A solid state switch fault monitoring method and system
Patent Information
- Application Number
- CN202610772248.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-01
- Publication Date
- 2026-08-18
AI Technical Summary
[0004]本发明的目的在于提供一种固态开关故障监测方法及系统，以解决上述背景技术中提出“如何利用已运行的固态开关历史数据，为新安装的固态开关的模型构建提供数据支撑”的问题
[0041] By dividing the system into high- and low-process groups, a fault evolution system can be constructed, and the operating status of solid-state switches can be dynamically evaluated. This provides a data foundation for fault evolution determination of target equipment, greatly improving the model's sensitivity to early anomalies. By constructing a fault analysis model, the coupling relationship between multiple parameters can be established, enabling cross-device generalized monitoring and improving the fault monitoring accuracy of solid-state switches. By generating a transfer learning model, adaptive transfer of fault monitoring experience can be achieved, breaking down data silos and improving the modeling capability of low-process group solid-state switches. Without sharing the original data, further convergence of model parameters can be achieved. Thus, while inheriting the fault evolution characteristics of the high-process group, the generalization ability and prediction accuracy of the low-process group model can be improved, enhancing the low-process group's ability to identify fault characteristics and providing a more stable data foundation for subsequent fault identification.
Smart Images
Figure CN122595030A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of fault monitoring technology, and in particular to a method and system for monitoring faults in solid-state switches. Background Technology
[0002] Solid-state switches (SSDs) are widely used in industrial automation control systems due to their advantages such as fast response speed, long lifespan, and lack of mechanical contacts. With the continuous development of big data technology, fault monitoring methods for SSDs are gradually evolving from traditional experience-based judgment and threshold determination to modeling and analysis methods. By constructing a mapping model between operating parameters and fault states, intelligent assessment and fault prediction of SSD operating status can be achieved. However, these big data models require extensive training with large amounts of operating data. This leads to difficulties in establishing reliable fault diagnosis models and timely identification of potential faults in the initial stages after SSD installation and commissioning due to a lack of sufficient operating data accumulation.
[0003] Therefore, "how to utilize historical data from existing solid-state switches to provide data support for model building of newly installed solid-state switches" is the technical problem that this invention aims to solve. Summary of the Invention
[0004] The purpose of this invention is to provide a solid-state switch fault monitoring method and system to solve the problem mentioned in the background art of "how to use historical data of already running solid-state switches to provide data support for the model construction of newly installed solid-state switches".
[0005] To achieve the above objectives, the present invention provides the following technical solution:
[0006] A method for monitoring faults in a solid-state switch, the method comprising:
[0007] Identify the application scenarios of solid-state switches, extract the characteristic parameters of each solid-state switch, wherein the characteristic parameters include at least: application scenario, load type and operating parameters, cluster the solid-state switches into several groups, edit the partitioning rules, cluster the solid-state switches in each group into high-process groups and low-process groups, collect abnormal events of solid-state switches in high-process groups, and obtain the operating parameters in the abnormal events, wherein the operating parameters include at least: temperature change rate, current fluctuation and voltage drop change.
[0008] Using time as the horizontal axis and each operating parameter as the vertical axis, several time series curves are plotted, integrated to generate a risk change trend chart, several sample data segments are extracted, test data is obtained, the sample data segment with the highest similarity to the test data is found, the failure probability is set and written into the test data, a training set is generated, and the pre-created failure analysis model is trained.
[0009] The system collects data on solid-state switches that need to be monitored for faults and defines them as target devices. It determines whether the target device is in a high-process group. If so, it collects the operating parameters of the target device and inputs them into the fault analysis model to output the fault probability. If not, it extracts the model parameters from the fault analysis model and performs manual fine-tuning to generate a transfer learning model. The operating parameters of the target device are then input into the transfer learning model to obtain the fault probability.
[0010] Furthermore, the step of clustering the solid-state switches into several groups, editing the partitioning rules, and clustering the solid-state switches in each group into high and low process groups includes:
[0011] Set the device number for each solid-state switch and generate a number set;
[0012] Using the group as the parent node and the process group as the child node, generate a state binary tree and write the device number corresponding to the solid-state switch under the child node.
