Quantitative Evaluation of Human-Reliability Based on Fuzzy-Clonal Selection

Human reliability analysis (HRA) is necessary for system safety assessment as well as equipment reliability analysis. The cognitive reliability and error analysis method (CREAM) as a representative HRA method provides nine common performance conditions (CPCs) to represent the contextual conditions under which a given action is performed. With a scarcity of empirical data, a high uncertainty in analyzing results is produced by subjective judgment. To obtain more objective and effective HRA results, this paper presents an optimized quantification method to evaluate the human error probability (HEP) according to CREAM, and its linguistic variables. The starting point for the quantification is an introduced fuzzy version of CREAM. However, many fuzzy rules are redundant, and occupy extensive computing time. The clonal selection algorithm combined with the fuzzy model is proposed to optimize the rule set. The evaluation method using the fuzzy-clonal selection algorithm is aimed to support possible applications for prospective and retrospective studies in the domain of safety assessment of power systems. The simulations are carried out on four presumed contexts, and a practical power system. The conclusions illuminate the feasibility of the proposed quantitative method.