Combining imperfect coverage with digraph models

Abstract

We present a prototype implementation of a program to compute the unreliability of a system based on the digraph model of the system and coverage models for individual components. The C program we have written takes as input a system description describing failure modes in terms of a digraph model. This, as well as coverage probability information are used to produce a quantitative unreliability result. The problem being addressed is an important one. The goal is not only to improve the validity of the model being used, but to keep the framework simple, usable and adaptable. A more complete model allows for more realistic analysis. It is essential that life critical systems meet their required level of accuracy. Excluding any of the factors discussed here could result in serious miscalculations. One benefit of performing a quantitative analysis is the digraph models could be used to help analyze the dependability of the system being designed so as to facilitate tradeoff analysis when alternative designs are considered. Another attractive feature of the proposed approach is that it could be used in conjunction with pre-existing tools to enhance the diagnosis process that already exists without significantly affecting the time, money or effort involved. Within the concept of fault diagnosis, a quantitative analysis could allow a prioritization of lists of possible failure causes based on the probabilities associated with those events. In other words, the paths of the digraphs would be weighted so that most likely causes could be considered first.