CN-121008858-B - Fault prediction and repair method, system, electronic equipment and storage medium

CN121008858BCN 121008858 BCN121008858 BCN 121008858BCN-121008858-B

Abstract

The application discloses a fault prediction and repair method, a system, electronic equipment and a storage medium, wherein the method comprises the steps of obtaining an image stream of a server terminal by utilizing an embedded KVM; the image stream comprises a plurality of acquired images, AI visual identification is carried out based on the acquired images to obtain the running state of the server terminal, state diagram information is constructed based on the running state, a processing strategy is determined based on state diagram information mapping, remote control operation is carried out on the server terminal by utilizing embedded KVM based on the processing strategy, feedback checksum is carried out based on the result of the remote control operation, and AI visual identification is optimized and adjusted. According to the application, based on the image-level remote access realized by the embedded KVM, the AI visual identification and the strategy mapping are further introduced to realize the remote control operation, and the automatic identification, response and repair capability of the server in the fault state can be effectively improved. The application can be widely applied to the technical field of data processing.

Inventors

Zeng Zhaonai
WU QINGHUA

Assignees

广东美凯技术有限公司

Dates

Publication Date: 20260512
Application Date: 20250708

Claims (9)

1. A method of fault prediction and repair, the method comprising the steps of: acquiring an image stream of a server terminal by using an embedded KVM, wherein the image stream comprises a plurality of acquired images; performing AI visual recognition based on the acquired image to obtain the running state of the server terminal, and constructing state diagram information based on the running state; Determining a processing strategy based on the state diagram information mapping; based on the processing strategy, utilizing the embedded KVM to remotely control the server terminal; performing feedback checksum on the basis of the remote control operation result to perform optimization adjustment on the AI visual recognition; the operation state comprises a stage state and a time sequence state, the AI visual identification is carried out based on the acquired image, and the operation state of the server terminal is obtained, and the method comprises the following steps: Based on the acquired image, performing state classification and identification by using a multi-task convolutional neural network to obtain a phase state corresponding to the acquired image; The multi-task convolutional neural network comprises a lightweight trunk, a stage classification branch and an error type identification branch, the state classification and identification are carried out by using the multi-task convolutional neural network to obtain a stage state corresponding to the acquired image, and the method comprises the following steps: the method comprises the steps of acquiring an acquisition image, wherein the acquisition image is subjected to feature extraction by utilizing the lightweight trunk to obtain multi-level image features, and the stage classification and the error type with the highest confidence are taken as the results of corresponding branches according to the confidence scores of stage classification branches on each stage classification and the confidence scores of error type identification branches on each error type; based on the multi-level image characteristics, mapping and outputting stage classification by utilizing the stage classification branches; when the stage is classified as an error stage, based on the multi-level image characteristics, identifying a branch mapping output error type by utilizing the error type; carrying out key text recognition on the acquired image to obtain text information; According to the error type, carrying out table lookup by combining the text information to determine the abnormal state of the server terminal as the stage state; and judging the time sequence state of the image sequence of the acquired images of the continuous frames in the image stream by using a time sequence convolution network to obtain the time sequence state.
2. The method according to claim 1, wherein the capturing the image stream of the server terminal using the embedded KVM comprises the steps of: acquiring an original image stream of the server terminal based on a preset frequency by using the embedded KVM through a high-speed frame buffer mechanism; performing image preprocessing on each original acquisition image in the original image stream, and finishing to obtain a target image stream; Wherein the image preprocessing includes color space conversion and image enhancement processing.
3. The method according to claim 1, wherein said building state diagram information based on said operating state comprises the steps of: collecting state data of the server terminal based on a state diagram model of a finite state machine; and integrating the running state and the state data to obtain the state diagram information.
4. The method of claim 1, wherein said determining a processing policy based on said state diagram information map comprises the steps of: outputting the processing strategy through the neural network mapping of the fusion rule based on the state diagram information; the neural network fusing the rules learns the mapping relation between different state diagram information and preset business rules in advance.
5. The method according to claim 1, wherein the remote control operation of the server terminal using the embedded KVM based on the processing policy comprises the steps of: And calling an embedded bottom power control API by utilizing the virtual USB technology of the embedded KVM based on the processing strategy so as to trigger a target control instruction to perform the remote control operation on the server terminal.
6. The method according to any one of claims 1 to 5, wherein the optimally adjusting the AI visual recognition based on the feedback checksum of the result of the remote control operation comprises the steps of: Responding to the completion signal of the remote control operation, starting a feedback checking mechanism to acquire the image of the server terminal again by using the embedded KVM and carrying out the AI visual identification to obtain the processing result of the server terminal; determining an operation result of the remote control operation based on the processing result, and controlling the server terminal to start a rollback mechanism or adjusting an identification confidence threshold of the model invoked by the AI visual identification when the operation result is operation failure; uploading the identification result corresponding to the AI visual identification, the operation log corresponding to the remote control operation and the processing result to a training data set of a cloud server; The training data set is used for carrying out iterative optimization on the model called by the AI visual recognition.
7. A fault prediction and repair system, the system comprising: the system comprises a data acquisition module, a server terminal, a data processing module and a data processing module, wherein the data acquisition module is used for acquiring an image stream of the server terminal by utilizing an embedded KVM; The visual identification module is used for carrying out AI visual identification based on the acquired image to obtain the running state of the server terminal, and constructing state diagram information based on the running state; A policy mapping module, configured to determine a processing policy based on the state diagram information mapping; the remote control module is used for performing remote control operation on the server terminal by utilizing the embedded KVM based on the processing strategy; The feedback optimization module is used for carrying out feedback checksum on the basis of the remote control operation result and carrying out optimization adjustment on the AI visual recognition; the operation state comprises a stage state and a time sequence state, the AI visual identification is carried out based on the acquired image, and the operation state of the server terminal is obtained, and the method comprises the following steps: Based on the acquired image, performing state classification and identification by using a multi-task convolutional neural network to obtain a phase state corresponding to the acquired image; The multi-task convolutional neural network comprises a lightweight trunk, a stage classification branch and an error type identification branch, the state classification and identification are carried out by using the multi-task convolutional neural network to obtain a stage state corresponding to the acquired image, and the method comprises the following steps: the method comprises the steps of acquiring an acquisition image, wherein the acquisition image is subjected to feature extraction by utilizing the lightweight trunk to obtain multi-level image features, and the stage classification and the error type with the highest confidence are taken as the results of corresponding branches according to the confidence scores of stage classification branches on each stage classification and the confidence scores of error type identification branches on each error type; based on the multi-level image characteristics, mapping and outputting stage classification by utilizing the stage classification branches; when the stage is classified as an error stage, based on the multi-level image characteristics, identifying a branch mapping output error type by utilizing the error type; carrying out key text recognition on the acquired image to obtain text information; According to the error type, carrying out table lookup by combining the text information to determine the abnormal state of the server terminal as the stage state; and judging the time sequence state of the image sequence of the acquired images of the continuous frames in the image stream by using a time sequence convolution network to obtain the time sequence state.
8. An electronic device comprising a memory storing a computer program and a processor implementing the method of any of claims 1 to 6 when the computer program is executed by the processor.
9. A computer readable storage medium storing a computer program, characterized in that the computer program, when executed by a processor, implements the method of any one of claims 1 to 6.

