A communication device monitoring method, apparatus and medium

Abstract

The application is a kind of communication equipment monitoring method, device and medium, relating to communication security technical field, including obtaining real-time monitoring data of communication equipment, extracting features of real-time monitoring data and carrying out cluster analysis, obtaining a variety of classification index data, stacking time series analysis and deep learning, combining with the obtained historical abnormal data of multiple periods, constructing abnormal operation monitoring model; the classification index data is input into the abnormal operation monitoring model, whether the classification index result is in the set threshold range is judged, when the identification result shows abnormal, the running mainboard of the communication equipment is switched to the standby mainboard; the target function is set as minimizing the false alarm rate, the abnormal operation monitoring model of the abnormal mainboard is adjusted according to the abnormal identification result, the classification index data is input into the abnormal operation monitoring model of the adjusted mainboard, if the output result is normal, the adjusted abnormal mainboard is used as the standby mainboard, if not, the abnormal identification result is used as the abnormal alarm.

Description

Technical Field

[0001] This invention relates to the field of communication security, and in particular to a method, apparatus and medium for monitoring communication equipment. Background Technology

[0002] With the rapid development of the Internet of Things (IoT) and communication technologies, communication equipment is a core component for maintaining the normal operation of communication systems. Monitoring its operational status allows for the timely detection and resolution of potential faults, ensuring the stability and reliability of the communication system. Communication equipment is handling increasingly larger data volumes and becoming more complex in structure. The use of communication devices (such as switches, routers, and smart boxes), environmental devices (such as temperature, humidity, water immersion, and smoke monitoring), power supply devices (such as switching power supplies, uninterruptible power supplies (UPS), and battery banks), and air conditioning equipment is also increasing. These devices operate in complex environments and may be affected by various human and natural factors (such as lightning strikes, static electricity, and power outages), leading to equipment failures and shutdowns. Furthermore, these devices are relatively independent, and if a fault occurs in one device, it will be impossible to send an external warning or indicate the fault.

[0003] Traditional communication equipment management systems are primarily based on maintaining and managing equipment on a per-site basis. Each site is assigned maintenance personnel for inspections, equipment upkeep, and repairs. However, existing communication monitoring devices process massive amounts of data, making them highly susceptible to anomalies caused by operational data, leading to false alarms or missed alarms. Furthermore, insufficient monitoring accuracy makes it difficult for administrators to accurately assess the equipment's operational status and promptly detect anomalies. This can result in situations where equipment is already damaged and malfunctioning, hindering normal operations and preventing the need for repair or replacement. This approach fails to provide early warnings and proactive maintenance. Summary of the Invention

[0004] To overcome the shortcomings of existing communication monitoring devices, which process a large amount of data and are prone to anomalies due to operational data, leading to false alarms or missed alarms, and whose insufficient monitoring accuracy makes it difficult for administrators to accurately judge the operating status of the equipment and detect abnormal phenomena in a timely manner, resulting in situations where the equipment is already damaged and unable to work properly, thus failing to provide early warning and maintenance, the main objective of this invention is to provide a communication equipment monitoring method, device, and medium.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for monitoring communication equipment, comprising the following steps:

[0006] Acquire real-time monitoring data of communication equipment, extract features of real-time monitoring data and perform cluster analysis to obtain multiple classification index data, and store the multiple classification index data in a distributed manner.

[0007] By stacking time series analysis and deep learning, and combining historical anomaly data from multiple periods, an anomaly monitoring model is constructed.

[0008] Input the classification index data into the abnormal operation monitoring model to determine whether the classification index result is within the set threshold range. When the identification result shows an abnormality, switch the operating motherboard of the communication equipment to the backup motherboard.

[0009] The objective function is set to minimize the false alarm rate. The parameters of the abnormal operation monitoring model of the abnormal motherboard are adjusted based on the anomaly identification results to obtain the adjusted motherboard. The classification index data is input into the abnormal operation monitoring model of the adjusted motherboard. If the output result is normal, the adjusted abnormal motherboard is used as the backup motherboard. If it is abnormal, the anomaly identification result is issued as an anomaly alarm.

[0010] Real-time data includes equipment operating status, performance parameters, and sensor data. The classification index results include data collection volume, collection time for a set amount of data collected at one time, and processing time per unit of data volume.

