Electric vehicle electric wheel fault diagnosis method based on physical information

Abstract

The invention provides an electric vehicle electric wheel fault diagnosis method based on physical information, and belongs to the field of electric vehicle fault diagnosis, and the method comprises the following steps: collecting a clean planetary gear vibration signal and a clean bearing vibration signal of an electric wheel; applying physical noise to the clean planetary gear vibration signal and the clean bearing vibration signal input signal to obtain a planetary gear vibration signal and a bearing vibration signal containing the physical noise; extracting vibration signal features, fusing the features of the clean planetary gear vibration signals and the features of the clean bearing vibration signals to obtain MF, and fusing the features of the planetary gear vibration signals and the features of the bearing vibration signals containing physical noise to obtain MF *; and capturing the depth global feature of the MF and the physical noise depth feature of the MF *, fusing the depth features of the MF and the MF *, and identifying the fault type according to the fused features. According to the invention, the fault diagnosis precision can be improved.

Description

Technical Field

[0001] The present invention belongs to the field of electric vehicle fault diagnosis, and particularly relates to a method for diagnosing faults of electric wheels of electric vehicles based on physical information. Background Art

[0002] As a key component of electric vehicles, the health status of electric wheels directly affects the reliable and safe operation of electric vehicles. However, due to the precision characteristics of electric wheels, it is very difficult to directly observe the health status of planetary gears and bearings in electric wheels. However, even for electric wheels produced in the same batch, their planetary gears and bearings will show significantly different remaining service lives due to factors such as working environment, driving habits, and load changes. This uncertainty leads to the complexity of the maintenance strategy for electric wheels. Regularly detecting planetary gears and bearings in electric wheels that have not reached the expected life will consume a large amount of resources and increase unnecessary downtime; while checking when approaching the life limit may face serious failure risks, leading to driving safety problems. Secondly, since the evolution law of the remaining life of planetary gears and bearings is closely related to early weak faults, once an early fault occurs in a bearing, its remaining service life often decreases sharply. Therefore, timely identification and handling of these early faults are crucial for ensuring the stable operation of electric wheels.

[0003] Developing a new method for early fault diagnosis of electric wheels. Through early fault diagnosis, appropriate preventive maintenance measures can be taken to avoid potential major faults and extend the service life of planetary gears and bearings in electric wheels. This not only helps to improve the overall reliability and safety of electric vehicles, reduce the accident risk caused by sudden electric wheel faults, but also effectively reduces the high maintenance costs and downtime losses caused by sudden faults, thereby reducing the overall operating cost.

[0004] Although existing technologies have achieved high-precision fault diagnosis of planetary gears or bearings of electric wheels by using deep learning models and machine learning models, there are still the following defects:

[0005] 1. The running signals of early weak faults of planetary gears are very similar to those of early weak faults of bearings. Existing methods are difficult to distinguish whether it is a planetary gear or a bearing that fails in an electric wheel.

[0006] 2. Existing methods often only conduct experiments on standard ideal vibration data collected in the laboratory, considering less whether there are different types of noise interferences in the data. In fact, due to the different operating environments of electric vehicles, the noise will have a greater impact on the vibration data, further reducing the fault diagnosis accuracy of planetary gears and bearings of electric wheels.

[0007] In summary, the existing fault diagnosis models cannot accurately detect the fault types. Summary of the Invention

[0008] To overcome the deficiencies of the above-mentioned existing technologies, the present invention provides a method for diagnosing faults in electric wheels of electric vehicles based on physical information, including the following steps:

[0009] Collect the clean planetary gear vibration signal and the clean bearing vibration signal of the electric wheel of the electric vehicle through experiments. The clean planetary gear vibration signal and the clean bearing vibration signal are signals that do not contain noise;

[0010] Apply physical noise to the clean planetary gear vibration signal and the clean bearing vibration signal to obtain a planetary gear vibration signal containing physical noise and a bearing vibration signal containing physical noise respectively;

[0011] Extract the features of the clean planetary gear vibration signal, the clean bearing vibration signal, the planetary gear vibration signal containing physical noise, and the bearing vibration signal. Fuse the features of the extracted clean planetary gear vibration signal and the clean bearing vibration signal to obtain the fused feature MF of the clean vibration signal, and fuse the features of the extracted planetary gear vibration signal containing physical noise and the bearing vibration signal to obtain the fused feature MF of the physical noise vibration signal * ;

[0012] Fuse the depth features of MF and MF * and identify the fault type according to the fused features.

