A method and system for transmitting circuit path fault condition data

Abstract

The present application relates to the technical field of fault data identification, in particular to a circuit channel fault state data transmission method and system, the method comprises: sending circuit data to the preset end device, edge gateway and cloud server in turn, obtaining the fault judgment result of the end device, edge gateway and cloud server, when the fault rate difference between the edge gateway fault judgment result and the cloud server fault judgment result is greater than the preset difference threshold, adjusting the first judgment model to obtain the third judgment model, identifying the target feature that is inconsistent in the third judgment model and the preliminary fault judgment result, adjusting the preset feature judgment threshold corresponding to the target feature in the end device to preferentially transmit the reported fault state data to the user end; the present application can effectively improve the fault reporting efficiency and fault identification accuracy, so as to facilitate the staff to take maintenance measures in time.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fault data identification, in particular to a circuit channel fault state data transmission method and system. BACKGROUND

[0002] In the field of intelligent manufacturing and key facility monitoring, real-time fault diagnosis and prediction of electrical equipment operation state is of great importance. Traditional circuit fault data processing systems usually adopt a three-level architecture of end-edge-cloud. In this architecture, the end device is responsible for collecting raw data and performing simple rule judgment, the edge gateway is deployed with a lightweight artificial intelligence model, responsible for real-time or near real-time preliminary fault analysis, and the cloud server is responsible for gathering multi-party data and running complex deep learning models for deep fault root cause analysis, trend prediction and model optimization.

[0003] However, this architecture has certain defects. Firstly, the complexity of the edge-side model is insufficient, the judgment accuracy is not high, and the update of the edge model usually depends on the periodic update package issued by the cloud, which has strong hysteresis and cannot respond in time, causing false negatives and false positives of faults. Secondly, the judgment standard coordination between the end, edge and cloud is poor. For example, after the edge model is updated, the internal decision logic and feature importance have changed, but the fixed rules of the end cannot be adaptively adjusted, the fault recognition ability is poor, the preliminary filtering effect is also poor, and the fault data cannot be preferentially uploaded, affecting the overall fault diagnosis efficiency. SUMMARY

[0004] To solve the above technical problems, the present application provides a circuit channel fault state data transmission method and system, which adjusts the fault judgment rules of the front end device by the fault judgment results of the model, effectively improves the fault reporting efficiency and fault recognition accuracy, so as to facilitate the staff to take maintenance measures in time.

[0005] According to the first aspect of the present application, a circuit channel fault state data transmission method is provided, comprising the following steps:

[0006] S1, the circuit data is sent to the preset end device, edge gateway and cloud server in turn, and the preliminary fault judgment result generated by the end device based on a plurality of preset feature judgment thresholds, the first fault judgment result generated by the edge gateway based on the first judgment model and the second fault judgment result generated by the cloud server based on the second judgment model are obtained respectively.

[0007] S2, when the difference between the failure rates corresponding to the first failure judgment result and the second failure judgment result is greater than a preset difference threshold value within a preset time period, adjusting the first judgment model in the virtual edge node based on the second failure judgment result, generating a third judgment model, and updating the third judgment model to the edge gateway; wherein the virtual edge node refers to a backup node of the edge gateway and a node that synchronizes data with the edge gateway.

[0008] S3, based on the third judgment model, re-performing failure judgment on the same data within the preset time period through the virtual edge node to obtain a third failure judgment result.

[0009] S4, identifying target features that are inconsistent in the third failure judgment result and the preliminary failure judgment result, for each target feature, adjusting the preset feature judgment threshold corresponding to each target feature in the end device according to the feature weight adjusted by the third judgment model for each target feature, and generating the final judgment rule of the end device.

[0010] S5, when receiving new circuit data, performing failure judgment based on the final judgment rule of the end device and the third judgment model of the edge gateway, and transmitting the judged failure state data to the user end.

[0011] According to the second aspect of the present application, a circuit channel failure state data transmission system is provided, which comprises:

[0012] The first failure judgment module is configured to sequentially send data to the preset end device, the edge gateway and the cloud server, and obtain the preliminary failure judgment result generated by the end device based on a plurality of preset feature judgment thresholds, the first failure judgment result generated by the edge gateway based on the first judgment model, and the second failure judgment result generated by the cloud server based on the second judgment model.

