利用資料探勘技術制定封測廠的機台保養維修計畫

A study for developing the machine maintenance plan of IC packaging and testing plant by used of data mining technology

10.6495/CIMBM.2015.016

林雅祺(Ya-Chi Lin)；王政鴻(Cheng-hung Wang)；蔡勝閎(Sheng-Hong Tsai)；蕭淑君(Shu-Jun Xiao)；彭定國(Ting-Kuo Peng)

資料探勘 ； 關聯法則 ； 機台篩選 ； Data Mining ； Association Rule ； Machines Selection

科際整合管理研討會

2015第18屆（2015 / 06 / 07）

230 - 240

繁體中文

在眾多的半導體產品測試作業下，除了產出客戶所需要的產品外，也產出了許多有關封裝測試的品質資料。由於半導體產業所使用的機台成本高，不可能輕易置換，使得廠商非常看重機台的保養與維修。傳統上，廠商要制定機台的保養與維修計劃時，受限於人員素質以及時間急迫等因素，大都只看機台的測試數據以及機台稼動率等數值，以量化的方式來找出有問題的機台。然而，半導體測試作業所牽涉到的機台多，機台之間也存在某些不易察覺的關聯性，如果只以測試數據來找出有問題的機台，只能改善到個別獨立的機台，而對彼此影響的機台卻成了漏網之魚，在提升測試效率上也會出現瓶頸。本研究以半導體封裝測試廠商的測試資料為主體，進行機台間潛在關聯法則的探勘。藉由資料探勘技術所得到的關聯法則，考慮到存在於機台間的相依性，彌補原先獨立有問題的機台保養維修的不足，使廠商能制定更完善的保養維修計畫，以提昇機台測試品質。

There are many testing information resulted from semiconductor product testing operations. Due to the high cost of the machine used in the semiconductor industry, many companies focus on developing a very high-value machine maintenance and repair plan instead of replacement. Traditionally, subject to the quality of personnel and time urgency factors, most decision makers build the machine maintenance and repair plan based on the test data and the machine utilization rate. However, the semiconductor test job involves a lot of equipments. There exists some imperceptible relevance between the machines. Only the individual machine can be improved independently if only considering the test data of individual machine. The mutual influence of the machine has been ignored to lower the efficiency of the machine maintenance and repair plan. This paper proposes an association rule mining approach for semiconductor packaging and testing manufacturers to find the main potentially relevant between machines. By used of the association rule data mining techniques, dependencies existing between machines will be taken into account so that manufacturers can develop a more comprehensive maintenance plan to enhance the test quality.