Industrial system fault diagnosis and state monitoring system and method
By introducing deep learning technology and multi-module collaborative working methods in industrial system fault diagnosis and status monitoring systems, the problem of low fault diagnosis accuracy in traditional systems is solved, and higher fault identification accuracy and real-time maintenance support are achieved.
Patent Information
- Application Number
- CN202510211372.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-25
- Publication Date
- 2025-06-13
AI Technical Summary
Due to the differences in equipment operating conditions and operating environment factors in traditional industrial system fault diagnosis and status monitoring systems, the learned non-causal characteristics have poor generalization capabilities, resulting in errors in fault feature extraction, hindering the accuracy of system diagnosis.
An industrial system fault diagnosis and status monitoring system is designed, including sensor module, fault diagnosis module, status monitoring module and display module. The fault diagnosis module performs deep processing and feature recognition through Bayesian convolutional neural analysis submodule, deep residual convolutional neural analysis submodule, meta-learning analysis submodule and transfer learning analysis submodule to improve the accuracy of fault diagnosis.
Through this system, it can accurately identify the abnormal status and potential faults of the equipment, improve the accuracy of equipment fault diagnosis, and help users correct maintenance errors in time and improve the reliability of equipment operation through real-time three-dimensional display and maintenance plan display.
Smart Images
Figure CN120145185A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of industrial big data, and specifically to an industrial system fault diagnosis and condition monitoring system and method. Background Art
[0002] With the development of modern industry, whether the equipment can operate safely, reliably and in the best state is of great significance for ensuring product quality, improving the production capacity of enterprises and ensuring safe production. Among them, there is a process of occurrence and development from normal to faulty for industrial equipment. Therefore, the operating conditions of the equipment should be checked and measured in a daily, continuous and standardized working state, that is, condition monitoring or state monitoring, which is a part of equipment management work. The current industrial system fault diagnosis and condition monitoring system specifically includes that the basic composition of the industrial equipment operating state detection system mainly includes sensors, data acquisition modules, data processing modules and user interfaces. Sensors are the "eyes" of the system, responsible for collecting various operating data of the equipment, such as temperature, pressure, vibration, etc. These data are transmitted to the data processing module through the data acquisition module for analysis. The data processing module uses algorithms to analyze the data in real time and displays the results on the user interface for reference by management personnel.
[0003] However, the traditional industrial system fault diagnosis and condition monitoring system has the following disadvantages:
[0004] During the process of fault diagnosis by the traditional industrial system fault diagnosis and condition monitoring system, due to factors such as equipment working conditions and operating environments, the data generated during the operation of mechanical equipment are different, resulting in non-causal features learned can only show good performance on the training set, with poor generalization ability, making the system have errors in extracting fault features of the equipment, which hinders the accuracy of system diagnosis. Summary of the Invention
[0005] The purpose of the present invention is to provide an industrial system fault diagnosis and condition monitoring system and method to solve the problem that during the process of fault diagnosis by the traditional industrial system fault diagnosis and condition monitoring system, due to factors such as equipment working conditions and operating environments, the data generated during the operation of mechanical equipment are different, resulting in non-causal features learned can only show good performance on the training set, with poor generalization ability, making the system have errors in extracting fault features of the equipment, which hinders the accuracy of system diagnosis as mentioned in the above background art.
[0006] To achieve the above object, the present invention provides the following technical solutions: an industrial system fault diagnosis and condition monitoring system, including a monitoring system, the monitoring system includes a sensor module, a fault diagnosis module, a condition monitoring module and a display module, the sensor module is connected to the fault diagnosis module, the fault diagnosis module is connected to the condition monitoring module, and the condition monitoring module is connected to the display module;
[0007] The sensor module monitors various parameters of industrial equipment and transmits them to the fault diagnosis module. The fault diagnosis module deeply processes the collected equipment data to identify abnormal states and potential faults of the equipment. The condition monitoring module simulates the equipment maintenance process based on the data analysis results and adjusts the deficiencies after simulation. The display module displays the maintenance process and maintenance plan in real-time three dimensions.
[0008] As a preferred technical solution of the present invention, the sensor module includes an equipment sensing sub-module and an information transmission sub-module;
[0009] The equipment sensing sub-module monitors various parameters and characteristics of the equipment in real-time through temperature sensors, pressure sensors, and vibration sensors. The information transmission sub-module transmits the real-time data collected by the sensors to the fault diagnosis module.