[0011] By stacking a time series analysis model and a random forest model, and using historical anomaly data from multiple periods as input, an anomaly monitoring model is constructed, including the following steps:

[0012] The time series analysis model selects ARIMA to fit the acquired historical anomaly data from multiple periods to obtain the residual between the predicted and actual values.

[0013] Features are extracted from the residual sequence, and features prior to the occurrence of anomalies are extracted using historical anomaly data to obtain residual features and anomaly event features predicted by the model.

[0014] The residual features and abnormal event features predicted by the model are used as a feature set. A random forest model is trained using the feature set to construct an abnormal operation monitoring model.

[0015] The process of extracting features from real-time data and performing cluster analysis to obtain various classification index data includes the following steps:

[0016] Multiple features are applied to the preprocessed real-time data to obtain various features of the real-time data, including time domain, frequency domain, time-frequency domain, statistical and dynamic features;

[0017] Principal component analysis was used to reduce the dimensionality of the extracted features, resulting in a variety of dimensionality-reduced feature data.

[0018] Hierarchical clustering is performed on the dimensionality-reduced feature data to obtain multiple classification feature data, which are used as indicators.

[0019] The step of inputting classification index data into the abnormal operation monitoring model and determining whether the classification index results meet the set threshold range includes the following steps:

[0020] Preprocessed classification index data is obtained from the distributed storage system and input into the abnormal operation monitoring model;

[0021] In the abnormal operation monitoring model, the threshold range of the classification indicators is preset and compared with the actual observed values;

[0022] Each classification indicator is compared with its corresponding threshold to determine whether it exceeds the set range. If it exceeds the range, it is marked as abnormal data.

[0023] For the tagged anomalous data, perform anomaly identification to determine the type and severity of the anomaly.

[0024] It also includes recording abnormal data, including the time when the abnormality occurred, the type of abnormality, and the value of the abnormal indicator.

[0025] The objective function is set to minimize the false alarm rate. Based on the anomaly identification results, parameters of the faulty motherboard are adjusted, including the following steps:

[0026] The objective function is set to minimize the false alarm rate, and the anomaly identification results are used as input;

[0027] Based on the anomaly identification results and the set target function, obtain the target function value under the current parameter settings;

[0028] Using selected simulated annealing, search in the parameter space for the optimal combination of parameters that optimizes the objective function to obtain iterative results;

[0029] Based on the iterative results of the optimization algorithm, update the parameter values ​​of the abnormal operation monitoring model for the abnormal motherboard.

[0030] The simulated annealing includes:

[0031] Randomly initialize the abnormal operation monitoring model parameter values ​​and operation strategy of the abnormal motherboard, and set the initial temperature and cooling rate;

[0032] Iterative optimization is performed within a certain temperature range until the stopping condition is reached. The current parameter values ​​and operating strategies are then fine-tuned or randomly varied to generate new parameter values ​​and operating strategies.

[0033] Based on the new parameter values ​​and operating strategies, calculate the objective function value, including indicators such as anomaly occurrence rate and equipment performance stability;

[0034] Determine whether to accept the new parameter values ​​and operating strategy; if the new objective function value is better, accept it directly; otherwise, accept it with a certain probability, the probability of which is related to the difference between the old and new objective function values ​​and the current temperature.

[0035] The iteration ends when the set stopping condition is met and the maximum number of iterations is reached;

[0036] Return the final converged parameter values ​​and operating strategy as the adjusted abnormal motherboard settings.

[0037] Compared with existing technologies, the beneficial effects of this invention are as follows: This application, through real-time monitoring data acquisition and feature extraction, can continuously monitor the operating status of communication equipment, ensuring a rapid response to any potential problems. Cluster analysis further helps identify key performance indicators from large amounts of data, enabling more accurate monitoring of equipment status. The abnormal operation monitoring model utilizes time series analysis and deep learning techniques (stacked random forest models) to enhance the ability to identify abnormal patterns, improving the accuracy and reliability of detection. When the monitoring model detects an anomaly, it can automatically switch the operating motherboard of the device to the backup motherboard, thereby reducing system downtime and maintaining service continuity and stability. Setting the objective function to minimize the false alarm rate and optimizing the parameters of the monitoring model through simulated annealing algorithm can significantly reduce false alarms and missed alarms, thereby improving the overall efficiency of the system and user trust. By using simulated annealing algorithm, the monitoring parameters can be self-adjusted in a dynamic environment, allowing the anomaly detection model to continuously adapt to new operating conditions and abnormal patterns. This adaptive adjustment strategy improves the robustness and flexibility of the model in a changing environment.