[0013] Preferably, the fusing of the depth features of MF and MF * and the identification of the fault type according to the fused features are specifically as follows: respectively capture the depth global feature DF of MF and the physical noise depth feature PF of MF * , perform dot product multiplication fusion on the elements in DF and the elements in PF, reduce the dimension of the fused feature, and convert the feature with reduced dimension into a fault diagnosis result.

[0014] Preferably, the fault types include the fault types of bearings and the fault types of planetary gears. The fault types of bearings include inner ring faults, outer ring faults, and ball faults; the fault types of planetary gears include tooth surface wear, tooth root cracks, and pitting.

[0015] Preferably, the applying of physical noise to the clean planetary gear vibration signal and the clean bearing vibration signal is specifically as follows: apply impact noise and electromagnetic noise to the clean planetary gear vibration signal, and apply friction noise and electromagnetic noise to the clean bearing vibration signal, specifically through the following formula:

[0016]

[0017] Among them, data1 is the clean planetary gear vibration signal collected in the laboratory, and data2 is the clean bearing vibration signal collected in the laboratory. is the planetary gear vibration signal containing physical noise, is the bearing vibration signal containing physical noise, γ1 is the impact noise, γ2 is the friction noise, and γ3 is the electromagnetic noise.

[0018] The present invention also provides an electric vehicle electric wheel fault diagnosis system based on physical information, comprising:

[0019] A data acquisition module, used for acquiring a clean planetary gear vibration signal and a clean bearing vibration signal of an electric wheel of an electric vehicle through an experiment, wherein the clean planetary gear vibration signal and the clean bearing vibration signal are signals that do not contain noise;

[0020] A vibration physical noise synthesis module, used for applying physical noise to a clean planetary gear vibration signal and a clean bearing vibration signal, to obtain a planetary gear vibration signal containing physical noise and a bearing vibration signal containing physical noise, respectively;

[0021] The feature extraction and fusion module is used to extract the features of the clean planetary gear vibration signal, the clean bearing vibration signal, and the planetary gear vibration signal and bearing vibration signal containing physical noise, fuse the extracted features of the clean planetary gear vibration signal and the clean bearing vibration signal to obtain the clean vibration signal fusion feature MF, and fuse the extracted features of the planetary gear vibration signal containing physical noise and the bearing vibration signal to obtain the physical noise vibration signal fusion feature MF. * ;

[0022] Fault detection module, used to detect MF and MF * The deep features of the proposed method are fused and the fault type is identified based on the fused features.

[0023] Preferably, the fault detection module includes a backbone network, a branch network and a prediction output network. Specifically, the backbone network includes 5 convolution layers, and after each convolution layer, a ReLU activation function is applied to perform layer-by-layer nonlinear feature transformation on MF and reduce the dimension of the feature, and finally outputs the deep global feature DF of MF; the branch network includes an RNN layer and 4 layers of deep separable convolution layers, the RNN layer extracts the spatial dependency relationship between different physical noises, and after the deep separable convolution layer, a ReLU activation function is applied to MF *Perform layer-by-layer non-linear feature transformation and reduce the dimension of the features, and finally output the physical noise depth feature PF; the prediction output network includes a dot product layer, a fully connected layer and a softmax activation function layer. Input DF and PF into the dot product layer to obtain the fused feature. The fully connected layer is used to reduce the dimension of the fused feature, and the softmax activation function layer converts the feature with reduced dimension into the fault diagnosis result.

[0024] The present invention also provides a computer device, including a memory and a processor; the memory stores a computer program, and the processor is used to run the computer program in the memory to execute the method for diagnosing faults of an electric wheel of an electric vehicle based on physical information.

[0025] The present invention also provides a computer-readable storage medium, which stores a computer program. The computer program is suitable for being loaded by a processor to execute the method for diagnosing faults of an electric wheel of an electric vehicle based on physical information.

[0026] The method for diagnosing faults of an electric wheel of an electric vehicle based on physical information provided by the present invention has the following beneficial effects:

[0027] 1. By adding real physical noise to the clean planetary gear vibration signal and the clean bearing vibration signal, the present invention can take into account the influence of noise on vibration data in the real environment, thereby improving the fault diagnosis accuracy of the electric wheel under real working conditions.

