Practical aspects of machine diagnostics accuracy

Abstract

In any activity connected with monitoring, diagnostics and prognostics, the accuracy and believability of results is of paramount importance. Two main kinds of diagnostic errors, called Type I and Type II or False Positives and False Negatives are encountered whenever classification, such as “good” vs “bad” is considered. This has been thoroughly discussed, especially in the field of medicine, where the consequences of false judgements can be particularly grave. Other disciplines highly interested in this topic are financial forecasting and insurance. Industrial equipment diagnostics deals with these issues as well. In this paper we consider various factors influencing the quality of data, upon which diagnostic classification is based. We also look at the consequences of diagnostic errors to demonstrate why they are bad. We consider various ways employed to mitigate both the probability and the consequences of false judgements.