Intelligent fault diagnosis method and system for welding equipment

Abstract

The invention discloses an intelligent fault diagnosis method and system for welding equipment. The method comprises the steps that a large language model analyzes according to input fault description to obtain a core retrieval dimension; based on a pre-constructed welding equipment fault knowledge graph, performing primary screening and link completion on the core retrieval dimension to obtain one or more fault link strengths and case success rates; according to the fault link strength and the case success rate, diagnosing the credibility of the link, and comparing the credibility of each link to obtain an optimal diagnosis result; wherein when a new fault case is added to the welding equipment fault knowledge graph, the new fault case is automatically updated through a large language model; according to the method, the problems of knowledge acquisition and semantic understanding are solved through the large language model, the problems of knowledge storage and link tracing are solved through the knowledge graph, full-process automatic diagnosis from fuzzy input to landing reporting is achieved, the technical pain point of a traditional scheme is solved, and the practical requirement of an industrial welding scene is met.

Description

TECHNICAL FIELD

[0001] The application relates to a welding equipment intelligent fault diagnosis method and system based on a knowledge graph and a large language model, and belongs to the technical field of welding equipment fault diagnosis. BACKGROUND

[0002] The core is to meet the high demand of industrial production for welding quality and efficiency, and to gradually form a technical system relying on multidisciplinary technology development in view of the limitations of traditional troubleshooting. The core goal is to realize accurate, efficient and early identification of welding equipment faults. The current welding equipment fault diagnosis technology is developing towards intelligence and automation, aiming to solve the pain points of traditional methods such as dependence on manual work and difficulty in knowledge reuse.

[0003] In the prior art, although there are fault diagnosis schemes based on rules or machine learning and knowledge graphs, the rule scheme lacks a unified structured storage paradigm, the machine learning scheme requires a large amount of labeled data and cannot directly reuse the analysis framework and knowledge of historical faults, the single knowledge graph content acquisition efficiency is low, the fault matching is rigid, the complex fault static link is unclear, and the scene adaptability is poor. SUMMARY

[0004] The problems to be solved by the application are that the fault diagnosis scheme in the prior art lacks a unified structured storage paradigm, cannot directly reuse the analysis framework and knowledge of historical faults, and the single knowledge graph content acquisition efficiency is low.

[0005] To solve the above technical problems, the application is implemented by using the following technical solutions.

[0006] On the one hand, the application provides a welding equipment intelligent fault diagnosis method, and the steps include:

[0007] The large language model obtains the core retrieval dimension according to the input fault description;

[0008] Based on the pre-constructed welding equipment fault knowledge graph, the core retrieval dimension is preliminarily screened and link completion is performed to obtain one or more fault link strengths and case success rates;

[0009] According to the fault link strength and the case success rate, the credibility of the link is diagnosed, the credibility of each link is compared, and the optimal diagnosis result is obtained.

[0010] The large language model solves the knowledge acquisition and semantic understanding problem, the knowledge graph solves the knowledge storage and link tracing problem, and the whole process automation diagnosis from fuzzy input to landing report is realized, which not only solves the technical pain points of traditional schemes, but also meets the practical needs of industrial welding scenes.

[0011] The step of constructing the welding equipment fault knowledge graph comprises:

[0012] The large language model collects welding fault documents, extracts welding exclusive entities, and labels entity attributes;

[0013] According to the entity attributes, the large language model combines welding field rules to establish core fault link relationships;

[0014] The entities, core fault link relationships, and entity attributes are stored in a graph database and a vector database;

[0015] Wherein, when the welding equipment fault knowledge graph adds new fault cases, the large language model is automatically updated.

[0016] Fragmented fault handling cases and experience are integrated to provide a unified structured storage paradigm, facilitating direct reuse of historical fault analysis frameworks and case experience, and shortening the fault handling cycle.

[0017] The entity attributes include equipment information, fault description, and solution;

[0018] The equipment information includes equipment model and core components;

[0019] The fault description includes fault phenomenon and fault code;

[0020] The solution includes fault cause, inspection method, solution measure, and fault case.

[0021] The core fault link relationship includes fault propagation link relationship and fault resource mapping relationship.

[0022] The calculation formula of the credibility is as follows:

[0023]

[0024] Wherein, C is the credibility, S is the total link strength, L is the link length, R is the average success rate of the case, W1 is the weight of the total link strength divided by the link length, W2 is the weight of the average success rate of the case, and W1+W2=1.

[0025] Quantitative diagnosis credibility can directly and clearly prioritize troubleshooting direction, without wasting time on low credibility conclusions, directly focusing on high credibility direction, and shortening fault repair time.

[0026] After obtaining the optimal diagnosis result, the large language model generates a diagnosis report according to the diagnosis result.

