Manufacturing flow time estimation: case study of IC back-end assembly

• Main Authors: CHEN, YU-WEI, 陳昱惟
• Other Authors: GUO, SHIN-MING
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/fq7f72

碩士 === 國立高雄第一科技大學 === 運籌管理系碩士班 === 105 === The manufacturing processes become more and more complicated nowadays such as semiconductor manufacturing. The materials and components used in the processes also increase and they would be different according to which product types are used. There are lots of materials need to take hours or even one day to defrost, the expired date is short and easily overdue. It is a big challenge for production line to decide when to thaw out the multiple materials. This study aims to develop the models for predicting the manufacturing flow time to suggest production line when to defrost the materials and avoid to waste too much materials. The regression tree approach is used and analyzed the historical data from an IC back-end assembly company. The target is to predict the manufacturing time to Molding stage when a product moves in to the assembly process. The model can directly predict the flow time from each stage to Molding stage, and there is also another model can forecast the flow time segmented. For example, when a product move into Rivet stage, the model can evaluate the process time from Rivet to Wire Bonding and Wire Bonding to Molding based on the process status at the moment. It can help not only Molding to prepare the material but also Wire Bonding stage. Finally, the research develops the different models for each stages, and the models estimate the manufacturing flow time according to work in process (WIP), machine numbers in each stage, order quantities and the slack time to due date. The forecast accuracy for predicting directly is similar to use the segmented models. The result shows after predicting the flow time to Molding in Rivet stage, when the product moves into next major stage, it should update the process status and re-predict the flow time, and it can improve the forecast accuracy.