China Mechanical Engineering ›› 2016, Vol. 27 ›› Issue (01): 105-108.

Previous Articles     Next Articles

Data Mining for Orders' LT Forecasting in Wafer Fabrication

Wang Junliang;Qin Wei;Zhang Jie   

  1. Shanghai Jiao Tong University,Shanghai,200240
  • Online:2016-01-10 Published:2016-01-08

基于数据挖掘的晶圆制造交货期预测方法

汪俊亮;秦威;张洁   

  1. 上海交通大学,上海,200240
  • 基金资助:
    国家自然科学基金资助重点项目(51435009)

Abstract:

The accurate prediction of LT plays an important role to help semiconductor manufacturers keep the promises of an accurate and steady delivery-time. However, the large production scale, and long cycle time significantly substantiated the complexity of such a problem. Based on large amounts of manufacturing data, a regression-based model which took account of thousands of parameters was proposed to obtain the correlation among 1669 manufacturing variables and LT. To select “LT-related” variables which had high mean Z-transformed correlations, the Fisher Z-transformation was applied, and the case-based reasoning method was designed to forecast orders' LT accurately.

Key words: wafer fabrication, data mining, lead-time (LT) forecasting, case-based reasoning

摘要:

晶圆订单的交货期预测对于保证订单交付的准时性和平顺性,具有重要的意义。然而,晶圆制造中的在制品数量多、生产周期长的特点加剧了交货期预测的复杂性。基于海量晶圆制造数据,设计回归模型来对1669个晶圆加工过程参数与订单交货期指标之间的关联关系进行分析,并采用费舍尔Z变换筛选得到强相关变量,对所得到的强相关变量采用案例推理方法实现了晶圆制造订单交货期的精准预测。

关键词: 晶圆制造, 数据挖掘, 交货期预测, 案例推理

CLC Number: