify potential faults and their occurrence probabilities based on the stator current data sent by the vehicle in real time. In addition, in order to ensure the accuracy of the model, the model can be retrained regularly to adapt to new fault categories or improve performance. For example, the stator current data is a current signal sequence I seq ,The training data can be obtained by OEM or supplier through collection and ,labeling.

[0078] A 1D convolution-based neural network is used to classify the fault types of permanent magnet synchronous motors. The 1D convolution-based neural network consists of several 1D convolution modules, a fully connected module and an output module. The 1D convolution module is constructed by sequentially constructing 1D convolution layers, pooling layers and batch normalization layers.

[0079] The input of the 1D convolutional layer is the current signal sequence I seq , whose output consists of a length of L window The sliding window has a step size of s window is 1 in the current signal sequence I seq Each convolution operation is followed by an activation operation, and the activation function in the activation operation can be the ReLU function.

[0080] The mathematical expression of the output of a 1D convolutional layer is as follows:

[0081]

[0082] Among them, z j is the output of the jth 1D convolution module; z i is the eigenvalue calculated after the i-th convolution operation and activation operation, where i∈[1, L sample -L window +1]; ω is the length L window The one-dimensional weight matrix of ; b is the bias; ReLU is the nonlinear activation function.

[0083] The pooling layer can choose the maximum pooling operation, that is, the length is L window，p The sliding window has a step size of s window，p The output sequence z of the 1D convolutional layer j Slide up, and take the maximum value in the window as the output each time. The batch normalization layer normalizes the output of the pooling layer to meet the specified mean and variance. The fully connected module is composed of several fully connected layers stacked together.

[0084] The input of the fully connected module is the output of the batch normalization layer in the last 1D convolution module. Its output module includes an output layer. The output layer uses a softmax function to classify motor faults. The categories of motor faults may include but are not limited to: ① no fault, ② stator phase winding short circuit, ③ all-phase open circuit, ④ magnet damage, ⑤ shaft bending, ⑥ bolt loosening, ⑦ bearing fault, ⑧ air gap eccentricity, ⑨ demagnetization fault. Additionally, the output layer outputs L Category The category corresponding to the maximum value is the result of fault diagnosis; L Category is the number of motor fault categories.

[0085] Further, in combination with the determination of the fault result in step 102, the following steps 103A-103D may be performed:

[0086] 103A: Check whether the occurrence probabilities are all less than a preset probability threshold.

[0087] Among them, the preset probability threshold is used to distinguish between high-probability fault categories and low-probability fault categories. Specifically, it can be set according to the equipment manufacturer or user's custom settings, or it can be calculated from the data performance of the occurrence probability of each fault category in the historical diagnosis results in combination with a statistical algorithm. This embodiment does not limit this. After the occurrence probability of each fault category is obtained, each occurrence probability can be compared with the preset probability threshold in turn to determine whether each occurrence probability is less than the preset probability threshold. If so, step 103B is executed, otherwise step 103C is executed.

[0088] 103B. The fault category with the highest probability of occurrence is regarded as the potential fault category.

[0089] Since each occurrence probability is less than the preset probability threshold, it means that the occurrence probability of each fault category is low at this time. Therefore, it is only necessary to take the fault category with the highest occurrence probability as the potential fault category.

[0090] 103C. The fault category whose occurrence probability is greater than or equal to a preset probability threshold is regarded as a potential fault category.

[0091] Since there are fault categories with a probability greater than or equal to the preset probability threshold among all fault categories, and there may be multiple fault categories. In other words, the permanent magnet synchronous motor may have multiple fault categories with a high probability of occurrence. Therefore, at this time, all fault categories with a probability greater than or equal to the preset probability threshold can be regarded as potential fault categories, so that the vehicle end can promptly know the potential faults with a higher probability of the permanent magnet synchronous motor, so as to check whether there are compound faults or fault states with high uncertainty.

[0092] 103D. Generate a fault result based on the potential fault category and the occurrence probability corresponding to the potential fault category.

[0093] It should be noted that the fault result includes one or more potential fault categories and their corresponding occurrence probabilities, which can enable the vehicle end to know the potential faults with a higher probability of the permanent magnet synchronous motor in a timely manner, which is conducive to subsequent early warning and maintenance.

[0094] For example, it is assumed that there are four possible fault categories and their occurrence probabilities are:

[0095] Fault A: 25%; Fault B: 15%; Fault C: 35%; Fault D: 20%.

[0096] If the preset probability threshold is set to 30% at this time, then all probabilities are checked and it is found that the probabilities of faults A / B / D are all lower than 30%, and the probability of fault C is 35%, which exceeds the threshold of 30%. Then, since there is a fault category (fault C) with a probability higher than or equal to the preset probability threshold, the branch of "whether all probabilities are lower than the preset threshold" is not satisfied, so fault C is regarded as a potential fault category. And the fault result of the permanent magnet synchronous motor is generated based on the potential fault category of "fault C" and its accompanying probability of occurrence of "35%".