[0013] The model updating module is configured to, when the difference between the failure rates corresponding to the first failure judgment result and the second failure judgment result is greater than a preset difference threshold value within a preset time period, adjust the first judgment model in the virtual edge node based on the second failure judgment result, generate a third judgment model, and update the third judgment model to the edge gateway; wherein the virtual edge node refers to a backup node of the edge gateway and a node that synchronizes data with the edge gateway.

[0014] The second failure judgment module is configured to, based on the third judgment model, re-perform failure judgment on the same data within the preset time period through the virtual edge node to obtain a third failure judgment result.

[0015] The rule adjusting module is configured to identify target features that are inconsistent in the third fault judgment result and the preliminary fault judgment result, adjust, for each target feature, a preset feature judgment threshold corresponding to each target feature in the end device according to a feature weight adjusted by the third judgment model for each target feature, and generate a final judgment rule of the end device.

[0016] The fault data reporting module is configured to perform fault judgment based on the final judgment rule of the end device and the third judgment model of the edge gateway when new data is received, and transmit and report fault state data to the user end.

[0017] The present application has at least the following beneficial effects:

[0018] The present application provides a circuit channel fault state data transmission method. First, circuit data is sequentially sent to a preset end device, an edge gateway and a cloud server to obtain fault judgment results of the end device, the edge gateway and the cloud server. When a fault rate difference between the fault judgment result of the edge gateway and the fault judgment result of the cloud server is greater than a preset difference threshold, a third judgment model is obtained by adjusting the first judgment model based on the fault judgment result of the cloud server. The parameters of the first judgment model are optimized based on the judgment result of the cloud server, so that the optimized first judgment model is more reliable and can improve the fault judgment accuracy of the first judgment model. Then, target features that are inconsistent in the third judgment model and the preliminary fault judgment result are identified, and a preset feature judgment threshold corresponding to the target features in the end device is adjusted. In this process, the change of the feature weight before and after the first judgment model is updated is used as an adjustment reference of the judgment rule of the end device, so that the judgment rule of the end device becomes sensitive to these key feature signals, effectively improving the fault judgment accuracy of the judgment rule of the end device. When new circuit data is received, fault judgment is performed based on the final judgment rule of the end device and the third judgment model of the edge gateway, which can identify faults in advance and report them preferentially, improving the fault reporting efficiency and fault identification accuracy, so that the staff can take maintenance measures or automatic shutdown measures in time. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0020] Figure 1 The flowchart of the circuit channel fault state data transmission method provided by the first embodiment of the present application is shown in the figure.

[0021] Figure 2 The structural diagram of the transmission system of the circuit channel fault state data provided by the second embodiment of the present application is shown. DETAILED DESCRIPTION

[0022] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person skilled in the art without creative work fall within the protection scope of the present application.

[0023] Embodiment one

[0024] The first embodiment of the present application provides a circuit channel fault state data transmission method, as shown in the figure, the method comprises the following steps: Figure 1

[0025] S1, the circuit data is sent to the preset end device, edge gateway and cloud server in turn, respectively obtaining the preliminary fault judgment result generated by the end device based on a plurality of preset feature judgment thresholds, the first fault judgment result generated by the edge gateway based on the first judgment model and the second fault judgment result generated by the cloud server based on the second judgment model. In a specific implementation, the circuit data is a plurality of pre-acquired monitoring data, such as voltage, current, temperature, humidity, etc.

[0026] Specifically, the end device is embedded with a plurality of preset sensors, wherein the preset sensors correspond one by one to the preset feature judgment thresholds; it can be understood that each preset sensor is configured with a corresponding preset feature judgment threshold.

[0027] Further, the first judgment model and the second judgment model are both pre-trained machine learning models for fault judgment, and a person skilled in the art knows the specific training process of the machine learning model for fault judgment, which will not be described here.

[0028] S2, within a preset time period, when the fault rate difference corresponding to the first fault judgment result and the second fault judgment result is greater than a preset difference threshold, the first judgment model in the virtual edge node is adjusted based on the second fault judgment result, a third judgment model is generated, and the third judgment model is updated to the edge gateway. A person skilled in the art sets the preset difference threshold according to actual needs.

[0029] Specifically, the virtual edge node refers to a backup node of the edge gateway and a node synchronized with the edge gateway data.

[0030] ​In another case, when the difference between the failure rates corresponding to the first failure judgment result and the second failure judgment result is not greater than the preset difference threshold, the first judgment model does not need to be updated.

