From Imprecise Probability Laws to Fault Tree Analysis

Abstract

Reliability studies and system health predictions are mostly based on the use of probability laws to model the failure of components. Behavior of the components of the system under study is represented by probability distributions, derived from failure statistics. The parameters of these laws are assumed to be precise and well known, which is not always true in practice. Impact of such imprecision on the end result can be crucial, and requires adequate sensitivity analysis. One way to tackle this imprecision is to bound such parameters within an interval. This paper investigates the impact of the uncertainty pervading the values of law parameters, specifically in fault tree based Safety analysis.