[0038] The optimized parameter settings can be applied to further updates and upgrades of the monitoring system, ensuring long-term stable operation and reducing maintenance requirements. Detailed anomaly data recorded by the system, such as the time, type, and index values ​​of anomalies, provides rich data support for subsequent analysis and decision-making. This data can be used for trend analysis, performance evaluation, and risk management, helping managers make more informed and scientific decisions.

[0039] In summary, this technical solution improves the stability, reliability, and operational efficiency of communication equipment through highly integrated monitoring, diagnostic, and recovery functions. At the same time, it reduces the time cost of manual intervention and fault handling, enhances the real-time performance and intelligence of the system, and strengthens the ability to respond to complex environments and emergencies. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the structural process of the present invention;

[0041] Figure 2 This is a flowchart illustrating the data classification process of this invention;

[0042] Figure 3 This is a schematic diagram of the anomaly detection model of the present invention;

[0043] Figure 4 This is a flowchart illustrating the process of adjusting a faulty motherboard according to the present invention. Detailed Implementation

[0044] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0045] Example:

[0046] See Figure 1 A communication equipment monitoring method is used to monitor relevant parameters of the communication equipment, ensure its normal operation, and promptly handle any abnormal situations that occur. Specifically, it includes:

[0047] The process includes real-time data collection and preprocessing of communication equipment, generation of classification index results, identification and optimization training of abnormal operation monitoring models, and handling of abnormal situations.

[0048] Among these steps, using sensors and monitoring devices to collect data from communication equipment in real time, and real-time collection and preprocessing of equipment operating status, performance parameters, and sensor data are crucial steps to ensure data quality and availability. Streaming technologies such as Apache Kafka and Apache Flink can be used to collect data including operating status, performance parameters such as temperature, voltage, and frequency, and sensor data. Preprocessing of the collected data includes data cleaning, noise reduction, and normalization to ensure data quality and accuracy.

[0049] Features are extracted from the preprocessed data, such as device operating frequency, temperature variation trends, and voltage fluctuations. Cluster analysis methods, such as K-means clustering and DBSCAN, are used to discover potential patterns and cluster structures in the feature data, yielding multiple classification indicators, such as device status (normal, warning, abnormal), and performance scores, for subsequent anomaly detection and operational optimization. The extracted classification indicator data is stored in a distributed storage system, such as Hadoop HDFS or Apache Cassandra, for fast access and processing.

[0050] By combining time series analysis with deep learning and integrating historical anomaly data from multiple periods, an anomaly monitoring model is constructed. Classification index results are input into this model for identification and judgment. The model determines whether the classification index results are abnormal based on preset threshold ranges. When the anomaly monitoring model identifies an anomaly, it immediately switches to a backup motherboard to ensure continuous and reliable operation of the equipment and guarantee the stable operation of the communication system.

[0051] The abnormal data portion is used as the training set to optimize and train the abnormal operation monitoring model, thereby improving its ability and accuracy in identifying abnormal situations.

[0052] The objective function is set to minimize the false alarm rate. Based on the anomaly identification results, parameters and operating strategies are adjusted for abnormal motherboards to try to resolve the abnormal situation and optimize the motherboard's operating performance and stability.

[0053] If the problem persists after adjustment and testing again, an error alert will be sent to the remote control center to notify the administrator for further processing and scheduling.

[0054] Throughout the process, the collection and processing of real-time data can promptly detect anomalies in communication equipment, while the generation of classification index results and the identification and optimization training of monitoring models can effectively monitor and respond to anomalies, ensuring the stable operation of the communication system.

[0055] The basic principle of a communication equipment monitoring device is as follows: real-time data collection and processing can promptly detect abnormal situations of communication equipment; the generation of classification index results and the identification and optimization training of monitoring models can effectively monitor and respond to abnormal situations, ensuring the stable operation of the communication system.

