Multi-fault intelligent diagnosis method and system for permanent magnet synchronous motor

Abstract

The invention discloses a multi-fault intelligent diagnosis method and system for a permanent magnet synchronous motor, relates to the field of motor fault identification, and solves the problem that a composite fault of a turn-to-turn short circuit fault and a high-resistance connection fault of the permanent magnet synchronous motor cannot be accurately diagnosed in the prior art. The method comprises the steps of obtaining simulated motor operation parameters based on a simulation model of the motor, and inputting the simulated motor operation parameters into a prediction model to obtain predicted three-phase stator current; inputting the predicted three-phase stator current and the actual three-phase stator current of the permanent magnet synchronous motor into a classification network to diagnose and distinguish the types and occurrence positions of turn-to-turn short circuit faults and high-resistance connection faults, wherein the motor operation parameters comprise Ud and Uq signals and an electrical angular velocity signal omega e of a rotor; according to the prediction network based on the channel attention mechanism, the speed and generalization ability of network training are improved, and the classification network can accurately diagnose and distinguish the types and occurrence positions of the turn-to-turn short circuit fault and the high-resistance connection fault.

Description

Technical Field

[0001] The present invention relates to the field of motor fault diagnosis and identification, and in particular to a method and system for intelligent diagnosis of multiple faults of a permanent magnet synchronous motor. Background Art

[0002] In the context of the research and application of new energy vehicles, permanent magnet synchronous motor (PMSM) as the core drive motor has become one of the key technologies to promote the development of electric vehicles. With the transformation of the global energy structure, the promotion of environmental protection policies and the rapid growth of the electric vehicle market, new energy vehicles have higher and higher requirements for electric drive systems. Permanent magnet synchronous motors have been widely used in the field of new energy vehicles with their high efficiency, high power density and excellent control performance. In new energy vehicles, permanent magnet synchronous motors not only drive vehicles as a power source, but also directly affect the performance and safety of vehicles. Therefore, it is increasingly important to ensure the safety and reliable operation of its motor system itself. During the operation of permanent magnet synchronous motors, different types of motor faults may occur. Among them, turn-to-turn short circuit faults and high resistance connection faults are the two most common early faults in the stator winding faults of permanent magnet synchronous motors, and most of the fault characteristics after the faults are very similar. However, most of the current research on the stator winding faults of permanent magnet synchronous motors focuses on single fault detection, and specific diagnostic methods are only effective for specific faults, which seriously affects their diagnostic results. Therefore, in addition to fault diagnosis, it is also necessary to distinguish the fault types so as to take targeted preventive measures.

[0003] In existing research, the diagnosis methods for turn-to-turn short circuit faults and high-resistance connection faults are mainly divided into the following three types: fault signal-based methods, model analysis-based methods, and data-driven methods.

[0004] The fault signal-based method mainly collects various fault signals, including stator current, zero-sequence voltage signal, and motor torque, and then uses a specific signal processing method to extract fault features. However, this method is too dependent on operating conditions and is difficult to apply in actual industry. The principle of the diagnosis method based on model analysis is to establish an accurate fault motor model based on the mathematical logical relationship between the parameters of the permanent magnet synchronous motor, calculate the difference between the output of the fault model and the actual output of the motor, and compare the difference with the predefined diagnostic threshold to detect the motor fault. Although the model-based method is easy to analyze, the diagnostic accuracy is highly dependent on the relevant parameters of the model. As the motor system becomes more and more complex, the accurate modeling and diagnosis of motor faults face great challenges. Data-driven methods often ignore the engineering experience and physical knowledge in traditional fault diagnosis, and fail to effectively combine the two. This may lead to the limitation of the diagnostic recognition ability of the model in some complex fault situations, resulting in misdiagnosis. Summary of the invention

[0005] In order to solve the problems existing in the above-mentioned prior art, the present invention provides a permanent magnet synchronous motor multi-fault intelligent diagnosis method and system to solve the problem that the prior art cannot accurately diagnose the combined fault of the inter-turn short circuit fault and the high-resistance connection fault of the permanent magnet synchronous motor.

[0006] A method for intelligent diagnosis of multiple faults of a permanent magnet synchronous motor, comprising:

[0007] Based on the simulation model of the motor, the simulated motor operating parameters are obtained and input into the prediction model to obtain the predicted three-phase stator current. The predicted three-phase stator current and the actual three-phase stator current of the permanent magnet synchronous motor are input into the classification network to diagnose and identify the type and location of the inter-turn short circuit fault and the high-resistance connection fault. The motor operating parameters include U d , U q Signal, rotor electrical angular velocity signal ω e .