[0031] In one specific embodiment, the third judgment model is generated by the following steps:

[0032] S201, input the circuit data sent to the edge gateway in a preset time period into the virtual edge node, obtain a prediction result through the first judgment model in the virtual edge node, take the second failure judgment result corresponding to the cloud server as a true label, and calculate a loss value between the prediction result and the true label. Those skilled in the art know the specific calculation method of the loss value, such as the L1 loss function or the cross-entropy loss function, which will not be described here.

[0033] S202, based on the loss value, calculate the gradient of the current parameters of the first judgment model with respect to the loss value through a back propagation algorithm, and use an optimization algorithm to iteratively update the parameters of the first judgment model in the virtual edge node according to the gradient.

[0034] S203, verify the first judgment model with updated parameters using a preset verification set, and when the verification result meets the preset requirement, determine the updated first judgment model as the third judgment model; it can be understood that the verification result meeting the preset requirement means that the difference between the prediction result and the true value is less than a preset threshold.

[0035] As mentioned above, since the cloud server can run a complex deep learning model to perform failure root cause analysis, its failure judgment accuracy is very high. The parameters of the first judgment model are optimized based on the judgment result of the cloud server, so that the optimized first judgment model is more reliable, the failure judgment accuracy of the first judgment model is improved, and the cloud server as a subsequent node of the edge gateway is used to update the edge gateway with the model of the cloud server, so that the failure judgment result of the edge gateway is close to the judgment result of the remote server, and the failure data identified is reported to the cloud server in priority before being sent to the cloud server.

[0036] Further, the S2 step further includes the following steps:

[0037] S210, during the updating of the third judgment model to the edge gateway, stop sending the data stream received by the end device to the edge gateway, and send the data stream to the virtual edge node.

[0038] S220, when the third judgment model of the edge gateway is updated, the transmission target point of the data stream received by the end device is switched back from the virtual edge node to the edge gateway, and the edge gateway and the virtual edge node are run in parallel and synchronized for data processing within a preset data synchronization time length, until the preset data synchronization time length is reached, and the data processing of the virtual edge node is stopped.

[0039] In the above, when the edge gateway performs model updating, the virtual edge node is started to replace the edge gateway to work, avoiding the influence of model updating of the edge gateway on data processing, and after synchronization with the virtual edge node for a period of time after the model updating is completed, the data processing of the virtual edge node is closed, which is beneficial to ensure the stability of the system.

[0040] S3, based on the third judgment model, the same data in a preset time period is re-judged for failure through the virtual edge node, and a third failure judgment result is obtained; it can be understood that the same data in the preset time period refers to the circuit data sent by the end device to the edge gateway in the preset time period.

[0041] In the above, the third judgment model in the virtual edge node is used to re-judge the data for failure, instead of using the edge gateway, which can avoid affecting the normal data processing of the edge gateway and reduce the occupation of the bandwidth of the edge gateway. The third failure judgment result obtained by the virtual edge node can be used as a reference to adjust the failure judgment rule of the end device.

[0042] S4, identify the target features that are inconsistent in the third failure judgment result and the preliminary failure judgment result, for each target feature, adjust the preset feature judgment threshold corresponding to each target feature in the end device according to the feature weight adjusted by the third judgment model for each target feature, and generate the final judgment rule of the end device. In specific implementation, the input features of the third judgment model are various sensor readings or indicators calculated from sensor readings, such as voltage, current, temperature, vibration amplitude, speed, pressure value, and statistical quantity of features, etc., wherein the statistical quantity of features can be average value, variance, peak value, etc. of the past 5 minutes; the output is a failure type label obtained based on comprehensive analysis of all input features, such as overheating anomaly, insulation aging, bearing wear, etc.

[0043] Specifically, adjusting the preset feature judgment threshold corresponding to each target feature in the end device according to the feature weight adjusted by the third judgment model for each target feature includes the following steps:

[0044] S401, for any target feature, calculate the change in feature weight corresponding to the target feature before and after the first judgment model is updated, and determine the adjustment direction of the preset feature judgment threshold according to the positive or negative sign of the change in feature weight; it can be understood that: the change in feature weight refers to the difference between the weight corresponding to the target feature in the updated first judgment model and the weight corresponding to the target feature in the unupdated first judgment model.

[0045] It should be noted that when the change in feature weight is positive, it indicates that the target feature corresponding to the third judgment model is more important. Therefore, the judgment rule of the terminal device should be more sensitive to the target feature, and the preset feature judgment threshold should be lowered to make it easier to trigger alarms. When the change in feature weight is negative, the preset feature judgment threshold should be raised to reduce the false alarm rate.

