An Efficient Expert System for Air Compressor Troubleshooting

Abstract

In this paper, an expert system for automobile air compressor troubleshooting (ACTS) is presented. This system can assist users to conduct an efficient and effective diagnosis on air compressor failures. Unlike most diagnosis expert systems, ACTS uses a new control strategy to enhance the efficiency of the diagnostic process. This control strategy attempts to spend the least amount of time to detect the compressor fault accurately by investigating only portions of the knowledge base. ACTS first constructs a diagnostic tree based on the functions or connectivity of the air compressor's devices. A fuzzy multiple‐attribute decision‐making method is used to determine the priority of the nodes (devices) in the diagnostic tree. The prioritized result creates a ‘meta knowledge base’ to control the diagnostic process. In addition, each node possesses its own knowledge base for hypothesizing the possible faults for the node. ACTS, written in MS Visual BASIC, has been successfully developed and implemented in MS‐Windows environment on a PC. To validate the system performance, ACTS is compared to EXACT, an expert system for automobile air compressor troubleshooting, using 50 sample cases. The evaluation results indicate that ACTS performs better than EXACT by reducing the number of queries and the diagnosis time by 20.7% and 24.9%, respectively.