Fault diagnosis methods and devices for motor systems

Abstract

This invention provides a method and apparatus for fault diagnosis of a motor system. The method includes: collecting actual characteristic data of the motor system to be observed; obtaining the motor system fault type corresponding to the actual characteristic data based on the actual characteristic data and a pre-established motor system fault diagnosis model; the motor system fault diagnosis model is trained based on feature data generated by a joint simulation model of the motor drive system corresponding to the motor system to be observed, and the feature data corresponds to a preset fault type. This application utilizes deep neural networks to diagnose motor system faults, possessing advantages such as good real-time performance, low cost, simple operation, and high fault diagnosis accuracy. This application utilizes simulation software for motor drive system modeling and reinforcement learning for training deep neural networks, solving the problem of obtaining a large amount of reliable and analyzable fault data of various types in existing model training.

Description

Technical Field

[0001] This invention relates to the field of motor system fault diagnosis, and more specifically to a method and apparatus for fault diagnosis of motor systems based on reinforcement learning. Background Technology

[0002] Various types of AC and DC motors are widely used in industrial manufacturing and daily life. In aerospace, high-speed rail, and automotive fields, motor control systems have extremely high requirements for the safety and reliability of motors. When a motor system operates under extreme conditions or a motor has already failed, if the motor control system does not take appropriate measures, it may induce more serious faults in the generator, leading to equipment damage and even endangering the lives of personnel. Therefore, for fields with high requirements for the safety and reliability of motor systems, accurate and reliable motor system fault diagnosis is a prerequisite for the motor control system to correctly respond to motor system fault conditions and a necessary means to ensure the safe and stable operation of motor equipment systems.

[0003] Currently, scholars have proposed various fault diagnosis schemes for motor systems. For example:

[0004] Option 1: Install a magnetic sensor inside the motor. Since motors are high-speed rotating devices, and large motors also have characteristics such as high voltage and high current, modifying the motor structure to install a sensor is not only difficult and uneconomical, but may also affect the motor's design performance.

[0005] Option 2: Establish a corresponding mathematical model for motor system faults, and use the mathematical model to solve for relevant information to determine the motor system fault. However, because there are many types of potential motor faults, and the location of faults is often random, it is difficult to establish an effective and unified mathematical model for motor system faults in practical applications.

[0006] Option 3: Collect information such as motor current, voltage, and sound, and process it using methods such as spectrum analysis and neural networks to diagnose motor system faults based on the processed information. This option requires pre-judgment of fault characteristics and setting relevant thresholds, but in practical applications, it is difficult to artificially create a large number of various faults and obtain a large amount of analyzable fault data of various types.

[0007] It is evident that existing motor system fault diagnosis schemes have drawbacks such as altering motor structure, affecting motor performance, increasing costs, difficulty in establishing effective and unified fault mathematical models, difficulty in obtaining a large amount of reliable and analyzable fault data of various types, and poor real-time performance. Summary of the Invention

[0008] To address the problems existing in the prior art and to facilitate more accurate fault diagnosis of motor systems, this application provides a method for fault diagnosis of motor systems, comprising:

[0009] Collect actual characteristic data of the motor system under observation;

[0010] Based on the actual feature data and the pre-established motor system fault diagnosis model, the motor system fault type corresponding to the actual feature data is obtained;

[0011] The fault diagnosis model of the motor system is trained based on the feature data generated by the joint simulation model of the motor drive system corresponding to the motor system to be observed, and the feature data corresponds to the preset fault type.

[0012] In one embodiment, the fault diagnosis method for the motor system further includes:

[0013] Establish a joint simulation model of the motor drive system based on the rated parameters of the motor system to be observed;

[0014] Based on the preset fault type and the joint simulation model of the motor drive system, the feature data corresponding to the preset fault type are obtained;

[0015] The deep neural network model is trained using the feature data to obtain the fault diagnosis model of the motor system.

[0016] In one embodiment, the motor system under observation includes a motor body and a drive system. The step of establishing a joint simulation model of the motor drive system based on the rated parameters of the motor system under observation includes:

[0017] The rated parameters of the motor body in the motor system to be observed are input into the finite element simulation software to obtain the finite element simulation model of the motor body.

[0018] A motor drive control algorithm model is established based on the rated parameters of the drive system in the motor system to be observed.

[0019] The finite element simulation model of the motor and the motor drive control algorithm model are combined to obtain the joint simulation model of the motor drive system.

[0020] In one embodiment, the preset fault type includes a drive system fault;

[0021] The step of obtaining feature data corresponding to the preset fault type based on the preset fault type and the joint simulation model of the motor drive system includes:

[0022] The drive system fault is input into the motor drive control algorithm model in the motor drive system co-simulation model, so that the motor drive system co-simulation model outputs the feature data corresponding to the drive system fault.

[0023] In one embodiment, the preset fault type includes a motor body fault; obtaining the feature data corresponding to the preset fault type based on the preset fault type and the joint simulation model of the motor drive system includes:

[0024] The motor body fault is input into the motor finite element simulation model in the motor drive system co-simulation model, so that the motor drive system co-simulation model outputs the feature data corresponding to the motor body fault.

[0025] In one embodiment, when inputting the drive system fault into the motor drive control algorithm model in the motor drive system co-simulation model, the method further includes:

[0026] The first noise and the drive system fault are input together into the motor drive control algorithm model in the joint simulation model of the motor drive system to obtain the feature data corresponding to the drive system fault.