[0097] Correspondingly, the embodiment of the present application provides another method for fault diagnosis of an electric vehicle drive motor, by which the consumption of vehicle-side computing resources can be reduced and the efficiency of motor fault diagnosis can be improved. The specific execution steps are as follows: Figure 2 As shown, it is applied to the vehicle side, including:

[0098] 201. Collect stator current data corresponding to the permanent magnet synchronous motor.

[0099] In this step, the stator current data is collected by a current sensor connected or fixed to the stator of the vehicle permanent magnet synchronous motor. It should be noted that when selecting the current sensor, factors such as the rated current, working environment temperature, measurement accuracy and output mode of the permanent magnet synchronous motor must be considered to ensure that the final selected current sensor can stably and accurately collect the stator current data.

[0100] The collection of the stator current data can be performed according to the set time measurement or mileage measurement, that is, the data is collected once every certain period of time or every certain mileage. The time measurement or mileage measurement can be customized by the equipment manufacturer or the user to set an initial time measurement or initial mileage measurement; or a corresponding prediction model can be set in the cloud to predict a corrected time measurement or corrected mileage measurement, and fed back to the vehicle side for execution. This embodiment does not limit this.

[0101] It should be emphasized that the stator current data refers to a current signal sequence, which is composed of current signals continuously sampled at a preset sampling frequency within a preset total sampling time, arranged from left to right in the order of sampling time. Through the current signal sequence, the change of stator current over time can be captured, which is helpful for fault detection and diagnosis in the cloud.

[0102] Specifically, the collected stator current value is output as an original current signal sequence I seq，org . Original current signal sequence I seq，org Length L sample The calculation method is as follows:

[0103] L sample =T sample fsample

[0104] Among them, T sample is the total sampling time (in seconds), that is, the time from the start of sampling to the end of sampling; f ample is the sampling frequency (in Hz).

[0105] Original current signal sequence I seq，org It can be expressed as The elements are arranged from left to right in the order of sampling time.

[0106] 202. Send the stator current data to the cloud so that the cloud can perform multi-classification fault diagnosis on the permanent magnet synchronous motor according to the stator current data.

[0107] In this step, the collected stator current data can be stored in the storage medium on the vehicle side, and then uploaded to the cloud by the storage medium through the vehicle-mounted communication device connected to the cloud communication such as TBOX (Telematics Box), which can upload the stator current data to the cloud through communication methods such as GPS, GPRS, and WIFI. The storage medium on the vehicle side can be the electronic control unit (ECU) of the vehicle or other computer systems. Among them, after measuring the stator current data of several cycles, the storage medium on the vehicle side can delete the past stator current data and only retain the stator current data of the most recent cycles, thereby reducing the occupation of the vehicle-side storage resources.

[0108] It should be noted that before sending the stator current data to the cloud, the stator current data can be preprocessed, that is, the original current signal sequence I seq，org Amplify and filter the current signal sequence I in turn. seq , and the current signal sequence I seq As the "stator current data" subsequently sent to the cloud. Among them, amplification processing is a processing method that uses a current amplifier to expand the amplitude of the current signal. The current amplifier may include but is not limited to an ordinary current amplifier, a differential current amplifier or a current transmission amplifier. The filtering processing is a processing method that uses a notch filter to suppress the base frequency of the power system in the current spectrum. The notch filter is a filter based on a circuit or digital algorithm, which is used to weaken the oscillation waveform within a certain frequency range in the signal, including but not limited to an active notch filter or a digital notch filter based on an RC circuit.

[0109] After receiving the stator current data, the cloud can perform multi-classification fault diagnosis on the permanent magnet synchronous motor according to the stator current data, and then feed back the fault results obtained by diagnosis to the vehicle end. Specifically, it can be implemented by setting a multi-classification non-fault diagnosis model; it can also be implemented by using multiple different fault diagnosis models for a specific fault, including but not limited to analytical mathematical models (Analytical Mathematical Model), magnetic equivalent circuit models (Magnetic Equivalent Circuit Model, MEC Model), digital simulation models (Digital Simulation Model), etc. This embodiment does not limit this.

[0110] 203. Determine whether the permanent magnet synchronous motor has a fault tendency based on the fault result corresponding to the permanent magnet synchronous motor received from the cloud.