[0013] Furthermore, the step of collecting abnormal events of solid-state switches in the high-process group and obtaining the operating parameters in the abnormal events includes:
[0014] Create a multi-parameter coupling analysis model, input the operating parameters of the target device into the multi-parameter coupling analysis model, and output the judgment result;
[0015] The failure probability is corrected using the judgment result.
[0016] Furthermore, the steps of setting the fault probability and writing it into the test data to generate a training set for training the pre-created fault analysis model include:
[0017] Configure edge devices in the usage scenario and deploy multi-parameter coupled analysis models and fault analysis models to the edge devices;
[0018] The operating parameters of the target device are uploaded to the edge device to update the fault probability.
[0019] Furthermore, the step of extracting model parameters from the fault analysis model and manually fine-tuning them to generate a transfer learning model includes:
[0020] Create a weight value corresponding to each solid-state switch, use the weight value to perform a weighted average of the model parameters to obtain the correction parameters, and then perform manual fine-tuning.
[0021] The manually fine-tuned correction parameters are deployed into the initialized fault analysis model to generate a transfer learning model.
[0022] Furthermore, the method also includes:
[0023] The failure probability is divided into several intervals, each interval corresponding to a sampling frequency;
[0024] The operating parameters are updated based on the sampling frequency.
[0025] Furthermore, the system includes:
[0026] The identification module is used to identify the application scenarios of solid-state switches, extract the feature parameters of each solid-state switch, wherein the feature parameters include at least: application scenario, load type and operating parameters, cluster the solid-state switches into several groups, edit the partitioning rules, cluster the solid-state switches in each group into high-process groups and low-process groups, collect abnormal events of solid-state switches in high-process groups, and obtain the operating parameters in the abnormal events, wherein the operating parameters include at least: temperature change rate, current fluctuation and voltage drop change.
[0027] The training module is used to plot several time series curves with time as the horizontal axis and each operating parameter as the vertical axis, integrate them to generate a risk change trend chart, extract several sample data segments, obtain test data, find the sample data segment with the highest similarity to the test data, set the failure probability, and write it into the test data to generate a training set for training the pre-created failure analysis model.
[0028] The learning module is used to collect data on solid-state switches that need to be monitored for faults and define them as target devices. It determines whether the target device is in a high-process group. If so, it collects the operating parameters of the target device and inputs them into the fault analysis model, outputting the fault probability. If not, it extracts the model parameters from the fault analysis model, performs manual fine-tuning, generates a transfer learning model, and inputs the operating parameters of the target device into the transfer learning model to obtain the fault probability.
[0029] Furthermore, the identification module includes:
[0030] The configuration unit is used to set the device number for each solid-state switch and generate a number set.
[0031] The write unit is used to generate a state binary tree with the group as the parent node and the process group as the child node, and write the device number corresponding to the solid-state switch to the child node.
[0032] The output unit is used to create a multi-parameter coupled analysis model. The operating parameters of the target device are input into the multi-parameter coupled analysis model, and the judgment result is output.
[0033] A correction unit is used to correct the fault probability using the determination result.
[0034] Furthermore, the training module includes:
[0035] The configuration unit is used to configure edge devices in the usage scenario and deploy multi-parameter coupled analysis models and fault analysis models to the edge devices;
[0036] The update unit is used to upload the operating parameters of the target device to the edge device and update the fault probability.
[0037] Furthermore, the learning module includes:
[0038] The fine-tuning unit is used to create a weight value corresponding to each solid-state switch, and to use the weight value to perform a weighted average of the model parameters to obtain the correction parameters, which are then manually fine-tuned.
[0039] The deployment unit is used to deploy the manually fine-tuned correction parameters into the initialized fault analysis model to generate a transfer learning model.