[0008] Furthermore, the prediction model is obtained by training and verification of a first data set, and the classification network is obtained by training and verification of a second data set. The first data set includes motor operating parameters and corresponding three-phase stator currents obtained from a drive system simulation model, and the second data set includes three-phase current signals and fault labels in healthy and fault states under different operating conditions.

[0009] Furthermore, the simulation model is built based on the action mechanism of the inter-turn short-circuit fault and the high-resistance connection fault of the permanent magnet synchronous motor, and the building process is as follows: the action mechanism of the inter-turn short-circuit fault and the high-resistance connection fault of the permanent magnet synchronous motor is analyzed, and according to the action mechanism obtained by the analysis, a simulation model of the drive system under the fault state of the permanent magnet synchronous motor is built through simulation software, and a simulation experiment is carried out to verify it.

[0010] Furthermore, the production process of the first data set is as follows:

[0011] Collect the motor’s U from the simulation model and the real motor respectively d , U q Signal, rotor electrical angular velocity signal ω e and three-phase stator currents to obtain a first data set;

[0012] The production process of the second data set is as follows:

[0013] The three-phase stator current of the motor is collected from the simulation model and the real motor respectively, and a fault label is given, and then preprocessed to obtain a second data set, wherein the fault label includes a fault location and a fault type, the fault location includes phase A, phase B, and phase C, and the fault type includes turn-to-turn short circuit fault, high resistance connection fault, and turn-to-turn short circuit fault mixed with high resistance connection fault;

[0014] The preprocessing includes first fusing multi-source and multi-dimensional information on the data, then normalizing the fused data to generate periodic data samples; expanding the sample data using a random overlapping sampling method, and finally dividing the data samples into a training set, a cross-validation set, and a test set according to a certain ratio.

[0015] Furthermore, the prediction model includes a data input module, a data preprocessing module, and a prediction network, and the prediction network includes a convolution module, a channel attention mechanism module, a flat layer and a fully connected layer. The convolution module consists of a convolution layer, a batch normalization layer, an activation function layer and a pooling layer.

[0016] The fault features are first extracted by the convolution layer, and then the extracted features are standardized using the batch normalization layer. The rectified linear unit is selected as the activation function to realize the nonlinear feature extraction of the network. The maximum pooling layer effectively reduces the size of the feature map while retaining the most important fault features of the data. The channel attention mechanism module is used to weight the importance of the feature channels to enhance the network's attention to key features, thereby improving the generalization ability and recognition accuracy of the model, helping to reduce the interference of redundant information, suppress irrelevant or noisy feature channels, enhance meaningful feature channels, and make the feature representation more concise and effective.

[0017] Finally, the extracted features are converted into prediction results through the flat layer and the fully connected layer, and the results are transmitted to the output layer to obtain the predicted three-phase stator current signal.

[0018] Furthermore, the motor operating parameters include: d Signal, U q Signal and permanent magnet synchronous motor rotor electrical angular velocity signal ω e , calculate the current signal i on the dq axis according to the motor operating parameters d ,i q , get the three-phase stator current, the calculation process is as follows:

[0019]

[0020] The solution is:

[0021]

[0022] Then, the current signal i on the dq axis is transformed by Park coordinate transformation and Clarke coordinate transformation. d 、i q Converted to three-phase stator current:

[0023]

[0024] Where: U d and U q is the stator voltage of the d-axis and q-axis; R s is the resistance of the stator winding; L d and L q is the inductance of the d-axis and q-axis; λ d and λ q is the magnetic flux of the d-axis and q-axis; ω e is the electrical angular velocity of the motor rotor.

[0025] Furthermore, the prediction model obtains the three-phase stator current at time t+1 by simulating the motor operating parameters at time t.

[0026] A permanent magnet synchronous motor multi-fault intelligent diagnosis system, comprising a data acquisition module, a simulation module, a prediction module and a classification module;

[0027] The data input processing module is used to obtain the actual operating parameters of the permanent magnet synchronous motor and perform preprocessing;

[0028] The simulation module is used to output the simulated operating parameters of the motor;

[0029] The prediction module is used to generate predicted three-phase stator currents according to the processing of the simulated operation parameters of the motor;

[0030] The classification module is used to diagnose and identify the types and locations of turn-to-turn short-circuit faults and high-resistance connection faults based on the predicted three-phase stator current and the actual three-phase stator current.

