Development of maintenance knowledge management system for power plant

This paper describes the knowledge management framework to systematize maintenance knowledge in power plant. The features of the framework are (1) to construct a decision tree of trouble cases incrementally based on Ripple Down Rules (RDR) method and (2) to decide the timing of installing new knowledge based on several metrics obtained by evaluating a structure of a decision tree. We are developing the experimental prototype of maintenance knowledge management system. The system is integrated with plant database (ex. the operation journals and maintenance records). The management of trouble cases is based on the format of XML. For practical purpose, each trouble case is linked with related data (ex. maintenance manuals, equipment configuration, records). Plant workers can easily search for maintenance knowledge and relevant data by specifying observed status. On the other hand, knowledge engineers can monitor occurrence counts of exceptional cases to each node. The metrics calculation gives a cue to locate the effective node for improving the efficiency of retrieval. We are evaluating the functionality and the effectiveness of the system through the application to maintenance work of controller systems. We believe that this framework will be helpful to share common knowledge among plant sites as well as to manage site-specific knowledge. (author)