A Method for Industrial Equipment Fault Prediction and Intelligent Maintenance Work Order Generation Based on a Large Model

By using a closed-loop feedback mechanism that generates root cause ranking through health degradation modeling and large language modeling, the problem of disconnect between fault prediction and operation and maintenance response is solved. This enables intelligent and executable equipment health assessment, fault prediction, operation and maintenance work order generation, and personnel dispatch, thereby improving the efficiency and accuracy of operation and maintenance.

CN123118060APending Publication Date: 2026-11-17SHANGHAI CAIJIANG INTELLIGENT TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611268610.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-20
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In existing technologies, fault prediction and maintenance response are disconnected, work orders lack guidance, personnel dispatch relies on experience, feedback loop is incomplete, and there is a lack of health prediction and root cause ranking of degradation baselines for the same model of equipment. As a result, the prediction results are difficult to directly translate into executable, scalable and correctable maintenance actions.

Method used

By assessing the equipment degradation stage through health degradation modeling, the system uses a large language model to generate root cause ranking and structured work orders, and combines personnel skills and location for intelligent dispatch, forming a closed-loop feedback mechanism of prediction, execution and correction. This includes equipment health scoring, failure time window and type probability distribution, failure mode inference, maintenance plan generation and personnel dispatch.

Benefits of technology

It achieves closed-loop management from fault prediction to operation and maintenance response, improves the guidance of work orders and the intelligence of personnel dispatch, ensures the feasibility of operation and maintenance actions and the timeliness of feedback, and enhances the accuracy of equipment health prediction and the effectiveness of maintenance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure FT_1
    Figure FT_1
  • Figure FT_2
    Figure FT_2
  • Figure FT_3
    Figure FT_3
Patent Text Reader

Abstract

This invention discloses a method for industrial equipment fault prediction and intelligent maintenance work order generation based on a large model, belonging to the field of intelligent maintenance technology for industrial equipment. The method collects multi-dimensional operating sequence indicators, maintenance history, and service life information of the equipment; constructs an equipment health degradation model to output health scores, fault time windows, and fault type probability distributions; uses a large language model combined with a fault mode knowledge base to infer candidate root causes and generate structured maintenance work orders; then, intelligent dispatch is performed based on maintenance personnel skills, workload, and geographical location, and the maintenance results are fed back to the prediction model and knowledge base. This method can improve the response efficiency and work order executability of predictive maintenance.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of intelligent operation and maintenance technology for industrial equipment, specifically involving a method that integrates equipment health degradation modeling, root cause inference of large language models, generation of structured operation and maintenance work orders, and intelligent dispatch of maintenance personnel. Background Technology

[0002] Existing equipment maintenance technologies can predict failure probabilities based on operational status and historical data, and automatically create or assign maintenance work orders based on the prediction results; some solutions also utilize large language models to generate structured work orders or select handlers according to skill requirements. However, failure prediction, root cause analysis, maintenance plan generation, personnel assignment, and result feedback are often completed by independent modules, lacking a closed loop for health prediction based on the degradation baseline of the same model of equipment and work order generation centered on candidate root cause ranking. As a result, prediction results are difficult to directly translate into executable, scalable, and correctable maintenance actions. Summary of the Invention

[0003] To address the issues of disconnect between fault prediction and maintenance response, insufficient work order guidance, reliance on experience for personnel dispatch, and incomplete feedback loops, this invention provides a full-chain method from health degradation modeling to work order generation, intelligent dispatch, and execution feedback. This method assesses the current equipment degradation stage by analyzing typical degradation trajectories of a group of devices of the same model, outputting fault time windows and fault type probabilities. It generates root cause ranking and structured work orders using a large language model and a fault mode knowledge base. Dispatch is then performed by integrating personnel skills, workload, and location, and escalates to backup personnel if there is no response within the timeout period. Maintenance results are used to correct the degradation model and knowledge base. Attached Figure Description Figure 1 This is a schematic diagram of the overall process of the industrial equipment fault prediction and intelligent operation and maintenance work order generation method of the present invention.