[0111] In this step, after receiving the fault result corresponding to the permanent magnet synchronous motor fed back by the cloud, the vehicle side will parse the fault result, which includes the potential fault category obtained by the cloud through fault diagnosis and the probability of occurrence indicating the possibility of the fault category. The potential fault category can be one or more, and the multi-category fault diagnosis on the cloud includes the category corresponding to no fault. In other words, the fault result may also include no fault and the probability of no fault. However, when the fault result includes no fault and the probability of no fault, it means that the permanent magnet synchronous motor may be healthy, and no early warning operation is required at this time. Therefore, in order to avoid the synchronous permanent magnet motor from mistakenly performing early warning operations when the vehicle is fault-free. The fault result can be used to determine whether the permanent magnet synchronous motor has a fault tendency, and the fault tendency is used to characterize the possibility of a fault in the permanent magnet synchronous motor. It can be determined according to the number of potential fault categories and the size of the probability of occurrence. For example, assuming that there is one potential fault category, if it is not fault-free, it has a fault tendency, otherwise it does not have a fault tendency. Assuming that there are multiple potential fault categories, if the probability of no fault is the largest, it does not have a fault tendency, otherwise it has a fault tendency, etc. When the fault result determines that the permanent magnet synchronous motor has a fault tendency, step 203 is executed, otherwise the storage operation of the fault result is triggered, that is, each fault result of the permanent magnet synchronous motor is recorded.

[0112] In addition, since each potential probability category has different effects on the overall health status of the permanent magnet synchronous motor, when there are multiple potential fault categories, a corresponding sensitivity coefficient can be set in advance for each potential fault category, and the sensitivity coefficient can be set based on the severity, frequency and possible economic losses of the fault. The sensitivity coefficient is superimposed on the occurrence probability corresponding to each potential fault category, so that it can more accurately reflect the contribution of the potential fault category to the overall health status of the permanent magnet synchronous motor, thereby improving the timeliness and accuracy of subsequent warning operations.

[0113] 204. Trigger a warning operation corresponding to the fault result.

[0114] In this step, after determining the potential fault category of the permanent magnet synchronous motor and its probability of occurrence, the vehicle end can trigger the corresponding early warning operation to prompt the user of the current fault state of the permanent magnet synchronous motor, so as to facilitate early detection of potential faults of the permanent magnet synchronous motor. Early warning operations include but are not limited to driver takeover, automatic pull-over parking, instrument panel warning lights on or audio warnings, etc. Among them, the early warning operation can be classified and set corresponding to different potential fault categories, or it can be uniformly set corresponding to all potential fault categories, which is not limited in this embodiment.

[0115] Based on the above Figure 2 It can be seen from the implementation method that when it is necessary to diagnose the fault of the permanent magnet synchronous motor, the stator current data corresponding to the permanent magnet synchronous motor is collected by the vehicle end and sent to the cloud. The cloud performs multi-classification fault diagnosis based on the stator current data, and obtains the probability of occurrence of each fault category of the permanent magnet synchronous motor in the multi-classification fault, and then determines the fault result corresponding to the permanent magnet synchronous motor according to the probability of occurrence corresponding to each fault category, and synchronizes the fault result to the vehicle end. The vehicle end determines whether the permanent magnet synchronous motor has a fault tendency based on the fault result corresponding to the permanent magnet synchronous motor fed back by the cloud, and determines to trigger the early warning operation corresponding to the fault result when there is a fault tendency. Through the technical solution provided in this application, the vehicle side is mainly responsible for collecting the stator current data of the permanent magnet synchronous motor and sending it to the cloud without performing complex fault diagnosis calculations, which effectively reduces the consumption of computing resources on the vehicle side and uses multi-classification fault diagnosis on the cloud to achieve more accurate and comprehensive fault detection. After obtaining the fault results including potential fault categories and their occurrence probabilities, the vehicle side can further determine whether the permanent magnet synchronous motor has a fault tendency, and determine to trigger a warning operation when it has a fault tendency, and do not perform any operation when it does not have a fault tendency, thereby effectively avoiding the erroneous execution of the warning operation and ensuring the safety of vehicle driving.

[0116] Further, in combination with the description of the stator current data collection in step 201, in this embodiment, it is preferred to use the mileage measurement to control the collection of the stator current data. However, since the mileage measurement includes the initial mileage measurement and the corrected mileage measurement, the following steps 201A-201C can be performed for the selection of the two:

[0117] 201A. Detect whether the corrected mileage measurement fed back from the cloud is received.

[0118] In this step, the corrected mileage measurement is used to characterize the next mileage measurement when the vehicle side collects the stator current data corresponding to the permanent magnet synchronous motor. It is predicted by inputting the current mileage measurement and the fault result corresponding to the current mileage measurement into the mileage prediction model trained on the cloud side. After obtaining the corrected mileage measurement, the cloud side sends it together with the fault result or sends it separately to the vehicle side. Therefore, the vehicle side can detect whether it has received the corrected mileage measurement fed back by the cloud side. If it has been received, step 201B is executed, otherwise step 201C is executed.

[0119] 201B. Collecting stator current data of a permanent magnet synchronous motor based on corrected mileage measurement.

[0120] In this step, since the corrected mileage measurement is obtained based on the current mileage measurement and the corresponding fault result prediction of the current mileage measurement, that is, in the prediction process, the relationship between the vehicle's fault probability and the mileage measurement is considered. Therefore, the corrected mileage measurement reflects the best time to collect the stator current data of the permanent magnet synchronous motor under the current fault state, that is, the corrected mileage measurement has a higher priority than the initial mileage measurement. That is, when the corrected mileage measurement exists, the stator current data is collected according to the corrected mileage measurement first.

