A method and system for analyzing operation and maintenance asset failure based on AI big model

Abstract

The present invention provides an operation and maintenance asset fault analysis method based on an AI big model and a system thereof, wherein the method comprises: obtaining multimodal status information of the operation and maintenance assets; performing fault analysis based on the multimodal status information based on the pre-trained AI big model; and outputting the fault analysis result. The operation and maintenance asset fault analysis method based on an AI big model and a system thereof of the present invention obtain multimodal status information of the operation and maintenance assets, perform fault analysis based on the multimodal status information based on the pre-trained AI big model, and output the fault analysis result, eliminating the need for staff to manually perform fault analysis on the operation and maintenance assets, thereby reducing labor costs and improving the timeliness, comprehensiveness, and efficiency of fault analysis.

Description

Technical Field

[0001] The present invention relates to the field of computer data processing technology, and in particular to an operation and maintenance asset failure analysis method and system based on an AI big model. Background Art

[0002] Operation and maintenance assets refer to the various resources and facilities owned and managed by an organization or enterprise in the process of operating and maintaining its business, such as hardware equipment including servers and network equipment, and software systems including operating systems and database management systems. However, operation and maintenance assets may fail during their operation. In order to ensure the stability of operation and maintenance assets, continuous and comprehensive fault analysis is required. However, when performing fault analysis on operation and maintenance assets, it is mostly done manually by multiple staff members, which has a high labor cost. In addition, the timeliness, comprehensiveness and efficiency of manual fault analysis by staff members are insufficient.

[0003] Therefore, a solution is urgently needed. Summary of the invention

[0004] One of the purposes of the present invention is to provide an operation and maintenance asset fault analysis method based on an AI big model, to obtain multimodal status information of the operation and maintenance assets, to perform fault analysis based on the multimodal status information based on a pre-trained AI big model, and to output the fault analysis results. This eliminates the need for staff to manually perform fault analysis on the operation and maintenance assets, thereby reducing labor costs and improving the timeliness, comprehensiveness and efficiency of fault analysis.

[0005] An embodiment of the present invention provides an operation and maintenance asset fault analysis method based on an AI big model, comprising:

[0006] Obtain multimodal status information of operation and maintenance assets;

[0007] Based on the pre-trained AI big model, fault analysis is performed according to multi-modal status information;

[0008] Output fault analysis results.

[0009] Preferably, the pre-training steps of the AI ​​large model include:

[0010] Obtain attribute information of operation and maintenance assets;

[0011] Generate sample evaluation template based on attribute information;

[0012] Obtain multiple training samples;

[0013] Based on the sample evaluation template, each training sample is evaluated in turn to obtain the evaluation value of each training sample;

[0014] Train the AI ​​big model based on training samples whose evaluation values ​​are greater than or equal to the evaluation threshold.

[0015] Preferably, generating a sample evaluation template based on the attribute information includes:

[0016] Performing characterization processing on the attribute information to obtain an attribute feature set;

[0017] Determining generation instruction knowledge corresponding to the attribute feature set from the generation instruction map; the generation instruction knowledge includes: multiple groups of one-to-one corresponding formulation basis mining rules and template formulation rules;

[0018] Traverse each formulation basis mining rule in turn, and mine the formulation basis from the attribute information based on the traversed formulation basis mining rule each time;

[0019] Based on the traversed formulation basis, the template formulation rule corresponding to the mining rule is formulated, and the sub-template is generated according to the formulation basis;

[0020] After traversing each basis mining rule, the sub-template generated by each traversal to the basis mining rule is integrated to obtain a sample evaluation template.

[0021] Preferably, the outputting of the fault analysis result includes:

[0022] Perform volume analysis on the fault analysis results to obtain the volume value;

[0023] When the volume value is less than or equal to the volume threshold, the fault analysis result is displayed to the user;

[0024] When the volume value is greater than the volume threshold, a visualization model is generated based on the visualization template and the fault analysis results;

[0025] Display the visual model to the user.

