Study on human reliability considering dependence among tasks in operation under accumulative stress condition

For analysis of human error rate in nuclear power plant safety, THERP (Technique for Human Error Rate Prediction) is mainly used. In the THERP, an engineering judgment is required to determine a level of dependence and stress. A mechanical and theoretical approach is more useful to determine the dependence. Furthermore, a stress level will change through tasks in one operation. This paper aims to evaluate the dependence quantitatively and to modify the stress level during the operation so as to analyze human error rate more particularly. Firstly, we propose a method for determining level of dependence mechanistically. Here the dependence is defined as a factor that comes from similarity among tasks and affects human error rate of current task depending on the result of previous task. In this paper, we assume that dependence is determined by three factors; personnel, time and place, and that level of dependence is determined by a weighted average of a level of similarity of these three factors. In order to determine weight of these factors, we apply AHP (Analytic Hierarchy Process) method. In the AHP, we set a purpose, an evaluation criteria and an alternative as each hierarchy and determine weight by comparing with importance of two elements in the hierarchy from the view point of contents of upper hierarchy. Then we modify the stress model so as to take into account the change of the level through tasks. Here the stress is a factor that is caused by environmental changes due to accidents and raises human error rate. In this paper, we assume that a level of stress is determined by not only scale of the accident but also a change for the worse during operation for the accident. Then we assume that a level of stress rises when personnel fails in a task. Furthermore, we set difference stress model in accordance with the skill of the personnel (beginner and veteran). In the paper, an optimization of personnel distribution with beginner and veteran in one operation is also discussed based on the present method. (author)