In this dissertation, we describe an implementation of a diagnosis expert system for emergency diesel electric generators in nuclear plant. To construct the diagnosis expert system, the classification and the analysis on causal relationship in diagnosis domain is conducted in the first place. Based on the knowledge on the causal relationship of a human expert in the domain, knowledge base is constructed using JRules which is an object-oriented and rue-based expert system development tool.
The system consists of a diagnosis module, a measuring module and a monitoring module. The diagnosis module is composed of a fault-diagnosis submodule which is activated by alarm occurrences and an early-diagnosis submodule which analyses measured data in a periodical manner. The measuring module measures data necessary for diagnosis from the diesel engine. Finally, the monitoring module provides graphic user interface. In this study, the measuring module is simulated by using data of a trial run obtained from DMDS, and the monitoring module is implemented by making use of a SCADA system, In-Touch.
To test the operation of the whole system, a set of simulated data is fed into the measuring module. The correct operation of the system could be confirmed by observing induced diagnosis results through the monitoring module.