[0026] Preferably, after displaying the visualization model to the user, the method further includes:

[0027] Continuously update the visualization model;

[0028] If the user switches the model viewing perspective N times in succession when viewing the visualization model, and the model viewing perspectives switched each time meet the standard perspective relationship, and the sum of the relationship weights of the standard perspective relationship that meets the standard perspective relationship is greater than or equal to the target threshold, the visual content of the model viewing perspective switched each time is characterized to obtain a content feature set; the target threshold is the product of N and the standard coefficient;

[0029] Determine the preconfigured knowledge corresponding to the content feature set from the preconfigured knowledge base;

[0030] Based on the pre-configured knowledge, the model display scheme of the visualization model is pre-configured in the future preset first time;

[0031] Based on the preconfigured model display scheme, the visualization model is controlled to display the corresponding model within the preset time in the future.

[0032] Preferably, after displaying the visualization model to the user, the method further includes:

[0033] When the user inputs an autonomous analysis instruction while viewing the visualization model, obtaining operation behavior information of the user within a preset second time before and after the user inputs the autonomous analysis instruction;

[0034] Serialize the operation behavior information to obtain a behavior type sequence;

[0035] Matching the behavior type sequence with multiple standard behavior type sequences respectively to obtain matching degrees;

[0036] Based on the operation behavior extraction rules corresponding to the standard behavior type sequence with the maximum matching degree, the operation behavior is extracted from the operation behavior information;

[0037] Match the operation behavior with multiple standard operation behaviors, and when the match is met, obtain the knowledge triggering conditions and analysis auxiliary strategies corresponding to the matched standard operation behaviors;

[0038] Determine whether other operation behaviors other than the operation behavior meet the knowledge triggering conditions;

[0039] When applicable, users are assisted in conducting independent analysis based on analysis assistance strategies.

[0040] An operation and maintenance asset fault analysis system based on an AI big model provided by an embodiment of the present invention includes:

[0041] The acquisition module is used to obtain multimodal status information of operation and maintenance assets;

[0042] The analysis module is used to perform fault analysis based on pre-trained AI large models and multi-modal status information;

[0043] Output module, used to output fault analysis results.

[0044] Preferably, the pre-training steps of the AI ​​large model include:

[0045] Obtain attribute information of operation and maintenance assets;

[0046] Generate sample evaluation template based on attribute information;

[0047] Obtain multiple training samples;

[0048] Based on the sample evaluation template, each training sample is evaluated in turn to obtain the evaluation value of each training sample;

[0049] Train the AI ​​big model based on training samples whose evaluation values ​​are greater than or equal to the evaluation threshold.

[0050] Preferably, generating a sample evaluation template based on the attribute information includes:

[0051] Performing characterization processing on the attribute information to obtain an attribute feature set;

[0052] Determining generation instruction knowledge corresponding to the attribute feature set from the generation instruction map; the generation instruction knowledge includes: multiple groups of one-to-one corresponding formulation basis mining rules and template formulation rules;

[0053] Traverse each formulation basis mining rule in turn, and mine the formulation basis from the attribute information based on the traversed formulation basis mining rule each time;

[0054] Based on the traversed formulation basis, the template formulation rule corresponding to the mining rule is formulated, and the sub-template is generated according to the formulation basis;

[0055] After traversing each basis mining rule, the sub-template generated by each traversal to the basis mining rule is integrated to obtain a sample evaluation template.

[0056] Preferably, the output module outputs the fault analysis result, including:

[0057] Perform volume analysis on the fault analysis results to obtain the volume value;

[0058] When the volume value is less than or equal to the volume threshold, the fault analysis result is displayed to the user;

[0059] When the volume value is greater than the volume threshold, a visualization model is generated based on the visualization template and the fault analysis results;

[0060] Display the visual model to the user.

[0061] Preferably, after displaying the visualization model to the user, the method further includes:

[0062] Continuously update the visualization model;

[0063] If the user switches the model viewing perspective N times in succession when viewing the visualization model, and the model viewing perspectives switched each time meet the standard perspective relationship, and the sum of the relationship weights of the standard perspective relationship that meets the standard perspective relationship is greater than or equal to the target threshold, the visual content of the model viewing perspective switched each time is characterized to obtain a content feature set; the target threshold is the product of N and the standard coefficient;

[0064] Determine the preconfigured knowledge corresponding to the content feature set from the preconfigured knowledge base;

[0065] Based on the pre-configured knowledge, the model display scheme of the visualization model is pre-configured in the future preset first time;

[0066] Based on the preconfigured model display scheme, the visualization model is controlled to display the corresponding model within the preset time in the future.

