Fault self-diagnosis method and device for concentrator

By evaluating and partitioning the topology network of the devices connected to the concentrator, and combining it with a fault self-diagnosis model, the concentrator can monitor and generate fault warning signals in real time, thus solving the problem of low efficiency in concentrator fault self-diagnosis and achieving efficient and rapid fault identification.

CN118802475BActive Publication Date: 2024-11-19JIANGSU DADIAN ENERGY TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411281753.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-13
Publication Date
2024-11-19
Estimated Expiration
2044-09-13

AI Technical Summary

Technical Problem

Existing self-diagnostic methods for concentrator faults are inefficient and slow to respond, making it difficult to meet the high-efficiency, precision, and safety requirements of modern industrial production.

Method used

By acquiring basic information about multiple devices connected to the concentrator, a device topology network is established, network environment assessment and partitioning are performed, a fault self-diagnosis model is trained, and fault early warning signals are generated in real time.

Benefits of technology

It improves the accuracy and response speed of fault identification, and enables efficient diagnosis and management of concentrator faults.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118802475B_ABST
    Figure CN118802475B_ABST
Patent Text Reader

Abstract

The present invention discloses a fault self-diagnosis method and device for a concentrator, which relates to the technical field of fault diagnosis, including: obtaining multiple devices connected to the concentrator and establishing a device topology network; identifying basic information of each device in the multiple devices; evaluating the network environment of the multiple devices according to the basic information, and outputting multiple network environment indicators; partitioning the device topology network according to the multiple network environment indicators, and outputting K partitions; training K fault self-diagnosis models corresponding to the K partitions; inputting K groups of real-time operation data sets corresponding to the K partitions into the K fault self-diagnosis models, and obtaining K fault identification results; generating a fault warning signal according to the K fault identification results, and sending it to a host computer for reminder. The present invention solves the technical problems of low efficiency and slow response in the prior art for the fault self-diagnosis of the concentrator, and achieves the technical effect of improving the accuracy of fault identification and the response speed.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of fault diagnosis, and in particular to a fault self-diagnosis method and device for a concentrator. Background Art

[0002] With the rapid development of modern network technology and industrial automation, concentrators play a vital role in data collection, processing and distribution. Concentrators are widely used in various network environments and are responsible for aggregating and distributing large amounts of sensor data, device status information and user requests. However, with the continuous expansion of network scale and the diversification of business needs, the management and maintenance of concentrators are also facing increasing challenges.

[0003] Although the traditional concentrator management method has achieved a certain degree of automation, there are still many problems in practical application. With the continuous expansion of production scale and the increasing complexity of process flow, the fault self-diagnosis method of concentrator faces increasing challenges. The traditional concentrator fault self-diagnosis method has problems such as low efficiency and slow response, which can hardly meet the requirements of modern industrial production for high efficiency, accuracy and safety. Summary of the invention

[0004] The present application provides a fault self-diagnosis method and device for a concentrator, which is used to solve the technical problems of low efficiency and slow response in the prior art for fault self-diagnosis of a concentrator.

[0005] In view of the above problems, the present application provides a fault self-diagnosis method and device for a concentrator.

[0006] In a first aspect of the present application, a fault self-diagnosis method for a concentrator is provided, the method comprising:

[0007] Acquire multiple devices connected to the concentrator and establish a device topology network; identify basic information of each of the multiple devices, including physical location, network configuration, connection type, IP address and subnet mask; evaluate the network environment of the multiple devices according to the basic information, and output multiple network environment indicators, wherein each device corresponds to a network environment indicator; partition the device topology network according to the multiple network environment indicators, and output K partitions, wherein the network environment similarity of each partition is greater than a preset similarity, and K is a positive integer greater than or equal to 2, representing the number of partitions; train K fault self-diagnosis models corresponding to the K partitions; input the K groups of real-time operation data sets corresponding to the K partitions into the K fault self-diagnosis models to obtain K fault identification results; generate a fault warning signal according to the K fault identification results, and send the fault warning signal to the host computer for reminder.

[0008] A second aspect of the present application provides a fault self-diagnosis device for a concentrator, the device comprising:

[0009] A topology network establishment module, the topology network establishment module acquires multiple devices connected to the concentrator and establishes a device topology network; a device information identification module, the device information identification module identifies the basic information of each of the multiple devices, including the physical location, network configuration, connection type, IP address and subnet mask; a network environment evaluation module, the network environment evaluation module evaluates the network environment of the multiple devices according to the basic information and outputs multiple network environment indicators, wherein each device corresponds to a network environment indicator; a partitioning module, the partitioning module partitions the device topology network according to the multiple network environment indicators and outputs K partitions districts, wherein the network environment similarity of each partition is greater than a preset similarity, K is a positive integer greater than or equal to 2, representing the number of partitions; a fault self-diagnosis model training module, the fault self-diagnosis model training module trains K fault self-diagnosis models corresponding to the K partitions; a fault identification result acquisition module, the fault identification result acquisition module inputs K groups of real-time operation data sets corresponding to the K partitions into the K fault self-diagnosis models to obtain K fault identification results; a fault warning signal generation module, the fault warning signal generation module generates a fault warning signal according to the K fault identification results, and sends the fault warning signal to the host computer for reminder.