[0027] The diagnosis report includes core conclusions, cause analysis, inspection steps, solution measures, and process parameter adjustment suggestions.

[0028] The core search dimension includes a device dimension, a fault dimension and an exclusion dimension.

[0029] The fault dimension includes a fault phenomenon and a process parameter.

[0030] The vector database performs preliminary screening on the core search dimension, and sequentially performs priority search on the device, the process parameter and the fault phenomenon to obtain one or more fault entities with high similarity;

[0031] The graph database performs link completion on the core search dimension based on the fault entity subjected to preliminary screening, and outputs one or more links with high fault link strength and case success rate.

[0032] In another aspect, the present application provides a welding equipment intelligent fault diagnosis system, comprising a processor and a memory, wherein the memory stores program instructions, and when the program instructions stored in the memory are executed by the processor, the welding equipment intelligent fault diagnosis method is realized.

[0033] Compared with the prior art, the present application has the following beneficial effects:

[0034] The present application solves the knowledge acquisition and semantic understanding problem with a large language model, and solves the knowledge storage and link tracing problem with a knowledge graph, realizes the full-process automatic diagnosis from fuzzy input to landing report, solves the technical pain points of the traditional scheme, and meets the practical needs of the industrial welding scene.

[0035] The present application integrates fragmented fault handling cases and experience, provides a unified structured storage paradigm, facilitates direct reuse of historical fault analysis framework and case experience, shortens the fault handling period, quantifies the diagnosis credibility, directly and clearly determines the priority of troubleshooting direction, does not need to waste time on low credibility conclusions, directly focuses on high credibility direction, and shortens the fault repair time. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 The flowchart of the fault diagnosis method of the present application embodiment 1 is shown in the figure.

[0037] Figure 2 The full-process link diagram of the welding equipment fault link diagnosis of the present application embodiment 1 is shown in the figure. DETAILED DESCRIPTION

[0038] The technical scheme of the present application will be described in detail below with the help of the accompanying drawings and specific embodiments. It should be understood that the specific features in the embodiments and the specific features in the embodiments are detailed descriptions of the technical scheme of the present application, and are not limitations of the technical scheme of the present application. In the case of no conflict, the technical features in the embodiments and the embodiments can be combined with each other.

[0039] Embodiment 1

[0040] As Figure 1 , Figure 2 shown, the embodiment introduces a welding equipment intelligent fault diagnosis method, the steps include:

[0041] The large language model parses the core retrieval dimension according to the input fault description;

[0042] Based on the pre-constructed welding equipment fault knowledge graph, the core retrieval dimension is preliminarily screened and the link is completed, and one or more fault link strengths and case success rates are obtained;

[0043] According to the fault link strength and the case success rate, the reliability of the link is diagnosed, the reliabilities of each link are compared, and the optimal diagnosis result is obtained.

[0044] The steps of constructing the welding equipment fault knowledge graph include:

[0045] The large language model collects welding fault documents, extracts welding exclusive entities, and labels entity attributes;

[0046] According to the entity attributes, the large language model combines the welding field rules to establish the core fault link relationship;

[0047] The entity, core fault link relationship and entity attribute are stored in the graph database and vector database;

[0048] Among them, when the welding equipment fault knowledge graph adds a new fault case, it is automatically updated by the large language model.

[0049] The entity attribute includes equipment information, fault description and solution;

[0050] The equipment information includes equipment model and core components;

[0051] The fault description includes fault phenomenon and fault code;

[0052] The solution includes fault reason, inspection method, solution measure and fault case.

[0053] The core fault link relationship includes fault propagation link relationship and fault resource mapping relationship.

[0054] The calculation formula of the reliability is as follows:

[0055]

[0056] Where C is the reliability, S is the total link strength, L is the link length, R is the average case success rate, W1 is the weight of the total link strength divided by the link length, W2 is the weight of the average case success rate, and W1+W2=1.

[0057] After obtaining the optimal diagnosis result, the large language model generates a diagnosis report according to the diagnosis result.

[0058] The diagnosis report includes core conclusions, cause analysis, inspection steps, solution measures, and process parameter adjustment suggestions.

[0059] The core search dimensions include equipment dimensions, fault dimensions, and exclusion dimensions.

[0060] The fault dimensions include fault phenomena and process parameters.

[0061] The vector database performs preliminary screening on the core search dimensions, and sequentially performs priority search according to equipment, process parameters, and fault phenomena to obtain one or more fault entities with high similarity.

[0062] The graph database performs link completion on the core search dimensions based on the preliminary screened fault entities, and outputs one or more links with high fault link strength and case success rate.

[0063] In this embodiment, the user inputs the fault description, such as MIG-500 welding with weak and strong arcs, welding current 200A, and wire feeding speed 5m / min. The large language model analyzes the core search dimensions, including model: MIG-500; phenomenon: weak and strong arcs; process parameters: current 200A, wire feeding speed 5m / min.