[0040] Compared with the prior art, the beneficial effects of the present invention are:
[0041] By dividing the system into high- and low-process groups, a fault evolution system can be constructed, and the operating status of solid-state switches can be dynamically evaluated. This provides a data foundation for fault evolution determination of target equipment, greatly improving the model's sensitivity to early anomalies. By constructing a fault analysis model, the coupling relationship between multiple parameters can be established, enabling cross-device generalized monitoring and improving the fault monitoring accuracy of solid-state switches. By generating a transfer learning model, adaptive transfer of fault monitoring experience can be achieved, breaking down data silos and improving the modeling capability of low-process group solid-state switches. Without sharing the original data, further convergence of model parameters can be achieved. Thus, while inheriting the fault evolution characteristics of the high-process group, the generalization ability and prediction accuracy of the low-process group model can be improved, enhancing the low-process group's ability to identify fault characteristics and providing a more stable data foundation for subsequent fault identification. Attached Figure Description
[0042] Figure 1 A flowchart illustrating the solid-state switch fault monitoring method provided in an embodiment of the present invention;
[0043] Figure 2 This is a first sub-flowchart of the solid-state switch fault monitoring method provided in an embodiment of the present invention;
[0044] Figure 3 This is a second sub-flowchart of the solid-state switch fault monitoring method provided in an embodiment of the present invention;
[0045] Figure 4 This is a third sub-flowchart of the solid-state switch fault monitoring method provided in this embodiment of the invention;
[0046] Figure 5This is a block diagram of the solid-state switch fault monitoring system provided in an embodiment of the present invention;
[0047] Figure 6 This is a block diagram of the identification module in the solid-state switch fault monitoring system provided in an embodiment of the present invention;
[0048] Figure 7 This is a block diagram of the training module in the solid-state switch fault monitoring system provided in an embodiment of the present invention;
[0049] Figure 8 This is a block diagram of the learning module in the solid-state switch fault monitoring system provided in an embodiment of the present invention. Detailed Implementation
[0050] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.
[0051] In Example 1, Figure 1 The implementation flow of the solid-state switch fault monitoring method provided in this embodiment of the invention is illustrated below, and is described in detail below:
[0052] S100: Identify the application scenarios of solid-state switches, extract the characteristic parameters of each solid-state switch, wherein the characteristic parameters include at least: application scenario, load type and operating parameters, cluster the solid-state switches into several groups, edit the partitioning rules, cluster the solid-state switches in each group into high-process groups and low-process groups, collect abnormal events of solid-state switches in high-process groups, and obtain the operating parameters in the abnormal events, wherein the operating parameters include at least: temperature change rate, current fluctuation and voltage drop change.
[0053] Identify the application scenarios of solid-state switches, which refer to the industrial environment in which the solid-state switches operate (such as production line control, heating control, motor drive, or lighting adjustment). Based on the data from the connected production system, equipment ledger information, and on-site operation logs, determine the characteristic parameters of each solid-state switch and standardize these parameters. The characteristic parameters include: application scenario, load type, and operating parameters. The application scenario characterizes the specific process and operating environment in which the solid-state switch operates (such as high temperature, high humidity, or high frequency start-stop conditions). The load type describes the load attributes controlled by the solid-state switch (such as resistive, inductive, or capacitive loads). The operating parameters refer to the specific equipment type (such as heater, motor, or capacitor bank), operating voltage, current, switching frequency, conduction time, turn-off time, and temperature rise.
[0054] Based on characteristic parameters, solid-state switches with similar operating conditions and load characteristics are divided into several groups. Within each group, solid-state switches are further divided into high- and low-process groups based on factors such as operating time, number of fault samples, and data stability. The high-process group refers to solid-state switches with a longer service life, sufficient fault sample accumulation, and that have entered the later stage of their equipment lifespan. Furthermore, solid-state switches with a shorter operating time, scarce historical data, or that are in the initial commissioning stage are classified as low-process groups. The classification rules are the specific classification criteria. For example, a classification rule might be: solid-state switches with a cumulative operating time ≥ 10,000 hours or a number of historical fault samples ≥ 50 are classified as high-process groups.
[0055] Historical operating data and corresponding fault records of solid-state switches in the high-process group are collected and aligned. The time interval of fault occurrence is defined, and a time window of preset length is constructed centered on this time interval. Multi-source data within the time window are aggregated and fused. Operating parameters (such as voltage, current, and temperature), equipment status information (such as conduction status, load conditions, and control commands), associated handling measures, and external intervention information (such as protection actions and adjustment strategies) are structurally integrated to generate abnormal events that characterize the fault evolution process. Operating parameters are extracted from these abnormal events, including temperature change rate, current fluctuation, and voltage drop change. These operating parameters are the status monitoring indicators of the solid-state switches and should be defined by the management personnel within the application scenario.