[0010] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0011] The present application obtains multiple devices connected to the concentrator and establishes a device topology network; identifies the basic information of each device in the multiple devices, including physical location, network configuration, connection type, IP address and subnet mask; evaluates the network environment of the multiple devices according to the basic information, and outputs multiple network environment indicators, wherein each device corresponds to a network environment indicator; partitions the device topology network according to the multiple network environment indicators, and outputs K partitions, wherein the network environment similarity of each partition is greater than the preset similarity, and K is a positive integer greater than or equal to 2, representing the number of partitions; trains K fault self-diagnosis models corresponding to the K partitions; inputs the K groups of real-time operation data sets corresponding to the K partitions into the K fault self-diagnosis models, and obtains K fault identification results; generates a fault warning signal according to the K fault identification results, and sends the fault warning signal to the host computer for reminder. The present invention solves the technical problems of low efficiency and slow response of the fault self-diagnosis of the concentrator in the prior art, and comprehensively evaluates and diagnoses the network environment of the device by real-time monitoring of the key parameters of the multiple devices connected to the concentrator, and combines the partition management of the device topology network with the customized fault self-diagnosis model to achieve the technical effect of improving the accuracy of fault identification and the response speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0013] Figure 1 A schematic flow chart of a fault self-diagnosis method for a concentrator provided in an embodiment of the present application.

[0014] Figure 2 A schematic diagram of the structure of a fault self-diagnosis device for a concentrator provided in an embodiment of the present application.

[0015] Explanation of the reference numerals: topology network establishing module 11 , device information identifying module 12 , network environment evaluating module 13 , partitioning module 14 , fault self-diagnosis model training module 15 , fault identification result acquiring module 16 , fault warning signal generating module 17 . DETAILED DESCRIPTION

[0016] The present application provides a fault self-diagnosis method and device for a concentrator, aiming to solve the technical problems of low efficiency and slow response in the prior art for fault self-diagnosis of a concentrator. By real-time monitoring of key parameters of multiple devices connected to the concentrator, a comprehensive assessment and diagnosis of the network environment of the devices is performed. Combined with the partition management of the device topology network and a customized fault self-diagnosis model, the application achieves the technical effect of improving the accuracy of fault identification and the response speed.

[0017] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0018] It should be noted that any variations of the terms "include" and "have" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0019] Embodiment 1

[0020] like Figure 1As shown, the present application provides a fault self-diagnosis method for a concentrator, the method comprising:

[0021] Step S100: Acquire multiple devices connected to the concentrator and establish a device topology network.

[0022] In an embodiment of the present application, all devices currently connected to the concentrator are identified through a network scanning tool such as Nmap. In this process, the network is scanned by Nmap, and by specifying a subnet range, Nmap scans all IP addresses within the range and lists the responding devices. The connection relationship between devices is identified using protocols such as LLDP or CDP. For example, LLDP and CDP establish a connection relationship by exchanging data packets containing device ports and adjacent device information. Based on the identified connection relationship, a network topology drawing tool such as Graphviz is used to draw a device topology network.

[0023] Step S200: Identify basic information of each device in the plurality of devices, including physical location, network configuration, connection type, IP address and subnet mask.

[0024] In an embodiment of the present application, when the basic information of each device in multiple devices is identified, the geographic location information of the device is obtained through the built-in GPS. The network configuration information of the device is obtained by sending an SNMP request, including interface configuration, routing table, etc. For the identification of the connection type, the LLDP or CDP protocol is enabled, and the connection information is exchanged between devices to identify the connection type, such as Ethernet, fiber optic or wireless connection. Finally, the identification of the IP address and subnet mask is achieved through the SNMP or SSH protocol and the DHCP server record. For example, use the SNMP management tool to send a request to directly obtain the IP address and subnet mask information of the device. This information can also be obtained by remotely logging into the device through SSH and executing the network configuration query command.

[0025] Step S300: Evaluate the network environment of the multiple devices according to the basic information, and output multiple network environment indicators, wherein each device corresponds to one network environment indicator.

[0026] In an embodiment of the present application, when evaluating the network environment of multiple devices, the network environment indicators of each device are first calculated based on basic information. These indicators include network delay, bandwidth utilization, packet loss rate, signal strength, etc.

[0027] For network latency, calculate the average round-trip time from the device to the concentrator based on the results of the Ping command. For bandwidth utilization, use SNMP to obtain interface traffic data and calculate bandwidth utilization. For packet loss rate, calculate packet loss within a certain period of time based on the results of the Ping command. For signal strength, obtain signal strength data through SNMP or the device's built-in management interface.

