Article ID: | iaor1991964 |
Country: | United Kingdom |
Volume: | 28 |
Issue: | 12 |
Start Page Number: | 2259 |
End Page Number: | 2276 |
Publication Date: | Dec 1990 |
Journal: | International Journal of Production Research |
Authors: | Naruo Nobuyuki, Lehto Mark, Salvendy Gavriel |
Keywords: | artificial intelligence, artificial intelligence: decision support |
This paper presents a case study documenting the development of an expert system for diagnosing the malfunctions of a machine used by the NEC Corporation to mount chips on integrated circuit boards. Development of the expert system was justified by the inability of operators to efficiently diagnose many malfunctions of the chip-mounting machine, the associated cost of production delays, and the disruption incurred when experts were forced to leave unrelated tasks to help operators troubleshoot malfunctions. The first step in development of the expert system was to elicit and organize the machine designer’s knowledge. This process resulted in a hierarchical classification of malfunction symptoms and causes, a set of 15 flow diagrams documenting the designer’s troubleshooting procedures for particular malfunction symptoms, and a matrix documenting design information. The flow diagrams were translated into a large logic network diagram, which was directly translated into a set of 94 rules. An additional set of 270 rules were derived from the design matrix. The resulting 364 rules were then implemented in an expert system using the KES shell. On-site validation revealed that 92% of the chip-mounting machine’s malfunctions occurring in 1988-1989 were successfully diagnosed by the expert system. Future directions of this research will be oriented toward the development of a general purpose expert system capable of diagnosing the malfunctions of other similar production equipment.