[0056] S200: Plot several time series curves with time as the horizontal axis and each operating parameter as the vertical axis, integrate them to generate a risk change trend chart, extract several sample data segments, obtain test data, find the sample data segment with the highest similarity to the test data, set the failure probability, and write it into the test data to generate a training set for training the pre-created failure analysis model.
[0057] Establish a correspondence between time and each operating parameter. Plot time series curves with time on the x-axis and the operating parameters at the corresponding time on the y-axis. Each operating parameter corresponds to a time series curve. Integrate all time series curves corresponding to each solid-state switch to generate a risk change trend chart, which is essentially a collection of time series curves. From the risk change trend chart, extract several sample data segments. For example, if an over-temperature alarm of the solid-state switch is detected at 12:05, then all operating parameters from 11:41 to 11:50 can be defined as the first sample data segment, and 11:51 to 11:55 as the second sample data segment. The specific time period division is determined by the management personnel.
[0058] One sample data segment is extracted from all sample data segments as test data, and a corresponding feature vector representation is constructed. Matching analysis is completed by calculating the similarity between the test data feature vector and other sample feature vectors; the similarity evaluation can be quantified using the cosine similarity calculation method. The test data and the corresponding sample data segment with the highest similarity are integrated and written into the fault probability to generate a training set. The fault probability should be determined by management personnel. Specifically, the closer to the fault time point, the higher the fault probability. Multiple training sets are created using the above method, and the fault analysis model is trained. The fault analysis model is built based on a deep learning algorithm and can evaluate the operating status of solid-state switches, identify anomalies, and output fault probabilities.
[0059] S300: Collects data on solid-state switches that need to be monitored for faults and defines them as target devices. It determines whether the target device is in a high-process group. If so, it collects the operating parameters of the target device and inputs them into the fault analysis model, outputting the fault probability. If not, it extracts the model parameters from the fault analysis model, performs manual fine-tuning, generates a transfer learning model, inputs the operating parameters of the target device into the transfer learning model, and obtains the fault probability.
[0060] Define the solid-state switch that needs to be monitored for faults as the target device. Based on the characteristic parameters of the target device, determine whether the target device is in the high process group. If the target device is determined to be in the high process group, collect the operating parameters of the target device according to the preset time step (e.g., 1 minute), preprocess and standardize the above operating parameters, and then input the processed operating parameters into the fault analysis model to output the fault probability. If the target device is not in the high-process group, it indicates that the target device lacks sufficient operational data and historical fault data, making it impossible to directly train the fault analysis model. Instead, model parameters (such as weight parameters, feature mapping layer parameters, and discrimination layer thresholds) are extracted from the fault analysis models already trained in the high-process group. It's important to note that when there are multiple usage scenarios, there are also multiple corresponding fault analysis models. Each fault analysis model corresponds to a set of model parameters. Administrators set a weight value for each set of model parameters, and then use this weight value to perform a weighted average of the model parameters to obtain the average value. This average value is then fine-tuned using a small amount of operational data. The fine-tuned average value is then deployed to the initialized fault analysis model to generate a transfer learning model. The operational parameters collected during the actual operation of the target device are input into this transfer learning model, which outputs the corresponding fault probability.
[0061] In Example 2, Figure 2The first sub-flowchart of the solid-state switch fault monitoring method provided in this embodiment of the invention is shown. The following details the steps of clustering the solid-state switches into several groups, editing the partitioning rules, and clustering the solid-state switches in each group into high and low process groups:
[0062] S101: Set the device number for each solid-state switch and generate a number set.
[0063] According to the preset numbering rules, each solid-state switch is assigned a corresponding device number. All solid-state switches and their corresponding device numbers are integrated to generate a number set, which is the collection of device numbers.
[0064] S102: Generate a state binary tree with the group as the parent node and the process group as the child node, and write the device number corresponding to the solid-state switch to the child node.
[0065] Parent nodes are created to correspond one-to-one with groups, and child nodes are created to correspond one-to-one with process groups. Following a binary tree data structure, groups and process groups are integrated to generate a state binary tree. Both parent and child nodes are logical nodes, primarily used to represent the relationships between groups and process groups. The state binary tree is similar to existing binary trees. Secondary child nodes are then attached to the child nodes, each corresponding to a device number. The advantage of this method is that it can establish relationships between groups, process groups, and solid-state switches, thus providing a data foundation for the evolutionary analysis of solid-state switches.