[0121] 201C. Collecting stator current data of a permanent magnet synchronous motor based on initial mileage measurement.

[0122] In this step, the initial mileage measurement can be the initially set mileage measurement or the last executed mileage measurement, which is not limited in this embodiment. Since the corrected mileage measurement is not received, but the collection process still needs to be continued, the stator current data is collected according to the initial mileage measurement at this time.

[0123] Further, in combination with the description of determining the fault tendency in step 203, the following steps 203A-203B may be specifically performed:

[0124] 203A. Check whether the potential fault category in the fault result is unique.

[0125] 203B. When the potential fault category is not no fault, it is determined that the permanent magnet synchronous motor has a fault tendency.

[0126] In this step, if the potential fault category is unique and the potential fault category is not fault-free, the vehicle end determines that the permanent magnet synchronous motor has a fault tendency.

[0127] 203C. When the potential fault category with the highest probability of occurrence among multiple potential fault categories is not fault-free, it is determined that the permanent magnet synchronous motor has a fault tendency.

[0128] In this step, if there are multiple potential fault categories, even if there is no fault, the permanent magnet synchronous motor may also have a complex fault or a high uncertainty fault state. Therefore, the vehicle end can compare the probability of occurrence in each potential fault category at this time. If the potential fault category with the highest probability is not no fault, the vehicle end can determine that the permanent magnet synchronous motor has a fault tendency. In addition, it is also possible to directly determine that the permanent magnet synchronous motor has a fault tendency when it is determined that there are multiple potential fault categories in the fault result, or to set a sensitivity coefficient that can reflect it for each potential fault category in advance. The sensitivity coefficient can be used to characterize the degree of influence of each potential fault category on the permanent magnet synchronous motor entering a fault state, which can be specifically set based on the severity, frequency and possible economic losses of the fault. The sensitivity coefficient is superimposed on the probability of occurrence corresponding to each potential fault category and then compared with the probability to determine whether the permanent magnet synchronous motor has a fault tendency.

[0129] Further, in combination with the above description of the sensitivity coefficient, in order to improve the timeliness and accuracy of subsequent warning operations when there are multiple potential fault categories. When the potential fault category with the highest probability of occurrence among multiple potential fault categories is not fault-free, the specific execution process of determining that the permanent magnet synchronous motor has the fault tendency is: obtaining the sensitivity coefficient pre-set for each potential fault category, the sensitivity coefficient is used to characterize the degree of influence of the potential fault category on the permanent magnet synchronous motor entering the fault state; multiplying the probability of occurrence corresponding to each potential fault category by the respective sensitivity coefficients, respectively, to obtain the weighted probability corresponding to each potential fault category; when the potential fault category with the highest weighted probability among multiple potential fault categories is not fault-free, determining that the permanent magnet synchronous motor has a fault tendency.

[0130] Among them, the weighted probability is used to characterize the contribution of potential fault categories to the health of the permanent magnet synchronous motor. By comparing the weighted probabilities in each potential fault category, if the potential fault category with the largest weighted probability is not fault-free, the vehicle side determines that the permanent magnet synchronous motor has a tendency to fail. This improves the timeliness and accuracy of subsequent warning operations.

[0131] Furthermore, combined with the above specific description of the warning operation, in actual applications, the driver taking over the vehicle is usually assumed to be the safest way to deal with the vehicle while it is driving, and automatic pull-over parking is to stop the vehicle from continuing to drive, which is also a very safe way to deal with it. Therefore, the warning operation at least includes reminding the driver to take over and automatically pull over. However, since the user may be more inclined to continue driving and park at a suitable place while the vehicle is driving, in order to better meet the user's driving needs, the specific execution process is: if the driver is not detected to take over within the specified time, the vehicle is controlled to automatically pull over by default.

[0132] Among them, the user can customize the specified time, such as 5 seconds, 10 seconds, etc. The reminder of the driver to take over can remind the user through visual, auditory or tactile warnings (such as dashboard warning lights, sound alarms or seat vibrations), and clearly instruct the user to take action through the infotainment system or voice prompts, etc., and whether the driver takes over can be determined by steering wheel torque detection, pedal activity detection, line of sight tracking or hand position detection. For example, by monitoring the torque change on the steering wheel or monitoring the action of the accelerator or brake pedal, if the user starts to turn the steering wheel or the user steps on the pedal, it means that the user is trying to control the vehicle, that is, it is determined that the driver's takeover is responded to. If the user does not respond within the specified time, automatic pull-over parking is performed. To avoid continuing to drive and reduce safety hazards.