[0067] Preferably, after displaying the visualization model to the user, the method further includes:

[0068] When the user inputs an autonomous analysis instruction while viewing the visualization model, obtaining operation behavior information of the user within a preset second time before and after the user inputs the autonomous analysis instruction;

[0069] Serialize the operation behavior information to obtain a behavior type sequence;

[0070] Matching the behavior type sequence with multiple standard behavior type sequences respectively to obtain matching degrees;

[0071] Based on the operation behavior extraction rules corresponding to the standard behavior type sequence with the maximum matching degree, the operation behavior is extracted from the operation behavior information;

[0072] Match the operation behavior with multiple standard operation behaviors, and when the match is met, obtain the knowledge triggering conditions and analysis auxiliary strategies corresponding to the matched standard operation behaviors;

[0073] Determine whether other operation behaviors other than the operation behavior meet the knowledge triggering conditions;

[0074] When applicable, users are assisted in conducting independent analysis based on analysis assistance strategies.

[0075] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings.

[0076] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0078] Figure 1Schematic diagram of a method for analyzing operation and maintenance asset failures based on an AI big model in an embodiment of the present invention;

[0079] Figure 2 This is a schematic diagram of an operation and maintenance asset fault analysis system based on an AI big model in an embodiment of the present invention. DETAILED DESCRIPTION

[0080] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0081] The embodiment of the present invention provides an operation and maintenance asset fault analysis method based on an AI big model, such as Figure 1 As shown, including:

[0082] S1. Obtain multimodal status information of operation and maintenance assets;

[0083] S2: Based on the pre-trained AI big model, fault analysis is performed according to multi-modal status information;

[0084] S3. Output fault analysis results.

[0085] The multimodal status information of operation and maintenance assets includes: the current working status, work plan and historical working parameters of the operation and maintenance assets; the pre-trained AI large model is used to perform fault analysis on the operation and maintenance assets based on the multimodal status information, and output the fault analysis results after the fault analysis is completed.

[0086] This application obtains multimodal status information of operation and maintenance assets, performs fault analysis based on the pre-trained AI large model according to the multimodal status information, and outputs the fault analysis results. There is no need for staff to manually perform fault analysis on operation and maintenance assets, which reduces labor costs and improves the timeliness, comprehensiveness and efficiency of fault analysis.

[0087] In one embodiment, the pre-training step of the AI ​​big model includes:

[0088] Obtain the attribute information of the operation and maintenance assets; the attribute information includes: the model, manufacturer, working hours, historical fault types, etc. of the operation and maintenance assets;

[0089] Based on the attribute information, a sample evaluation template is generated; the sample evaluation template is used to evaluate the degree to which the training sample trains the AI ​​model so that the AI ​​big model is suitable for fault analysis of operation and maintenance assets, that is, the evaluation value;

[0090] Obtain multiple training samples; the training samples may be the fault types, fault causes, and fault analysis processes generated by different operation and maintenance assets in history;

[0091] Based on the sample evaluation template, each training sample is evaluated in turn to obtain the evaluation value of each training sample; the larger the evaluation value, the higher the above applicability;

[0092] Based on the training samples whose evaluation values ​​are greater than or equal to the evaluation threshold, the AI ​​big model is trained; the evaluation threshold may be 8. When the evaluation value is greater than or equal to the evaluation threshold, it means that the above applicability is high, and the corresponding training samples are used to train the AI ​​big model;

[0093] The generating of a sample evaluation template based on the attribute information includes:

[0094] Performing feature processing on the attribute information to obtain an attribute feature set; the attribute features in the attribute feature set include: operation and maintenance asset model, operation and maintenance asset manufacturer, etc.;

