Tianwan Nuclear Power Advances AI-Empowered Intelligent Upgrade of Maintenance Operations

In recent years, the maintenance domain of Tianwan Nuclear Power has relied on platforms such as "Hezhi·Longling" to address pain points in efficiency, quality, and management across the entire maintenance business chain, promoting the integration of artificial intelligence with nuclear power maintenance scenarios. To date, applications including personnel profiling, work order generation, compliance inspection, and auxiliary generation of electric actuator maintenance plans have achieved phased results, while projects such as the intelligent lifting command platform and marine tunnel cleaning robot are also under development.

In personnel management, Tianwan Nuclear Power has established a seven-dimensional personnel capability scoring model, setting differentiated evaluation indicators for different maintenance positions to form position-based, data-driven personnel capability profiles. The system can integrate historical data, support multi-period data retention and trend comparison, and display personnel capability distribution through radar charts, bar charts, and trend graphs. After batch data import, the backend automatically completes normalization calculation, multi-dimensional scoring, and ranking, generating materials that include overall overview, position comparison, special analysis of responsibility performance, and improvement suggestions, providing a basis for training resource allocation, position matching, and differentiated management.

To address the issues of time-consuming maintenance work order preparation and reliance on manual experience, the relevant team has developed an intelligent work order preparation assistant based on the Longhuitong intelligent agent platform, forming an application solution of "1 intelligent agent, 1 Skill capability, and 7 categories of professional knowledge bases." The system integrates the historical work order library, maintenance procedure library, equipment ledger library, risk and safety measure library, spare parts and tools library, personnel qualification library, and management procedure library. It can dynamically generate work package catalogs based on historical operation data and procedural requirements, automatically fill in work order business content, and output downloadable Excel work package drafts that can interface with the EPM system.

In terms of maintenance quality control, Tianwan Nuclear Power has developed an AI intelligent agent for maintenance compliance inspection. Based on large language models and RAG retrieval-augmented generation technology, this agent can batch receive maintenance procedures, preventive maintenance program lists, and maintenance completion reports, automatically parse document content, and conduct semantic comparison. The system can check item by item the coverage of maintenance procedure implementation steps by the program, while also identifying issues in completion reports such as missing items, omissions, data anomalies, missing signatures, and logical contradictions. The structured inspection reports it outputs can be traced back to specific procedure steps or report entries, and provide rectification suggestions.

Intelligent assistance is also being introduced into maintenance of key equipment. Given the large number of electric actuators in nuclear power plants, their complex models, and high maintenance requirements, Tianwan Nuclear Power has developed an auxiliary generation system for electric actuator maintenance plans. After engineering personnel input equipment information and fault descriptions, the system can call upon the electric actuator maintenance knowledge base, historical plans, and equipment data to automatically generate a draft maintenance plan, covering cause analysis, maintenance steps, technical key points, spare parts and tool preparation, and safety precautions. Verified by maintenance engineers, the relevant plans have basically met the conditions for on-site application and can save preparation time for emergency repairs.

In addition to applications already deployed, Tianwan Nuclear Power is also advancing the construction of an intelligent lifting command platform. Given the high professionalism and numerous control links involved in crane lifting operations at nuclear power plants, the platform is planned to support both web and API interface usage, and to manage processes such as development, training, and online monitoring through AI models. According to the design, the platform can assist in lifting planning after inputting lifting object parameters and site conditions, recommending equipment and lifting point positions, and alerting to risks such as overload and collision. It can also generate lifting plans based on cargo information, provide handling steps, required materials, and contact person information for emergency situations, and be used for training question bank generation, operation record analysis, and rectification report output.

Tianwan Nuclear Power stated that it will continue to expand AI application scenarios in the maintenance domain based on actual production needs, promote the digitalization and intelligent upgrade of maintenance operations, and strengthen network security management while advancing intelligent applications, providing support for the safe and stable operation of the units.