[0031] Furthermore, the prediction model includes a data input module, a data preprocessing module, and a prediction network, the prediction network includes a convolution module, a channel attention mechanism module, a flat layer and a fully connected layer, and the convolution module consists of a convolution layer, a batch normalization layer, an activation function layer and a pooling layer.

[0032] Furthermore, the simulation module includes a simulation model, which is built based on the action mechanism of the inter-turn short circuit fault and the high-resistance connection fault of the permanent magnet synchronous motor.

[0033] The beneficial effects of the present invention include:

[0034] 1. Aiming at the problem of insufficient permanent magnet synchronous motor fault data set, a prediction network and a classification network were constructed to diagnose the combined fault of inter-turn short circuit fault and high resistance connection fault;

[0035] 2. The prediction network based on the channel attention mechanism improves the speed and generalization ability of network training. It can predict the three-phase stator current in advance according to the operating parameters of the permanent magnet synchronous motor, and diagnose and warn it according to the predicted three-phase stator current signal;

[0036] 3. The diagnosis and identification of predicted current based on the convolution classification network can identify the types and locations of inter-turn short-circuit faults and high-resistance connection faults of permanent magnet synchronous motors in advance and accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is an equivalent circuit diagram of a turn-to-turn short-circuit fault involved in an embodiment of the present application.

[0038] Figure 2 This is an equivalent circuit diagram of a high-resistance connection fault involved in an embodiment of the present application.

[0039] Figure 3 This is a diagram of the simulation results of the inter-turn short-circuit fault involved in the embodiment of the present application.

[0040] Figure 4 This is a diagram of a high-resistance connection fault simulation model involved in an embodiment of the present application.

[0041] Figure 5 This is a schematic diagram of the prediction model structure involved in the embodiment of the present application.

[0042] Figure 6 This is a permanent magnet synchronous motor fault diagnosis flowchart involved in an embodiment of the present application. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0044] A method for intelligent diagnosis of multiple faults of a permanent magnet synchronous motor, comprising:

[0045] Based on the simulation model of the motor, the simulated operating parameters are obtained and input into the prediction model to obtain the predicted three-phase stator current. The predicted three-phase stator current and the actual three-phase stator current of the permanent magnet synchronous motor are input into the classification network to diagnose and identify the types and locations of turn-to-turn short-circuit faults and high-resistance connection faults.

[0046] In another embodiment, the prediction model obtains the three-phase stator current at time t+1 by simulating the motor operating parameters at time t.

[0047] By collecting the motor operating parameters under different working conditions of the fault simulation model, the motor operating parameters include Ud , U q Signal, rotor electrical angular velocity signal ω e。 It should be noted that the sampling frequency of the motor operating parameters in the simulation model is less than the motor's base frequency, and the base frequency calculation formula of the permanent magnet synchronous motor is as follows:

[0048]

[0049] Where: n s is the speed of the permanent magnet synchronous motor, f is the fundamental frequency of the motor, and p is the number of pole pairs.

[0050] When collecting motor operating parameters under different speed conditions, different sampling frequencies should be set. When the sampling frequency is less than the motor's base frequency, it can ensure that the simulated motor parameters obtained at time t contain complete feature information, so that the prediction network can fully learn its features and train the three-phase stator current at time t+1.

[0051] In another embodiment, the prediction model is obtained by training and verification of a first data set, and the classification network is obtained by training and verification of a second data set, the first data set includes motor operating parameters and corresponding three-phase stator currents obtained from a drive system simulation model, and the second data set includes three-phase current signals and fault labels in healthy and fault states under different operating conditions.

[0052] In another embodiment, the simulation model is built based on the action mechanism of the inter-turn short-circuit fault and the high-resistance connection fault of the permanent magnet synchronous motor, and the building process is as follows: the action mechanism of the inter-turn short-circuit fault and the high-resistance connection fault of the permanent magnet synchronous motor is analyzed, and according to the action mechanism obtained by the analysis, a simulation model of the drive system under the permanent magnet synchronous motor fault state is built through MATLAB / Simulink simulation software, and a simulation experiment is performed to verify it.