[0066] In Example 3, Figure 2 The first sub-flowchart of the solid-state switch fault monitoring method provided in this embodiment of the invention is shown. The steps of collecting abnormal events of solid-state switches in high-process groups and obtaining the operating parameters in the abnormal events are described in detail below:
[0067] S103: Create a multi-parameter coupling analysis model, input the operating parameters of the target device into the multi-parameter coupling analysis model, and output the judgment result.
[0068] By utilizing deep learning algorithms, a multi-parameter coupled analysis model is constructed to uniformly model and process the operating parameters involved in the actual operation of solid-state switches, thereby establishing the coupling relationship between multiple parameters. The operating parameters of the target device are input into the multi-parameter coupled analysis model, and the output is a judgment result, which is used to characterize whether the current operating status of the target device is normal, whether there are potential fault risks or performance degradation trends, etc.
[0069] S104: Use the determination result to correct the fault probability.
[0070] The failure probability is corrected using the judgment results, whereby the correction includes increasing or decreasing the failure probability.
[0071] In Example 4, Figure 3 The second sub-flowchart of the solid-state switch fault monitoring method provided in this embodiment of the invention is shown. The following details the steps of setting the fault probability, writing it into the test data, generating a training set, and training the pre-created fault analysis model:
[0072] S201: Configure edge devices in the usage scenario and deploy multi-parameter coupled analysis models and fault analysis models to the edge devices.
[0073] Identify the edge devices in the usage scenarios, including edge servers or industrial gateways, and deploy the multi-parameter coupling analysis model and fault analysis model to the edge devices.
[0074] S202: Upload the operating parameters of the target device to the edge device and update the fault probability.
[0075] According to the preset frequency, the operating parameters are uploaded to the edge device, and the failure probability at the corresponding time is updated.
[0076] In Example 5, Figure 4 The third sub-flowchart of the solid-state switch fault monitoring method provided in this embodiment of the invention is shown. The following details the steps of extracting model parameters from the fault analysis model, manually fine-tuning them, and generating a transfer learning model:
[0077] S301: Create a weight value corresponding to each solid-state switch, use the weight value to perform a weighted average of the model parameters to obtain the correction parameters, and perform manual fine-tuning.
[0078] Each solid-state switch is assigned a corresponding weight value, which should be determined based on factors such as equipment importance, operational stability, and load level. Using these weight values, the model parameters of each fault analysis model are weighted and averaged to obtain a set of average values, i.e., correction parameters. These correction parameters are then fine-tuned by administrators. The advantage of this approach is that, when constructing multiple fault analysis models, federated learning can be used to collaboratively fuse model parameters from different solid-state switches without sharing original operational data, providing a data foundation for the construction of transfer learning models.
[0079] S302: Deploy the manually fine-tuned correction parameters into the initialized fault analysis model to generate a transfer learning model.
[0080] The fine-tuned correction parameters are deployed to the initialized fault analysis model to obtain a transfer learning model. The transfer learning model can retrain the fault analysis model through transfer learning, thereby obtaining a data model that is more suitable for low-process groups.
[0081] In Example 6, unlike Example 1, the method further includes:
[0082] The failure probability is divided into several intervals, each interval corresponding to a sampling frequency;
[0083] The operating parameters are updated based on the sampling frequency.
[0084] The failure probability is divided into multiple intervals, and a corresponding sampling frequency is set for each interval. For example, 50-59% can be defined as the first interval with a sampling frequency of once every 5 seconds, and 60-69% can be defined as the second interval with a sampling frequency of once every 3 seconds. Based on the sampling frequency, the operating parameters at the corresponding time are collected, and the failure probability is updated.
[0085] Figure 5 This diagram illustrates the structural block diagram of a solid-state switch fault monitoring system provided in an embodiment of the present invention. The solid-state switch fault monitoring system 1 includes:
[0086] The identification module 11 is used to identify the application scenarios of solid-state switches, extract the feature parameters of each solid-state switch, wherein the feature parameters include at least: application scenario, load type and operating parameters, cluster the solid-state switches into several groups, edit the partitioning rules, cluster the solid-state switches in each group into high and low process groups, collect abnormal events of solid-state switches in the high process group, and obtain the operating parameters in the abnormal events, wherein the operating parameters include at least: temperature change rate, current fluctuation and voltage drop change.
