(108d) Impact of Failure Data Especialization in Quantitative Risk Assessment of Process Plants

The aim of the paper is showing the bayesian inference in reliability studies, which are used to failures rates updating in safety analysis. It is developed the impact of its using in Quantitative Risks assessments (QRA) for industrial process plants and also with its applications in probabilistic risks assessments for nuclear installations. With this approach we find a structured and auditable way of showing the difference between an industrial installation with a good project and maintenance structure from another one that shows a low level of quality in these areas. In general the evidence from failures rates and as follow the frequency of occurrence from scenarios, which the risks taken in account in ERA, are taken from generics data banks, instead of, the installation in analysis. This is not the case of nuclear installations where the updating of data is a reality. Along of the paper is also showed some limitations and precautions to be taken in order to avoid mistakes in the use of this methodology, like choosing corrects probaility density functions. The results for QRA before and after the data especialization is ploted in isorisk curves from a real scenario of a liquid amonia leakage in a process plant.