Expert diagnostic system

The Expert Diagnostic System (EDS), an Automatic System Level Test Tool, is a hybrid knowledge based diagnostic system under development for use in diagnosing failures in Westinghouse Avionic Systems. The EDS contains both shallow and deep knowledge reasoning. Expert emperical (Shallow) Knowledge is captured and represented using rules in OPS5 [1,3] production system format. These rules are applied to the Fault Detection (FD) test result using OPS5 inferencing to localize failures to a failure class and subsequent to a failure subclass using Unit Under Test (UU'I') failure heirarchy. Once a failure subclass has been selected, control is transferred to deep reasoning inference engine. Deep reasoning loads and creates a frame based model pertaining to the failure subclass under consideration. This model describes the UUT signal flow and connectivity of the system components in relation to the failure subclass. This model is best-first searched using heuristics for best probing locations to perform more tests. This process continues until failure has been isolated to component. This paper is a description of the concepts, knowledge representations and n',ethodologies used by EDS. INTRODUCTION The rapid increase in the complexity, use and dependency of our society on state-of-the-art electronic systems has produced the need for better methods of maintaining the operational readiness of these complex systems. Diagnostics, i.e.~ the Fault Detection (FD), Fault Localization (FL), and Fault Isolation (FI), is the process of detecting, localizing, isolating and fixing the failures in such sophisticated systems. Present automated diagnostic tools are not an effective and efficient method for diagnosis of failures in these very complex electronic systems. These available automated diagnostic systems have some major limitations, such as; . Hard coded diagnostic rules and system knowledge in software make fault isolation fixed and procedural at all times, . Only a subset of the total system failures can be diagnosed due to the size of the software, 3. High cost of diagnostic software life cycle maintenance, 4. Systems can neither be 'taught' nor can they 'learn' and therefore are not adaptable, and 5. Lack of transportability across different UUTs. The purpose of this project has been to: 1. Study the problems with the current electronic diagnostic tools.