[0027] In one embodiment, the step of inputting the motor body fault into the motor finite element simulation model in the joint simulation model of the motor drive system further includes:

[0028] The second noise and the motor body fault are input together into the motor finite element simulation model in the joint simulation model of the motor drive system to obtain the feature data corresponding to the motor body fault.

[0029] In one embodiment, training the deep neural network model using the feature data includes:

[0030] Step 1: Input the feature data into a pre-constructed first deep neural network model to obtain the first fault type;

[0031] Step 2: Compare the first fault type with the preset fault type corresponding to the feature data according to the pre-established reward and punishment mechanism, and use the reinforcement learning algorithm to train the first deep neural network model based on the comparison results;

[0032] If the first deep neural network model does not converge, update the first deep neural network model and iteratively execute steps 1 to 2 until the first deep neural network model converges.

[0033] In one embodiment, the fault diagnosis method for the motor system further includes:

[0034] Establish a second deep neural network model with the same structure as the first deep neural network model;

[0035] Extract the parameters of the trained first deep neural network model and assign the parameters to the second deep neural network model to obtain the motor system fault diagnosis model.

[0036] Secondly, this application provides a fault diagnosis device for an electric motor system, comprising:

[0037] The feature data acquisition module is used to acquire the actual feature data of the motor system under observation.

[0038] The motor system fault type determination module is used to obtain the motor system fault type corresponding to the actual characteristic data based on the actual characteristic data and the pre-established motor system fault diagnosis model.

[0039] The fault diagnosis model of the motor system is trained based on the feature data generated by the joint simulation model of the motor drive system corresponding to the motor system to be observed, and the feature data corresponds to the preset fault type.

[0040] In one embodiment, the fault diagnosis device for the motor system further includes:

[0041] The co-simulation model building module is used to build a co-simulation model of the motor drive system based on the rated parameters of the motor system to be observed.

[0042] The feature data acquisition module is used to obtain feature data corresponding to the preset fault type based on the preset fault type and the joint simulation model of the motor drive system.

[0043] The motor system fault diagnosis model training module is used to train the deep neural network model using the feature data to obtain the motor system fault diagnosis model.

[0044] In one embodiment, the motor system to be observed includes a motor body and a drive system, and the co-simulation model building module includes:

[0045] The finite element simulation model establishment unit is used to input the rated parameters of the motor body in the motor system to be observed into the finite element simulation software to obtain the motor finite element simulation model corresponding to the motor body.

[0046] The motor drive control algorithm model establishment unit is used to establish a motor drive control algorithm model based on the rated parameters of the drive system in the motor system to be observed.

[0047] The model combination unit is used to combine the finite element simulation model of the motor and the motor drive control algorithm model to obtain the joint simulation model of the motor drive system.

[0048] In one embodiment, the preset fault types include motor body faults and drive system faults;

[0049] The feature data acquisition module includes:

[0050] A drive system fault injection unit is used to input the drive system fault into the motor drive control algorithm model in the motor drive system co-simulation model, so that the motor drive system co-simulation model outputs the feature data corresponding to the drive system fault.

[0051] The motor body fault injection unit is used to input the motor body fault into the motor finite element simulation model in the motor drive system co-simulation model, so that the motor drive system co-simulation model outputs the feature data corresponding to the motor body fault.

[0052] In one embodiment, the drive system fault injection unit is specifically used for:

[0053] The first noise and the drive system fault are input together into the motor drive control algorithm model in the joint simulation model of the motor drive system to obtain the feature data corresponding to the drive system fault.

[0054] In one embodiment, the motor body fault injection unit is specifically used for:

[0055] The second noise and the motor body fault are input together into the motor finite element simulation model in the joint simulation model of the motor drive system to obtain the feature data corresponding to the motor body fault.

[0056] In one embodiment, the motor system fault diagnosis model training module is specifically used to perform the following steps:

[0057] Step 1: Input the feature data into a pre-constructed first deep neural network model to obtain the first fault type;

[0058] Step 2: Compare the first fault type with the preset fault type corresponding to the feature data according to the pre-established reward and punishment mechanism, and use the reinforcement learning algorithm to train the first deep neural network model based on the comparison results;

[0059] If the first deep neural network model does not converge, update the first deep neural network model and iteratively execute steps 1 to 2 until the first deep neural network model converges.

[0060] In one embodiment, the fault diagnosis device for the motor system further includes a motor system fault diagnosis model establishment module, used for:

[0061] Establish a second deep neural network model with the same structure as the first deep neural network model;

[0062] Extract the parameters of the trained first deep neural network model and assign the parameters to the second deep neural network model to obtain the motor system fault diagnosis model.

[0063] Thirdly, this application also provides an electronic device, comprising:

[0064] The system includes a central processing unit, a memory, and a communication module. The memory stores a computer program, and the central processing unit can call the computer program. When the central processing unit executes the computer program, it implements any of the fault diagnosis methods for motor systems provided in this application.

[0065] Fourthly, this application also provides a computer storage medium for storing a computer program, which, when executed by a processor, implements any of the fault diagnosis methods for motor systems provided in this application.

[0066] The fault diagnosis method and apparatus for motor systems disclosed in this application utilize deep neural networks to diagnose motor system faults. This method offers advantages such as good real-time performance, low cost, ease of operation, and high fault diagnosis accuracy, all without requiring changes to the motor structure. Compared to other schemes using neural networks for motor system fault diagnosis, this application utilizes simulation software for modeling the motor drive system and training the deep neural network. This allows for the easy acquisition of millions or even tens of millions of fault data points suitable for training, solving the problem of obtaining large amounts of reliable and analyzable fault data in existing model training methods. Attached Figure Description

[0067] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0068] Figure 1 This is a schematic diagram of the fault diagnosis method for the motor system provided in this application.