[0095] Determine the generation indication knowledge corresponding to the attribute feature set from the generation indication map; the generation indication knowledge includes: multiple groups of one-to-one corresponding formulation basis mining rules and template formulation rules; an attribute feature set represents a situation where a sample evaluation template needs to be generated, and the generation indication map contains the generation indication knowledge corresponding to the situation, and the generation indication knowledge contains formulation basis mining rules and template formulation rules, the formulation basis mining rules are rules for mining formulation basis from attribute information in this situation, and the template formulation rules are corresponding rules for generating sub-templates according to the formulation basis in this situation, for example: the attribute features in the attribute feature set are operation and maintenance asset model A and operation and maintenance asset manufacturer B, then the formulation basis mining rules are to mine historical fault types C, D, etc. of different operation and maintenance assets with the same operation and maintenance asset model A from the attribute information, and the template formulation rules are to formulate the sub-template to evaluate whether the operation and maintenance asset with operation and maintenance asset model A in the training sample has experienced asset faults of fault type C or D, and if so, the output evaluation value is 2;

[0096] Traverse each formulation basis mining rule in turn, and mine the formulation basis from the attribute information based on the traversed formulation basis mining rule each time;

[0097] Based on the traversed formulation basis, the template formulation rule corresponding to the mining rule is generated, and the sub-template is generated according to the formulation basis; each time the traversal is reached, the above-mentioned mining and formulation principles are used to mine the formulation basis and generate the sub-template;

[0098] After traversing each formulation basis mining rule, the sub-templates generated by each traversal to the formulation basis mining rule are integrated to obtain a sample evaluation template. After traversing each formulation basis mining rule, the sub-templates are integrated to obtain a sample evaluation template.

[0099] When the AI ​​big model is pre-trained, the quality and applicability of the training samples directly determine the working ability of the AI ​​big model for fault analysis. Therefore, after obtaining the training samples, the embodiment of the present invention introduces a sample evaluation template, evaluates the training samples, and selects the training samples for training the AI ​​big model based on the evaluation values, which greatly improves the working ability of the pre-trained AI big model for fault analysis. In addition, when generating the sample evaluation template, the attribute feature set and the generation indication map are introduced to quickly determine the formulation basis mining rules and template formulation rules, which greatly improves the generation efficiency of the sample evaluation template and improves its accuracy for evaluation.

[0100] In one embodiment, the outputting of the fault analysis result includes:

[0101] Perform volumetric analysis on the fault analysis results to obtain a volume value; the volume value represents the volume of information in the fault analysis results. The larger the volume value, the larger the volume.

[0102] When the volume value is less than or equal to the volume threshold, the fault analysis result is displayed to the user; the volume threshold can be 5; when the volume value is less than or equal to the volume threshold, it means that the volume is small, and the fault analysis result is directly displayed to the user;

[0103] When the volume value is greater than the volume threshold, a visualization model is generated based on the visualization template and the fault analysis results; when the volume value is greater than the volume threshold, it means that the volume is large, and the fault analysis results are converted to generate a visualization model; the visualization template is a template for generating a visualization model based on the fault analysis results, which can be set by the technicians according to actual needs;

[0104] Display the visual model to the user. Display the visual model to the user, so that the user can understand the fault analysis results intuitively and conveniently.

[0105] In one embodiment, after displaying the visualization model to the user, the method further includes:

[0106] Continuously update the visualization model; when updating the visualization model, the fault analysis results obtained from the latest fault analysis can be added to the visualization model, which is convenient for users to compare and view from a historical perspective;

[0107] If the user switches the model viewing perspective N times in succession when viewing the visualization model, and the model viewing perspectives switched each time meet the standard perspective relationship, and the sum of the relationship weights of the standard perspective relationships that meet the requirements is greater than or equal to the target threshold, the visual content of the model viewing perspective switched each time is characterized to obtain a content feature set; the target threshold is the product of N and the standard coefficient; the visualization model is a three-dimensional model, and a model viewing perspective is generated when the user views the visualization model; when the user switches the model viewing perspective N times in succession when viewing the visualization model, and the model viewing perspectives switched each time meet the standard perspective relationship, and the sum of the relationship weights of the standard perspective relationships that meet the requirements is greater than or equal to the target threshold, it means that the visualization model needs to be characterized. The display scheme of the future model is pre-configured; the overlap between model viewing perspectives is greater than or equal to 65%, etc.; the standard viewing perspective relationship is preset with a corresponding relationship weight, and the larger the relationship weight, the greater the degree of pre-configuration is required if the model viewing perspective meets the corresponding standard viewing perspective relationship; the technician can pre-set the relationship weight for different standard viewing perspective relationships; the target threshold can be 20; when the user switches N times in a row and the overlap between model viewing perspectives is greater than or equal to 65%, it means that the user is focusing on the content in the visualization model, and the visual content of the model viewing perspective switched each time is characterized to obtain a content feature set; the content features in the content feature set can be content type, content volume, etc.

