新旧制御盤比較におけるタスク評価へのFLIAの応用 : 第2報, クラスター分析によるタスクの区分け

概要

論文の詳細を見る
Human reliability in a complex system has been studied to establish safety systems by analyzing the operator's performance in a control room of a nuclear power plant. In this paper, results of a mathematical model and a questionnaire given to plant designers and operators led to the proposal of a fuzzy tool for evaluating the quality of recent automated control systems. The first report described a method which is capable of calculating human performance by summing the weighted utility of attributes. The modified fuzzy measures learning identification algorithm (FLIA) reduces a set of attributes until human tasks are represented clearly. A change in the performance is illustrated on a two-dimensional map of the dominant attributes as a function of the automated level. The designers and the operators determined the balance of the manipulated/automated phase on the map after careful individual interviews. In the present paper, we attempt to interpret the boundary with a cluster-analysis theory, where the Euclidian square distance and the nearest-neighbor method are applied. The evaluated aspect of the boundary on the map can be divided into the manipulated/automated phase. It is shown that the calculated boundary is equal to the vertical bisector between the center of gravity of the clusters. The analytical boundary agrees precisely with the questionnaire result.
一般社団法人日本機械学会の論文
1996-08-25