Reliability data are very important for probabilistic safety assessment by fault tree analysis. Due to the limitations of historical failure data of the system being evaluated, generic data have been occupied. However, generic data comes with uncertainties and, hence, cannot represent its real performance. To overcome this limitation, fuzzy probability has been integrated into conventional fault tree analysis to evaluate the performance of nuclear reactors' safety systems. In fuzzy fault tree analysis, the qualitative judgments of experts are collected to assess the failure possibility of basic events. Furthermore, the membership functions of fuzzy numbers are used to convert those qualitative judgments into quantitative data. The assessment made by experts is a subjective judgment. Each expert has varying levels of expertise, so assessment of the same basic event will provide different judgments. The experts who are involved in the assessment process should be selected properly, because expert judgment is very influential on the final outcome of analysis. This study proposes a four stage expert selection process. Through the implementation of the proposed approach, five groups of experts has been selected. The five groups consisted of (1) head of division (reactor maintenance division, reactor operation division, and occupational and operation safety division); (2) head of subdivision (mechanical system subdivision, electrical system subdivision, instrumentation and control subdivision, operation reactor subdivision, and operation safety subdivision); (3) supervisor of reactor and supervisor of maintenance; (4) operator of reactor and maintenance technician; (5) radiation protection officer and staff of operation safety subdivision.