Automated Knowledge Acquisition for Intelligent Support of Diagnostic Reasoning

Abstract

An intelligent simulation-based system for presenting and supporting diagnostic exercises can resolve many of the constraints that currently limit the value of troubleshooting practice in the training setting. By focusing on highly contextualized and domain-specific reasoning issues rather than upon general troubleshooting principles, such a system can reveal gaps in system understanding and can address shortcomings in applied diagnostic reasoning. Moreover, by providing expert guidance in choosing and interpreting tests, such a system can essentially eliminate the impasses that so often typify troubleshooting exercises. Authoring such systems via conventional expert systems has proven to be exceedingly expensive. A more robust approach automatically inspects a domain-specific model of the target system to produce fault effect information capable of supporting diagnostic reasoning and hence intelligent support of diagnostic practice. Because the diagnostic reasoning processes are general rather than domain specific, only the model of the target system must be developed to support this process. The authoring process involves 1) representing the behaviors of the operational units of a system in terms of inputs received from functionally adjacent elements, 2) representing controls and indicators graphically, so that the learner may operate and inspect them during practice problems, and 3) invoking an automated function that generates fault effect information by executing the device simulation.