ارزیابی قابلیت اطمینان فازی وقایع پایه درخت های خطا از طریق پردازش داده های کیفی A fuzzy reliability assessment of basic events of fault trees through qualitative data processing

In this section, we describe how the proposed fuzzy reliability algorithm is validated. To investigate the feasibility of the proposed algorithm, the actual failure probabilities taken from the US Combustion Engineering reactor protection system (CERPS) during the period 1984 through 1998 operating experience, which are well documented in Wierman et al. [50], are compared to the failure probabilities generated by the algorithm. Many authors in the past have used this data source to validate their experimental studies. Bondavalli and Filippini [4] used this data source to validate their proposed stochastic Petri net to assess the availability and performance of the safety function of the reactor protection system. In the study by Bartha et al. [1], this data was used to validate their proposed periodic and outage testing methodology of the reactor protection systems in the Paks Nuclear Power Plant. Meanwhile, Kang and Han [23] used this data source to calculate alpha parameters to make the common cause failure event failure probabilities suitable for the emergency diesel generator for Ulchin Unit 3. Bickel [2] used this data set to evaluate the risk implications of the core protection calculator system failure in the reactor protection system. Component failure probabilities in Wierman et al. [50] are presented in three different values, i.e. best estimate, lower bound, and upper bound reliability values. The best estimate reliability value is the recommended reliability data to be used in the FTA. Meanwhile, the upper and the lower bound reliability values represent a range of reliability data estimation. To verify the feasibility and the applicability of the proposed algorithm, the basic event failure probabilities generated by the proposed algorithm have to be between the upper and the lower bound reliability values and as close as possible to the best estimate reliability value. If the results show that the generated failure probabilities are beyond the range of the reliability data estimation, the membership functions to represent the qualitative data need to be explored in more detail.