Using Bayesian Inference and System Simulation in Production Process Yield Prediction - A Case Study

• Main Authors: Chun-Yi Hsiao, 蕭俊逸
• Other Authors: Hen-Yi Jen
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/2s32ag

碩士 === 元智大學 === 工業工程與管理學系 === 106 === Avoiding excessive losses in the process has always been an important issue that many companies are highly concerned about. However, with the advancement of technology, it has become relatively easy to collect such relevant data in addition to drastically reducing the occurrence of failure events in the process. Therefore, how to effectively use the collected data to reduce production costs has suddenly become an important issue in this research. Since the failure event is a rare event, this study uses the historical data provided by the company to use Bayesian inference to update the parameters based on the mean time between failures, and then uses the computer simulation software Flexsim to establish the simulation of the polarizer production process. The scenario experiments are under a fixed production schedule and uses different equipment efficiency evaluations to observe the differences while the MTBF parameter is updated in the simulation model. It is hoped that the method and results of this study will support the company assessment whether they need to reschedule the production planning in an effort to improve the decision making analysis.