A human error detection system in nuclear power plant operations

Abstract

The most significant factor in nuclear power plant operations is safety. A lot of people in the nuclear industry have continued their unremitting efforts. After Three Miles Island accident, human factors came out into the open that it greatly contributes to the course of the accident of nuclear power plants. Thus, a lot of efforts have been made to reduce the human factor error. As nuclear power plant design developed, a new type of digitalized main control rooms has appeared, the conventional paper-based procedures have been left behind as backup. In advanced main control rooms (MCRs), computerized procedure system (CPS) is used to support human operators. Applying computer-based procedures in the main control room allows to reduce mental workload, enhance situation awareness, and produce lower errors of omission than paper-based procedure. However, current CPS does not yet utilize artificial intelligence technology. In order to reduce human errors, the framework which detects unsafe acts of human operators is suggested. The unsafe acts (UAs) detecting system implements Coloured Petri Nets, and deep neural networks to determine if an operating action is an error. The system uses two steps of filters to discover the effect of an operating action on the plant integrity. © 2018 Westinghouse Electric Company LLC All Rights Reserved