Sensitivity analysis for system design improvement

In order to help identify suitable design modifications for complex systems, there is a need to (1) determine the weakest elements of the system, (2) identify a better design alternative, and (3) evaluate the effect of the adopted solution on system safety/availability. This is a well-known procedure that is applied to a single system failure mode. In reality, several undesired system states have to be checked for. On the other hand, systems grow in complexity, and components are often multi-functional, so that they could affect several system states concurrently, and even in conflicting terms. This paper presents the sensitivity analysis module of the ASTRA (Advanced Software Tools for Reliability Analysis) package, based on component importance analysis techniques to be applied to all system failure states concurrently. The technique employs fault-tree analysis and Markov processes as the basic probabilistic models.