[0053] Here, the fault of phase A stator winding is taken as an example. The fault of phases B and C is the same.

[0054] like Figure 1 As shown in the figure, it is the equivalent circuit diagram of the permanent magnet synchronous motor with inter-turn short circuit. In the A-phase stator winding, the stator winding is short-circuited by the resistance R sf It is divided into two parts, one is the healthy state part shown in a1, and the other is the short circuit fault part shown in a2. sf Represents the resistance R sf The short-circuit current is defined as:

[0055]

[0056] Where: n is the number of short-circuit turns of the stator winding, and N is the total number of turns of the stator winding.

[0057] When a turn-to-turn short circuit fault occurs in phase A of the permanent magnet synchronous motor, the voltage equation of the permanent magnet synchronous motor is:

[0058]

[0059] Where R sf 、i sf 、V f 、e sf The expression is:

[0060]

[0061] i sf =[i a i b i c i f ] T

[0062] V f =[V a V b V c 0] T

[0063] e sf =[e a1 +e a2 e b e c e f ] T

[0064] Where: V f is the phase voltage matrix of the stator winding, R sf is the resistance matrix, i sf is the current matrix, e sf is the back EMF matrix of the three-phase stator winding, R a1 is the resistance of stator winding a1, R a2 is the resistance of stator winding a2, R b is the resistance of stator winding b; R c is the resistance of stator winding c, i a is the current through the A-phase stator winding, i b is the current through the B-phase stator winding, i c is the current through the C-phase stator winding, i f is the current flowing through the short-circuited wire; R f is the short-circuit resistance, V0 is the zero-sequence voltage; V a is the voltage across the A-phase stator winding, V b is the voltage across the B-phase stator winding, V cis the voltage across the C-phase stator winding, e a1 is the back electromotive force of stator winding a1, e a2 is the back electromotive force of stator winding a2, e b is the back electromotive force of stator winding b, e c is the back electromotive force of stator winding c, e f is the back electromotive force on the short-circuited wire.

[0065] The inductance L f The expression is:

[0066]

[0067] Where: L a1 , L a2 , L b and L c Represent the self-inductance of stator windings a1, a2, b and c respectively, and M j,k It represents the mutual inductance of stator windings j and k, where j∈{a1,a2,b,c}, k∈{a1,a2,b,c}).

[0068] When a permanent magnet synchronous motor has a turn-to-turn short circuit fault, the electromagnetic torque expression is:

[0069]

[0070] Where: r is the mechanical angular velocity of the motor, ω e is the electrical angular velocity of the rotor, n p is the number of magnetic pole pairs of the PMSM.

[0071] like Figure 2 As shown in the figure, it is an equivalent circuit diagram of a high-resistance connection fault of a permanent magnet synchronous motor. For a surface-mounted permanent magnet synchronous motor, its three-phase stator windings have the same resistance value, and the inductance between the three-phase stator windings is also the same, that is:

[0072] R a =R b =R c =R

[0073] M ab =M bc =M ca =M

[0074] When a high-resistance connection fault occurs in a permanent magnet synchronous motor, the stator resistance distribution is unbalanced. The high-resistance connection fault can be simulated by adding an additional resistor in series with the A-phase winding. The faulty PMSM equation with a high-resistance connection in phase A is expressed as:

[0075]

[0076] Where: u a 、u b and u c is the three-phase voltage of the stator winding, u0 is the voltage between the DC midpoint of the inverter and the neutral point center of the stator winding, i a 、i b and i c is the three-phase stator current, u xo (x=a、b、c) is u x The difference between u0 and R is the stator phase resistance, ΔR is the fixed resistance of the additional high resistance, L x (x=a, b, c) is the stator phase self-inductance of phases A, B, and C, M is the mutual inductance between the stator windings, and the other motion equations and magnetic flux equations are consistent with those of a healthy permanent magnet synchronous motor, which will not be elaborated here.

[0077] After analyzing the action mechanism of the inter-turn short-circuit fault and high-resistance connection fault of the permanent magnet synchronous motor, a simulation model of the inter-turn short-circuit fault and high-resistance connection fault of the permanent magnet synchronous motor was built in the Matlab / simulink simulation software, and the built simulation model was verified.