[0069] Figure 2 Another schematic diagram of the fault diagnosis method for the motor system provided in this application.

[0070] Figure 3 Another schematic diagram of the fault diagnosis method for the motor system provided in this application.

[0071] Figure 4This is a schematic diagram of the joint simulation model of the motor drive system of this application.

[0072] Figure 5 Another schematic diagram of the fault diagnosis method for the motor system provided in this application.

[0073] Figure 6 This is a schematic diagram of the feature data output from the finite element simulation model of the motor in this application.

[0074] Figure 7 A schematic diagram of the method for training the first deep neural network provided in this application.

[0075] Figure 8 A schematic diagram of the framework for training the first deep neural network provided in this application.

[0076] Figure 9 This is a schematic diagram illustrating the use of the motor system fault diagnosis model of this application for fault diagnosis.

[0077] Figure 10 A schematic diagram of the fault diagnosis device for the motor system provided in this application.

[0078] Figure 11 Another schematic diagram of the fault diagnosis device for the motor system provided in this application.

[0079] Figure 12 Another schematic diagram of the fault diagnosis device for the motor system provided in this application.

[0080] Figure 13 Another schematic diagram of the fault diagnosis device for the motor system provided in this application.

[0081] Figure 14 Another schematic diagram of the fault diagnosis device for the motor system provided in this application.

[0082] Figure 15 A schematic diagram of an electronic device provided in this application. Detailed Implementation

[0083] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0084] Firstly, such as Figure 1 As shown, this application provides a fault diagnosis method for a motor system, which includes the following steps:

[0085] Step S101: Collect actual characteristic data of the motor system to be observed.

[0086] Specifically, the actual characteristic data of the observed motor system refers to the operating parameter values ​​during the operation of the observed motor system, including but not limited to the command values ​​of motor angle position, motor speed, motor torque, actual values ​​of motor angle position, actual values ​​of motor speed, actual values ​​of motor current, the square of the actual value of motor current, actual values ​​of motor torque, and the historical values ​​of the above operating parameters counting n times backward from the current time t, that is, the historical values ​​of each operating parameter at the current time t, time t-1, ..., time tn, where n is a positive integer.

[0087] The motor system to be observed in this application can be any type of AC motor system or DC motor system.

[0088] Step S102: Based on the actual feature data and the pre-established motor system fault diagnosis model, obtain the motor system fault type corresponding to the actual feature data; wherein, the motor system fault diagnosis model is trained based on the feature data generated by the joint simulation model of the motor drive system corresponding to the observed motor system, and the feature data corresponds to the preset fault type.

[0089] Specifically, the input data of the motor system fault diagnosis model is the actual characteristic data of the motor system under observation, and the output data is the fault type of the motor system obtained based on the actual characteristic data of the motor system under observation. In practical applications, the actual characteristic data of the motor system under observation is input into the motor system fault diagnosis model to obtain the current fault state and fault type of the motor.

[0090] The motor system fault diagnosis model in this application is obtained by training a deep neural network model. The training data used in training the motor system fault diagnosis model is feature data generated by the co-simulation model of the motor drive system corresponding to the observed motor system, and this feature data corresponds to a preset fault type. This application generates training data through the co-simulation model of the motor drive system, easily obtaining millions or even tens of millions of fault data of various types that can be used for training. This solves the problem of obtaining a large amount of reliable and analyzable fault data of various types in existing schemes using neural networks for motor system fault diagnosis. Furthermore, the motor system fault diagnosis method of this application utilizes a deep neural network to diagnose motor system faults, possessing advantages such as good real-time performance, low cost, simple operation, and high fault diagnosis accuracy, without requiring any changes to the motor structure.

[0091] In one embodiment, such as Figure 2 As shown, the fault diagnosis method for the motor system further includes the step of training a fault diagnosis model for the motor system, specifically including the following steps:

[0092] Step S201: Establish a joint simulation model of the motor drive system based on the rated parameters of the motor system to be observed.

[0093] Specifically, the joint simulation model of the motor drive system includes two parts: a motor finite element simulation model and a motor drive control algorithm model. The motor finite element simulation model is a simulation model established based on the rated parameters of the motor body in the motor system under observation. Through the motor finite element simulation model, various operating states of the motor body can be simulated and the operating parameters of the motor body under different operating states can be obtained. The motor drive control algorithm model is a simulation model established based on the rated parameters of the drive system in the motor system under observation. It is used to drive the motor finite element simulation model. Through the motor drive control algorithm model, various operating states of the drive system can be simulated and the operating parameters of the drive system can be obtained, realizing the simulation of the drive system's drive control of the motor body.

[0094] The co-simulation model of the motor drive system allows for the simulation of the operating state of the observed motor system under arbitrary motor system faults by changing the input parameters of the simulation model. Compared to establishing complex mathematical models of motor system faults, the process of establishing the simulation model of the observed motor system in this application simplifies the model establishment and simulation operation procedures.

[0095] Step S202: Based on the preset fault type and the joint simulation model of the motor drive system, obtain the feature data corresponding to the preset fault type.

[0096] Specifically, the preset fault types include two categories: one is drive system faults, including but not limited to short circuit / open circuit faults in the drive system power circuit; the other is motor body faults, including but not limited to short circuit / open circuit faults between motor turns, short circuit faults between motor phases, phase loss faults, grounding short circuit faults, and demagnetization faults (for permanent magnet motors).