[0133] Furthermore, as a response to the above Figure 1 The implementation of the method embodiment shown in the figure, the embodiment of the present application provides a fault diagnosis device for an electric vehicle drive motor, which is used to reduce the consumption of vehicle-side computing resources and improve the efficiency of motor fault diagnosis. The embodiment of the device corresponds to the aforementioned method embodiment. For ease of reading, this embodiment will no longer repeat the details of the aforementioned method embodiment one by one, but it should be clear that the device in this embodiment can correspond to all the contents of the aforementioned method embodiment. Specifically, Figure 3 As shown, the device is applied to the cloud and includes:

[0134] The receiving unit 301 is used to receive the stator current data corresponding to the permanent magnet synchronous motor sent by the vehicle end;

[0135] The diagnosis unit 302 is used to input the stator current data obtained by the receiving unit 301 into the trained multi-classification fault diagnosis model to obtain a diagnosis result corresponding to the permanent magnet synchronous motor, wherein the diagnosis result includes an occurrence probability corresponding to each fault category;

[0136] A determination unit 303, configured to determine a fault result corresponding to the permanent magnet synchronous motor according to the occurrence probability corresponding to each of the fault categories obtained by the diagnosis unit 302;

[0137] The sending unit 304 is used to synchronize the fault result obtained by the determining unit 303 to the vehicle end, so that the vehicle end can trigger a corresponding warning operation.

[0138] Further, such as Figure 4 As shown, the determining unit 303 includes:

[0139] A detection module 3031 is used to detect whether the occurrence probabilities are all less than a preset probability threshold;

[0140] A confirmation module 3032 is configured to take the fault category with the highest occurrence probability as a potential fault category if the detection module 3031 detects that each occurrence probability is less than a preset probability threshold;

[0141] The confirmation module 3032 is further configured to, if the detection module 3031 detects that each occurrence probability is less than a preset probability threshold, take the fault category with the occurrence probability greater than or equal to the preset probability threshold as a potential fault category;

[0142] The generating module 3033 is used to generate the fault result based on the potential fault category and the occurrence probability corresponding to the potential fault category obtained by the confirming module 3032.

[0143] Further, such as Figure 4 As shown, the device also includes:

[0144] A setting unit 305, configured to pre-set a fault category corresponding to the permanent magnet synchronous motor before the determining unit 303, wherein the fault category is set according to faults that affect the stator current in different forms;

[0145] A marking unit 306 is used to extract target fault samples matching the fault category from the historical fault samples of the permanent magnet synchronous motor, and mark a category label for each target fault sample based on the fault category obtained by the setting unit 305;

[0146] A first training unit 307 is used to select target fault samples with different category labels obtained by the marking unit 306 as training samples according to a preset ratio, and train a 1D convolutional neural network model based on the training samples;

[0147] The confirmation unit 308 is used to use the trained 1D convolutional neural network model obtained by the first training unit 307 as the multi-classification fault diagnosis model.

[0148] Further, such as Figure 4 As shown,

[0149] The fault categories include at least no fault, stator phase winding short circuit, all-phase open circuit, magnet damage, shaft bending, bolt loosening, bearing fault, air gap eccentricity and demagnetization fault.

[0150] Further, such as Figure 4 As shown, the device also includes:

[0151] An acquisition unit 309 is used to acquire, after the determination unit 303, the current mileage measurement corresponding to the vehicle end when collecting the stator current data corresponding to the permanent magnet synchronous motor;

[0152] A prediction unit 310 is used to input the occurrence probability corresponding to each of the fault categories and the current mileage measurement obtained by the acquisition unit 309 into a trained mileage prediction model to obtain a prediction result, wherein the prediction result is used to characterize the next mileage measurement when the vehicle end collects the stator current data corresponding to the permanent magnet synchronous motor;

[0153] The sending unit 34 is further configured to use the next mileage measurement obtained by the prediction unit 310 as a corrected mileage measurement and feed it back to the vehicle end.

[0154] Further, such as Figure 4 As shown, the device also includes:

[0155] An extraction unit 311 is used to extract a specified number of historical fault results and a historical mileage measurement corresponding to the historical fault results;

[0156] A second training unit 312 is used to convert each historical fault result and historical mileage measurement obtained by the extraction unit 311 into graph sequence data, and train a graph convolution-long short-term memory neural network model based on the graph sequence data of the specified number of times;

[0157] The confirmation unit 308 is further configured to use the trained graph convolution-long short-term memory neural network model obtained by the second training unit 312 as the mileage prediction model.