[0046] S402, based on the absolute value of the change in feature weights, combined with the predefined weight-threshold sensitivity coefficient corresponding to the target feature, calculates the undetermined adjustment range of the preset feature judgment threshold; it can be understood as: the undetermined adjustment range is the product of the absolute value of the change in feature weights and the weight-threshold sensitivity coefficient.

[0047] Furthermore, the weight-threshold sensitivity coefficient corresponding to the target feature is obtained through the following steps:

[0048] S4021, Obtain the numerical sequence of the target feature within a historical time period, and the accurate fault label generated by the cloud server corresponding to each value in the numerical sequence; This can be understood as: the accurate fault label refers to the label that has been verified by the cloud server.

[0049] S4022, based on the accurate fault label generated by the cloud server corresponding to each value in the numerical sequence, statistically analyze the numerical distribution of the target feature under normal and fault states respectively. For example, the numerical distribution under normal state is P(x|normal), and the numerical distribution under fault state is P(x|fault).

[0050] S4023, Calculate the discrimination degree used to measure the ability of the target feature to distinguish faults based on the numerical distribution of the target feature under normal and fault conditions.

[0051] In one implementation, the discriminant strength is obtained by calculating the Bartholomew's distance or KL divergence between the numerical distributions under normal conditions and those under fault conditions. A larger Bartholomew's distance or KL divergence value indicates less overlap between the two distributions, meaning the target feature is more easily distinguished between normal and fault conditions, i.e., higher discriminant strength. For example, the discriminant strength meets the following condition:

[0052] wherein, L is the discrimination, n is the number of pre-divided intervals, P(i) is the probability value of the numerical distribution in the i-th interval under the normal state, and Q(i) is the probability value of the numerical distribution in the i-th interval under the fault state.

[0053] In another embodiment, a classifier such as logistic regression is trained using the target features themselves alone, and the accuracy of the classifier is used as the discrimination.

[0054] S4024, converting the discrimination into a weight-threshold sensitivity coefficient corresponding to the target feature through a preset mapping function; wherein the discrimination is directly proportional to the weight-threshold sensitivity coefficient.

[0055] Specifically, the mapping function satisfies the following condition:

[0056] S=a x D+b, wherein S refers to the weight-threshold sensitivity coefficient, a and b are both preset constants, and D is the discrimination.

[0057] In the above, when obtaining the weight-threshold sensitivity coefficient, the numerical distribution of each target feature under the normal state and the fault state is introduced for different target features, the sensitivity coefficient is obtained by calculating the discrimination, and the discrimination ability of each target feature for normal and fault is considered. Therefore, the sensitivity coefficient obtained by the above method is more reasonable and reliable, and the sensitivity coefficient is dynamically calculated according to the historical data instead of presetting an empirical value, which can greatly improve the adaptive ability and precision of the system.

[0058] S403, comparing the undetermined adjustment amplitude of the preset feature judgment threshold with the historical adjustment amplitude of the preset feature judgment threshold, applying a calibration algorithm to obtain a target adjustment amplitude of the preset feature judgment threshold.

[0059] In specific implementation, the calibration algorithm can use a clamp function or a PID controller.

[0060] S404, adjusting the preset feature judgment threshold according to the adjustment direction and the target adjustment amplitude of the preset feature judgment threshold.

[0061] The above changes in feature weights before and after updating the first judgment model are used as the adjustment reference of the judgment rule of the terminal device, which can improve the fault judgment reliability of the terminal device for each feature, and does not require the judgment rule of the terminal device to understand complex fault categories. Only the key signal judged as important by the model needs to be focused on, and by making the judgment rule of the terminal device sensitive to these key signals, the fault judgment accuracy of the judgment rule of the terminal device is effectively improved, and the early warning ability of the entire system to complex faults is also indirectly improved.

[0062] S5, when receiving new circuit data, based on the final judgment rule of the terminal device, the third judgment model of the edge gateway, the fault is judged, and the fault state data judged is transmitted and reported to the user end.

[0063] The above, since the fault data is sent by the terminal device to the edge gateway, and then sent to the remote server by the edge gateway, the data processing and fault judgment need to consume more time, after improving the judgment accuracy, the fault can be identified in advance and reported preferentially, the fault reporting efficiency and fault identification accuracy are improved, so as to facilitate the staff to take maintenance measures or automatic cutting measures in time, which is beneficial to the stable operation of the system.