[0028] Next, normalize the different network environment indicators and generate the comprehensive network environment indicators for each device. When normalizing network delay, divide the measured delay value by the maximum network delay value. When normalizing bandwidth utilization, divide the actual bandwidth utilization by the total bandwidth. When normalizing packet loss rate, use the packet loss rate value directly. When normalizing signal strength, divide the signal strength value by the maximum signal strength value.

[0029] Finally, the normalized indicators are combined through the weighted average method to generate the comprehensive network environment indicator of each device, in which each normalized indicator has the same weight.

[0030] Step S400: partitioning the device topology network according to the multiple network environment indicators, and outputting K partitions, wherein the network environment similarity of each partition is greater than a preset similarity, and K is a positive integer greater than or equal to 2, representing the number of partitions.

[0031] In an embodiment of the present application, when the device topology network is partitioned according to multiple network environment indicators, the K-means clustering algorithm is selected to partition the devices. In this process, the network environment indicators of K devices are first randomly selected as the initial cluster centers, and then the Euclidean distance of each device to each cluster center is calculated, and the device is assigned to the nearest cluster center to form an initial partition. Then, based on the devices in the current partition, the new cluster center of each partition is calculated, that is, the average value of the network environment indicators of all devices in the partition. The steps of allocating devices and calculating cluster centers are repeated until the cluster centers are stable, that is, the cluster centers of each partition no longer change significantly.

[0032] In the cluster analysis process, to ensure that the network environment similarity of each partition is greater than the preset similarity, the similarity of the devices in the partition is calculated. The similarity is measured by calculating the standard deviation or variance of the network environment indicators of the devices in the partition. If the similarity of a partition is lower than the preset threshold, adjust the K value or re-cluster analysis until the similarity of all partitions meets the requirements. Finally, the results of the cluster analysis are output as K partitions, each of which contains a number of devices, and the network environment similarity of each partition is marked.

[0033] Step S500: training K fault self-diagnosis models corresponding to the K partitions.

[0034] In an embodiment of the present application, operation data and fault data are collected from the equipment in each partition through network management tools, such as SNMP and fault log system. The collected data is preprocessed, mainly including data cleaning, feature extraction and data standardization. The preprocessed data set is divided into a training set and a test set. The division is divided according to the ratio of 8:2, that is, 80% is used for training the model and 20% is used for testing the performance of the model. A machine learning or deep learning model suitable for fault self-diagnosis is selected, such as a random forest. The model is trained using the training set data, and the performance of the model is evaluated using the test set data. Through this process, K fault self-diagnosis models corresponding to K partitions are obtained.

[0035] Step S600: inputting K groups of real-time operation data sets corresponding to the K partitions into the K fault self-diagnosis models to obtain K fault identification results.

[0036] In an embodiment of the present application, the real-time operation data of the equipment in each partition is continuously collected through network management tools and protocols to obtain K groups of real-time operation data. The K groups of real-time operation data collected in real time are preprocessed, including data cleaning, feature extraction and standardization. The preprocessed K groups of real-time operation data are input into the corresponding fault self-diagnosis model, and the model performs fault identification according to the input data and outputs K fault identification results.

[0037] Step S700: Generate a fault warning signal according to the K fault identification results, and send the fault warning signal to a host computer for reminder.

[0038] In an embodiment of the present application, the fault identification results output by the fault self-diagnosis model of each partition are first summarized. These results include detailed information such as the fault type, fault probability, and the time and device when the fault occurred. According to the fault identification results, certain thresholds and rules are set to generate fault warning signals. For example, if the failure probability of a certain device exceeds a preset threshold, a fault warning signal is generated. According to the identified fault type, such as excessive network delay, excessive bandwidth utilization, excessive packet loss rate, etc., a corresponding warning signal is generated. If multiple devices in a partition fail at the same time, a high-level comprehensive fault warning signal is generated. Finally, the fault warning signal is sent to the host computer for reminder.

[0039] Furthermore, step S700 in the method provided in the application embodiment also includes:

[0040] The K fault indicators in the K fault identification results are judged; the number of partitions greater than a preset fault threshold is obtained, and if the number of partitions is greater than or equal to the preset number, a concentrator-fault warning signal is generated; after the host computer receives the concentrator-fault warning signal, the backup concentrator is activated, and the device topology network is updated with the backup concentrator.

[0041] In the embodiment of the present application, the fault identification result is first obtained from the fault self-diagnosis model of each partition. The fault identification result of each partition includes multiple fault indicators, which reflect the operating status and fault probability of the equipment in the partition. The fault indicators include network delay, bandwidth utilization, packet loss rate, signal strength, etc.

[0042] Next, the fault indicators of each partition are judged. The judgment standard is the preset fault threshold, which is pre-set by technical experts. For each partition, if a certain fault indicator exceeds the preset fault threshold, the partition is considered to have a fault. For example, the preset fault threshold of network delay is 200ms, the fault threshold of bandwidth utilization is 80%, the fault threshold of packet loss rate is 5%, and the fault threshold of signal strength is -70dBm. If the network delay of a partition exceeds 200ms, the partition is marked as a faulty partition.