[0028] 2. By extracting the features of the clean planetary gear vibration signal, the clean bearing vibration signal, the planetary gear vibration signal and the bearing vibration signal containing physical noise, and fusing the features of the extracted clean planetary gear vibration signal and the clean bearing vibration signal, and fusing the features of the extracted planetary gear vibration signal and the bearing vibration signal containing physical noise, the present invention can capture the differences between the planetary gear vibration signal and the bearing vibration signal, so as to distinguish the planetary gear vibration signal and the bearing vibration signal, and realize the simultaneous diagnosis of the planetary gear and the bearing of the electric wheel. Description of the Drawings

[0029] In order to more clearly illustrate the embodiments of the present invention and its design solutions, the following will briefly introduce the drawings required for this embodiment. The drawings described below are only partial embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0030] Figure 1 It is a structural diagram of an electric wheel of an electric vehicle according to an embodiment of the present invention;

[0031] Figure 2Flow chart of the fault diagnosis method for the electric wheel of an electric vehicle based on physical information according to an embodiment of the present invention;

[0032] Figure 3 It is the structure diagram of the fault detection module.

[0033] Explanation of reference numerals: 1 - permanent magnet, 2 - internal gear ring, 3 - planetary gear bearing, 4 - planet carrier bearing, 5 - output shaft, 6 - sun gear, 7 - planet carrier, 8 - electric wheel housing, 9 - planetary gear, 10 - motor bearing, 11 - armature winding, 12 - stator core, 13 - rotor core. Detailed implementation manners

[0034] In order to enable those skilled in the art to better understand the technical solution of the present invention and implement it, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0035] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "axial", "radial", "circumferential", etc. is based on the orientation or positional relationship shown in the accompanying drawings. It is only for the convenience of describing the technical solution of the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation to the present invention.

[0036] In addition, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In the description of the present invention, it should be noted that unless otherwise clearly specified or limited, the terms "connected" and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances. In the description of the present invention, unless otherwise specified, the meaning of "plurality" is two or more, which will not be elaborated here.

[0037] Embodiment

[0038] The electric wheel system applied in the present invention is as Figure 1As shown in the figure, it specifically includes: permanent magnet 1, internal gear ring 2, planetary gear bearing 3, planet carrier bearing 4, output shaft 5, sun gear 6, planet carrier 7, electric wheel housing 8, planetary gear 9, motor bearing 10, armature winding 11, stator core 12, and rotor core 13. The built-in V-type permanent magnet synchronous motor is coaxially connected to the planetary gear reducer and shares a set of housings. Its vibration transmission paths mainly include:

[0039] 1. Planetary gear 9 → internal gear ring 2 → electric wheel housing 8 → sensor;

[0040] 2. Planetary gear 9 → planetary gear bearing 3 → planet carrier 7 → planet carrier bearing 4 → electric wheel housing 8 → sensor;

[0041] 3. Sun gear 6 → planetary gear 9 → internal gear ring 2 → electric wheel housing 8 → sensor;

[0042] 4. Sun gear 6 → planetary gear 9 → planetary gear bearing 3 → planet carrier 7 → planet carrier bearing 4 → electric wheel housing 8 → sensor;

[0043] 5. Motor rotor → motor bearing 10 → transmission shaft → sun gear 6 → planetary gear 9 → external gear ring → electric wheel housing 8 → sensor;

[0044] 6. Motor rotor → motor bearing 10 → transmission shaft → sun gear 6 → planetary gear 9 → planetary gear bearing 3 → planet carrier 7 → planet carrier bearing 4 → electric wheel housing 8 → sensor;

[0045] Based on the above electric wheel system, the present invention provides a method for diagnosing faults of an electric wheel of an electric vehicle based on physical information, specifically as Figure 2 shown, including the following steps:

[0046] Step 1: Collect the clean planetary gear vibration signal and the clean bearing vibration signal of the electric wheel of the electric vehicle through experiments. The clean planetary gear vibration signal and the clean bearing vibration signal are signals without noise.