[0010] As a preferred technical solution of the present invention, the fault diagnosis module includes a data preprocessing sub-module, a data analysis sub-module, and a fault identification sub-module. The data preprocessing sub-module is connected to the data analysis sub-module, and the data analysis sub-module is connected to the fault identification sub-module;
[0011] The data preprocessing sub-module collects industrial system sensor data signals through the sensor module and performs normalization and segmentation processing to obtain multivariate time series segments. The data analysis sub-module deeply processes the collected equipment data to identify abnormal states and potential faults of the equipment. The fault identification sub-module determines and identifies equipment faults based on the data analysis results of the data analysis sub-module.
[0012] As a preferred technical solution of the present invention, the data analysis sub-module includes a Bayesian convolutional neural analysis sub-module, a deep residual convolutional neural analysis sub-module, a meta-learning analysis sub-module, and a transfer learning analysis sub-module;
[0013] The Bayesian convolutional neural analysis sub-module provides predictive uncertainty information after data analysis. The deep residual convolutional neural analysis sub-module captures the structural characteristics of industrial equipment better through multi-level feature learning, thereby improving the accuracy of prediction. The meta-learning analysis sub-module helps the model find suitable parameters faster on new equipment by learning the distribution of parameters on multiple equipment parameters. The transfer learning analysis sub-module reduces the model training time and computational cost by transferring the parameters of the pre-trained model.
[0014] As a preferred technical solution of the present invention, the status monitoring module includes an equipment maintenance simulation sub-module, an equipment parameter comparison sub-module, and an equipment anomaly warning sub-module;
[0015] The equipment maintenance simulation sub-module simulates the maintenance process of the equipment to be repaired. The equipment parameter comparison sub-module compares the equipment parameters after the simulated maintenance with the stored normal equipment parameters to determine whether the equipment maintenance is correct. The equipment anomaly warning sub-module warns about the anomalies of the equipment.
[0016] As a preferred technical solution of the present invention, the display module includes a three-dimensional dynamic display sub-module, a maintenance display sub-module, and a status display sub-module;
[0017] The three-dimensional dynamic display sub-module performs real-time three-dimensional dynamic display of the industrial equipment maintenance process. The maintenance display sub-module displays the maintenance plan and maintenance documents for the industrial equipment to be repaired. The status display sub-module displays the operating status of the industrial equipment in real time.
[0018] The usage method of the industrial system fault diagnosis and status monitoring system of the present invention includes the following steps:
[0019] Step 1, information acquisition: The sensor module monitors parameters and characteristics such as temperature and pressure of the operating equipment of the industrial system in real time and transmits the monitoring data to the fault diagnosis module;
[0020] Step 2, fault diagnosis: The fault diagnosis module deeply processes the collected equipment data to identify abnormal states and potential faults of the equipment;
[0021] Step 3, simulation maintenance: Perform equipment simulation maintenance according to the diagnosis result and correct the maintenance errors;
[0022] Step 4, plan display: The display module displays the efficient maintenance plan and the operating status of the industrial equipment in real-time three dimensions.
[0023] As a preferred technical solution of the present invention, in the fault diagnosis module in Step 2, the Bayesian convolutional neural analysis sub-module measures the expected value of the possibility by calculating the possibility cost. The calculation formula is as follows:
[0024] ELBO(D,θ)=KL(q(w|θ)|p(w))-E q(w|θ) (logp(D|w)),
[0025] where q(w|θ) is the variational distribution, p(w) is the weight, and the specific Bayesian formula in the Bayesian convolutional neural analysis sub-module is that if events A 1 ,......A n are mutually exclusive, When P(B)>0, we have The calculation cost formula of the depth convolution of the deep residual convolution neural analysis sub-module in the fault diagnosis module in the second step is as follows:
[0026]
[0027] where D f is the spatial dimension of the input feature map, M is the number of input channels, D K is the size of the convolution kernel, and the calculation formula of the standard convolution is as follows:
[0028]
[0029] D f is the spatial dimension of the input feature map, M is the number of input channels, N is the number of output channels, D K is the size of the convolution kernel, and the calculation formula of the loss function of the meta-learning analysis sub-module in the fault diagnosis module in the second step is as follows:
[0030]
[0031] Φ is a training hyperparameter, and the calculation formula of the feedback function of the transfer learning analysis sub-module in the fault diagnosis module in the second step is as follows:
[0032]
[0033] The superscripts s and t represent the source domain and the target domain respectively, represents a distribution metric function, r(s,a,s') represents that the state s becomes the state s' after the action a, Φ represents the corresponding feature, B j-1 and B j ; respectively represent the data of a batch in the (j-1)-th round and the j-th round of iteration. The MMD calculation formula is as follows:
[0034] MMD 2 =sup f∈ζ (E x-P [f(x)]-E y-Q [f(y)]) 2 ,
[0035] where F is a set of functions in RKHS, and E x-P [f(x)] and E y-Q [f(y)] respectively represent the expected values of the function f under the distributions P and Q.