[0078] The motor speed is set to 300r / min, the motor load torque is set to 10N·m, the turn-to-turn short circuit fault and the high-resistance connection fault are set to occur on the A-phase stator winding, the fault occurrence time is 0.3s, and the total simulation length is 1s;

[0079] The simulation results of the permanent magnet synchronous motor with inter-turn short circuit fault are as follows: Figure 3 As shown in Figure 2, the simulation results of a high-resistance connection failure are as follows: Figure 4 shown.

[0080] The simulation results show that when a turn-to-turn short circuit fault occurs in the permanent magnet synchronous motor, the amplitude of the A-phase current increases significantly; when a high-resistance connection fault occurs in the permanent magnet synchronous motor, the amplitude of the A-phase current decreases significantly.

[0081] The simulation results prove the accuracy of the constructed fault model and provide data simulation support for the subsequent permanent magnet synchronous motor fault diagnosis and classification.

[0082] In another embodiment, the first data set is prepared as follows:

[0083] Collect the motor’s U from the simulation model and the real motor respectively d , U q Signal, rotor electrical angular velocity signal ω e and three-phase stator currents to obtain a first data set;

[0084] The production process of the second data set is as follows:

[0085] The three-phase stator current of the motor is collected from the simulation model and the real motor respectively, and a fault label is given, and then preprocessed to obtain a second data set, wherein the fault label includes a fault location and a fault type, the fault location includes phase A, phase B, and phase C, and the fault type includes turn-to-turn short circuit fault, high resistance connection fault, and turn-to-turn short circuit fault mixed with high resistance connection fault;

[0086] The preprocessing includes first fusing multi-source and multi-dimensional information on the data, then normalizing the fused data to generate periodic data samples; expanding the sample data using a random overlapping sampling method, and finally dividing the data samples into a training set, a cross-validation set, and a test set according to a certain ratio.

[0087] Taking the first data set as an example, the U of the motor collected in the simulation model d , U q Signal and permanent magnet synchronous motor rotor electrical angular velocity signal ω e Multi-source and multi-dimensional information fusion is performed, and the minimum-maximum normalization method is used to plan the data within the range of [-1, +1] to eliminate the impact of working condition changes on diagnostic accuracy and performance, and generate periodic data samples; and the random overlapping sampling method is used to expand the sample data. After that, the data samples are divided into training sets, cross-validation sets, and test sets according to a certain ratio.

[0088] Specifically, after the multi-source and multi-dimensional information fusion of the acquired data, data preprocessing is performed on it, mainly normalizing the original collected data, using the minimum-maximum normalization method, and its formula is expressed as:

[0089]

[0090] Where: max(x) is the maximum value of the data, min(x) is the minimum value of the data;

[0091] After normalizing the original data, the range of the input data is scaled to a small interval, so that the optimization algorithm (such as gradient descent) converges faster during the training of the prediction network model, and improves the training speed and prediction accuracy of the prediction model; since model training requires a large amount of data, the random overlapping sampling method is used to expand the normalized data, and the formula is:

[0092]

[0093] Where: L is the given data length, N is the window size, and S is the set step size.

[0094] In another embodiment, if Figure 5 As shown in the figure, the prediction model includes a data input module, a data preprocessing module, and a prediction network. The prediction network mainly includes three convolution modules, a channel attention mechanism module, a flat layer, and a fully connected layer. The convolution module consists of a convolution layer, a batch normalization layer, an activation function layer, and a pooling layer. In detail, the fault features of the permanent magnet synchronous motor are first extracted by the convolution layer, and then the extracted features are standardized using the batch normalization layer, so that the network training is faster and more stable; in addition, the rectified linear unit (ReLU) is selected as the activation function to realize the nonlinear feature extraction of the network. This structure is simple and has the function of preventing gradient disappearance and gradient explosion. The mathematical expression of ReLU can be expressed as:

[0095] ReLU(x)=max(0,x)

[0096] After the convolution module extracts the permanent magnet synchronous motor features, the maximum pooling layer effectively reduces the size of the feature map while retaining the most important fault features of the data.

[0097] Afterwards, the channel attention mechanism module is used to weight the importance of feature channels, and the input feature map is compressed in the spatial dimension through the pooling layer to obtain a global description of each channel; then the fully connected layer is used to transform these global descriptions to obtain the weight of each channel; and then these weights are applied to each channel of the original feature map to re-weight the channel features. This can enhance the network's attention to key features, thereby improving the model's generalization ability and recognition accuracy, helping to reduce the interference of redundant information, suppress those irrelevant or noisy feature channels, enhance meaningful feature channels, and make feature representation more concise and effective.