[0108] Determine the preconfigured knowledge corresponding to the content feature set from the preconfigured knowledge base; the content feature set represents a situation that needs to be preconfigured, and the preconfigured knowledge base has the preconfigured knowledge corresponding to the situation; for example, if the content type in the content feature set is the server fault type and fault cause, then the preconfigured knowledge is to give priority to displaying other server faults in the future display content, etc.;

[0109] Based on the pre-configured knowledge, the model display scheme of the visualization model is pre-configured within the first preset time in the future; the first time may be 5 minutes;

[0110] Based on the preconfigured model display scheme, the visualization model is controlled to display the corresponding model within the preset time in the future.

[0111] The embodiment of the present invention pre-configures the model display scheme of the visualization model within the first preset time in the future, ensuring that during the continuous updating of the visualization model, the user can continuously view the content that is valuable for their fault analysis, thereby improving the humanization and the user's work efficiency. At the same time, it is also more intelligent and improves the applicability of the system.

[0112] In one embodiment, after displaying the visualization model to the user, the method further includes:

[0113] When the user enters an autonomous analysis instruction while viewing the visualization model, the operation behavior information within a preset second time before and after the user enters the autonomous analysis instruction is obtained; when the user enters the autonomous analysis instruction, it indicates that the user wants to perform a secondary autonomous analysis based on the visualization model; the second time can be 2 minutes; the operation behavior information includes the operation type, operation time, operation position, etc. of the user's operation on the visualization model;

[0114] Serializing the operation behavior information to obtain a behavior type sequence; extracting the generated operation behaviors and the corresponding generation sequence in the operation behavior information, and chronologically sorting the behavior types of the operation behaviors according to the generation sequence to obtain a behavior type sequence;

[0115] Matching the behavior type sequence with multiple standard behavior type sequences respectively to obtain matching degrees;

[0116] Based on the operation behavior extraction rules corresponding to the standard behavior type sequence with the maximum matching degree, the operation behavior is extracted from the operation behavior information;

[0117] The operation behavior is matched with multiple standard operation behaviors, and when the match is met, the knowledge triggering condition and analysis auxiliary strategy corresponding to the matched standard operation behavior are obtained; a standard behavior type sequence, its corresponding operation behavior extraction rules and its corresponding knowledge triggering condition and analysis auxiliary strategy are preset, for example: the behavior type in the standard behavior type sequence is to view different asset devices of the same fault type, then the corresponding operation behavior extraction sequence is to extract the fault severity of different asset devices of the same fault type viewed by the user, and the knowledge triggering condition is to browse the corresponding asset devices in order from the largest to the smallest fault severity, then the analysis auxiliary strategy is to sort the different asset devices of the same fault type viewed by the user in order from the largest to the smallest fault severity;

[0118] Determine whether other operation behaviors other than the operation behavior meet the knowledge triggering conditions;

[0119] When applicable, users are assisted in conducting independent analysis based on analysis assistance strategies.

[0120] Generally, when viewing a visualization model, a user may want to conduct an independent analysis. In the embodiment of the present invention, when a user wants to conduct an independent analysis, a standard behavior type sequence, its corresponding operation behavior extraction rules, and its corresponding knowledge triggering conditions and analysis auxiliary strategies are introduced to quickly help the user conduct an independent analysis, thereby further improving the humanization, improving the user's work efficiency, and at the same time, being more intelligent, thereby improving the applicability of the system.

[0121] The embodiment of the present invention provides an operation and maintenance asset fault analysis system based on an AI big model, such as Figure 2As shown, including:

[0122] Acquisition module 1, used to acquire multimodal status information of operation and maintenance assets;

[0123] Analysis module 2 is used to perform fault analysis based on the pre-trained AI large model and multi-modal state information;

[0124] The output module 3 is used to output the fault analysis result.

