| Most current machine fault diagnosis systems emphasize the correctness of the hypothesized result; however, in time constrained situations, the efficiency of the diagnostic process becomes more important and should not be overlooked. This dissertation presents an Integrated Machine Fault Diagnosis System (IMFDS) that enhances the efficiency of the diagnostic process, improves the completeness and consistency of the knowledge base, and assists users in developing and maintaining their diagnostic systems.; IMFDS consists of five modules: (1) a diagnostic tree module establishes the hierarchical structure regarding the function or connectivity of the diagnosis system, (2) a fuzzy multi-attribute decision-making module determines the most efficient diagnostic process and creates a "meta knowledge base" to control the diagnosis process, (3) a knowledge-base module captures human expertise and deep knowledge to diagnose the possible machine fault, (4) an inference-engine module controls the diagnosis process and deals with uncertainty from the user input and knowledge base itself, and (5) a learning module uses the failure-driven learning method to train the knowledge base from past actual cases.; This system has been successfully implemented in the MS-Windows environment and it is written in MS Visual BASIC. To validate the system performance, IMFDS is compared to EXACT, an expert system for automobile air-compressor troubleshooting, using fifty sample cases of actual repair records. The result shows that IMFDS can reduce the diagnosis time by 24.9%. |