[0036] Compared with the prior art, the beneficial effects of the present invention are:
[0037] 1. By setting up a fault diagnosis module, the data preprocessing module converts the industrial system fault problem into a graph classification problem, and obtains the corresponding feature maps through the methods of the Bayesian convolutional neural analysis sub-module, the deep residual convolutional neural analysis sub-module, the meta-learning analysis sub-module, and the transfer learning analysis sub-module, so as to obtain accurate device fault features and non-device fault features, and improve the accuracy of device fault diagnosis;
[0038] 2. By setting up a status monitoring module, the equipment maintenance simulation sub-module simulates the maintenance process of the equipment to be repaired, the equipment parameter comparison sub-module compares the equipment parameters after the simulated maintenance with the stored normal equipment parameters to judge whether the equipment maintenance is correct, and the equipment anomaly warning sub-module warns of the equipment anomalies, facilitating the timely correction of the errors in the maintenance plan;
[0039] 3. By setting up a display module, the three-dimensional dynamic display sub-module dynamically displays the industrial equipment maintenance process in real time, the maintenance display sub-module displays the maintenance plan and maintenance documents for the industrial equipment to be repaired, and the status display sub-module displays the operating status of the industrial equipment in real time, facilitating the user to view the monitoring results. BRIEF DESCRIPTION OF THE DRAWINGS
[0040] Figure 1 is a schematic structural diagram of the present invention;
[0041] Figure 2 is a schematic architecture diagram of the sensor module of the present invention;
[0042] Figure 3 is a schematic architecture diagram of the fault diagnosis module of the present invention
[0043] Figure 4 is a schematic architecture diagram of the data analysis sub-module of the present invention;
[0044] Figure 5 is a schematic architecture diagram of the status monitoring module of the present invention;
[0045] Figure 6 is a schematic architecture diagram of the display module of the present invention;
[0046] Figure 7 is a flow chart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0047] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.
[0048] Please refer to Figure 1-7 , the present invention provides an industrial system fault diagnosis and condition monitoring system, including a monitoring system, the monitoring system includes a sensor module, a fault diagnosis module, a condition monitoring module and a display module, the sensor module is connected to the fault diagnosis module, the fault diagnosis module is connected to the condition monitoring module, and the condition monitoring module is connected to the display module;
[0049] The sensor module monitors various parameters of industrial equipment and transmits them to the fault diagnosis module. The fault diagnosis module deeply processes the collected equipment data to identify abnormal states and potential faults of the equipment. The condition monitoring module simulates the equipment maintenance process based on the data analysis results and adjusts the deficiencies after simulation. The display module three-dimensionally displays the maintenance process and maintenance plan in real time.
[0050] The sensor module includes an equipment sensing sub-module and an information transmission sub-module;
[0051] The equipment sensing sub-module monitors various parameters and characteristics of the equipment in real time through temperature sensors, pressure sensors, and vibration sensors. The information transmission sub-module transmits the real-time data collected by the sensors to the fault diagnosis module.
[0052] The fault diagnosis module includes a data preprocessing sub-module, a data analysis sub-module, and a fault identification sub-module. The data preprocessing sub-module is connected to the data analysis sub-module, and the data analysis sub-module is connected to the fault identification sub-module;
[0053] The data preprocessing sub-module collects industrial system sensor data signals through the sensor module and performs normalization and segmentation processing to obtain multivariate time series segments. The data analysis sub-module deeply processes the collected equipment data to identify abnormal states and potential faults of the equipment. The fault identification sub-module determines and identifies equipment faults based on the data analysis results of the data analysis sub-module.
