Application of Mahalanobis-Taguchi System and Statistical Process Control Charts for the Alerts of Real-time Predictive Maintenance

• Main Authors: Cheng-YangChen, 陳政陽
• Other Authors: Yu-Ching Chang
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/3p569e

碩士 === 國立成功大學 === 工業與資訊管理學系 === 105 === This study proposes a real-time monitoring system for machine’s status based on Mahalanobis-Taguchi system (MTS), moving statistics, MaxMEWMA control charts and dynamic alert limits. First, we collect real-time data from numerous sensors of the SECS/GEM (SEMI Equipment Communication Standard/Generic Equipment Model) system to analyze the machine’s overall status, and then the developed system provides alerts for the engineers as the basis of the predictive maintenance if the machine is in an abnormal status. We use the grinding process of semiconductor assembly as an example. First, to reduce time and cost for analysis, we utilize MTS to discover significant machine characteristics which can reflect the machine’s status. Secondly, we apply moving variance to remove the discrepancy between multiple recipes running on the same machine; as a result, we are able to use one MaxMEWMA control chart to monitor a machine. Finally, we adjust the alert limits dynamically based on the number of engineers and their availability for solving the abnormal alerts. Using normal data as the reference to identify the variability of machine status, we provide empirical analysis for our model with historical production data. We then define two adjustable parameters and set the appropriate alert limits according to the availability of engineers to handle the problem machine.