[0097] The input data of the motor drive system co-simulation model is one or more of the aforementioned preset fault types, and the output data is the operating parameters of the observed motor system when encountering the corresponding preset fault type, obtained through simulation, which is also the feature data generated by the motor drive system co-simulation model mentioned in step S102 above.

[0098] Step S203: Use the feature data to train the deep neural network model to obtain the motor system fault diagnosis model.

[0099] In this embodiment, the training data used to train the motor system fault diagnosis model is generated by a pre-established motor drive system co-simulation model based on preset fault types. By generating training data through the motor drive system co-simulation model, millions or even tens of millions of fault data of various types that can be used for training can be easily obtained, solving the problem in existing schemes that use neural networks for motor system fault diagnosis that it is difficult to obtain a large amount of reliable and analyzable fault data of various types.

[0100] In one embodiment, such as Figure 3 As shown, the motor system under observation includes a motor body and a drive system. Step S201 involves establishing a joint simulation model of the motor drive system based on the rated parameters of the motor system under observation, including:

[0101] Step S2011: Input the rated parameters of the motor body in the motor system to be observed into the finite element simulation software to obtain the finite element simulation model of the motor body.

[0102] Specifically, existing finite element simulation software can be used, such as Ansys, Abaqus, and MSC. The specific parameters involved in simulating and modeling the motor body depend on the type of motor. For example, for a permanent magnet motor, the rated parameters of the motor body include, but are not limited to, stator outer diameter, stator inner diameter, stator large tooth count, stator small tooth count, rotor outer diameter, axial length, rotor tooth count, magnetic ring thickness, magnetic ring outer diameter, magnetic ring inner diameter, stator winding turns, number of motor phases, number of motor steps, and step angle.

[0103] After inputting the rated parameters of the motor body into the finite element simulation software, a finite element simulation model of the motor corresponding to the physical motor can be obtained. Since finite element simulation can accurately simulate actual systems, and this motor finite element simulation model has the same parameters as the motor body, the characteristic parameters obtained when using this motor finite element simulation model for simulation can be approximately considered as the actual characteristic parameters generated during motor operation.

[0104] Step S2012: Establish a motor drive control algorithm model based on the rated parameters of the drive system in the motor system to be observed.

[0105] The motor drive control algorithm model can perform electrical control on the motor finite element simulation model, changing the operating state of the motor body simulated by the finite element simulation model; at the same time, the motor drive control algorithm model will receive the motor operating state data fed back by the motor finite element simulation model, realizing closed-loop control of the motor.

[0106] Step S2013: Combine the finite element simulation model of the motor and the motor drive control algorithm model to obtain the joint simulation model of the motor drive system.

[0107] Figure 4 This is a schematic diagram of the joint simulation model of the motor drive system obtained through steps S2011 to S2013. The motor drive control algorithm model outputs a voltage signal to the motor finite element simulation model. The motor finite element simulation model feeds back the motor current, speed (or rotational position), and other parameter values ​​under the drive of the voltage signal to the motor drive control algorithm model, enabling it to achieve closed-loop control of the motor. After combining the motor drive control algorithm model and the motor finite element simulation model to establish the joint simulation model of the motor drive system, real-time joint simulation of the operating state of the motor system to be observed can be achieved.

[0108] In one embodiment, such as Figure 5 As shown, in step S202, based on the preset fault type and the joint simulation model of the motor drive system, the characteristic data corresponding to the preset fault type is obtained, including:

[0109] Step S2021: When the preset fault type is a drive system fault, the drive system fault is input into the motor drive control algorithm model in the motor drive system co-simulation model, so that the motor drive system co-simulation model outputs the feature data corresponding to the drive system fault.

[0110] In this context, drive system faults refer to faults occurring within the motor drive system. Therefore, to simulate a fault in the motor drive system, the drive system fault must be injected into the motor drive control algorithm model corresponding to the motor drive system. At this point, the motor drive system co-simulation model can output the characteristic data corresponding to the drive system fault.

[0111] Step S2022: When the preset fault type is a motor body fault, the motor body fault is input into the motor finite element simulation model in the motor drive system co-simulation model, so that the motor drive system co-simulation model outputs the feature data corresponding to the motor body fault.

[0112] Among these, motor body faults refer to faults occurring within the motor itself. Therefore, to simulate a fault in the motor body, the motor body fault must be injected into the finite element simulation model corresponding to the motor body. At this point, the co-simulation model of the motor drive system can output the characteristic data corresponding to the drive system fault.

[0113] The joint simulation model of the motor drive system will normalize the output feature data and establish the correspondence between the feature data and the preset fault types.

[0114] In this embodiment, steps S2021 and S2022 are not necessarily both required to be executed; this depends on the preset fault type to be injected. When the preset fault type to be injected is a drive system fault, executing only step S2021 is sufficient to obtain the corresponding feature data; when the preset fault type to be injected is a motor body fault, executing only step S2022 is sufficient to obtain the corresponding feature data; when the preset fault type to be injected includes both drive system faults and motor body faults, then both steps S2021 and S2022 must be executed. The terms "step S2021" and "step S2022" in this embodiment are for illustrative purposes only and are not intended to limit the execution order of steps S2021 and S2022.