[0047] The present invention uses triaxial acceleration sensors to collect the clean vibration data of the planetary gear and the bearing in the electric wheel in three directions of X ( Figure 1 measurement point X in Figure 1 ), Y ( Figure 1 measurement point Y in Figure 1 ), and Z ( Figure 1 measurement point Z in Figure 1 ), that is, the clean planetary gear vibration signal and the clean bearing vibration signal. Among them, the data used in the experiment includes a total of 4 types of faults according to the degree of fault, which are divided into early minor faults, minor faults, medium faults, and severe faults. This method takes the early minor faults for simulation experiments to verify the effectiveness of the method.

[0048] Step 2: Construct an electric wheel fault diagnosis model. The electric wheel fault diagnosis model includes a vibration physical noise synthesis module, a feature extraction and fusion module, and a fault detection module. The feature extraction and fusion module includes an LSTM network and a multi-head attention mechanism. The fault detection module includes a backbone network, a branch network, and a prediction output network.

[0049] (1) Use the vibration physical noise synthesis module to apply physical noise to the clean planetary gear vibration signal and the clean bearing vibration signal, respectively, to obtain a planetary gear vibration signal containing physical noise and a bearing vibration signal containing physical noise.

[0050] The vibration signals generated by the planetary gears and bearings in the electric wheel under real working conditions all contain complex noises. However, the vibration signals collected in the laboratory do not contain the noise information under various complex working conditions, resulting in low fault diagnosis accuracy of the neural network trained based on the data collected in the laboratory under real working conditions. Therefore, in the first step of this method, the vibration physical noise synthesis module is used to apply physical noise to the electric wheel planetary gear vibration signal data and bearing vibration signal data collected in the laboratory. When the electric vehicle is working, due to the manufacturing error, wear, or misalignment of the planetary gears, the planetary gear vibration signal contains impact noise. Due to poor bearing lubrication, the bearing vibration signal contains friction noise. Due to the electromagnetic force during the operation of the motor, which will cause vibration, the planetary gear vibration signal and the bearing vibration signal contain electromagnetic noise. In the present invention, impact noise and electromagnetic noise are applied to the planetary gear vibration signal to obtain a planetary gear vibration signal containing physical noise, and friction noise and electromagnetic noise are applied to the bearing vibration signal to obtain a bearing vibration signal containing physical noise. The specific calculation formulas are as follows:

[0051]

[0052] where data1 is the clean planetary gear vibration signal collected in the laboratory, data2 is the clean bearing vibration signal collected in the laboratory, is the planetary gear vibration signal containing physical noise, is the bearing vibration signal containing physical noise; γ1 is the impact noise, which is represented by a random noise that satisfies the exponential distribution; γ2 is the friction noise, which is represented by Gaussian white noise; γ3 is the electromagnetic noise, which is represented by a random noise that satisfies the Gaussian distribution.

[0053] (2) Use the feature extraction and fusion module to extract the features of the clean planetary gear vibration signal, the clean bearing vibration signal, the planetary gear vibration signal containing physical noise, and the bearing vibration signal. Then, fuse the features of the clean planetary gear vibration signal and the clean bearing vibration signal to obtain the fused feature MF of the clean vibration signal. Also, fuse the features of the planetary gear vibration signal containing physical noise and the bearing vibration signal to obtain the fused feature MF of the physical noise vibration signal. * 。

[0054] To effectively fuse the features of the planetary gear vibration signal and the bearing vibration signal, the present invention constructs a feature extraction and fusion module including an LSTM network and a multi-head attention mechanism. The LSTM is used to extract the features of data1, data2 and . The multi-head attention mechanism performs attention-weighted fusion on data1 and data2 to obtain the fused feature MF of the clean vibration signal. At the same time, it performs attention-weighted fusion on and to obtain the fused feature MF of the physical noise vibration signal. * 。

[0055] (3) Use the fault detection module to fuse the deep features of MF and MF * , and identify the fault type based on the fused features.

[0056] To improve the network's learning ability for the true physical noise features of the vibration signal, the present invention constructs a fault detection module, which consists of a backbone network, a branch network, and a prediction output network. Input MF into the backbone network to capture the deep global feature DF of the clean vibration signal, input MF * into the branch network to obtain the physical noise deep feature PF. Finally, input DF and PF into the prediction output network simultaneously to obtain the electric wheel fault prediction result.

