Aiding fault diagnosis under symptom masking in dynamic systems

Abstract

In large-scale dynamic systems such as power plants, fault diagnosis becomes more difficult owing to changing system states. Symptoms can be obscured by other symptoms or disappear as the operator proceeds with diagnosis or compensation. Such a symptom-masking phenomenon increases the operator's memory load, constrains diagnostic strategies and, hence, degrades the performance of the diagnosis. This study investigates the effects of a symptom bookkeeping aid (SB-aid) for diagnostic performance in a simulated sub-system of nuclear power plants. The results show that this SB-aid improved overall performance. Further analyses provide insights on the sources of cognitive difficulties and on the possible direction of supporting human diagnosis in dynamic plants.