An Empirical Study of Statistical Process Control and Data Science for Yield Enhancement in Machine Tool Industry

• Main Authors: Bo-KaiJang, 張博凱
• Other Authors: Chia-Yen Lee
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/2bxv52

碩士 === 國立成功大學 === 製造資訊與系統研究所 === 106 === In machine tool industry, it’s critical to maintain high-quality products during the manufacturing process while the variance in the process shows significantly impact on quality issue, especially in mass production. Typically, it’s necessary to monitor the real-time status of equipment control; that is, it’s urgent to develop a production system which has real-time monitoring for driving productivity and reducing variance. This study developed several statistical control charts via data mining framework. First, data was divided into many groups by cluster analysis, and the baseline of product can be defined for each cluster. Second, univariate control charts were developed based on the baselines of qualified products. In particular, the control charts were based on time series and profile monitoring since the monitored target will constantly change over time. Finally, we also extended to develop the multivariate control chart with Hotelling’s T2 control chart and kernel distance control chart for the comparison. An empirical study of machine tool manufacturer was conducted to validate the proposed framework. The result shows that the quality system embedded with data mining and machine learning technology significantly enhanced the monitoring reliability of control charts and improved the yield.