[0057] The multi-head attention mechanism assigns attention weights to the gear vibration signal and the bearing vibration signal, focusing on the intersection of the two signals, enabling the model to capture the features of the two types of data from different perspectives, which is described by the following formula:

[0058] MultiHead(Q,K,V)=Concat(head1,…,head h )W O ;

[0059] where each head i is:

[0060]

[0061] Among them, h represents the number of attention heads, which allows the model to learn information from two heterogeneous data spaces in parallel. is the weight matrix of each attention head, and Concat is to concatenate the outputs of multiple heads. W O can then transform the concatenated features back to a suitable dimension. Q is the query vector, K is the key vector, V is the value vector, and MultiHead(Q, K, V) is the weight vector of the gear vibration signal and the bearing vibration signal.

[0062] As a preferred implementation, the structure of the fault detection module is as Figure 3 shown, including a backbone network, a branch network, and a prediction output network. The backbone network consists of 5 convolutional layers. After each convolutional layer, the ReLU activation function is applied to perform a layer-by-layer non-linear feature transformation on the MF and reduce the dimension of the features. Finally, the backbone network outputs the deep global feature DF of the clean vibration signal. The branch network consists of an RNN layer and 4 depthwise separable convolutional layers. The RNN layer is responsible for extracting the spatial dependence relationship between different physical noises. After the depthwise separable convolutional layer, the ReLU activation function is applied to perform a layer-by-layer non-linear feature transformation on the MF * and reduce the dimension of the features. Finally, the branch network outputs the physical noise depth feature PF. The prediction output network consists of a dot product layer, a fully connected layer, and a softmax activation function layer. The DF and PF are input into the dot product layer to obtain the fused feature, which is described by the following formula:

[0063]

[0064] where DPF is the fused feature, DF n is the nth element in DF, and PF n is the nth element in PF.

[0065] Before inputting the clean planetary gear vibration signal and the clean bearing vibration signal into the electric wheel fault diagnosis model, it also includes training the electric wheel fault diagnosis model, which includes the following steps:

[0066] (1) Collect vibration noise data of planetary gears and bearings including different health state types.

[0067] Data collection is the first and crucial step in building an electric wheel fault diagnosis. A large number of vibration signals need to be collected, containing sufficient vibration data of planetary gears and bearings with different health state types to ensure that the feature extraction and fusion module and the fault detection module can learn rich fault diagnosis features. The main source of data collection is the bearing signals collected in the laboratory.

[0068] (2) Label the vibration noise data according to the fault type.

[0069] Data marking is to add meaningful labels to the preprocessed data, enabling the model to learn the inherent features and relationships of the data and perform mathematical calculations.

[0070] Label division: According to the three common faults of bearings, inner ring fault, outer ring fault, and ball fault, they are respectively marked as 0, 1, 2. According to the three common faults of planetary gears, tooth surface wear, tooth root crack, and pitting, they are respectively marked as 3, 4, 5. These faults are all artificial faults made by electrical discharge machining.

[0071] (3) Dataset construction.

[0072] In order to divide the labeled data into a training set, a validation set, and a test set to prepare for the training, evaluation, and testing of the model, a dataset needs to be constructed: Denote the labeled dataset as (D, L), and divide the training set, validation set, and test set according to the ratio of 8:1:1. The specific steps are as follows:

[0073] Divide the labeled dataset (D, L) according to the following ratio:

[0074] 80% of the data is used as the training set (Training Set).

[0075] 10% of the data is used as the validation set (Validation Set).

[0076] 10% of the data is used as the test set (Test Set).

[0077] Such a division can ensure that there is sufficient data for parameter adjustment and optimization during the model training process, and conduct the evaluation of the validation set and the final performance verification of the test set after the model training to ensure the generalization ability and effectiveness of the model.

[0078] The present invention uses the Adam optimizer for network training, selects the cross-entropy function as the loss function, sets the neuron random inactivation ratio Drop-out to 0.12, sets the learning rate to 0.001, the sample data includes 1024 sampling points, and uses 32 sampling points as the batch size for 100 epochs of iteration.

[0079] After adopting the above technical solution, the beneficial effects of the present invention are:

[0080] 1. The present invention uses the physical noise synthesis module to add real physical noise to the vibration signal data collected in the laboratory, enabling the neural network to fully learn the noise information contained in the vibration signal and improving the electric wheel fault diagnosis accuracy of the neural network under real working conditions.