[0004] Figure 2 This is a schematic diagram of the equipment health degradation modeling and fault prediction process in this invention.

[0005] Figure 3 This is a schematic diagram of the closed-loop process of intelligent dispatching, execution tracking and feedback of maintenance work orders in this invention. Detailed Implementation

[0006] In this embodiment, multi-dimensional time-series data such as vibration, temperature, current, and pressure are collected from the target equipment, and associated with the equipment model, service life, and historical maintenance records. The system extracts trend, fluctuation, and periodic characteristics from each indicator; then, it extracts typical degradation trajectories from historical pre-failure data of the same model of equipment as a reference baseline for assessing the current degradation stage of the equipment.

[0007] The health degradation model compares the current equipment characteristics with a reference baseline and outputs a health score, an expected failure time window, and a failure type probability distribution. When the output meets preset warning conditions, the equipment model, current operating parameters, predicted failure type, and time window are written into the processing context.

[0008] The large language model extracts fault modes, maintenance strategies, and safety requirements related to the prediction results from the context to be processed and the fault mode knowledge base, and infers one or more candidate root causes. For multiple candidate root causes, the corresponding maintenance solutions are ranked from high to low probability, and a structured maintenance work order is generated. The work order includes at least a fault description, severity, maintenance steps, spare parts list, estimated maintenance time, and safety precautions.

[0009] The dispatch module reads skill qualifications, current workload, and geographical location information from the maintenance personnel information, calculates a comprehensive ranking result corresponding to the work order requirements, and dispatches the work order to the most suitable personnel. For urgent work orders, a response time limit is set; if the designated personnel do not respond within the time limit, a backup personnel is automatically selected for escalation dispatch. Maintenance personnel can read the steps and candidate root cause ranking in the work order during execution.

[0010] After maintenance is completed, the system collects the actual fault cause, maintenance measures, actual time consumption and execution results, compares them with the predicted fault type and candidate root causes, and feeds the deviation information back to the health degradation model and fault mode knowledge base to form a continuous closed loop of prediction, execution and correction.

Claims

1. A method for industrial equipment fault prediction and intelligent operation and maintenance work order generation based on a large model, characterized in that, include: Collect multi-dimensional operating sequence indicators of the equipment, such as vibration frequency, temperature, current, and pressure, as well as maintenance history and equipment service life information; extract Based on trend characteristics, fluctuation characteristics, and periodic characteristics, a device health degradation model is constructed, outputting a health score, fault occurrence time window, and fault type probability distribution. The fault type, device model, and operating parameters are input into a large language model, and combined with a fault mode knowledge base, candidate root causes and impact ranges are inferred. A structured maintenance work order is generated, containing a fault description, severity, repair steps, spare parts list, estimated repair time, and safety precautions. The skill qualifications, current workload, and geographical location information of maintenance personnel are obtained, and they are comprehensively ranked according to skill matching degree, current workload, and response time requirements. The structured maintenance work order is then dispatched to the maintenance personnel corresponding to the ranking results.

2. The method according to claim 1, characterized in that, Statistical analysis is performed on historical fault data of equipment of the same model to obtain typical degradation trajectories of each fault type as reference baselines, and the current indicator trends of the equipment are compared with the reference baselines to assess the degradation stage.

3. The method according to claim 1, characterized in that, When the fault type corresponds to multiple candidate root causes, a corresponding maintenance plan is generated according to the root cause probability from high to low, so that maintenance personnel can check in sequence.

4. The method according to claim 1, characterized in that, Set response time limits for structured maintenance work orders with high urgency; if the dispatched maintenance personnel do not respond within the response time limit, the structured maintenance work order will be automatically upgraded to the backup maintenance personnel.

5. The method according to claim 1, characterized in that, After maintenance is completed, the actual cause of the fault, maintenance measures and execution results are collected, compared with the probability distribution of the fault type and candidate root causes, and the deviation information is fed back to the equipment health degradation model and fault mode knowledge base.