[0158] Furthermore, as a response to the above Figure 2 The implementation of the method embodiment shown in the figure, the embodiment of the present application provides another fault diagnosis device for the electric vehicle drive motor, which is used to reduce the consumption of vehicle-side computing resources and improve the efficiency of motor fault diagnosis. The embodiment of the device corresponds to the aforementioned method embodiment. For ease of reading, this embodiment will no longer repeat the details of the aforementioned method embodiment one by one, but it should be clear that the device in this embodiment can correspond to all the contents of the aforementioned method embodiment. Specifically, Figure 5 As shown, the device is applied to the vehicle end and includes:

[0159] The acquisition unit 401 is used to acquire the stator current data corresponding to the permanent magnet synchronous motor;

[0160] A sending unit 402, configured to send the stator current data obtained by the acquisition unit 401 to a cloud, so that the cloud can perform multi-classification fault diagnosis on the permanent magnet synchronous motor according to the stator current data;

[0161] The processing unit 403 is used to determine whether the permanent magnet synchronous motor has a fault tendency based on the fault result corresponding to the permanent magnet synchronous motor fed back by the cloud after being sent by the sending unit 402;

[0162] The early warning unit 404 is used to trigger an early warning operation corresponding to the fault result if the processing unit 403 determines that the permanent magnet synchronous motor has a fault tendency.

[0163] Further, such as Figure 6 As shown, the acquisition unit 401 includes:

[0164] A first detection module 4011 is used to detect whether the corrected mileage measurement fed back by the cloud is received;

[0165] The acquisition module 4012 is configured to acquire the stator current data of the permanent magnet synchronous motor based on the corrected mileage measurement fed back by the cloud if the first detection module 4011 detects and receives the corrected mileage measurement fed back by the cloud;

[0166] The acquisition module 4012 is further used to acquire the stator current data of the permanent magnet synchronous motor based on the initial mileage measurement if the detection module 4011 detects that the corrected mileage measurement fed back by the cloud is not received.

[0167] Further, such as Figure 6 As shown, the processing unit 403 includes:

[0168] The second detection module 4031 is used to detect whether the potential fault category in the fault result is unique;

[0169] A first determination module 4032 is configured to determine that the permanent magnet synchronous motor has the fault tendency if the potential fault category in the fault result detected by the second detection module 4031 is unique and the potential fault category is not fault-free;

[0170] The second determination module 4033 is used to determine that the permanent magnet synchronous motor has the fault tendency if the potential fault category in the fault result detected by the second detection module 4031 is not unique and the potential fault category with the highest probability of occurrence among the multiple potential fault categories is not fault-free.

[0171] Further, such as Figure 6 As shown, the second determination module 4033 is specifically used to:

[0172] Acquire a sensitivity coefficient pre-set for each of the potential fault categories, wherein the sensitivity coefficient is used to characterize the degree of influence of the potential fault category on the permanent magnet synchronous motor entering a fault state;

[0173] Multiplying the occurrence probability corresponding to each of the potential fault categories by the respective sensitivity coefficients to obtain a weighted probability corresponding to each of the potential fault categories;

[0174] When the potential fault category with the greatest weighted probability among the multiple potential fault categories is not fault-free, it is determined that the permanent magnet synchronous motor has the fault tendency.

[0175] Further, such as Figure 6 As shown, the warning operation at least includes reminding the driver to take over and automatically pull over; the warning unit 404 is specifically used to:

[0176] If the driver is not detected taking over within the specified time, the vehicle will be automatically pulled over by default.

[0177] Furthermore, the embodiment of the present invention also provides a fault diagnosis system for an electric vehicle drive motor, such as Figure 7 As shown, the system includes a cloud end 51 and a vehicle end 52, wherein:

[0178] The cloud 51 can be used to execute any of the above-mentioned methods for diagnosing the fault of the electric vehicle drive motor applied to the cloud when diagnosing the fault of the electric vehicle drive motor.

[0179] The vehicle end 52 can be used to execute any of the above-mentioned methods for diagnosing the fault of the electric vehicle drive motor applied to the vehicle end when diagnosing the fault of the electric vehicle drive motor.

[0180] The present application provides a fault diagnosis system for an electric vehicle drive motor. When it is necessary to diagnose a fault of a permanent magnet synchronous motor, the vehicle end sends the collected stator current data corresponding to the permanent magnet synchronous motor to the cloud. After the cloud receives the stator current data corresponding to the permanent magnet synchronous motor sent by the vehicle end, the stator current data is input into a trained multi-classification fault diagnosis model to obtain the occurrence probability corresponding to each fault category of the permanent magnet synchronous motor in the multi-classification faults, and then the fault result corresponding to the permanent magnet synchronous motor is determined according to the occurrence probability corresponding to each fault category, and the fault result is synchronized to the vehicle end, so that the vehicle end can judge whether the permanent magnet synchronous motor has a fault tendency according to the fault result, and trigger a corresponding early warning operation when it has a fault tendency. Through the technical solution provided in the present application, the motor fault diagnosis can be effectively transferred from the vehicle side to the cloud, thereby reducing the consumption of computing resources on the vehicle side. The cloud side does not need to deploy a corresponding fault diagnosis model for a specific fault. Instead, a multi-classification fault diagnosis model is used to diagnose the permanent magnet synchronous motor and obtain the probability of occurrence of each fault category, which effectively reduces the number of model deployments, further reduces the consumption of cloud computing resources, and ensures the efficiency of motor fault diagnosis. The cloud side then determines the final fault result based on the probability of occurrence of each fault category and feeds back the result to the vehicle side. The vehicle side can determine whether the permanent magnet synchronous motor has a fault tendency based on the fault result fed back by the cloud side, and decide whether to perform the corresponding early warning operation based on the fault tendency. That is, the timeliness and accuracy of motor fault diagnosis are improved through vehicle-cloud collaboration.

