Development of Reliability Function and Failure Rate Considering Information of Censored Data Available to Purchase

Generally, there are two methods to estimate reliability of system or component failure in engineering, or survival function of patient death in medicine: 1) parametric and 2) non-parametric methods. The parametric method has been used widely since there is few data on severe accidents in nuclear power plants. However, if we have a lot of data on system failure or patient death, the non-parametric method could be used to estimate reliability or survival function. In this paper we focus on the latter case in which one problem arises in treating the data. The problem is how to reflects efficiently the censoring information into reliability or survival function. In this study, two works have been done: 1) develop the model to reflect censoring information and 2) show how efficient the model is using the Monte-Carlo Simulation. In the WASH-1400 (NUREG-75/014), Failure to Operate of Motor Operated Valves from Nuclear Experience is suggested 1×10⁻³/d. The new developed method can estimate closer to the failure rate 1×10⁻³/d than the other existing methods. Also the failure rate using the new method satisfies the conservative limit.