Using Sequential Pattern Mining for Optimum Combination of Using Sequential Pattern Mining for Optimum Combination of Machines in LED Industry
碩士 === 國立高雄應用科技大學 === 資訊管理系碩士在職專班 === 97 === Because of the competition of manufacturing industry, high yield rate of LED (Light Emitting Diode) production is one of major aims for LED manufacturers. For the procedures of LED production, every LED must go through one or more complex and complicated...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | zh-TW |
Published: |
2008
|
Online Access: | http://ndltd.ncl.edu.tw/handle/35938598936004552978 |
Summary: | 碩士 === 國立高雄應用科技大學 === 資訊管理系碩士在職專班 === 97 === Because of the competition of manufacturing industry, high yield rate of LED (Light Emitting Diode) production is one of major aims for LED manufacturers. For the procedures of LED production, every LED must go through one or more complex and complicated machines during the manufacturing procedures. As a result, how to obtain a high yield rate and find out excellent combination of machines for LED production is a major developmental direction of LED manufacturers. The purpose of this research is to propose the Data Mining architecture for LED manufacturing process to discover the best combination of machines for obtaining higher yield rate of LED production.
A Data mining method is constructed on the MES (Manufacturing Execution System). An experiment of a LED manufacturing process will be used to illustrate and verify the feasibility of the proposed method. WAT (Wafer Acceptance Test) data are extracted randomly from a certain production of manufacturing line for the past six months. With the preprocessing of production data, data collected from different manufacturing machines and production sites are transformed and integrated for data analysis. A total of 1657 records collected from forty-one production sites are used as experimental data to discover best combination of machines by using the sequential mining method.
Sequential mining method is used in this study and the manufacturing process is divided into five processes for the purpose of sequential data analysis. With the minimum support value of 30% and confidence value of 70%, the frequent 4-sequence sets in each process can be found above the yield rate of 95%. The result of discovering sequential pattern shows that key machines will affect the yield rate of LED production. The result also can provide process engineers with important information for dealing with extraordinary problems in the manufacturing machine.
|
---|