[0181] Furthermore, an embodiment of the present application further provides a storage medium, wherein the storage medium is used to store a computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute the above Figure 1 or Figure 2 A fault diagnosis method for an electric vehicle drive motor as described in.

[0182] Furthermore, the embodiment of the present application also provides a processor, the processor is used to run a program, wherein the program executes the above Figure 1 or Figure 2 A fault diagnosis method for an electric vehicle drive motor as described in.

[0183] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0184] It is understandable that the related features in the above methods, devices and systems can be referenced to each other. In addition, the "first", "second" and the like in the above embodiments are used to distinguish the embodiments, but do not represent the advantages and disadvantages of the embodiments.

[0185] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0186] The algorithm and display provided herein are not inherently related to any particular computer, virtual system or other device. Various general purpose systems can also be used together with the teachings based on this. According to the above description, it is obvious that the structure required for constructing such systems. In addition, the application is not directed to any specific programming language either. It should be understood that various programming languages ​​can be utilized to realize the content of the application described herein, and the description of the specific language above is for the purpose of disclosing the best mode of implementation of the application.

[0187] In addition, the memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0188] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0189] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0190] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0191] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0192] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0193] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0194] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0195] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0196] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0197] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A fault diagnosis method for an electric vehicle drive motor, It is characterized in that Applied to the cloud, the method includes: Receive the stator current data of the permanent magnet synchronous motor sent by the vehicle end; Inputting the stator current data into a trained multi-classification fault diagnosis model to obtain a diagnosis result corresponding to the permanent magnet synchronous motor, wherein the diagnosis result includes an occurrence probability corresponding to each fault category; Determine a fault result corresponding to the permanent magnet synchronous motor according to the occurrence probability corresponding to each of the fault categories; The fault result is synchronized to the vehicle end so that the vehicle end can trigger a corresponding early warning operation.

2. The method according to claim 1, It is characterized in that Determining a fault result corresponding to the permanent magnet synchronous motor according to the occurrence probability of each of the fault categories includes: Detecting whether the occurrence probabilities are all less than a preset probability threshold; If so, the fault category with the highest probability of occurrence is taken as a potential fault category; If not, taking the fault category with the occurrence probability greater than or equal to the preset probability threshold as a potential fault category; The fault result is generated based on the potential fault category and the occurrence probability corresponding to the potential fault category.

3. The method according to claim 1, It is characterized in that Before inputting the stator current data into the trained multi-classification fault diagnosis model to obtain a diagnosis result corresponding to the permanent magnet synchronous motor, the method further includes: Presetting a fault category corresponding to the permanent magnet synchronous motor, wherein the fault category is set according to faults that affect the stator current in different forms; Extracting target fault samples matching the fault category from historical fault samples of the permanent magnet synchronous motor, and marking a category label for each target fault sample based on the fault category; Select target fault samples with different category labels as training samples according to a preset ratio, and train a 1D convolutional neural network model based on the training samples; The trained 1D convolutional neural network model is used as the multi-classification fault diagnosis model.

4. The method according to any one of claims 1 to 3, It is characterized in that The fault categories include at least no fault, stator phase winding short circuit, all-phase open circuit, magnet damage, shaft bending, bolt loosening, bearing fault, air gap eccentricity and demagnetization fault.

5. The method according to claim 1, It is characterized in that After inputting the stator current data into the trained multi-classification fault diagnosis model to obtain a diagnosis result corresponding to the permanent magnet synchronous motor, the method further includes: Obtaining the current mileage measurement corresponding to the vehicle end when collecting the stator current data corresponding to the permanent magnet synchronous motor; Inputting the occurrence probability corresponding to each of the fault categories and the current mileage measurement into a trained mileage prediction model to obtain a prediction result, wherein the prediction result is used to characterize the next mileage measurement when the vehicle end collects the stator current data corresponding to the permanent magnet synchronous motor; The next mileage measurement is used as a corrected mileage measurement and fed back to the vehicle end.

6. The method according to claim 5, It is characterized in that Before inputting the occurrence probability corresponding to each of the fault categories and the current mileage measurement into the trained mileage prediction model to obtain the prediction result, the method further includes: Extracting a specified number of historical fault results and a historical mileage measurement corresponding to the historical fault results; Convert each historical fault result and historical mileage measurement into graph sequence data respectively, and train a graph convolution-long short-term memory neural network model based on the graph sequence data of the specified number of times; The trained graph convolution-long short-term memory neural network model is used as the mileage prediction model.