[0081] 2. The present invention uses a multi-head attention mechanism to capture the differences between the planetary gear vibration signals and the bearing vibration signals, enabling the neural network to fully learn the differential features of the two, and realizing the simultaneous diagnosis of the planetary gears and bearings of the electric wheel.

[0082] 3. The present invention creatively designs an electric wheel fault diagnosis model with a double-branch structure. This model can fully learn the differences between the clean vibration signals and the physical noise under actual working conditions, and improve the fault diagnosis accuracy of the neural network under real working conditions.

[0083] The present invention also provides an electric wheel fault diagnosis system for electric vehicles based on physical information, including:

[0084] A data acquisition module for collecting clean planetary gear vibration signals and clean bearing vibration signals of the electric wheel of the electric vehicle through experiments. The clean planetary gear vibration signals and clean bearing vibration signals are signals without noise.

[0085] A vibration physical noise synthesis module for applying physical noise to the clean planetary gear vibration signals and clean bearing vibration signals to obtain planetary gear vibration signals containing physical noise and bearing vibration signals containing physical noise respectively.

[0086] A feature extraction and fusion module for extracting the features of the clean planetary gear vibration signals, clean bearing vibration signals, and planetary gear vibration signals and bearing vibration signals containing physical noise, fusing the features of the extracted clean planetary gear vibration signals and clean bearing vibration signals to obtain a clean vibration signal fusion feature MF, and fusing the features of the extracted planetary gear vibration signals and bearing vibration signals containing physical noise to obtain a physical noise vibration signal fusion feature MF * 。

[0087] A fault detection module for fusing the deep features of MF and MF * and identifying the fault type according to the fused features.

[0088] Among them, the fault detection module includes a backbone network, a branch network, and a prediction output network. Specifically, the backbone network includes 5 convolutional layers. After each convolutional layer, the ReLU activation function is applied to perform a layer-by-layer non-linear feature transformation on MF and reduce the dimension of the features, and finally output the deep global feature DF of MF; the branch network includes an RNN layer and 4 depthwise separable convolutional layers. The RNN layer extracts the spatial dependence relationship between different physical noises. After the depthwise separable convolutional layer, the ReLU activation function is applied to MF *Perform layer-by-layer non-linear feature transformation and reduce the dimension of features, and finally output the physical noise depth feature PF; the prediction output network includes a dot product layer, a fully connected layer, and a softmax activation function layer. Input DF and PF into the dot product layer to obtain a fused feature. The fully connected layer is used to reduce the dimension of the fused feature, and the softmax activation function layer converts the feature with reduced dimension into a fault diagnosis result.

[0089] The present invention also provides a computer device, including a memory and a processor; the memory stores a computer program, and the processor is configured to run the computer program in the memory to execute the physical information-based electric vehicle electric wheel fault diagnosis method.

[0090] The present invention also provides a computer-readable storage medium, which stores a computer program, and the computer program is suitable for being loaded by a processor to execute the physical information-based electric vehicle electric wheel fault diagnosis method.

[0091] The above embodiments are only preferred specific embodiments of the present invention, and the protection scope of the present invention is not limited thereto. Any simple changes or equivalent replacements of technical solutions that can be obviously obtained by those skilled in the art within the technical scope disclosed by the present invention all belong to the protection scope of the present invention.

Claims

1. A fault diagnosis method for an electric wheel of an electric vehicle based on physical information, characterized in that, It includes the following steps: Collect the clean planetary gear vibration signal and clean bearing vibration signal of the electric wheel of the electric vehicle through experiments. The clean planetary gear vibration signal and clean bearing vibration signal are signals without noise; Apply physical noise to the clean planetary gear vibration signal and clean bearing vibration signal to obtain the planetary gear vibration signal containing physical noise and the bearing vibration signal containing physical noise respectively; Extract the features of the clean planetary gear vibration signal, the clean bearing vibration signal, the planetary gear vibration signal containing physical noise, and the bearing vibration signal. Fuse the features of the extracted clean planetary gear vibration signal and the clean bearing vibration signal to obtain the fused feature MF of the clean vibration signal. Fuse the features of the extracted planetary gear vibration signal containing physical noise and the bearing vibration signal to obtain the fused feature MF of the physical noise vibration signal * ; Fuse the depth features of MF and MF * and identify the fault type based on the fused features.