[0043] After judging all partitions, count the number of partitions that are greater than the preset fault threshold. Set a preset number, and if the number of partitions that are greater than the preset fault threshold is greater than or equal to the preset number, generate a concentrator fault warning signal. For example, the preset number is 3, and if the number of partitions that are greater than the preset fault threshold is 3 or more, generate a concentrator fault warning signal.

[0044] The concentrator fault warning signal is sent to the host computer. After receiving the concentrator fault warning signal, the host computer activates the backup concentrator. The backup concentrator is a pre-configured concentrator that can take over the work of the current faulty concentrator. Specifically, the host computer is connected to the backup concentrator through the network and sends the configuration information of the current device topology network to the backup concentrator. The configuration information includes network parameters such as the device's IP address, subnet mask, and gateway. The host computer adjusts the network routing and switches the device's network traffic to the backup concentrator. Through these steps, the backup concentrator takes over the work of the original concentrator to ensure the normal operation of the network and the stable connection of the device.

[0045] Furthermore, in the method provided in the embodiment of the application, after obtaining the number of partitions greater than the preset fault threshold, it also includes:

[0046] If the number of partitions is less than the preset number, a partition-fault warning signal is generated; after the host computer receives the partition-fault warning signal, the fault source is located for the abnormal partition.

[0047] In an embodiment of the present application, the number of partitions greater than a preset fault threshold is counted. If the number of partitions greater than the preset fault threshold is less than the preset number, a partition fault warning signal is generated. For example, if the preset number is 3, and the number of partitions greater than the preset fault threshold is 2, a partition fault warning signal is generated. After the host computer receives the partition fault warning signal, it starts to locate the fault source of the abnormal partition. Using network management and fault diagnosis technology, the host computer gradually narrows the scope of the fault and determines the specific faulty device. Specifically, use tools such as SNMP and SSH to query the device status and check the device logs and performance indicators. Obtain the device's interface traffic, CPU utilization, memory usage, etc. through SNMP, log in to the device through SSH and check the detailed system logs and operating status. Finally, through the fault tree analysis method, the possible causes of the fault are gradually checked to complete the fault source location of the abnormal partition.

[0048] Furthermore, in the method provided in the embodiment of the application, after the host computer receives the concentrator-fault warning signal, the standby concentrator is activated, and the method further includes:

[0049] Connect to the data management center where the concentrator is located and output the concentrator distribution network; obtain the neighboring candidate concentrators of the concentrator according to the concentrator distribution network; identify the supportable protocol types, centralized processing performance and configuration attribute information of the concentrator; perform compatibility matching on the neighboring candidate concentrators according to the supportable protocol types, centralized processing performance and configuration attribute information of the concentrator, and output the backup concentrator.

[0050] In an embodiment of the present application, the data management center is a centralized management platform that communicates with the concentrator through network management protocols such as SNMP and system interfaces such as API. Through these tools, the location information, connection status and network topology of all concentrators are obtained, the obtained data are integrated, and the concentrator distribution network is output.

[0051] Next, according to the obtained concentrator distribution network, the neighbor candidate concentrators of the current concentrator are identified. The neighbor concentrators refer to the concentrators that have a physical or logical proximity relationship with the current concentrator in the network topology. By analyzing the network topology and using topology analysis tools or algorithms, such as the Dijkstra algorithm, it is determined which concentrators have direct or indirect connections with the current concentrator, and the neighbor candidate concentrators of the concentrator are obtained.

[0052] To obtain the supported protocol types, use the SNMP management tool to send a request to query the MIB object of the concentrator, read the OID value related to the protocol support, and determine the protocol types supported by the device, such as IP, TCP, UDP, etc. To obtain the centralized processing performance, query the OIDs of CPU usage and memory usage through SNMP to obtain the processing performance data of the device. Log in to the device remotely through SSH and execute commands to obtain real-time CPU and memory usage. To obtain configuration attribute information, use SNMP to query the OIDs of the number of interfaces and bandwidth to obtain the interface configuration and bandwidth information of the device. Use SNMP to query the OID of the IP address configuration to obtain the IP address information of the device. Log in to the device through SSH, read and parse the device configuration file, and obtain detailed network configuration parameters.

[0053] Compatibility matching is performed on neighboring candidate concentrators based on the concentrator's supported protocol types, centralized processing performance, and configuration attribute information. When performing compatibility matching, check whether each candidate concentrator supports all protocol types required by the current concentrator. If a candidate concentrator does not support any of the necessary protocol types, it is excluded from the candidate list. Compare the CPU and memory usage of the candidate concentrators and the current concentrator. Select concentrators with CPU and memory usage below a certain threshold, such as 80%, to ensure that the backup concentrator has sufficient processing power. Check the number of interfaces, bandwidth, and IP address configuration of the candidate concentrators to ensure that these attributes meet the requirements of the current concentrator. If a candidate concentrator does not meet the requirements in terms of the number of interfaces or bandwidth, it is excluded from the candidate list. After completing the compatibility matching, the neighboring candidate concentrators that meet the requirements are output as backup concentrators.