[0087] Training module 12 is used to plot several time series curves with time as the horizontal axis and each operating parameter as the vertical axis, integrate and generate a risk change trend chart, extract several sample data segments, obtain test data, find the sample data segment with the highest similarity to the test data, set the failure probability, write it into the test data, generate a training set, and train the pre-created failure analysis model.
[0088] Learning module 13 is used to collect data on solid-state switches that need to be monitored for faults and define them as target devices. It determines whether the target device is in a high-process group. If so, it collects the operating parameters of the target device and inputs them into the fault analysis model to output the fault probability. If not, it extracts the model parameters from the fault analysis model and performs manual fine-tuning to generate a transfer learning model. The operating parameters of the target device are then input into the transfer learning model to obtain the fault probability.
[0089] Figure 6 This diagram illustrates the composition of the identification module 11 in the solid-state switch fault monitoring system provided in an embodiment of the present invention. The identification module 11 includes:
[0090] Setting unit 111 is used to set the device number of each solid-state switch and generate a number set;
[0091] The write unit 112 is used to generate a state binary tree with the group as the parent node and the process group as the child node, and write the device number corresponding to the solid-state switch to the child node.
[0092] Output unit 113 is used to create a multi-parameter coupling analysis model, input the operating parameters of the target device into the multi-parameter coupling analysis model, and output the judgment result.
[0093] The correction unit 114 is used to correct the fault probability using the determination result.
[0094] Figure 7 This diagram illustrates the structural composition of the training module 12 in the solid-state switch fault monitoring system provided in an embodiment of the present invention. The training module 12 includes:
[0095] Configuration unit 121 is used to configure edge devices in the usage scenario and deploy multi-parameter coupled analysis model and fault analysis model to edge devices;
[0096] The update unit 122 is used to upload the operating parameters of the target device to the edge device and update the fault probability.
[0097] Figure 8 This diagram illustrates the structural block diagram of the learning module 13 in the solid-state switch fault monitoring system provided in an embodiment of the present invention. The learning module 13 includes:
[0098] The fine-tuning unit 131 is used to create a weight value corresponding to each solid-state switch, use the weight value to perform a weighted average of the model parameters to obtain the correction parameters, and perform manual fine-tuning.
[0099] Deployment unit 132 is used to deploy the manually fine-tuned correction parameters into the initialized fault analysis model to generate a transfer learning model.
[0100] The recognition module 11 is mainly used to complete step S100, the training module 12 is mainly used to complete step S200, and the learning module 13 is mainly used to complete step S300.
[0101] Setting unit 111 is mainly used to complete step S101, writing unit 112 is mainly used to complete step S102, output unit 113 is mainly used to complete step S103, and correction unit 114 is mainly used to complete step S104.
[0102] Configuration unit 121 is mainly used to complete step S201, and update unit 122 is mainly used to complete step S202;
[0103] The fine-tuning unit 131 is mainly used to complete step S301, and the deployment unit 132 is mainly used to complete step S302.
[0104] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0105] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.
[0106] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for monitoring faults in a solid-state switch, characterized in that, The method includes: Identify the application scenarios of solid-state switches, extract the characteristic parameters of each solid-state switch, wherein the characteristic parameters include at least: application scenario, load type and operating parameters, cluster the solid-state switches into several groups, edit the partitioning rules, cluster the solid-state switches in each group into high-process groups and low-process groups, collect abnormal events of solid-state switches in high-process groups, and obtain the operating parameters in the abnormal events, wherein the operating parameters include at least: temperature change rate, current fluctuation and voltage drop change. Using time as the horizontal axis and each operating parameter as the vertical axis, several time series curves are plotted, integrated to generate a risk change trend chart, several sample data segments are extracted, test data is obtained, the sample data segment with the highest similarity to the test data is found, the failure probability is set and written into the test data, a training set is generated, and the pre-created failure analysis model is trained. The system collects data on solid-state switches that need to be monitored for faults and defines them as target devices. It determines whether the target device is in a high-process group. If so, it collects the operating parameters of the target device and inputs them into the fault analysis model to output the fault probability. If not, it extracts the model parameters from the fault analysis model and performs manual fine-tuning to generate a transfer learning model. The operating parameters of the target device are then input into the transfer learning model to obtain the fault probability.