7. A fault diagnosis method for an electric vehicle drive motor, It is characterized in that Applied to the vehicle side, the method includes: Collect stator current data corresponding to the permanent magnet synchronous motor; Sending the stator current data to a cloud so that the cloud can perform multi-classification fault diagnosis on the permanent magnet synchronous motor according to the stator current data; Determining whether the permanent magnet synchronous motor has a fault tendency based on the fault result corresponding to the permanent magnet synchronous motor received from the cloud feedback; If so, a warning operation corresponding to the fault result is triggered.

8. The method according to claim 7, It is characterized in that Collect stator current data of permanent magnet synchronous motor, including: Detecting whether the corrected mileage measurement fed back by the cloud is received; If yes, collecting the stator current data of the permanent magnet synchronous motor based on the corrected mileage metric; If not, the stator current data of the permanent magnet synchronous motor is collected based on the initial mileage measurement.

9. The method according to claim 7, It is characterized in that Judging whether the permanent magnet synchronous motor has a fault tendency based on the fault result corresponding to the permanent magnet synchronous motor fed back from the cloud includes: Detecting whether the potential fault category in the fault result is unique; If so, when the potential fault category is not fault-free, determining that the permanent magnet synchronous motor has the fault tendency; If not, when the potential fault category with the highest occurrence probability among the multiple potential fault categories is not fault-free, it is determined that the permanent magnet synchronous motor has the fault tendency.

10. The method according to claim 9, It is characterized in that When the potential fault category with the highest occurrence probability among the multiple potential fault categories is not fault-free, determining that the permanent magnet synchronous motor has the fault tendency includes: Acquire a sensitivity coefficient pre-set for each of the potential fault categories, the sensitivity coefficient being used to characterize the degree of influence of the potential fault category on the permanent magnet synchronous motor entering a fault state; Multiplying the occurrence probability corresponding to each of the potential fault categories by the respective sensitivity coefficients to obtain a weighted probability corresponding to each of the potential fault categories; When the potential fault category with the greatest weighted probability among the multiple potential fault categories is not fault-free, it is determined that the permanent magnet synchronous motor has the fault tendency.

11. The method according to any one of claims 7 to 10, It is characterized in that The warning operation at least includes driver taking over and automatic side parking; Triggering the warning operation corresponding to the fault result includes: If the driver taking over is not detected within the specified time, the vehicle is controlled to perform the automatic pull-over parking by default.

12. A fault diagnosis device for an electric vehicle drive motor, It is characterized in that Applied to the cloud, the device comprises: A receiving unit, used for receiving stator current data corresponding to the permanent magnet synchronous motor sent by the vehicle end; A diagnosis unit, used for inputting the stator current data obtained by the receiving unit into a trained multi-classification fault diagnosis model to obtain a diagnosis result corresponding to the permanent magnet synchronous motor, wherein the diagnosis result includes an occurrence probability corresponding to each fault category; a determination unit, configured to determine a fault result corresponding to the permanent magnet synchronous motor according to the occurrence probability corresponding to each of the fault categories obtained by the diagnosis unit; The sending unit is used to synchronize the fault result obtained by the determination unit to the vehicle end, so that the vehicle end can trigger a corresponding early warning operation.

13. A fault diagnosis device for an electric vehicle drive motor, It is characterized in that Applied to the vehicle end, the device comprises: A collection unit, used for collecting stator current data corresponding to the permanent magnet synchronous motor; A sending unit, used to send the stator current data obtained by the acquisition unit to a cloud, so that the cloud can perform multi-classification fault diagnosis on the permanent magnet synchronous motor according to the stator current data; a processing unit, configured to determine whether the permanent magnet synchronous motor has a fault tendency based on a fault result corresponding to the permanent magnet synchronous motor received from the cloud feedback; The early warning unit is used to trigger an early warning operation corresponding to the fault result if the processing unit determines that the permanent magnet synchronous motor has a fault tendency.

14. A fault diagnosis system for an electric vehicle drive motor, It is characterized in that The system includes: a cloud side and a vehicle side, When the cloud performs fault diagnosis of the electric vehicle drive motor, the method for fault diagnosis of the electric vehicle drive motor according to any one of claims 1 to 6 is executed; When the vehicle end performs fault diagnosis of the electric vehicle drive motor, the electric vehicle drive motor fault diagnosis method described in any one of claims 7 to 11 is executed.

15. A storage medium, It is characterized in that The storage medium includes a stored program, wherein when the program is running, the device where the storage medium is located is controlled to execute the fault diagnosis method for the electric vehicle drive motor as described in any one of claims 1 to 6 or claims 7 to 11.

16. A processor, It is characterized in that The processor is used to run a program, wherein when the program is run, the fault diagnosis method for the electric vehicle drive motor as described in any one of claims 1 to 6 or claims 7 to 11 is executed.

Images