A knowledge-based approach to instrument fault detection and isolation

Abstract

A knowledge-based instrument fault detection and isolation (IFDI) technique is proposed and described. It is based on the duplication of measurement devices by means of suitable mathematical relationships and is implemented on an expert system. The latter allows a mathematical model to be substituted by integrating qualitative models with empirical knowledge, thereby reducing overall computer effort without any corresponding decrease in diagnostic capabilities. This technique is particularly efficient when applied to detect and isolate faults in measurement systems integrated in control architectures. It is applied to an automatic measurement station for induction motor testing to illustrate its characteristics and performance. The experimental results are also reported, illustrating IFDI performance mainly in terms of promptness, sensitivity, and selectivity. >