Evaluating diagnostic failures during system design

For the modern industry, diagnostic failures have been an area where huge sums of money is wasted through increasingly high levels of unscheduled removals. This is not only due to the continued tread in increasing system complexities and cost implications on maintenance programmes, but also attributed to design limitations associated with testability. As a result, No Fault Found (NFF) events continue to pose significant problems for achieving successful failure diagnosis within a system. This has a direct impact on the availability requirements, compromises reliability and also contributes to the cost of resolving an unknown failure. This article investigates how the diagnostic analysis of a system can be used to help recognise and reduce the number of NFF events. It is revealed that such a problem is difficult to resolve using traditional fault isolation calculations, and that Monte Carlo based methods provide a better alternative. A simulation-based evaluation is carried out on a UAV fuel system, to establish the relation between large ambiguity groups and the number of NFF events. It also demonstrates a method for estimating cumulative replacement costs due to false removals and hence, the overall system costs during the design stage. Such studies can be used for benchmarking, as well as a measure of how maintenance requirements might change over the intended lifetime of a system.