[0054] Furthermore, step S500 in the method provided in the application embodiment also includes:

[0055] Obtain a first network environment indicator corresponding to a first partition; collect a first group of fault operation data samples of each device in the first partition under the first network environment indicator; perform feature convolution on the first group of fault operation data samples to obtain feature operation data samples related to the concentrator, including forwarding-feature operation data samples, switching-feature operation data samples and integration-feature operation data samples in which the concentrator participates; train a first fault self-diagnosis model corresponding to the first partition according to the forwarding-feature operation data samples, switching-feature operation data samples and integration-feature operation data samples and identification information of the fault indicator; perform transfer learning with the first fault self-diagnosis model as the source model, and output K fault self-diagnosis models corresponding to the K partitions.

[0056] In an embodiment of the present application, network environment indicators of the first partition are obtained through a network management tool, such as SNMP. These indicators include network delay, bandwidth utilization, packet loss rate, signal strength, etc. Under the determined network environment indicators, a first group of fault operation data samples of each device in the first partition are collected from the fault log of the network management system. These data samples include operation data of the device under different fault conditions, such as device behavior under scenarios such as increased network delay and excessive bandwidth utilization.

[0057] The first group of fault operation data samples collected are subjected to feature convolution to extract feature operation data samples related to the concentrator. Feature convolution is achieved by using deep learning technologies such as convolutional neural networks. Specifically, it includes forwarding features, switching features, and integration features. Forwarding features involve the operation data of the concentrator during the packet forwarding process, such as forwarding delay, packet loss rate, etc. Switching features involve the operation data of the concentrator during the switching process, such as switching speed, bandwidth utilization, etc. Integration features are the operation data of the comprehensive forwarding and switching processes, reflecting the overall network performance. Through this process, feature operation data samples related to the concentrator are obtained, including forwarding-feature operation data samples, switching-feature operation data samples, and integration-feature operation data samples in which the concentrator participates. After that, the identification information of the corresponding fault indicator is assigned to each feature operation data sample, and the fault type or normal state corresponding to each data sample is marked through the network monitoring system, such as Nagios and Zabbix, combined with the historical fault log.

[0058] The training data set is constructed by combining the forwarding-feature operation data sample, the switching-feature operation data sample, the integration-feature operation data sample and the identification information of the fault indicator. Each record contains feature data and the corresponding fault label. A suitable machine learning model is selected for training. Here, the random forest model is used. First, the data set is divided into a training set and a test set, and then a random forest model is constructed and trained. During the training process, the model performance is optimized through cross-validation and parameter adjustment so that it can accurately identify faults. Through this process, the first fault self-diagnosis model corresponding to the first partition is obtained.

[0059] In order to perform fault self-diagnosis on other partitions, the first fault self-diagnosis model is used as the source model, and the fault self-diagnosis models of other partitions are trained through transfer learning. In this process, the trained feature extraction layers in the first fault self-diagnosis model are migrated to the new model. These layers contain the basic recognition ability of fault features. The shared layers in the source model are retained. These layers have strong universality for the data characteristics of different partitions. New layers that adapt to the data of the new partition are added to the migrated model. These layers are used to adjust and optimize the model's specific processing of the data of the new partition. The migrated model is fine-tuned using the data of the new partition. The fine-tuning process includes collecting network environment data and fault data of the new partition; extracting features from the data of the new partition to generate new feature data samples; combining the feature data samples of the new partition with the fault identification information to construct a new training data set; using the data set of the new partition to train the migrated model, and adjusting the model parameters to adapt it to the fault identification requirements of the new partition. Through the above steps, the first fault self-diagnosis model is migrated to other partitions, and corresponding K fault self-diagnosis models are generated.

[0060] Furthermore, in the method provided in the embodiment of the application, determining the first partition from the K partitions further includes:

[0061] Obtain K characteristic network environment indicators corresponding to the K partitions; identify the similarity of the network environment between each partition and other partitions based on the K characteristic network environment indicators, and obtain K mean similarities corresponding to the K partitions; and determine the first partition in the middle position based on the size of the K mean similarities.

[0062] In an embodiment of the present application, K characteristic network environment indicators corresponding to K partitions are obtained through tools such as SNMP. Next, the network environment similarity between each partition and other partitions is calculated through Euclidean distance or cosine similarity. For each partition i, the average similarity between it and all other partitions is calculated. Specifically, the Euclidean distance between partition i and all other partitions is calculated, and the average of these distances is taken to obtain the mean similarity of partition i. The mean similarities of all partitions are sorted, and the partition in the middle of the sort is found as the first partition in the in-order position.