[0054] The data analysis sub-module includes a Bayesian convolutional neural analysis sub-module, a deep residual convolutional neural analysis sub-module, a meta-learning analysis sub-module, and a transfer learning analysis sub-module;
[0055] The Bayesian convolutional neural analysis sub-module provides predictive uncertainty information after data analysis. The deep residual convolutional neural analysis sub-module captures the structural features of industrial equipment better through multi-level feature learning, thereby improving the accuracy of prediction. The meta-learning analysis sub-module helps the model find suitable parameters faster on new equipment by learning the distribution of parameters on multiple equipment parameters. The transfer learning analysis sub-module reduces the model training time and computational cost by transferring the parameters of the pre-trained model.
[0056] The condition monitoring module includes an equipment maintenance simulation sub-module, an equipment parameter comparison sub-module, and an equipment anomaly warning sub-module;
[0057] The equipment maintenance simulation sub-module simulates the maintenance process of the equipment that needs to be repaired. The equipment parameter comparison sub-module compares the equipment parameters after the simulated maintenance with the stored normal equipment parameters to determine whether the equipment maintenance is correct. The equipment anomaly warning sub-module warns about the equipment anomalies.
[0058] The display module includes a three-dimensional dynamic display sub-module, a maintenance display sub-module, and a status display sub-module;
[0059] The three-dimensional dynamic display sub-module three-dimensionally dynamically displays the industrial equipment maintenance process in real time. The maintenance display sub-module displays the maintenance plan and maintenance documents for the industrial equipment that needs to be repaired. The status display sub-module displays the operating status of the industrial equipment in real time.
[0060] The usage method of the industrial system fault diagnosis and status monitoring system of the present invention includes the following steps:
[0061] Step 1, information acquisition: The sensor module monitors the temperature, pressure and other parameters and characteristics of the operating equipment of the industrial system in real time and transmits the monitoring data to the fault diagnosis module;
[0062] Step 2, fault diagnosis: The fault diagnosis module deeply processes the collected equipment data to identify the abnormal status and potential faults of the equipment;
[0063] Step 3, simulated maintenance: Perform equipment simulated maintenance according to the diagnosed results and correct the maintenance errors;
[0064] Step 4, solution display: The display module three-dimensionally displays the efficient maintenance solution and the operating status of the industrial equipment in real time.
[0065] In step 2, the Bayesian convolutional neural analysis sub-module in the fault diagnosis module measures the expected value of the possibility by calculating the possibility cost. The calculation formula is as follows:
[0066] ELBO(D,θ)=KL(q(w|θ)|p(w))-E q(w|θ) (logp(D|w)),
[0067] where q(w|θ) is the variational distribution, p(w) is the weight. The specific Bayesian formula in the Bayesian convolutional neural analysis sub-module is that if events A 1 ,......A n are mutually exclusive, then when P(B)>0, there is In step 2, the calculation cost formula of the deep convolution of the deep residual convolutional neural analysis sub-module in the fault diagnosis module is as follows:
[0068]
[0069] where D f is the spatial dimension of the input feature map, M is the number of input channels, and D K is the size of the convolutional kernel. The calculation formula for standard convolution is as follows:
[0070]
[0071] D f is the spatial dimension of the input feature map, M is the number of input channels, N is the number of output channels, and D K is the size of the convolutional kernel. The calculation formula for the loss function of the meta - learning analysis sub - module in the fault diagnosis module in step 2 is as follows:
[0072]
[0073] Φ is a training hyperparameter. The calculation formula for the feedback function of the transfer - learning analysis sub - module in the fault diagnosis module in step 2 is as follows:
[0074]
[0075] The superscripts s and t represent the source domain and the target domain respectively, represents a distribution metric function, r(s, a, s') represents that the state s becomes the state s’ after the action a, Φ represents the corresponding feature, and B j-1 and B j ; represent the data of one batch in the (j - 1)-th round and the j - th round of iteration respectively. The calculation formula for MMD is as follows:
[0076] MMD 2 = sup f∈ζ (E x-P [f(x)] - E y-Q [f(y)]) 2 ,
[0077] where F is a set of functions in RKHS, and E x-P [f(x)] and E y-Q [f(y)] represent the expected values of the function f under the distributions P and Q respectively.
