Design for diagnosis using a diagnostic evaluation measure - Evaluating the potential cost of diagnosing a system when it fails

Abstract

The results demonstrated by the SAN system show how a real fault tolerant system can have more than one configuration with almost the same reliability. However, from a diagnosis perspective, the designs can have different costs of diagnosis at failure time. A method for measuring the cost of diagnosing diagnosis decision trees (DDT) was demonstrated, which is a specialized weighted mean depth of traversing the DDT. This evaluation method incorporates the qualitative (structure) and quantitative (reliability analysis) into one measure for predicting diagnosis cost. DDTs incorporate the same qualitative and quantitative data as the diagnosis cost measure, which makes this measure suitable for evaluating these decision trees. Evaluating and comparing alternative system designs using the diagnosis cost function as we demonstrated is an efficient and simple method for pursuing the design for diagnosis objective