[0056] Example 2

[0057] A communication monitoring method, in practical applications, uses sensors to collect real-time data from equipment, including temperature, humidity, and voltage. The collected data undergoes preprocessing, including data cleaning, noise reduction, and normalization, to ensure data quality and usability. Features, such as average temperature and maximum voltage, are extracted from the preprocessed data. Clustering analysis techniques, such as K-means clustering, are used to classify the equipment status into normal and abnormal categories. The extracted classification index data is stored in a distributed storage system, such as Hadoop HDFS. The classification index data is input into an anomaly monitoring model to detect the presence of anomalies. If the output result is abnormal, the communication equipment switches to a backup motherboard to ensure continuous operation. Simultaneously, the abnormal data is used as a training set to optimize and update the parameters of the anomaly monitoring model, improving its accuracy and robustness. Based on a multi-objective optimization algorithm, combined with the anomaly identification results, the parameters and operating strategies of the abnormal motherboard are adjusted to optimize the equipment's performance and stability. If the adjusted abnormal motherboard's output result is within the normal range, then the adjusted motherboard is used as the backup motherboard.

[0058] If the adjusted motherboard output remains abnormal, the monitoring device immediately issues an error alarm and notifies the remote control center so that timely measures can be taken to address the problem. This embodiment integrates real-time data acquisition, preprocessing, anomaly monitoring, and automatic switching into a complete fault detection and automatic switching system, ensuring reliable operation of communication equipment under monitoring and timely and effective fault handling.

[0059] More specifically, the classification index results include the amount of data collected, the collection time for a set amount of data collected at one time, and the processing time per unit of data.

[0060] To extract features from the preprocessed data and perform cluster analysis to obtain various classification index data, the following steps are included:

[0061] Multiple features are applied to the preprocessed real-time data to obtain various features of the real-time data, including time domain, frequency domain, time-frequency domain, statistical and dynamic features;

[0062] Principal component analysis was used to reduce the dimensionality of the extracted features, resulting in a variety of dimensionality-reduced feature data.

[0063] Hierarchical clustering is performed on the dimensionality-reduced feature data to obtain multiple classification feature data, which are used as indicators.

[0064] The process of inputting classification indicator data into the abnormal operation monitoring model and determining whether the classification indicator results meet the set threshold range includes the following steps:

[0065] Preprocessed classification index data is obtained from the distributed storage system and input into the abnormal operation monitoring model;

[0066] In the abnormal operation monitoring model, the threshold range of the classification indicators is preset so that it can be compared with the actual observed values;

[0067] Each classification metric is compared to its corresponding threshold to determine if it exceeds the set range. If it exceeds the range, it is marked as outlier data.

[0068] For the tagged anomalous data, perform anomaly identification to determine the type and severity of the anomaly.

[0069] Record abnormal data, including the time of occurrence, type of abnormality, and value of abnormal indicator.

[0070] Based on a multi-objective optimization algorithm and combined with anomaly identification results, the following steps are taken to adjust the parameters and operating strategies of the abnormal motherboard:

[0071] Using the anomaly identification results as input, the multi-objective optimization algorithm selects simulated annealing and defines reducing the motherboard anomaly occurrence rate as the optimization objective.

[0072] Based on the anomaly identification results and the set target function, calculate the target function value under the current parameter settings;

[0073] Using selected simulated annealing, search in the parameter space for the optimal combination of parameters that optimizes the objective function to obtain iterative results;

[0074] Based on the iterative results of the optimization algorithm, update the parameter values ​​and operating strategy of the abnormal motherboard.

[0075] The simulated annealing includes:

[0076] Randomly initialize the parameter values ​​and operating strategy of the abnormal motherboard, and set the initial temperature and cooling rate;

[0077] Iterative optimization is performed within a certain temperature range until the stopping condition is reached. The current parameter values ​​and operating strategies are then fine-tuned or randomly varied to generate new parameter values ​​and operating strategies.

[0078] Based on the new parameter values ​​and operating strategies, calculate the objective function value, including indicators such as anomaly occurrence rate and equipment performance stability;

[0079] Determine whether to accept the new parameter values ​​and operating strategy; if the new objective function value is better, accept it directly; otherwise, accept it with a certain probability, the probability of which is related to the difference between the old and new objective function values ​​and the current temperature.