[0115] Figure 6 This diagram illustrates the feature data obtained through the joint simulation model of the motor drive system in this application. The type of feature data here is the same as the type of actual feature data of the observed motor system in step S101, including but not limited to the motor angle position command value, motor speed command value, motor torque command value, actual motor angle position value, actual motor speed value, actual motor current value, the square of the actual motor current value, the actual motor torque value, and the historical values ​​of the above operating parameters counting n times backward from the current time t, i.e., the historical values ​​of each operating parameter at the current time t, time t-1, ..., time tn, where n is a positive integer. Among them, the motor angle position command value, motor speed command value, and motor torque command value are output by the motor drive control algorithm model, while the actual motor angle position value, actual motor speed value, actual motor current value, the square of the actual motor current value, and actual motor torque value are output by the motor finite element simulation model.

[0116] Figure 6 In the feature data shown, θ * (t), θ * (t-1)...θ * (tn) represent the command values ​​for the motor rotation angle at time t, time t-1, ..., time tn, respectively; ω * (t), ω * (t-1)...ω * (tn) represent the command values ​​for motor speed at time t, time t-1, ..., time tn, respectively; Let θ(t), θ(t-1), ..., θ(tn) be the commanded values ​​of the motor torque at time t, θ(t-1), ..., θ(tn) be the actual values ​​of the motor rotation angle at time t, θ(t-1), ..., θ(tn) be the actual values ​​of the motor rotation angle at time t, θ(t-1), ..., θ(tn) be the actual values ​​of the motor rotation speed at time t, θ(t-1), ..., θ(tn) be the actual values ​​of the motor rotation speed at time t, θ(t-1), ..., θ(tn) be the actual values ​​of the motor rotation angle at time t, θ(t-1), ..., θ(tn) be the actual values ​​of the motor rotation angle at time t, θ(t-1), ..., θ(tn) be the actual values ​​of the motor rotation current at time t, θ(t-1), ..., θ(tn) be the actual values ​​of the motor rotation angle ... 2 (t), I 2 (t-1)...I 2 (tn) represent the squares of the actual values ​​of the motor current at time t, t-1, ..., tn, respectively; T e (t), T e (t-1)...T e (tn) represents the actual value of the motor torque at time t, time t-1, ..., time tn, respectively; n is a positive integer.

[0117] In one embodiment, when inputting the drive system fault into the motor drive control algorithm model in the motor drive system co-simulation model, the method further includes:

[0118] The first noise and the drive system fault are input together into the motor drive control algorithm model in the joint simulation model of the motor drive system to obtain the feature data corresponding to the drive system fault.

[0119] In one embodiment, the step of inputting the motor body fault into the motor finite element simulation model in the joint simulation model of the motor drive system further includes:

[0120] The second noise and the motor body fault are input together into the motor finite element simulation model in the joint simulation model of the motor drive system to obtain the feature data corresponding to the motor body fault.

[0121] In one embodiment, step S2021, when inputting the drive system fault into the motor drive control algorithm model in the motor drive system co-simulation model, further includes:

[0122] The first noise and the drive system fault are input together into the motor drive control algorithm model in the joint simulation model of the motor drive system to obtain the feature data corresponding to the drive system fault.

[0123] Step S2022, the step of inputting the motor body fault into the motor finite element simulation model in the joint simulation model of the motor drive system, further includes:

[0124] The second noise and the motor body fault are input together into the motor finite element simulation model in the joint simulation model of the motor drive system to obtain the feature data corresponding to the motor body fault.

[0125] In this embodiment, during the process of obtaining feature data, the first noise and the drive system fault are input together into the motor drive control algorithm model, or the second noise and the motor body fault are input together into the motor finite element simulation model. In practical applications, noise can be added at each stage of the simulation. The noise in this application includes, but is not limited to, Gaussian noise. Adding noise can compensate for the error between the established ideal motor drive system co-simulation model and the actual motor drive system, and improve the robustness and generalization ability of the motor system fault diagnosis model trained with feature data in the later stage.

[0126] In one embodiment, step S203, training the deep neural network model using the feature data, includes:

[0127] Step 1: Input the feature data into a pre-constructed first deep neural network model to obtain the first fault type.

[0128] Step 2: Compare the first fault type with the preset fault type corresponding to the feature data according to the pre-established reward and punishment mechanism, and use the reinforcement learning algorithm to train the first deep neural network model based on the comparison results;

[0129] If the first deep neural network model does not converge, update the first deep neural network model and iteratively execute steps 1 to 2 until the first deep neural network model converges.

[0130] Specifically, it can be determined according to Figure 7 The process shown is executed as follows:

[0131] 1) Establish the first deep neural network model.

[0132] 2) Input the feature data obtained by the motor drive system co-simulation model in step S202 into the first deep neural network model, so that the first deep neural network model outputs the first fault type.

[0133] The first fault type is the fault determination result of the current first deep neural network model based on the feature data. The feature data input to the first neural network model is normalized feature data.

[0134] 3) Compare the first fault type output by the first deep neural network model with the preset fault type corresponding to the feature data, and determine whether the first deep neural network has converged based on the comparison result. If yes, proceed to step 6); if no, proceed to step 4).

[0135] 4) Using a pre-designed reward and punishment mechanism for reinforcement learning, the reinforcement learning algorithm is trained based on the comparison results to adjust the parameters of the first deep neural network model.

[0136] 5) Repeat steps 2) to 3 above.

[0137] 6) The first deep neural network model training is complete, and the process ends.

[0138] A schematic diagram of the above process can be found here. Figure 8 .

[0139] In one embodiment, the fault diagnosis method for the motor system further includes:

[0140] Establish a second deep neural network model with the same structure as the first deep neural network model;

[0141] Extract the parameters of the trained first deep neural network model and assign the parameters to the second deep neural network model to obtain the motor system fault diagnosis model.