[0125] The pre-training steps of the AI ​​large model include:

[0126] Obtain attribute information of operation and maintenance assets;

[0127] Generate sample evaluation template based on attribute information;

[0128] Obtain multiple training samples;

[0129] Based on the sample evaluation template, each training sample is evaluated in turn to obtain the evaluation value of each training sample;

[0130] Train the AI ​​big model based on training samples whose evaluation values ​​are greater than or equal to the evaluation threshold.

[0131] The generating of a sample evaluation template based on the attribute information includes:

[0132] Performing characterization processing on the attribute information to obtain an attribute feature set;

[0133] Determining generation instruction knowledge corresponding to the attribute feature set from the generation instruction map; the generation instruction knowledge includes: multiple groups of one-to-one corresponding formulation basis mining rules and template formulation rules;

[0134] Traverse each formulation basis mining rule in turn, and mine the formulation basis from the attribute information based on the traversed formulation basis mining rule each time;

[0135] Based on the traversed formulation basis, the template formulation rule corresponding to the mining rule is formulated, and the sub-template is generated according to the formulation basis;

[0136] After traversing each basis mining rule, the sub-template generated by each traversal to the basis mining rule is integrated to obtain a sample evaluation template.

[0137] The output module outputs the fault analysis result, including:

[0138] Perform volume analysis on the fault analysis results to obtain the volume value;

[0139] When the volume value is less than or equal to the volume threshold, the fault analysis result is displayed to the user;

[0140] When the volume value is greater than the volume threshold, a visualization model is generated based on the visualization template and the fault analysis results;

[0141] Display the visual model to the user.

[0142] After displaying the visual model to the user, it also includes:

[0143] Continuously update the visualization model;

[0144] If the user switches the model viewing perspective N times in succession when viewing the visualization model, and the model viewing perspectives switched each time meet the standard perspective relationship, and the sum of the relationship weights of the standard perspective relationship that meets the standard perspective relationship is greater than or equal to the target threshold, the visual content of the model viewing perspective switched each time is characterized to obtain a content feature set; the target threshold is the product of N and the standard coefficient;

[0145] Determine the preconfigured knowledge corresponding to the content feature set from the preconfigured knowledge base;

[0146] Based on the pre-configured knowledge, the model display scheme of the visualization model is pre-configured in the future preset first time;

[0147] Based on the preconfigured model display scheme, the visualization model is controlled to display the corresponding model within the preset time in the future.

[0148] After displaying the visual model to the user, it also includes:

[0149] When the user inputs an autonomous analysis instruction while viewing the visualization model, obtaining operation behavior information of the user within a preset second time before and after the user inputs the autonomous analysis instruction;

[0150] Serialize the operation behavior information to obtain a behavior type sequence;

[0151] Matching the behavior type sequence with multiple standard behavior type sequences respectively to obtain matching degrees;

[0152] Based on the operation behavior extraction rules corresponding to the standard behavior type sequence with the maximum matching degree, the operation behavior is extracted from the operation behavior information;

[0153] Match the operation behavior with multiple standard operation behaviors, and when the match is met, obtain the knowledge triggering conditions and analysis auxiliary strategies corresponding to the matched standard operation behaviors;

[0154] Determine whether other operation behaviors other than the operation behavior meet the knowledge triggering conditions;

[0155] When applicable, users are assisted in conducting independent analysis based on analysis assistance strategies.

[0156] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for analyzing operation and maintenance asset failures based on an AI big model, characterized in that: include: Obtain multimodal status information of operation and maintenance assets; Based on the pre-trained AI big model, fault analysis is performed according to multi-modal status information; Output fault analysis results; The pre-training steps of the AI ​​large model include: Obtain attribute information of operation and maintenance assets; Generate sample evaluation template based on attribute information; Obtain multiple training samples; Based on the sample evaluation template, each training sample is evaluated in turn to obtain the evaluation value of each training sample; Train the AI ​​big model based on training samples whose evaluation values ​​are greater than or equal to the evaluation threshold; The generating of a sample evaluation template based on the attribute information includes: Performing characterization processing on the attribute information to obtain an attribute feature set; Determine generation instruction knowledge corresponding to the attribute feature set from the generation instruction map; the generation instruction knowledge includes: multiple groups of one-to-one corresponding formulation basis mining rules and template formulation rules; Traverse each formulation basis mining rule in turn, and mine the formulation basis from the attribute information based on the traversed formulation basis mining rule each time; Based on the traversed formulation basis, the template formulation rule corresponding to the mining rule is formulated, and the sub-template is generated according to the formulation basis; After traversing each basis mining rule, the sub-template generated by each traversal to the basis mining rule is integrated to obtain a sample evaluation template.