[0080] The iteration ends when the set stopping condition is met and the maximum number of iterations is reached;

[0081] Return the final converged parameter values ​​and operating strategy as settings for the adjusted abnormal motherboard.

[0082] Example 3

[0083] A specific implementation case involves monitoring the health status of a communication base station's server. This case identifies and prevents potential faults by collecting and analyzing real-time monitoring data from the server. The entire process is described in detail below, including the specific steps of data collection, model building, anomaly detection, and parameter optimization:

[0084] Collect server CPU usage, memory usage, temperature readings, and network traffic data. Data is collected every 5 minutes.

[0085] The collected data undergoes time-domain, frequency-domain, and time-frequency-domain analysis to extract statistical and dynamic features. Principal component analysis (PCA) is used to reduce the dimensionality of the features. Hierarchical clustering is then applied to the dimensionality-reduced features to generate various classification index data.

[0086] The ARIMA model was used to fit the time series of historical outlier data.

[0087] Calculate the residual between the predicted and actual values, and extract features from the residual sequence. Combine historical anomaly data to extract features prior to the occurrence of the anomaly. Use the output of the time series model and the anomaly features as a feature set to train a random forest model.

[0088] The model's output is categorized into two classes: abnormal and non-abnormal. Classification index data is retrieved from distributed storage and input into the anomaly monitoring model. Data exceeding the model's preset threshold range is marked as abnormal. If data is marked as abnormal, the server automatically switches to the backup motherboard.

[0089] Furthermore, the system automatically corrects malfunctioning motherboards, setting the objective function to minimize the false alarm rate. Based on the anomaly identification results, the system adjusts the parameters of the malfunctioning motherboard, specifically setting the initial temperature and cooling rate, and performing iterative optimization until a stopping condition is met. The monitoring model parameters for the malfunctioning motherboard are updated based on the iterative results obtained from simulated annealing. Classification index data is input into the adjusted motherboard's anomaly monitoring model. If the output is normal, the adjusted malfunctioning motherboard is used as a backup motherboard. If it is abnormal, the anomaly identification result is sent as an anomaly alarm to the remote control center, requiring further adjustments or replacement of the malfunctioning motherboard by staff.

[0090] The results of the implementation case study show that the false positive rate of the model was 5% in the initial stage, which was reduced to 1% after optimization. The formula for calculating the false positive rate is:

[0091] The health monitoring system for communication equipment can identify potential problems in real time, respond to abnormal events promptly, and continuously improve the stability and reliability of the system through intelligent optimization.

[0092] It should be noted that, in this invention, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus.

[0093] The above embodiments are merely illustrative examples of the present invention and do not constitute a limitation on the scope of protection of the present invention. Any designs that are the same as or similar to the present invention are within the scope of protection of the present invention.

Claims

1. A method for monitoring communication equipment, characterized in that, Includes the following steps: Acquire real-time monitoring data of communication equipment, extract features of real-time monitoring data and perform cluster analysis to obtain multiple classification index data, and store the multiple classification index data in a distributed manner. By stacking time series analysis models and random forest models, and combining them with historical anomaly data from multiple periods, an anomaly monitoring model is constructed. Input the classification index data into the abnormal operation monitoring model to determine whether the classification index result is within the set threshold range. When the identification result shows an abnormality, switch the operating motherboard of the communication equipment to the backup motherboard. The objective function is set to minimize the false alarm rate. The parameters of the abnormal motherboard are adjusted based on the anomaly identification results to obtain the adjusted motherboard. The classification index data of the adjusted motherboard is input into the abnormal operation monitoring model. If the output result is normal, the adjusted abnormal motherboard is used as the backup motherboard. If it is abnormal, the anomaly identification result is issued as an anomaly alarm.

2. The communication equipment monitoring method according to claim 1, characterized in that, The real-time monitoring data includes equipment operating status, performance parameters, and sensor data. The classification index results include data collection volume, collection time for a set amount of data collected at one time, and processing time per unit of data volume.

3. The communication equipment monitoring method according to claim 1, characterized in that, The method of stacking a time series analysis model and a random forest model, and using historical anomaly data from multiple periods as input to construct an anomaly monitoring model, includes the following steps: The time series analysis model selects ARIMA to fit the acquired historical anomaly data from multiple periods to obtain the residual between the predicted and actual values. Features are extracted from the residual sequence, and features prior to the occurrence of anomalies are extracted using historical anomaly data to obtain residual features and anomaly event features predicted by the model. The residual features and abnormal event features predicted by the model are used as a feature set. A random forest model is trained using the feature set to construct an abnormal operation monitoring model.