2. The method for diagnosing faults of an electric wheel of an electric vehicle based on physical information according to claim 1, wherein, The depth features of MF and MF * are fused, and the fault type is identified according to the fused features. Specifically: the depth global feature DF of MF and the physical noise depth feature PF of MF * are respectively captured, the elements in DF and the elements in PF are fused by dot product multiplication, the dimension of the fused feature is reduced, and the feature with reduced dimension is converted into a fault diagnosis result.

3. The method for diagnosing faults of an electric wheel of an electric vehicle based on physical information according to claim 1, characterized in that The fault types include the fault types of the bearing and the planetary gear. The fault types of the bearing include inner ring fault, outer ring fault and ball fault; The fault types of the planetary gear include tooth surface wear, tooth root crack and pitting.

4. The method for diagnosing faults of an electric wheel of an electric vehicle based on physical information according to claim 1, wherein The application of physical noise to the clean planetary gear vibration signal and clean bearing vibration signal is specifically as follows: impact noise and electromagnetic noise are applied to the clean planetary gear vibration signal, and friction noise and electromagnetic noise are applied to the clean bearing vibration signal, which is specifically carried out through the following formula: Among them, data1 is the clean planetary gear vibration signal collected in the laboratory, and data2 is the clean bearing vibration signal collected in the laboratory. is the planetary gear vibration signal containing physical noise. is the bearing vibration signal containing physical noise, γ1 is the impact noise, γ2 is the friction noise, and γ3 is the electromagnetic noise.

5. A fault diagnosis system for an electric wheel of an electric vehicle based on physical information, characterized in that, It includes: A data acquisition module for collecting the clean planetary gear vibration signal and clean bearing vibration signal of the electric wheel of the electric vehicle through experiments. The clean planetary gear vibration signal and clean bearing vibration signal are signals without noise; A vibration physical noise synthesis module for applying physical noise to the clean planetary gear vibration signal and clean bearing vibration signal to obtain the planetary gear vibration signal containing physical noise and the bearing vibration signal containing physical noise respectively; Feature extraction and fusion module, which is used to extract the features of the clean planetary gear vibration signal, the clean bearing vibration signal, the planetary gear vibration signal containing physical noise and the bearing vibration signal, fuse the features of the extracted clean planetary gear vibration signal and the clean bearing vibration signal to obtain the fused feature MF of the clean vibration signal, and fuse the features of the extracted planetary gear vibration signal containing physical noise and the bearing vibration signal to obtain the fused feature MF of the physical noise vibration signal * ; A fault detection module for fusing the deep features of MF and MF * and identifying the fault type according to the fused features.

6. The physical-information-based electric vehicle electric wheel fault diagnosis system according to claim 5, characterized in that, The fault detection module includes a backbone network, a branch network and a prediction output network. Specifically, the backbone network includes 5 convolutional layers. After each convolutional layer, the ReLU activation function is applied to perform layer-by-layer non-linear feature transformation on the MF and reduce the dimension of the features, and finally output the depth global feature DF of the MF; The branch network includes an RNN layer and four layers of depthwise separable convolutional layers. The RNN layer extracts the spatial dependence relationships between different physical noises. After the depthwise separable convolutional layers, the ReLU activation function is applied to perform a layer-by-layer non-linear feature transformation on the MF * and reduce the dimension of the features, and finally output the physical noise depth features PF. The prediction output network includes a dot product layer, a fully connected layer, and a softmax activation function layer. The DF and PF are input into the dot product layer to obtain the fused features. The fully connected layer is used to reduce the dimension of the fused features, and the softmax activation function layer converts the features with reduced dimension into the fault diagnosis results.

7. A computer device, characterized in that, It includes a memory and a processor; The memory stores a computer program, and the processor is used to run the computer program in the memory to execute the physical information-based fault diagnosis method for the electric wheel of the electric vehicle according to any one of claims 1-4.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, and the computer program is suitable for being loaded by the processor to execute the physical information-based fault diagnosis method for the electric wheel of the electric vehicle according to any one of claims 1-4.

Images