Building a Rule-Based Expert System to Enhance the Hard Disk Drive Manufacturing Processes
Suppakrit Kirdponpattara,Pitikhate Sooraksa,Veera Boonjing
DOI: https://doi.org/10.1109/access.2024.3369443
IF: 3.9
2024-03-02
IEEE Access
Abstract:The manufacturing of hard disk drives involves the intricate assembly of numerous components, making the testing process time-consuming and resource intensive. To optimize the manufacturing process and increase testing efficiency, the development of a rule-based expert system is proposed. This system leverages predictive models constructed from assembly process data to identify potentially defective hard drives before undergoing extensive testing. By preemptively identifying defects, this approach substantially reduces testing time and enhances tester capacity. Given the categorical and imbalanced nature of assembly data, Decision Trees are employed as the prediction model. Specifically, three Decision Tree algorithms are explored: ID3, C4.5, and CART. In addition, four feature selection techniques, namely Information Gain, Gain Ratio, Chi-Square, and Symmetrical Uncertainty, are utilized to identify high-impact features. Our experimental findings reveal that Information Gain coupled with the C4.5 algorithm yields the most favorable results in terms of prediction accuracy, modeling efficiency, and rule generation. Moreover, our study establishes that setting the failure probability threshold between 0.15 and 0.70 provides the shortest total test time for the proposed process, as supported by a 95% confidence level. This achievement represents a statistically significant enhancement compared with the existing manufacturing process.
computer science, information systems,telecommunications,engineering, electrical & electronic