4. The communication equipment monitoring method according to claim 1, characterized in that, The process of extracting features from real-time monitoring data and performing cluster analysis to obtain various classification index data includes the following steps: Multiple features are extracted from the preprocessed real-time monitoring data to obtain various features of the real-time monitoring data, including time domain, frequency domain, time-frequency domain, statistical and dynamic features; Principal component analysis was used to reduce the dimensionality of the extracted features, resulting in a variety of dimensionality-reduced feature data. Hierarchical clustering is performed on the dimensionality-reduced feature data to obtain multiple classification feature data, which are used as indicators.

5. The communication equipment monitoring method according to claim 1, characterized in that, The step of inputting classification index data into the abnormal operation monitoring model and determining whether the classification index results meet the set threshold range includes the following steps: Preprocessed classification index data is obtained from the distributed storage system and input into the abnormal operation monitoring model; In the abnormal operation monitoring model, the threshold range of the classification indicators is preset and compared with the actual observed values; Each classification indicator is compared with its corresponding threshold to determine whether it exceeds the set range. If it exceeds the range, it is marked as abnormal data. For the tagged anomalous data, perform anomaly identification to determine the type and severity of the anomaly.

6. The communication equipment monitoring method according to claim 1, characterized in that, It also includes recording abnormal data, including the time when the abnormality occurred, the type of abnormality, and the value of the abnormal indicator.

7. The communication equipment monitoring method according to claim 1, characterized in that, The objective function is set to minimize the false alarm rate. Based on the anomaly identification results, parameters of the faulty motherboard are adjusted, including the following steps: The objective function is set to minimize the false alarm rate, and the anomaly identification results are used as input; Based on the anomaly identification results and the set target function, obtain the target function value under the current parameter settings; Using selected simulated annealing, search in the parameter space for the optimal combination of parameters that optimizes the objective function to obtain iterative results; Based on the iterative results of the optimization algorithm, update the parameter values ​​and operating strategy of the abnormal motherboard.

8. The communication equipment monitoring method according to claim 7, characterized in that, The simulated annealing includes: Randomly initialize the abnormal operation monitoring model parameter values ​​and operation strategy of the abnormal motherboard, and set the initial temperature and cooling rate; Iterative optimization is performed within a certain temperature range until the stopping condition is reached. The current parameter values ​​and operating strategies are then fine-tuned or randomly varied to generate new parameter values ​​and operating strategies. Based on the new parameter values ​​and operating strategies, calculate the objective function value, including the anomaly rate and equipment performance stability indicators; Determine whether to accept the new parameter values ​​and operating strategy; if the new objective function value is better, accept it directly; otherwise, accept it with a certain probability, the probability of which is related to the difference between the old and new objective function values ​​and the current temperature. The iteration ends when the set stopping condition is met and the maximum number of iterations is reached; Return the final converged parameter values ​​and operating strategy as the adjusted abnormal motherboard settings.

9. A communication equipment monitoring system, characterized in that, include: The data processing module is used to acquire real-time monitoring data of communication equipment, extract features of real-time monitoring data and perform cluster analysis to obtain multiple classification index data, and distribute the multiple classification index data for storage. The anomaly detection model building module is used to stack the time series analysis model and the random forest model, and combine the acquired historical anomaly data from multiple periods to build an anomaly operation monitoring model. The anomaly handling module is used to input classification index data into the anomaly operation monitoring model, determine whether the classification index result is within the set threshold range, and switch the operating motherboard of the communication device to the backup motherboard when the identification result shows an anomaly. The monitoring and early warning module is used to set the objective function to minimize the false alarm rate, and adjust the parameters of the abnormal motherboard based on the anomaly identification results to obtain the adjusted motherboard. The classification index data of the adjusted motherboard is input into the abnormal operation monitoring model. If the output result is normal, the adjusted abnormal motherboard is used as the backup motherboard. If it is abnormal, the anomaly identification result is issued as an anomaly alarm.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the method described in any one of claims 1 to 8.

Images