Description

Fault prediction and repair method, system, electronic equipment and storage medium Technical Field The present application relates to the field of data processing technologies, and in particular, to a fault prediction and repair method, a system, an electronic device, and a storage medium. Background In the traditional server maintenance process, an administrator relies on operating system layer logs, SNMP protocol alarms, or human observation of a remote desktop to identify faults. However, in the event of an operating system crash, or BIOS stay, the conventional approach fails. Although some existing communication technologies can realize remote access to a server, the existing communication technologies lack an intelligent judgment and automatic response mechanism, and still require manual operation analysis judgment, so that the existing technology has low maintenance efficiency for the server. Disclosure of Invention The embodiment of the application mainly aims to provide a fault prediction and repair method, a system, electronic equipment and a storage medium, and aims to solve at least one problem in the prior art. In order to achieve the above objective, an aspect of an embodiment of the present application provides a fault prediction and repair method, including: Acquiring an image stream of a server terminal by using the embedded KVM, wherein the image stream comprises a plurality of acquired images; performing AI visual recognition based on the acquired image to obtain the running state of the server terminal, and constructing state diagram information based on the running state; determining a processing strategy based on the state diagram information mapping; Based on the processing strategy, utilizing the embedded KVM to remotely control the server terminal; And carrying out feedback checksum on the basis of the result of the remote control operation to carry out optimization adjustment on the AI visual recognition. In some embodiments, capturing an image stream of a server terminal using an embedded KVM includes the steps of: acquiring an original image stream of a server terminal based on a preset frequency through a high-speed frame buffer mechanism by utilizing an embedded KVM; Performing image preprocessing on each original acquisition image in the original image stream, and finishing to obtain a target image stream; wherein the image preprocessing includes color space conversion and image enhancement processing. In some embodiments, the operation state includes a phase state and a time sequence state, and AI visual recognition is performed based on the acquired image, so as to obtain the operation state of the server terminal, including the following steps: Based on the acquired image, performing state classification and identification by utilizing a multitasking convolutional neural network to obtain a phase state corresponding to the acquired image; The multi-task convolutional neural network comprises a lightweight trunk, a stage classification branch and an error type identification branch, and the multi-task convolutional neural network is utilized to carry out state classification and identification to obtain a stage state corresponding to an acquired image, and the method comprises the following steps: extracting features of the acquired images by using a lightweight trunk to obtain multi-level image features; Based on the multi-level image characteristics, outputting stage classification by stage classification branch mapping; When the stage is classified as an error stage, based on multi-level image characteristics, the error type is used for identifying branch mapping output error types; Carrying out key text recognition on the acquired image to obtain text information; According to the error type and the text information, table lookup is carried out to determine the abnormal state of the server terminal as a stage state; and judging the time sequence state of the image sequence of the collected images of the continuous frames in the image stream by using the time sequence convolution network to obtain the time sequence state. In some embodiments, building state diagram information based on operating states includes the steps of: Collecting state data of a server terminal based on a state diagram model of a finite state machine; and integrating the running state and the state data to obtain state diagram information. In some embodiments, determining a processing policy based on a state diagram information map includes the steps of: Based on the state diagram information, a processing strategy is mapped and output through a neural network of the fusion rule; the neural network fusing the rules learns the mapping relation between different state diagram information and preset business rules in advance. In some embodiments, the remote control operation of the server terminal with the embedded KVM based on the processing policy includes the steps of: Based on the processing strate