[0142] In this embodiment, the second deep neural network model has the same structure and parameters as the first deep neural network model. Therefore, the second deep neural network model can be used as a fault diagnosis model for motor system to diagnose motor system faults. Figure 9 This is a schematic diagram illustrating the use of a motor system fault diagnosis model for motor system fault diagnosis. For example... Figure 9 As shown, the observed motor system includes a motor body and a drive system. When the observed motor system is running, the drive system sends a voltage signal to the motor body to drive its operation. The motor body feeds back parameters such as current and speed during operation to the motor drive system to achieve closed-loop control. In practical applications, real-time characteristic data of the observed motor system is collected and input into the motor system fault diagnosis model. The motor system fault diagnosis model can then output the motor system fault type corresponding to the actual characteristic data online in real time, thus diagnosing the motor system fault. The motor system fault type output by the motor system fault diagnosis model is consistent with the preset fault type used when generating the characteristic data for training the motor system fault diagnosis model.

[0143] This application re-establishes a second deep neural network model, assigning it the same parameters and structure as the trained first deep neural network model. The second deep neural network model is then used for motor system fault diagnosis. Training of the first deep neural network model can be performed offline, while the fault diagnosis process using the motor system fault diagnosis model (second deep neural network model) is conducted online in real time. If subsequent updates to the parameters of the motor system fault diagnosis model are needed, the first deep neural network model can be trained without affecting the use of the second deep neural network model. Then, the parameters of the second deep neural network model can be updated at a specific time.

[0144] In summary, the fault diagnosis method for motor systems proposed in this application utilizes deep neural networks to diagnose motor system faults, offering advantages such as good real-time performance, low cost, ease of operation, and high fault diagnosis accuracy, all without requiring changes to the motor structure. Compared to other schemes using neural networks for motor system fault diagnosis, this application utilizes simulation software for modeling the motor drive system and training the deep neural network, easily obtaining millions or even tens of millions of fault data points suitable for training. This solves the problem of obtaining large amounts of reliable and analyzable fault data for various types in existing model training.

[0145] Based on the same inventive concept, this application also provides a fault diagnosis device for a motor system, which can be used to implement the method described in the above embodiments, as shown in the following embodiments. Since the principle of the fault diagnosis device for a motor system in solving problems is similar to that of the fault diagnosis method for a motor system, the implementation of the fault diagnosis device for a motor system can refer to the implementation of the fault diagnosis method for a motor system, and repeated details will not be elaborated further. As used below, the terms "unit" or "module" can refer to a combination of software and / or hardware that implements a predetermined function. Although the system described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0146] Secondly, this application provides a fault diagnosis device for a motor system, such as... Figure 10 As shown, the fault diagnosis device for this motor system includes:

[0147] The feature data acquisition module 901 is used to acquire the actual feature data of the motor system under observation.

[0148] The motor system fault type determination module 902 is used to obtain the motor system fault type corresponding to the actual characteristic data based on the actual characteristic data and the pre-established motor system fault diagnosis model.

[0149] The fault diagnosis model of the motor system is trained based on the feature data generated by the joint simulation model of the motor drive system corresponding to the motor system to be observed, and the feature data corresponds to the preset fault type.

[0150] In one embodiment, such as Figure 11 As shown, the fault diagnosis device for the motor system further includes:

[0151] The co-simulation model establishment module 903 is used to establish a co-simulation model of the motor drive system based on the rated parameters of the motor system to be observed.

[0152] The feature data acquisition module 904 is used to obtain feature data corresponding to the preset fault type based on the preset fault type and the joint simulation model of the motor drive system.

[0153] The motor system fault diagnosis model training module 905 is used to train the deep neural network model using the feature data to obtain the motor system fault diagnosis model.

[0154] In one embodiment, the motor system to be observed includes a motor body and a drive system, such as... Figure 12 As shown, the co-simulation model building module 903 includes:

[0155] The finite element simulation model establishment unit 9031 is used to input the rated parameters of the motor body in the motor system to be observed into the finite element simulation software to obtain the motor finite element simulation model corresponding to the motor body.

[0156] The motor drive control algorithm model establishment unit 9032 is used to establish a motor drive control algorithm model based on the rated parameters of the drive system in the motor system to be observed.

[0157] The model combination unit 9033 is used to combine the motor finite element simulation model and the motor drive control algorithm model to obtain the joint simulation model of the motor drive system.

[0158] In one embodiment, the preset fault types include motor body faults and drive system faults; such as Figure 13 As shown, the feature data acquisition module 904 includes:

[0159] The drive system fault injection unit 9041 is used to input the drive system fault into the motor drive control algorithm model in the motor drive system co-simulation model, so that the motor drive system co-simulation model outputs the feature data corresponding to the drive system fault.

[0160] The motor body fault injection unit 9042 is used to input the motor body fault into the motor finite element simulation model in the motor drive system co-simulation model, so that the motor drive system co-simulation model outputs the feature data corresponding to the motor body fault.

[0161] In one embodiment, the drive system fault injection unit 9041 is specifically used for:

[0162] The first noise and the drive system fault are input together into the motor drive control algorithm model in the joint simulation model of the motor drive system to obtain the feature data corresponding to the drive system fault.

[0163] In one embodiment, the motor body fault injection unit 9042 is specifically used for:

[0164] The second noise and the motor body fault are input together into the motor finite element simulation model in the joint simulation model of the motor drive system to obtain the feature data corresponding to the motor body fault.