[0098] Finally, the extracted features are converted into predicted results through the flat layer and the fully connected layer, and this result is transmitted to the output layer to obtain the predicted i abc *Signal.

[0099] In another embodiment, the motor operating parameters include: d Signal, U q Signal and permanent magnet synchronous motor rotor electrical angular velocity signal ω e , get the three-phase stator current according to the motor operating parameters, and calculate the current signal i on the dq axis d ,i q , the calculation process is as follows:

[0100]

[0101] The solution is:

[0102]

[0103] Then, the current signal i on the dq axis is transformed by Park coordinate transformation and Clarke coordinate transformation. d 、i q Converted to three-phase stator current:

[0104]

[0105] Where: U d and U q is the stator voltage of the d-axis and q-axis; R s is the resistance of the stator winding; L d and L q is the inductance of the d-axis and q-axis; λ d and λ q is the magnetic flux of the d-axis and q-axis; ω e is the electrical angular velocity of the motor rotor.

[0106] In another embodiment, the virtual current value output by the prediction model layer and the current value collected in the physical model are input into the diagnosis network model to determine whether the motor fails, and to determine the type of the motor failure and the location of the failure.

[0107] Firstly, the three-phase current signals and corresponding fault labels under healthy and faulty conditions collected in the simulation model are input into the classification model. After relevant training and verification, the network structure of the classification model is finally determined to obtain the classification network for subsequent classification diagnosis.

[0108] Table 1 Corresponding fault labels for turn-to-turn short circuit fault and high-resistance connection fault

[0109]

[0110] In order to facilitate the subsequent diagnosis network to diagnose the motor fault type, different types of fault types and the location of occurrence data are sampled, and the corresponding labels are defined: as shown in Table 1: Motor fault labels include A-phase ISF, A-phase HRC, A-phase ISF mixed HRC; B-phase ISF, B-phase HRC, B-phase ISF mixed HRC; C-phase ISF, C-phase HRC, C-phase ISF mixed HRC; Health status. Among them, ISF refers to inter-turn short circuit fault, and HRC refers to high-resistance connection fault.

[0111] like Figure 6 As shown, after the above, the prediction model output i abc * Current signal and real-time acquisition of three-phase stator current i of the real motor abc Signal and the trained classification network are used to diagnose permanent magnet synchronous motor faults.

[0112] A permanent magnet synchronous motor multi-fault intelligent diagnosis system, comprising a data acquisition module, a simulation module, a prediction module and a classification module;

[0113] The data input processing module is used to obtain the actual operating parameters of the permanent magnet synchronous motor and perform preprocessing;

[0114] The simulation module is used to output the simulated operating parameters of the motor;

[0115] The prediction module is used to generate predicted three-phase stator currents based on processing of simulated operation parameters;

[0116] The classification module is used to diagnose and identify the types and locations of turn-to-turn short-circuit faults and high-resistance connection faults based on the predicted three-phase stator current and the actual three-phase stator current.

[0117] In another embodiment, the prediction model includes a prediction model, which includes a convolution module, an attention mechanism module, a flat layer and a fully connected layer, and the convolution module includes a convolution layer, a batch normalization layer, an activation function layer and a pooling layer.

[0118] In another embodiment, the simulation module includes a simulation model, which is built based on the action mechanism of the inter-turn short circuit fault and the high-resistance connection fault of the permanent magnet synchronous motor.

[0119] The above-mentioned embodiments only express the specific implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the protection scope of the present application. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the technical solution concept of the present application, and these all belong to the protection scope of the present application.

Claims

1. A method for intelligent diagnosis of multiple faults of permanent magnet synchronous motor, characterized in that: include: The simulated motor operating parameters are obtained based on the motor simulation model, and are input into the prediction model to obtain the predicted three-phase stator current. The predicted three-phase stator current and the actual three-phase stator current of the permanent magnet synchronous motor are input into the classification network to diagnose and identify the types and locations of turn-to-turn short-circuit faults and high-resistance connection faults. The motor operating parameters include U d , U q Signal, rotor electrical angular velocity signal ω e .

