Expert System-Based Implementation of Failure Detection

In light of the need to improve MMIS of NPPs, the advanced I&C research team of KAERI has embarked on developing an Alarm and Diagnosis-Integrated Operator Support System, called ADIOS. ADIOS was composed of two modules such as alarm process module and fault diagnostic module. The alarm process module can offer key alarms to operator using dynamic alarm filtering methods and the fault diagnostic module was able to diagnose the component failure caused by a sensor fault and a hardware fault. ADIOS has been built in an object-oriented AI environment of G2 expert system software tool. In the G2 environment, every alarm and diagnostic component have been treated as an object of class. The attributes of each alarm object include activation status, alarm message, process value, activated time, priority, acknowledgment state, interlocked equipment and caused alarm. If an alarm is activated, its icon color on an overview process mimic diagram changes corresponding to its priority, which can be set initially or determined dynamically by reasoning rules and procedures. The process conditions, such as plant or equipment status and correlated alarms’ states have determined the priority of the activated alarm and the fault diagnostic condition. The knowledge base and the inference engines of ADIOS have been constructed by process analysis of the plant and discussion with operators and nuclear plant experts.