[0078] In the present invention, the sensor module monitors the temperature, pressure, various parameters and characteristics of the operating equipment in the industrial system in real time and transmits the monitoring data to the fault diagnosis module; the device sensing sub-module monitors the various parameters and characteristics of the equipment in real time through temperature sensors, pressure sensors and vibration sensors, and the information transmission sub-module transmits the real-time data collected by the sensors to the fault diagnosis module. The fault diagnosis module deeply processes the collected device data to identify the abnormal state and potential faults of the device. The data preprocessing sub-module collects the sensor data signals of the industrial system sensors through the sensor module and performs normalization and segmentation processing to obtain multi-source time series segments. The data analysis sub-module deeply processes the collected device data to identify the abnormal state and potential faults of the device. The fault identification sub-module determines and identifies the device faults according to the data analysis results of the data analysis sub-module, performs device simulation maintenance according to the diagnosed results, and corrects the maintenance errors; the device maintenance simulation sub-module simulates the maintenance process of the equipment to be repaired. The device parameter comparison sub-module compares the device parameters after the simulated maintenance with the stored normal device parameters to judge whether the device maintenance is correct. The device abnormal warning sub-module warns the abnormal parts of the device. The display module displays the efficient maintenance plan and the operation status of the industrial equipment in real-time three dimensions. The three-dimensional dynamic display sub-module dynamically displays the industrial equipment maintenance process in real time in three dimensions. The maintenance display sub-module displays the maintenance plan and maintenance documents for the industrial equipment to be repaired. The status display sub-module displays the operation status of the industrial equipment in real time;
[0079] The hydraulic system of the stamping equipment of the industrial equipment failed due to abnormal increase in oil temperature. After investigation, it was found that the cooler in the hydraulic system failed, resulting in poor heat dissipation, and finally leading to the failure of the hydraulic system. The fault diagnosis module deeply processes the collected device data to identify the abnormal state and potential faults of the device. The data preprocessing sub-module collects the sensor data signals of the industrial system sensors through the sensor module and performs normalization and segmentation processing to obtain multi-source time series segments. The data analysis sub-module deeply processes the collected device data to identify the abnormal state and potential faults of the device. The fault identification sub-module determines and identifies the device faults according to the data analysis results of the data analysis sub-module, performs device simulation maintenance according to the diagnosed results, and corrects the maintenance errors; the device maintenance simulation sub-module simulates the maintenance process of the equipment to be repaired. The device parameter comparison sub-module compares the device parameters after the simulated maintenance with the stored normal device parameters to judge whether the device maintenance is correct.
[0080] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.
Claims
1. An industrial system fault diagnosis and condition monitoring system, including a monitoring system, characterized in that: The monitoring system comprises a sensor module, a fault diagnosis module, a state monitoring module and a display module, wherein the sensor module is connected to the fault diagnosis module, the fault diagnosis module is connected to the state monitoring module, and the state monitoring module is connected to the display module; The sensor module monitors various parameters of industrial equipment and transmits them to the fault diagnosis module. The fault diagnosis module deeply processes the collected equipment data to identify abnormal conditions and potential faults of the equipment. The status monitoring module simulates the equipment maintenance process according to the data analysis results and adjusts the deficiencies after the simulation. The display module displays the maintenance process and maintenance plan in real time in three dimensions.
2. The industrial system fault diagnosis and status monitoring system according to claim 1, characterized in that: The sensor module includes a device sensing submodule and an information transmission submodule; The equipment sensing submodule monitors the various parameters and characteristics of the equipment in real time through temperature sensors, pressure sensors, and vibration sensors. The information transmission submodule transmits the real-time data collected by the sensors to the fault diagnosis module.
3. The industrial system fault diagnosis and status monitoring system according to claim 1, characterized in that: The fault diagnosis module includes a data preprocessing submodule, a data analysis submodule and a fault identification submodule, wherein the data preprocessing submodule is connected to the data analysis submodule, and the data analysis submodule is connected to the fault identification submodule; The data preprocessing submodule collects the sensor data signals of the industrial system through the sensor module, normalizes and segments them to obtain multivariate time series segments. The data analysis submodule performs in-depth processing on the collected equipment data to identify abnormal conditions and potential faults of the equipment. The fault identification submodule determines and identifies equipment faults based on the data results analyzed by the data analysis submodule.
