Sequencing reliability growth tasks using multiattribute utility functions

In both hardware and software engineering, the reliability of systems improve over the Test, Analyse and Fix (TAAF) cycle as reliability tasks are performed and faults are designed out of the system. There are many possible tasks which could be carried out and a large numbers of sequences of these tasks possible. In this paper we consider the sequencing problem, taking into account the fact that the testing will be stopped once a reliability target is reached. We solve the problem by maximising the expectation of a two attribute utility function over cost and time on test. All marginal utilities are set to be risk averse. A reliability growth model based on the underlying engineering process is used. The method is illustrated with an example grounded in work with the aerospace industry