Reliability data acquisition and evaluation in process plants

Based on the practical experience of assessing reliability data in two plants of the explosives industry, the organization and scope of the data collection, the component delimitation, the procedures of quality assurance and the data evaluation for their final use in PSA studies are presented. An example is given for the type of detection and repair, etc. The mathematical background for evaluating observed component lifetimes is discussed. Both frequentist and Bayesian methods are addressed. The resulting failure rate distributions and their approximations by log-normal distributions for several key components are presented. They were obtained using a Bayesian approach with a non-informative prior. Remarks on uncertainty, their treatment and a proposal for transferring data to plants other than the ones investigated conclude the presentation.

► Reliability data for components in two existing process plants were collected and processed.
► A quality assurance system was devised and implemented.
► Frequentist and Bayesian evaluations were made and the impact of different prior distributions is shown.