2. A method for intelligent diagnosis of multiple faults of a permanent magnet synchronous motor according to claim 1, characterized in that: The prediction model is obtained by training and verification of the first data set, and the classification network is obtained by training and verification of the second data set. The first data set includes the motor operating parameters and the corresponding three-phase stator currents obtained in the drive system simulation model, and the second data set includes the three-phase current signals and fault labels in the healthy state and fault state under different working conditions.

3. A method for intelligent diagnosis of multiple faults of a permanent magnet synchronous motor according to claim 1, characterized in that: The simulation model is built based on the action mechanism of the inter-turn short-circuit fault and the high-resistance connection fault of the permanent magnet synchronous motor. The construction process is as follows: the action mechanism of the inter-turn short-circuit fault and the high-resistance connection fault of the permanent magnet synchronous motor is analyzed, and according to the action mechanism obtained by the analysis, a simulation model of the drive system under the fault state of the permanent magnet synchronous motor is built through simulation software, and a simulation experiment is carried out to verify it.

4. The method for intelligent diagnosis of multiple faults of a permanent magnet synchronous motor according to claim 1, characterized in that: The construction process of the first data set is as follows: Collect the motor’s U from the simulation model and the real motor respectively d , U q Signal, rotor electrical angular velocity signal ω e and three-phase stator currents to obtain a first data set; The construction process of the second data set is as follows: The three-phase stator current of the motor is collected from the simulation model and the real motor respectively, and a fault label is given, and then preprocessed to obtain a second data set, wherein the fault label includes a fault location and a fault type, the fault location includes phase A, phase B, and phase C, and the fault type includes turn-to-turn short circuit fault, high resistance connection fault, and turn-to-turn short circuit fault mixed with high resistance connection fault; The preprocessing includes first fusing multi-source and multi-dimensional information on the data, then normalizing the fused data to generate periodic data samples; expanding the sample data using a random overlapping sampling method, and finally dividing the data samples into a training set, a cross-validation set, and a test set according to a certain ratio.

5. The method for intelligent diagnosis of multiple faults of a permanent magnet synchronous motor according to claim 1, characterized in that: The prediction model includes a data input module, a data preprocessing module and a prediction network. The prediction network includes a convolution module, an attention mechanism module, a flat layer and a fully connected layer. The convolution module includes a convolution layer, a batch normalization layer, an activation function layer and a pooling layer.

6. The method for intelligent diagnosis of multiple faults of a permanent magnet synchronous motor according to claim 1, characterized in that: Calculate the current signal i on the dq axis according to the motor operating parameters d ,i q , get the three-phase stator current, the calculation process is as follows: The solution is: Then, the current signal i on the dq axis is transformed by Park coordinate transformation and Clarke coordinate transformation. d 、i q Converted to three-phase stator current: Where: U d and U q is the stator voltage of the d-axis and q-axis; R s is the resistance of the stator winding; L d and L q is the inductance of the d-axis and q-axis; λ d and λ q is the magnetic flux of the d-axis and q-axis; ω e is the electrical angular velocity of the motor rotor.

7. The method for intelligent diagnosis of multiple faults of a permanent magnet synchronous motor according to claim 1, characterized in that: The prediction model obtains the three-phase stator current at time t+1 by simulating the motor operation parameters at time t.

8. A permanent magnet synchronous motor multi-fault intelligent diagnosis system, characterized in that: It includes data acquisition module, simulation module, prediction module and classification module; The data input processing module is used to obtain the actual operating parameters of the permanent magnet synchronous motor and perform preprocessing; The simulation module is used to output the simulated operating parameters of the motor; The prediction module is used to generate predicted three-phase stator currents based on processing of simulated operation parameters; The classification module is used to diagnose and identify the types and locations of turn-to-turn short-circuit faults and high-resistance connection faults based on the predicted three-phase stator current and the actual three-phase stator current.

9. A method for intelligent diagnosis of multiple faults of a permanent magnet synchronous motor according to claim 8, characterized in that: The prediction module includes a data input module, a data preprocessing module, and a prediction network. The prediction network includes a convolution module, an attention mechanism module, a flat layer, and a fully connected layer. The convolution module consists of a convolution layer, a batch normalization layer, an activation function layer, and a pooling layer.

10. The permanent magnet synchronous motor multi-fault intelligent diagnosis system according to claim 8, characterized in that: The simulation module includes a simulation model, which is built based on the action mechanism of inter-turn short circuit fault and high-resistance connection fault of the permanent magnet synchronous motor.

Images