2. The operation and maintenance asset failure analysis method based on the AI ​​big model according to claim 1 is characterized in that: The output fault analysis result includes: Perform volume analysis on the fault analysis results to obtain the volume value; When the volume value is less than or equal to the volume threshold, the fault analysis result is displayed to the user; When the volume value is greater than the volume threshold, a visualization model is generated based on the visualization template and the fault analysis results; Display the visual model to the user.

3. The operation and maintenance asset failure analysis method based on AI big model as claimed in claim 2 is characterized in that: After displaying the visual model to the user, it also includes: Continuously update the visualization model; If the user switches the model viewing perspective N times in succession when viewing the visualization model, and the model viewing perspectives switched each time meet the standard perspective relationship, and the sum of the relationship weights of the standard perspective relationship that meets the standard perspective relationship is greater than or equal to the target threshold, the visual content of the model viewing perspective switched each time is characterized to obtain a content feature set; the target threshold is the product of N and the standard coefficient; Determine the preconfigured knowledge corresponding to the content feature set from the preconfigured knowledge base; Based on the pre-configured knowledge, the model display scheme of the visualization model is pre-configured in the future preset first time; Based on the preconfigured model display scheme, the visualization model is controlled to display the corresponding model within the preset time in the future.

4. The method for analyzing failures of operation and maintenance assets based on an AI big model according to claim 2, characterized in that: After displaying the visual model to the user, it also includes: When the user inputs an autonomous analysis instruction while viewing the visualization model, obtaining operation behavior information of the user within a preset second time before and after the user inputs the autonomous analysis instruction; Serialize the operation behavior information to obtain a behavior type sequence; Matching the behavior type sequence with multiple standard behavior type sequences respectively to obtain matching degrees; Based on the operation behavior extraction rules corresponding to the standard behavior type sequence with the maximum matching degree, the operation behavior is extracted from the operation behavior information; Match the operation behavior with multiple standard operation behaviors, and when the match is met, obtain the knowledge triggering conditions and analysis auxiliary strategies corresponding to the matched standard operation behaviors; Determine whether other operation behaviors other than the operation behavior meet the knowledge triggering conditions; When applicable, users are assisted in conducting independent analysis based on analysis assistance strategies.

5. An operation and maintenance asset fault analysis system based on AI big model, characterized in that: include: The acquisition module is used to obtain multimodal status information of operation and maintenance assets; The analysis module is used to perform fault analysis based on pre-trained AI large models and multi-modal status information; Output module, used for outputting fault analysis results; The pre-training steps of the AI ​​large model include: Obtain attribute information of operation and maintenance assets; Generate sample evaluation template based on attribute information; Obtain multiple training samples; Based on the sample evaluation template, each training sample is evaluated in turn to obtain the evaluation value of each training sample; Train the AI ​​big model based on training samples whose evaluation values ​​are greater than or equal to the evaluation threshold; The generating of a sample evaluation template based on the attribute information includes: Performing characterization processing on the attribute information to obtain an attribute feature set; Determine generation instruction knowledge corresponding to the attribute feature set from the generation instruction map; the generation instruction knowledge includes: multiple groups of one-to-one corresponding formulation basis mining rules and template formulation rules; Traverse each formulation basis mining rule in turn, and mine the formulation basis from the attribute information based on the traversed formulation basis mining rule each time; Based on the traversed formulation basis, the template formulation rule corresponding to the mining rule is formulated, and the sub-template is generated according to the formulation basis; After traversing each basis mining rule, the sub-template generated by each traversal to the basis mining rule is integrated to obtain a sample evaluation template.

6. The operation and maintenance asset fault analysis system based on AI big model as claimed in claim 5, characterized in that: The output module is used to output the fault analysis result, including: Perform volume analysis on the fault analysis results to obtain the volume value; When the volume value is less than or equal to the volume threshold, the fault analysis result is displayed to the user; When the volume value is greater than the volume threshold, a visualization model is generated based on the visualization template and the fault analysis results; Display the visual model to the user.

Images