[0165] In one embodiment, the motor system fault diagnosis model training module 905 is specifically used to perform the following steps:

[0166] Step 1: Input the feature data into a pre-constructed first deep neural network model to obtain the first fault type;

[0167] Step 2: Compare the first fault type with the preset fault type corresponding to the feature data according to the pre-established reward and punishment mechanism, and use the reinforcement learning algorithm to train the first deep neural network model based on the comparison results;

[0168] If the first deep neural network model does not converge, update the first deep neural network model and iteratively execute steps 1 to 2 until the first deep neural network model converges.

[0169] In one embodiment, such as Figure 14 As shown, the fault diagnosis device for the motor system further includes a motor system fault diagnosis model establishment module 906, used for:

[0170] Establish a second deep neural network model with the same structure as the first deep neural network model;

[0171] Extract the parameters of the trained first deep neural network model and assign the parameters to the second deep neural network model to obtain the motor system fault diagnosis model.

[0172] The fault diagnosis device for motor systems presented in this application utilizes deep neural networks to diagnose motor system faults. It offers advantages such as good real-time performance, low cost, ease of operation, and high fault diagnosis accuracy, all without requiring changes to the motor structure. Compared to other solutions using neural networks for motor system fault diagnosis, this application utilizes simulation software for modeling the motor drive system and training the deep neural network. This allows for the easy acquisition of millions or even tens of millions of fault data points suitable for training, solving the problem of obtaining large amounts of reliable and analyzable fault data in existing model training methods.

[0173] Thirdly, the present invention also provides an electronic device, see [link to relevant documentation]. Figure 15 The electronic device 100 specifically includes:

[0174] The system includes a central processing unit (CPU) 110, a memory 120, a communication module 130, an input unit 140, an output unit 150, and a power supply 160.

[0175] The memory 120, communication module 130, input unit 140, output unit 150, and power supply 160 are all connected to the central processing unit 110. The memory 120 stores a computer program, which the central processing unit 110 can call. When the central processing unit 110 executes the computer program, it implements all the steps in the fault diagnosis method for any of the motor systems described in the above embodiments.

[0176] Fourthly, embodiments of this application also provide a computer storage medium for storing a computer program that can be executed by a processor. When the computer program is executed by the processor, it implements any of the fault diagnosis methods for motor systems provided by this invention.

[0177] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, the embodiments of this specification can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, the embodiments of this specification can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The various embodiments in this specification are described in a progressive manner, and similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for system embodiments, since they are basically similar to method embodiments, the description is relatively simple, and relevant parts can be referred to the description of the method embodiments. In the description of this specification, the reference to the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., means that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the embodiments of this specification.

[0178] In this specification, the illustrative expressions of the terms used do not necessarily refer to the same embodiments or examples. Furthermore, those skilled in the art can combine and integrate different embodiments or examples described in this specification, as well as features of different embodiments or examples, without contradiction. The above descriptions are merely embodiments of this specification and are not intended to limit the embodiments of this specification. Various modifications and variations can be made to the embodiments of this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the embodiments of this specification should be included within the scope of the claims of the embodiments of this specification.

Claims

1. A method of diagnosing a failure of an electric motor system, characterized by, The method comprises the following steps: Collecting actual characteristic data of the motor system to be observed, the actual characteristic data including but not limited to motor angle position command value, motor speed command value, motor torque command value, motor angle position actual value, motor speed actual value, motor current actual value, square value of the motor current actual value, motor torque actual value, and historical values of each operating parameter from the current time t to n time points in the past; According to the actual characteristic data and a pre-established motor system fault diagnosis model, the motor system fault type corresponding to the actual characteristic data is obtained; The motor system fault diagnosis model is generated based on the characteristic data of the motor drive system joint simulation model corresponding to the motor system to be observed, and the characteristic data corresponds to a preset fault type; The method further comprises the following steps: Establishing a motor drive system joint simulation model according to the rated parameters of the motor system to be observed; According to the preset fault type and the motor drive system joint simulation model, the characteristic data corresponding to the preset fault type is obtained; The characteristic data is used to train a deep neural network model to obtain the motor system fault diagnosis model, wherein noise is added to each link of the simulation, and the noise includes Gaussian noise; The training of the deep neural network model using the characteristic data comprises the following steps: Step 1: input the characteristic data into a pre-constructed first deep neural network model to obtain a first fault type; Step 2: compare the first fault type with the preset fault type corresponding to the characteristic data according to a pre-established reward and punishment mechanism, and use a reinforcement learning algorithm to train the first deep neural network model according to the comparison result; If the first deep neural network model does not converge, update the first deep neural network model, and iteratively execute steps 1 and 2 until the first deep neural network model converges; The motor system to be observed comprises a motor body and a drive system, and the establishment of the motor drive system joint simulation model according to the rated parameters of the motor system to be observed comprises the following steps: Input the rated parameters of the motor body in the motor system to be observed into a finite element simulation software to obtain a motor finite element simulation model corresponding to the motor body; Establish a motor drive control algorithm model according to the rated parameters of the drive system in the motor system to be observed; Combine the motor finite element simulation model and the motor drive control algorithm model to obtain the motor drive system joint simulation model.

2. The failure diagnostic method of an electric motor system according to claim 1, characterized by, The preset fault type includes a drive system fault; The characteristic data corresponding to the preset fault type is obtained according to the preset fault type and the motor drive system joint simulation model, which comprises the following steps: Input the drive system fault into the motor drive control algorithm model in the motor drive system joint simulation model, so that the motor drive system joint simulation model outputs the characteristic data corresponding to the drive system fault.