[0063] Furthermore, in the method provided in the embodiment of the application, determining the first partition from the K partitions further includes:

[0064] Migrate the fault feature extraction layer, the fault diagnosis output layer, and the model parameters of the first fault self-diagnosis model from the first fault self-diagnosis model; obtain K-1 groups of fault operation data samples corresponding to K-1 partitions; perform transfer learning on the fault feature extraction layer, the fault diagnosis output layer and the model parameters with the K-1 groups of fault operation data samples, and output K-1 fault self-diagnosis models; output K fault self-diagnosis models with the first fault self-diagnosis model and the K-1 fault self-diagnosis models.

[0065] In an embodiment of the present application, in order to migrate the fault self-diagnosis model of the first partition to other partitions and generate corresponding K fault self-diagnosis models, a transfer learning method is adopted. First, the key layers and model parameters are extracted from the fault self-diagnosis model of the first partition through deep learning technology. These key layers include convolutional layers, pooling layers, fully connected layers and output layers. The convolutional layers and pooling layers have been trained on the first partition data and can effectively extract fault-related features. The fully connected layers and output layers can map the extracted features to specific fault classification results.

[0066] Next, we collect the fault operation data samples of K-1 partitions through network management tools, such as SNMP and historical fault data collection tools. When collecting network environment data, we use SNMP and SSH protocols to obtain indicators such as network delay, bandwidth utilization, packet loss rate, and signal strength for each partition. The fault data is combined with historical fault logs through network monitoring systems such as Nagios and Zabbix to mark the fault type or normal state corresponding to each data sample.

[0067] Use transfer learning technology to fine-tune the model of the first partition to adapt to the data of other partitions. First, copy the feature extraction layer and the diagnostic output layer and their parameters from the first fault self-diagnosis model to initialize the new migration model. Divide the data of K-1 partitions into training sets and validation sets, use these data to extract features, and generate feature data samples. Then use the training data of K-1 partitions to fine-tune the migration model. During the fine-tuning process, use the new data samples to fine-tune the convolution layer and the pooling layer to adapt to the new data features; use the new data samples to fine-tune the fully connected layer and the output layer to ensure that the model can accurately classify the new fault type; according to the new data features and fault types, adjust the model's hyperparameters to optimize the model performance. In the process of fine-tuning the model, use cross-validation technology to evaluate the performance of the model to prevent overfitting. Evaluate the model performance on the validation set, and the main evaluation indicators include accuracy, recall rate, and F1 value. According to the evaluation results, further optimize the model parameters to ensure the consistency and reliability of the model on the new partition data. Through the above transfer learning process, a corresponding fault self-diagnosis model is generated for each partition, and finally K-1 models corresponding to K-1 partitions are output.

[0068] Finally, the first fault self-diagnosis model and the K-1 fault self-diagnosis models after transfer learning and fine-tuning are integrated together to form a set of K complete fault self-diagnosis models.

[0069] In the embodiments of the present application, in summary, the embodiments of the present application have at least the following technical effects:

[0070] The present application obtains multiple devices connected to the concentrator and establishes a device topology network; identifies the basic information of each device in the multiple devices, including physical location, network configuration, connection type, IP address and subnet mask; evaluates the network environment of the multiple devices according to the basic information, and outputs multiple network environment indicators, wherein each device corresponds to a network environment indicator; partitions the device topology network according to the multiple network environment indicators, and outputs K partitions, wherein the network environment similarity of each partition is greater than the preset similarity, and K is a positive integer greater than or equal to 2, representing the number of partitions; trains K fault self-diagnosis models corresponding to the K partitions; inputs the K groups of real-time operation data sets corresponding to the K partitions into the K fault self-diagnosis models, and obtains K fault identification results; generates a fault warning signal according to the K fault identification results, and sends the fault warning signal to the host computer for reminder. The present invention solves the technical problems of low efficiency and slow response of the fault self-diagnosis of the concentrator in the prior art, and comprehensively evaluates and diagnoses the network environment of the device by real-time monitoring of the key parameters of the multiple devices connected to the concentrator, and combines the partition management of the device topology network with the customized fault self-diagnosis model to achieve the technical effect of improving the accuracy of fault identification and the response speed.

[0071] Embodiment 2

[0072] Based on the same inventive concept as the fault self-diagnosis method for the concentrator in the aforementioned embodiment, Figure 2 As shown, the present application provides a fault self-diagnosis device for a concentrator, and the device and method embodiments in the present application are based on the same inventive concept. The device includes:

[0073] A topology network establishing module 11, wherein the topology network establishing module 11 acquires a plurality of devices connected to the concentrator and establishes a device topology network;

[0074] A device information identification module 12, wherein the device information identification module 12 identifies basic information of each device in the plurality of devices, including physical location, network configuration, connection type, IP address, and subnet mask;

[0075] A network environment evaluation module 13, wherein the network environment evaluation module 13 evaluates the network environment of the plurality of devices according to the basic information and outputs a plurality of network environment indicators, wherein each device corresponds to a network environment indicator;

