The design and development of a causal reasoning based machine fault diagnosis shell

摘要

This research focuses on two main issues related to the design, development, implementation, and application of machine fault diagnosis expert systems: (1) investigating the actual cognitive process of human diagnosticians, and (2) studying the current practices used in the development of machine fault diagnosis expert systems. The investigation of a human expert's diagnostic reasoning process is used to abstract and capture in generalized modules the human ability to understand, learn and diagnose different machinery belonging to a particular class. The analysis of current practices followed in the development of machine fault diagnosis expert systems is used to design and develop a generalized machine fault diagnosis shell which reduces the burden of custom designing and developing each application diagnosis expert system separately. The designed shell reduces the development time, effort and skill by making use of generalized modules for knowledge acquisition, knowledge verification, application system generation, learning, explanation, etc. The shell takes as its input the design descriptions of a particular type of machine to be diagnosed and generates a machine fault diagnosis expert system by effectively combining application knowledge with the built-in generalized diagnosis modules. The generated application diagnosis expert systems use shallow reasoning for diagnosing simple, known faults, and causal reasoning for diagnosing complex faults. The shell has been validated by generating a prototype fault diagnosis expert system for a Cincinnati Milacron 786 robot.