[0064] Embodiment two

[0065] The embodiment two of the application provides a transmission system of circuit channel fault state data, as shown in Figure 2 The transmission system comprises:

[0066] A first fault judgment module 100 is used for sequentially sending circuit data to a preset terminal device, an edge gateway and a cloud server, respectively acquiring a preliminary fault judgment result generated by the terminal device based on a plurality of preset feature judgment thresholds, a first fault judgment result generated by the edge gateway based on a first judgment model and a second fault judgment result generated by the cloud server based on a second judgment model.

[0067] A model updating module 200 is used for adjusting the first judgment model in the virtual edge node based on the second fault judgment result when the fault rate difference corresponding to the first fault judgment result and the second fault judgment result is greater than a preset difference threshold in a preset time period, generating a third judgment model, and updating the third judgment model to the edge gateway, wherein the virtual edge node is a standby node of the edge gateway and a node synchronized with the edge gateway.

[0068] A second fault judgment module 300 is used for re-judging the fault of the same data in the preset time period through the virtual edge node based on the third judgment model, and obtaining a third fault judgment result.

[0069] A rule adjusting module 400 is used for identifying target features inconsistent in the third fault judgment result and the preliminary fault judgment result, adjusting the preset feature judgment threshold corresponding to each target feature in the terminal device for each target feature according to the feature weight adjusted by the third judgment model for each target feature, and generating a final judgment rule of the terminal device.

[0070] A fault data reporting module 500 is used for judging the fault based on the final judgment rule of the terminal device and the third judgment model of the edge gateway when receiving new circuit data, and transmitting and reporting the fault state data judged to the user end.

[0071] It should be noted that the information interaction between the above modules, the execution process and the like, since the same concept based on the method embodiments of the present application, its specific functions and the resulting technical effects, specific can refer to the method embodiments part, this will not be repeated here.

[0072] Although some specific embodiments of the present application have been described in detail by examples, those skilled in the art should understand that the above examples are only for illustration, not for limiting the scope of the present application. Those skilled in the art should also understand that various modifications can be made to the embodiments without departing from the scope and spirit of the present application. The scope of the present application is defined by the appended claims.

Claims

1. A method for transmitting circuit channel fault status data, characterized in that, The method includes the following steps: S1, the circuit data is sent sequentially to the preset end device, edge gateway and cloud server, and the preliminary fault judgment result generated by the end device based on several preset feature judgment thresholds, the first fault judgment result generated by the edge gateway based on the first judgment model and the second fault judgment result generated by the cloud server based on the second judgment model are obtained respectively. S2, within a preset time period, when the difference in failure rates between the first fault judgment result and the second fault judgment result is greater than a preset difference threshold, the first judgment model in the virtual edge node is adjusted based on the second fault judgment result to generate a third judgment model, and the third judgment model is updated to the edge gateway; wherein, the virtual edge node refers to the backup node of the edge gateway and the node that is synchronized with the edge gateway data; S3, based on the third judgment model, re-judges the same data within a preset time period through virtual edge nodes to obtain the third fault judgment result; S4, identify target features that are inconsistent between the third fault judgment result and the preliminary fault judgment result. For each target feature, adjust the preset feature judgment threshold corresponding to each target feature in the terminal device according to the feature weight adjusted by the third judgment model for each target feature, and generate the final judgment rule of the terminal device. S5, when receiving new circuit data, performs fault judgment based on the final judgment rules of the end device and the third judgment model of the edge gateway, and transmits the judged fault status data to the user terminal.

2. The method for transmitting circuit channel fault status data according to claim 1, characterized in that, The terminal device is embedded with several preset sensors, and each preset sensor corresponds one-to-one with a preset feature judgment threshold.

3. The method for transmitting circuit channel fault status data according to claim 1, characterized in that, In step S2, the third judgment model is generated through the following steps: S201, input the circuit data sent to the edge gateway within the preset time period into the virtual edge node, obtain the prediction result through the first judgment model in the virtual edge node, take the second fault judgment result corresponding to the cloud server as the real label, and calculate the loss value between the prediction result and the real label. S202, based on the loss value, the gradient of the current parameters of the first judgment model relative to the loss value is calculated by the backpropagation algorithm, and the parameters of the first judgment model in the virtual edge node are iteratively updated according to the gradient using an optimization algorithm. S203, use a preset verification set to verify the first judgment model after parameter update. When the verification result meets the preset requirements, the updated first judgment model is determined as the third judgment model.