[0076] A partitioning module 14, wherein the partitioning module 14 partitions the device topology network according to the multiple network environment indicators, and outputs K partitions, wherein the network environment similarity of each partition is greater than a preset similarity, and K is a positive integer greater than or equal to 2, representing the number of partitions;

[0077] A fault self-diagnosis model training module 15, wherein the fault self-diagnosis model training module 15 trains K fault self-diagnosis models corresponding to the K partitions;

[0078] A fault identification result acquisition module 16, wherein the fault identification result acquisition module 16 inputs the K groups of real-time operation data sets corresponding to the K partitions into the K fault self-diagnosis models to obtain K fault identification results;

[0079] The fault warning signal generating module 17 generates a fault warning signal according to the K fault identification results, and sends the fault warning signal to the host computer for reminder.

[0080] Furthermore, the device is also used to achieve the following functions:

[0081] Judging K fault indicators in the K fault identification results;

[0082] Obtaining the number of partitions greater than a preset fault threshold, and generating a concentrator-fault warning signal if the number of partitions is greater than or equal to the preset number;

[0083] After receiving the concentrator-fault warning signal, the host computer activates the backup concentrator and updates the device topology network with the backup concentrator.

[0084] Furthermore, the device is also used to achieve the following functions:

[0085] If the number of partitions is less than the preset number, generating a partition-fault warning signal;

[0086] After receiving the partition-fault warning signal, the host computer locates the fault source of the abnormal partition.

[0087] Furthermore, the device is also used to achieve the following functions:

[0088] Connecting to the data management center where the concentrator is located, and outputting the concentrator distribution network;

[0089] According to the concentrator distribution network, obtaining a neighboring candidate concentrator of the concentrator;

[0090] Identify the supported protocol types, centralized processing capabilities, and configuration attribute information of the concentrator;

[0091] According to the supported protocol type, centralized processing performance and configuration attribute information of the concentrator, compatibility matching is performed on the neighboring candidate concentrators, and a backup concentrator is output.

[0092] Furthermore, the device is also used to achieve the following functions:

[0093] Obtaining a first network environment indicator corresponding to the first partition;

[0094] Collecting a first group of fault operation data samples of each device in the first partition under the first network environment indicator;

[0095] Performing feature convolution on the first group of fault operation data samples to obtain feature operation data samples related to the concentrator, including forwarding-feature operation data samples, switching-feature operation data samples, and integration-feature operation data samples in which the concentrator participates;

[0096] Training a first fault self-diagnosis model corresponding to the first partition according to the forwarding-feature operation data sample, the switching-feature operation data sample, the integration-feature operation data sample and the identification information of the identification fault indicator;

[0097] Transfer learning is performed using the first fault self-diagnosis model as a source model, and K fault self-diagnosis models corresponding to the K partitions are output.

[0098] Furthermore, the device is also used to achieve the following functions:

[0099] Obtain K characteristic network environment indicators corresponding to the K partitions;

[0100] According to the K characteristic network environment indicators, identify the similarity of the network environment between each partition and other partitions, and obtain K mean similarities corresponding to the K partitions;

[0101] According to the order of the K mean similarities, the first partition in the middle position is determined.

[0102] Furthermore, the device is also used to achieve the following functions:

[0103] Migrating a fault feature extraction layer, a fault diagnosis output layer, and model parameters of the first fault self-diagnosis model from the first fault self-diagnosis model;

[0104] Obtain K-1 groups of fault operation data samples corresponding to K-1 partitions;

[0105] Using the K-1 groups of fault operation data samples, transfer learning is performed on the fault feature extraction layer, the fault diagnosis output layer and the model parameters, and K-1 fault self-diagnosis models are output;

[0106] K fault self-diagnosis models are output using the first fault self-diagnosis model and the K-1 fault self-diagnosis models.

[0107] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0108] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0109] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A fault self-diagnosis method for a concentrator, characterized in that: The method comprises: Acquire multiple devices connected to the concentrator and establish a device topology network; identifying basic information of each of the plurality of devices, including physical location, network configuration, connection type, IP address, and subnet mask; Evaluate the network environment of the multiple devices according to the basic information, and output multiple network environment indicators, wherein each device corresponds to a network environment indicator; Partition the device topology network according to the multiple network environment indicators, and output K partitions, wherein the network environment similarity of each partition is greater than a preset similarity, and K is a positive integer greater than or equal to 2, representing the number of partitions; Training K fault self-diagnosis models corresponding to the K partitions, including: collecting operation data and fault data from devices in each partition through a network management tool; preprocessing the collected operation data and fault data; dividing the preprocessed operation data and fault data into a training set and a test set; training a machine learning or deep learning model using the training set data, and evaluating the performance of the model using the test set data, to obtain K fault self-diagnosis models corresponding to the K partitions; Inputting K groups of real-time operation data sets corresponding to the K partitions into the K fault self-diagnosis models to obtain K fault identification results; A fault warning signal is generated according to the K fault identification results, and the fault warning signal is sent to a host computer for reminder.