2. The solid-state switch fault monitoring method according to claim 1, characterized in that, The step of clustering the solid-state switches into several groups, editing the partitioning rules, and clustering the solid-state switches in each group into high and low process groups includes: Set the device number for each solid-state switch and generate a number set; Using the group as the parent node and the process group as the child node, generate a state binary tree and write the device number corresponding to the solid-state switch under the child node.
3. The solid-state switch fault monitoring method according to claim 1, characterized in that, The steps of collecting abnormal events of solid-state switches in the high-process group and obtaining the operating parameters in the abnormal events include: Create a multi-parameter coupling analysis model, input the operating parameters of the target device into the multi-parameter coupling analysis model, and output the judgment result; The failure probability is corrected using the judgment result.
4. The solid-state switch fault monitoring method according to claim 3, characterized in that, The steps of setting the fault probability, writing it into the test data, generating a training set, and training the pre-created fault analysis model include: Configure edge devices in the usage scenario and deploy multi-parameter coupled analysis models and fault analysis models to the edge devices; The operating parameters of the target device are uploaded to the edge device to update the fault probability.
5. The solid-state switch fault monitoring method according to claim 4, characterized in that, The steps of extracting model parameters from the fault analysis model, performing manual fine-tuning, and generating a transfer learning model include: Create a weight value corresponding to each solid-state switch, use the weight value to perform a weighted average of the model parameters to obtain the correction parameters, and then perform manual fine-tuning. The manually fine-tuned correction parameters are deployed into the initialized fault analysis model to generate a transfer learning model.
6. The solid-state switch fault monitoring method according to claim 4, characterized in that, The method further includes: The failure probability is divided into several intervals, each interval corresponding to a sampling frequency; The operating parameters are updated based on the sampling frequency.
7. A solid-state switch fault monitoring system, characterized in that, The system includes: The identification module is used to identify the application scenarios of solid-state switches, extract the feature parameters of each solid-state switch, wherein the feature parameters include at least: application scenario, load type and operating parameters, cluster the solid-state switches into several groups, edit the partitioning rules, cluster the solid-state switches in each group into high-process groups and low-process groups, collect abnormal events of solid-state switches in high-process groups, and obtain the operating parameters in the abnormal events, wherein the operating parameters include at least: temperature change rate, current fluctuation and voltage drop change. The training module is used to plot several time series curves with time as the horizontal axis and each operating parameter as the vertical axis, integrate them to generate a risk change trend chart, extract several sample data segments, obtain test data, find the sample data segment with the highest similarity to the test data, set the failure probability, and write it into the test data to generate a training set for training the pre-created failure analysis model. The learning module is used to collect data on solid-state switches that need to be monitored for faults and define them as target devices. It determines whether the target device is in a high-process group. If so, it collects the operating parameters of the target device and inputs them into the fault analysis model, outputting the fault probability. If not, it extracts the model parameters from the fault analysis model, performs manual fine-tuning, generates a transfer learning model, and inputs the operating parameters of the target device into the transfer learning model to obtain the fault probability.
8. The solid-state switch fault monitoring system according to claim 7, characterized in that, The identification module includes: The configuration unit is used to set the device number for each solid-state switch and generate a number set. The write unit is used to generate a state binary tree with the group as the parent node and the process group as the child node, and write the device number corresponding to the solid-state switch to the child node. The output unit is used to create a multi-parameter coupled analysis model. The operating parameters of the target device are input into the multi-parameter coupled analysis model, and the judgment result is output. A correction unit is used to correct the fault probability using the determination result.
9. The solid-state switch fault monitoring system according to claim 8, characterized in that, The training module includes: The configuration unit is used to configure edge devices in the usage scenario and deploy multi-parameter coupled analysis models and fault analysis models to the edge devices; The update unit is used to upload the operating parameters of the target device to the edge device and update the fault probability.
10. The solid-state switch fault monitoring system according to claim 9, characterized in that, The learning module includes: The fine-tuning unit is used to create a weight value corresponding to each solid-state switch, and to use the weight value to perform a weighted average of the model parameters to obtain the correction parameters, which are then manually fine-tuned. The deployment unit is used to deploy the manually fine-tuned correction parameters into the initialized fault analysis model to generate a transfer learning model.