[0064] Specifically, in the vector database, three fault entities with high similarity are screened out, and the graph database is used to complete the links to output two links with high link strength and case success rate, such as link 1: weak and strong arcs → wire feeding speed fluctuation → motor stall → carbon brush wear, total strength 2.4; link 2: weak and strong arcs → power supply voltage fluctuation → power supply module contact failure, total strength 1.8. The credibility of the two links is calculated, the total strength of link 1 is 2.4, the link length is 3, and the average case success rate is 90%. The following formula is used:

[0065]

[0066] The credibility of link 1 is 84%, and the credibility of link 2 is 65%. The diagnosis result of link 1 is preferentially recommended.

[0067] The large language model generates a diagnosis report according to the diagnosis result.

[0068] Specifically, a welding equipment failure knowledge graph is constructed, and a large language model is selected as Qwen or Deepseek. After establishing the failure propagation link relationship, the strength value of each relationship is calculated. The causal relationship explicitly recorded in the document, such as the manual record that the wire feeding plug is caused by carbon brush wear, the strength is taken as 0.8-1.0. The correlation relationship based on case statistics, such as 80% of the wire feeding plug cases are caused by carbon brush problems, the strength is taken as 0.5-0.7. The strength of the common sense inference relationship is taken as 0.2-0.4.

[0069] Embodiment 2

[0070] Based on the same inventive concept as Embodiment 1, this embodiment introduces a welding equipment intelligent fault diagnosis system, which includes a processor and a memory. The memory stores program instructions, and when the program instructions stored in the memory are executed by the processor, the welding equipment intelligent fault diagnosis method is realized.

Claims

1. A method for intelligent fault diagnosis of welding equipment, characterized in that the steps include: include: The large language model parses the input fault description to obtain the core retrieval dimensions; Based on a pre-built welding equipment fault knowledge graph, the core retrieval dimensions are initially screened and the links are completed to obtain the strength of one or more fault links and the success rate of cases. Based on the strength of the faulty link and the success rate of the cases, the reliability of the link is diagnosed, and the reliability of each link is compared to obtain the optimal diagnosis result.

2. The intelligent fault diagnosis method for welding equipment according to claim 1, characterized in that, The steps for constructing the welding equipment fault knowledge graph include: The large language model collects welding fault documents, extracts welding-specific entities, and labels entity attributes; Based on the entity attributes, the large language model, combined with welding domain rules, establishes core fault link relationships; The entities, core fault chain relationships, and entity attributes are stored in a graph database and a vector database. When a new fault case is added to the welding equipment fault knowledge graph, it is automatically updated through the large language model.

3. The intelligent fault diagnosis method for welding equipment according to claim 2, characterized in that, The entity attributes include equipment information, fault description, and solutions; The equipment information includes the equipment model and core components; The fault description includes the fault phenomenon and the fault code; The solution includes the cause of the failure, the inspection method, the solution, and the failure case.

4. The intelligent fault diagnosis method for welding equipment according to claim 2, characterized in that, The core fault link relationships include fault propagation link relationships and fault resource mapping relationships.

5. The intelligent fault diagnosis method for welding equipment according to claim 1, characterized in that, The formula for calculating the credibility is as follows: ， Where C is the credibility, S is the total link strength, L is the link length, R is the average success rate of cases, W1 is the weight of the total link strength divided by the link length, W2 is the weight of the average success rate of cases, and W1+W2=1.

6. The intelligent fault diagnosis method for welding equipment according to claim 1, characterized in that, After obtaining the optimal diagnostic result, the large language model generates a diagnostic report based on the diagnostic result.

7. The intelligent fault diagnosis method for welding equipment according to claim 6, characterized in that, The diagnostic report includes core conclusions, cause analysis, inspection steps, solutions, and recommendations for adjusting process parameters.

8. The intelligent fault diagnosis method for welding equipment according to claim 1, characterized in that, The core search dimensions include device dimension, fault dimension, and troubleshooting dimension; The fault dimensions include fault phenomena and process parameters.

9. The intelligent fault diagnosis method for welding equipment according to claim 1, characterized in that, The vector database performs an initial screening of the core search dimensions, prioritizing searches by equipment, process parameters, and fault phenomena to obtain one or more fault entities with high similarity. The graph database performs link completion on the core search dimensions based on the initially screened faulty entities, and outputs one or more links with high faulty link strength and high case success rate.

10. An intelligent fault diagnosis system for welding equipment, characterized in that, It includes a processor and a memory, wherein the memory stores program instructions, and when the program instructions stored in the memory are executed by the processor, the intelligent fault diagnosis method for welding equipment according to any one of claims 1-9 is implemented.

Images