3. The method of claim 1, wherein The preset fault type includes a motor body fault; The characteristic data corresponding to the preset fault type is obtained according to the preset fault type and the motor drive system joint simulation model, which comprises the following steps: The motor body fault is input into a motor finite element simulation model in the motor drive system co-simulation model, so that the motor drive system co-simulation model outputs characteristic data corresponding to the motor body fault.

4. The failure diagnostic method of an electric motor system according to claim 2, characterized by, When the drive system fault is input into the motor drive control algorithm model in the motor drive system co-simulation model, the motor drive system co-simulation model further comprises: The first noise and the drive system fault are input into the motor drive control algorithm model in the motor drive system co-simulation model, and characteristic data corresponding to the drive system fault is obtained.

5. The method of claim 3, wherein The motor body fault is input into a motor finite element simulation model in the motor drive system co-simulation model, so that the motor drive system co-simulation model outputs characteristic data corresponding to the motor body fault. The second noise and the motor body fault are input into the motor finite element simulation model in the motor drive system co-simulation model, and characteristic data corresponding to the motor body fault is obtained.

6. The failure diagnostic method of an electric motor system according to claim 1, characterized by, Further comprising: A second deep neural network model with the same structure as the first deep neural network model is established; Parameters of the trained first deep neural network model are extracted and assigned to the second deep neural network model, and the motor system fault diagnosis model is obtained.

7. A failure diagnosing device of an electric motor system characterized by comprising: Comprise: A characteristic data acquisition module is configured to acquire actual characteristic data of a motor system to be observed, which includes but is not limited to motor angle position command value, motor speed command value, motor torque command value, motor angle position actual value, motor speed actual value, motor current actual value, square value of the motor current actual value, motor torque actual value, and historical values of the above-mentioned operating parameters from the current time t to n time points in the past; A motor system fault type determination module is configured to obtain a motor system fault type corresponding to the actual characteristic data according to the actual characteristic data and a pre-established motor system fault diagnosis model; The motor system fault diagnosis model is trained based on characteristic data generated by a motor drive system co-simulation model corresponding to the motor system to be observed, and the characteristic data corresponds to a preset fault type; A co-simulation model establishment module is configured to establish a motor drive system co-simulation model according to rated parameters of the motor system to be observed; A characteristic data acquisition module is configured to obtain characteristic data corresponding to the preset fault type according to the preset fault type and the motor drive system co-simulation model; A motor system fault diagnosis model training module is configured to train a deep neural network model using the characteristic data, and obtain the motor system fault diagnosis model, wherein noise is added in each link of the simulation, and the noise includes Gaussian noise; The motor system fault diagnosis model training module is specifically configured to perform the following steps: Step 1: input the characteristic data into a pre-constructed first deep neural network model to obtain a first fault type; Step 2: compare the first fault type with a preset fault type corresponding to the characteristic data according to a pre-established reward and punishment mechanism, and use a reinforcement learning algorithm to train the first deep neural network model according to a comparison result. If the first deep neural network model does not converge, updating the first deep neural network model, iteratively performing steps 1 to 2 until the first deep neural network model converges; The motor system to be observed includes a motor body and a driving system, and the joint simulation model establishing module includes: A finite element simulation model establishing unit, configured to input rated parameters of the motor body in the motor system to be observed into finite element simulation software to obtain a motor finite element simulation model corresponding to the motor body; A motor driving control algorithm model establishing unit, configured to establish a motor driving control algorithm model according to rated parameters of the driving system in the motor system to be observed; A model combining unit, configured to combine the motor finite element simulation model and the motor driving control algorithm model to obtain the motor driving system joint simulation model.

8. The fault diagnostic apparatus of the motor system according to claim 7, characterized by The preset fault types include motor body faults and driving system faults; The feature data obtaining module includes: A driving system fault injection unit, configured to input the driving system fault into the motor driving control algorithm model in the motor driving system joint simulation model, so that the motor driving system joint simulation model outputs feature data corresponding to the driving system fault; A motor body fault injection unit, configured to input the motor body fault into the motor finite element simulation model in the motor driving system joint simulation model, so that the motor driving system joint simulation model outputs feature data corresponding to the motor body fault.

9. The fault diagnostic apparatus of an electric motor system according to claim 8, characterized by The driving system fault injection unit is specifically configured to: Input the first noise and the driving system fault into the motor driving control algorithm model in the motor driving system joint simulation model to obtain the feature data corresponding to the driving system fault.

10. The fault diagnostic apparatus of the motor system according to claim 8, characterized by, The motor body fault injection unit is specifically configured to: Input the second noise and the motor body fault into the motor finite element simulation model in the motor driving system joint simulation model to obtain the feature data corresponding to the motor body fault.

11. The fault diagnostic apparatus of an electric motor system according to claim 7, characterized by Further including a motor system fault diagnosis model establishing module, configured to: establish a second deep neural network model with the same structure as the first deep neural network model; extract parameters of the trained first deep neural network model and assign the parameters to the second deep neural network model to obtain the motor system fault diagnosis model.

12. An electronic device, comprising: It includes: A central processing unit, a memory, and a communication module, the memory stores a computer program, the central processing unit can call the computer program, and the central processing unit implements the motor system fault diagnosis method according to any one of claims 1 to 6 when executing the computer program.

13. A computer storage medium storing a computer program, characterized in that, The computer program is executed by the processor to implement the motor system fault diagnosis method according to any one of claims 1 to 6.

Images