4. The industrial system fault diagnosis and status monitoring system according to claim 3, characterized in that: The data analysis submodule includes a Bayesian convolutional neural analysis submodule, a deep residual convolutional neural analysis submodule, a meta-learning analysis submodule and a transfer learning analysis submodule; The Bayesian convolutional neural analysis submodule provides prediction uncertainty information after data analysis; the deep residual convolutional neural analysis submodule better captures the structural characteristics of industrial equipment through multi-level feature learning, thereby improving the accuracy of prediction; the meta-learning analysis submodule helps the model find suitable parameters faster on new equipment by learning the distribution of parameters on multiple device parameters; the transfer learning analysis submodule reduces model training time and computing costs by migrating the parameters of pre-trained models.
5. The industrial system fault diagnosis and status monitoring system according to claim 1, characterized in that: The status monitoring module includes an equipment maintenance simulation submodule, an equipment parameter comparison submodule and an equipment abnormality warning submodule; The equipment maintenance simulation submodule simulates the maintenance process of the equipment that needs to be repaired. The equipment parameter comparison submodule compares the equipment parameters after the simulated maintenance with the stored normal equipment parameters to determine whether the equipment maintenance is correct. The equipment abnormality warning submodule warns of equipment abnormalities.
6. The industrial system fault diagnosis and status monitoring system according to claim 1, characterized in that: The display module includes a three-dimensional dynamic display submodule, a maintenance display submodule and a status display submodule; The three-dimensional dynamic display submodule displays the maintenance process of industrial equipment in three dimensions in real time. The maintenance display submodule displays the maintenance plan and maintenance documents for industrial equipment that needs maintenance. The status display submodule displays the operating status of industrial equipment in real time.
7. The method for using the industrial system fault diagnosis and status monitoring system according to any one of claims 1 to 6, characterized in that: The following steps are involved: Step 1: Information acquisition: The sensor module monitors the temperature, pressure and other parameters and characteristics of the industrial system operating equipment in real time and transmits the monitoring data to the fault diagnosis module; Step 2: Fault diagnosis: The fault diagnosis module performs in-depth processing on the collected equipment data to identify abnormal conditions and potential faults of the equipment; Step 3: Simulated maintenance: Perform simulated maintenance on the equipment based on the diagnosis results and correct maintenance errors; Step 4: Solution display: The display module displays the efficient maintenance solution and the operating status of industrial equipment in real time in three dimensions.
8. The method for using the industrial system fault diagnosis and status monitoring system according to claim 7, characterized in that: In the second step, the Bayesian convolutional neural analysis submodule in the fault diagnosis module measures the expected value of the possibility by calculating the possibility cost as follows: ELBO(D,θ)=KL(q(w|θ)|p(w))-E q(w|θ) (logp(D|w)), Where q(wθ) is the variational distribution, p(w) is the weight, and the Bayesian formula in the Bayesian convolutional neural analysis submodule is as follows: If event A1, ... A n Incompatible, Then when P(B)>0, we have 1≤j≤n, the calculation cost formula of the deep convolution of the deep residual convolution neural analysis submodule in the fault diagnosis module in step 2 is as follows: Calculation cost = D f 2 xD k 2 , Where D f is the spatial dimension of the input feature map, M is the number of input channels, and D K is the size of the convolution kernel. The calculation formula for standard convolution is as follows: D f is the spatial dimension of the input feature map, M is the number of input channels, N is the number of output channels, and D K is the size of the convolution kernel. The loss function calculation formula of the meta-learning analysis submodule in the fault diagnosis module in step 2 is as follows: Φ is a training hyperparameter. The calculation formula of the feedback function of the transfer learning analysis submodule in the fault diagnosis module in step 2 is as follows: The superscripts s and t represent the source domain and target domain respectively. represents a distribution metric function, r(s,a,s') indicates that state s becomes state s' after action a, Φ represents the corresponding feature, B j-1 and B j ; They represent a batch of data in the j-1th round and the jth round of iteration respectively. The MMD calculation formula is as follows: MMD 2 =sup f∈ζ (E x-P [f(x)]-E y-Q [f(y)]) 2 , Where F is a set of functions in RKHS, E x-P [f(x)] and E y-Q [f(y] represents the expected value of function f under distribution P and Q respectively.