刘元修, 史峥, 张培勇. 一种基于随机森林的可寻址WAT良率诊断方法[J]. 微电子学与计算机, 2019, 36(9): 94-98.
引用本文: 刘元修, 史峥, 张培勇. 一种基于随机森林的可寻址WAT良率诊断方法[J]. 微电子学与计算机, 2019, 36(9): 94-98.
LIU Yuan-xiu, SHI Zheng, ZHANG Pei-yong. A yield diagnosis method basedon random forest for addressable WAT[J]. Microelectronics & Computer, 2019, 36(9): 94-98.
Citation: LIU Yuan-xiu, SHI Zheng, ZHANG Pei-yong. A yield diagnosis method basedon random forest for addressable WAT[J]. Microelectronics & Computer, 2019, 36(9): 94-98.

一种基于随机森林的可寻址WAT良率诊断方法

A yield diagnosis method basedon random forest for addressable WAT

  • 摘要: WAT (Wafer Acceptance Test, 晶圆允收测试)是晶圆在完成制造后必须通过的一项电学测试.本文提出了一种基于随机森林的良率诊断方法, 使用随机森林算法对来自可寻址WAT的测试数据建立分类模型, 并从中提取关键规则.提取出的规则集合可以帮助分析人员快速、准确地对生产中导致低良率的根本原因进行定位、分析, 对于良率提升有着重要意义.在真实可寻址WAT测试数据集上应用本文提出的方法, 得到的规则集合在分类性能上较基于决策树的方法有着更好的表现.

     

    Abstract: WAT(Wafer Acceptance Test) is an electrical test that a wafer must pass upon completion of fabrication. An add ressable WATyield diagnosis method is proposed in this paper, which uses the random forest algorithm to establish a classification model on test data and extractskey rules.The extracted rules can help analysts quickly and accurately locate the rootcause of low yield in production, which plays a key role in yield enhancement. The proposed method hasbeen experimented on dataset from arealaddressable WAT, and the obtained rule set has better classification performance than the rule set obtained by method based on decision tree.

     

/

返回文章
返回