4. The method for transmitting circuit channel fault status data according to claim 1, characterized in that, Step S2 also includes the following steps: S210, during the process of updating the third judgment model to the edge gateway, stop sending the data stream received by the end device to the edge gateway, and send the data stream to the virtual edge node; S220: After the third judgment model of the edge gateway is updated, the transmission target point of the data stream received by the end device is switched from the virtual edge node back to the edge gateway. Within the preset data synchronization time, the edge gateway and the virtual edge node are run in parallel and data processing is performed synchronously until the preset data synchronization time is reached, and then the data processing of the virtual edge node is stopped.

5. The method for transmitting circuit channel fault status data according to claim 1, characterized in that, In step S4, adjusting the preset feature judgment threshold corresponding to each target feature in the terminal device based on the feature weights adjusted according to the third judgment model for each target feature includes the following steps: S401, For any target feature, calculate the change in feature weight corresponding to the target feature before and after the first judgment model is updated, and determine the adjustment direction of the preset feature judgment threshold according to the positive or negative sign of the change in feature weight. S402, based on the absolute value of the change in feature weights, combined with the predefined weight-threshold sensitivity coefficient corresponding to the target feature, calculate the undetermined adjustment range of the preset feature judgment threshold; S403, compare the undetermined adjustment range of the preset feature judgment threshold with the historical adjustment range of the preset feature judgment threshold, and apply a calibration algorithm to obtain the target adjustment range of the preset feature judgment threshold; S404, Adjust the preset feature judgment threshold according to the adjustment direction and target adjustment range of the preset feature judgment threshold.

6. The method for transmitting circuit channel fault status data according to claim 5, characterized in that, The weight-threshold sensitivity coefficient corresponding to the target feature is obtained through the following steps: S4021, Obtain the numerical sequence of the target feature within a historical time period, and the accurate fault label generated by the cloud server corresponding to each value in the numerical sequence. S4022, based on the accurate fault label generated by the cloud server corresponding to each value in the numerical sequence, the numerical distribution of the target feature under normal and fault states is statistically analyzed respectively. S4023, Calculate the discrimination degree used to measure the ability of the target feature to distinguish faults based on the numerical distribution of the target feature under normal and fault conditions. S4024, through a preset mapping function, converts the discrimination into the weight-threshold sensitivity coefficient corresponding to the target feature; wherein, the discrimination is proportional to the weight-threshold sensitivity coefficient.

7. The method for transmitting circuit channel fault status data according to claim 6, characterized in that, The distinguishability meets the following conditions: Where L is the discrimination index, n is the number of pre-divided intervals, P(i) is the probability value of the value distributed in the i-th interval under normal conditions, and Q(i) is the probability value of the value distributed in the i-th interval under fault conditions.

8. A system for transmitting circuit channel fault status data, characterized in that, The transmission system includes: The first fault judgment module is used to send circuit data sequentially to a preset end device, an edge gateway, and a cloud server, and respectively obtain the preliminary fault judgment result generated by the end device based on several preset feature judgment thresholds, the first fault judgment result generated by the edge gateway based on the first judgment model, and the second fault judgment result generated by the cloud server based on the second judgment model. The model update module is used to adjust the first judgment model in the virtual edge node based on the second fault judgment result within a preset time period when the difference in the failure rate corresponding to the first fault judgment result and the second fault judgment result is greater than a preset difference threshold, generate a third judgment model, and update the third judgment model to the edge gateway; wherein, the virtual edge node refers to the backup node of the edge gateway and the node that is synchronized with the edge gateway data; The second fault judgment module is used to re-judge the same data within a preset time period based on the third judgment model and through virtual edge nodes to obtain the third fault judgment result. The rule adjustment module is used to identify target features that are inconsistent between the third fault judgment result and the preliminary fault judgment result. For each target feature, the module adjusts the preset feature judgment threshold corresponding to each target feature in the terminal device according to the feature weight adjusted for each target feature in the third judgment model, and generates the final judgment rule of the terminal device. The fault data reporting module is used to determine faults based on the final judgment rules of the end device and the third judgment model of the edge gateway when new circuit data is received, and then transmit and report the determined fault status data to the user terminal.

Images