2. The fault self-diagnosis method for a concentrator according to claim 1, characterized in that: Generating a fault warning signal according to the K fault identification results, the method further includes: Judging K fault indicators in the K fault identification results; Obtaining the number of partitions greater than a preset fault threshold, and generating a concentrator-fault warning signal if the number of partitions is greater than or equal to the preset number; After receiving the concentrator-fault warning signal, the host computer activates the backup concentrator and updates the device topology network with the backup concentrator.

3. The fault self-diagnosis method for a concentrator according to claim 2, characterized in that: After obtaining the number of partitions greater than a preset failure threshold, the method further includes: If the number of partitions is less than the preset number, generating a partition-fault warning signal; After receiving the partition-fault warning signal, the host computer locates the fault source of the abnormal partition.

4. The fault self-diagnosis method for a concentrator according to claim 2, characterized in that: After the host computer receives the concentrator-failure warning signal, the backup concentrator is activated, and the method includes: Connecting to the data management center where the concentrator is located, and outputting the concentrator distribution network; According to the concentrator distribution network, obtaining a neighboring candidate concentrator of the concentrator; Identify the supported protocol types, centralized processing capabilities, and configuration attribute information of the concentrator; According to the supported protocol type, centralized processing performance and configuration attribute information of the concentrator, compatibility matching is performed on the neighboring candidate concentrators, and a backup concentrator is output.

5. The fault self-diagnosis method for a concentrator according to claim 1, characterized in that: Training K fault self-diagnosis models corresponding to the K partitions, the method comprising: Obtaining a first network environment indicator corresponding to the first partition; Collecting a first group of fault operation data samples of each device in the first partition under the first network environment indicator; Performing feature convolution on the first group of fault operation data samples to obtain feature operation data samples related to the concentrator, including forwarding-feature operation data samples, switching-feature operation data samples, and integration-feature operation data samples in which the concentrator participates; Training a first fault self-diagnosis model corresponding to the first partition according to the forwarding-feature operation data sample, the switching-feature operation data sample, the integration-feature operation data sample and the identification information identifying the fault indicator; Transfer learning is performed using the first fault self-diagnosis model as a source model, and K fault self-diagnosis models corresponding to the K partitions are output.

6. The fault self-diagnosis method for a concentrator according to claim 5, characterized in that: The method for determining the first partition from the K partitions includes: Obtain K characteristic network environment indicators corresponding to the K partitions; According to the K characteristic network environment indicators, identify the similarity of the network environment between each partition and other partitions, and obtain K mean similarities corresponding to the K partitions; According to the order of the K mean similarities, the first partition in the middle position is determined.

7. The fault self-diagnosis method for a concentrator according to claim 5, characterized in that: The first fault self-diagnosis model is used as a source model for transfer learning, and the method includes: Migrating a fault feature extraction layer, a fault diagnosis output layer, and model parameters of the first fault self-diagnosis model from the first fault self-diagnosis model; Obtain K-1 groups of fault operation data samples corresponding to K-1 partitions; Using the K-1 groups of fault operation data samples, transfer learning is performed on the fault feature extraction layer, the fault diagnosis output layer and the model parameters, and K-1 fault self-diagnosis models are output; K fault self-diagnosis models are output using the first fault self-diagnosis model and the K-1 fault self-diagnosis models.

8. A fault self-diagnosis device for a concentrator, characterized in that: The device comprises: A topology network establishment module, wherein the topology network establishment module acquires a plurality of devices connected to the concentrator and establishes a device topology network; a device information identification module, the device information identification module identifying basic information of each of the plurality of devices, including physical location, network configuration, connection type, IP address, and subnet mask; A network environment evaluation module, wherein the network environment evaluation module evaluates the network environment of the plurality of devices according to the basic information and outputs a plurality of network environment indicators, wherein each device corresponds to a network environment indicator; A partitioning module, wherein the partitioning module partitions the device topology network according to the multiple network environment indicators, and outputs K partitions, wherein the network environment similarity of each partition is greater than a preset similarity, and K is a positive integer greater than or equal to 2, representing the number of partitions; A fault self-diagnosis model training module, wherein the fault self-diagnosis model training module trains the K fault self-diagnosis models corresponding to the K partitions, including: collecting operation data and fault data from the equipment in each partition through a network management tool; preprocessing the collected operation data and fault data; dividing the preprocessed operation data and fault data into a training set and a test set; using the training set data to train a machine learning or deep learning model, and using the test set data to evaluate the performance of the model, to obtain K fault self-diagnosis models corresponding to the K partitions; A fault identification result acquisition module, wherein the fault identification result acquisition module inputs K groups of real-time operation data sets corresponding to the K partitions into the K fault self-diagnosis models to obtain K fault identification results; A fault warning signal generation module generates a fault warning signal according to the K fault identification results, and sends the fault warning signal to a host computer for reminder.