数据驱动的晶圆图缺陷模式识别方法
|
摘要
针对晶圆生产过程中晶圆图数据角度与维度多样性和数量不平衡的特点,提出了基于对抗生成网络的晶圆图缺陷模式识别方法。设计了Radon变换,实现了数据多角度的统一;采用重采样机制实现数据多维度的缩放,得到了标准晶圆缺陷数据。提出了基于对抗生成网络的晶圆缺陷分类方法,利用生成机制平衡各缺陷模式的样本数量,以提升缺陷模式识别精度。试验结果表明,该方法可大幅提升少类样本的识别精度,且在整体识别率上远优于支持向量机和Adaboost算法。
Aiming at the characteristics of wafer map data angle and dimension diversity and quantity imbalance during wafer production processes,a wafer pattern defect recognition method was proposed based on generative adversarial networks.The two-stage wafer defect data pre-processing method was proposed to obtain standard wafer defect data,where Radon transform was designed to solve the multi-angle characteristics of the wafer map,and a resampling mechanism was used to realize the scaling of various data dimensions.The proposed wafer defect classification method used ageneration mechanism to balance the number of samples of each defect type based on a generative adversarial networks,which could improve the defect pattern recognition accuracy.The experimental results show that this method may greatly improve the accuracy of small class samples,and the overall recognition rate is much better than the support vector machine and Adaboost algorithm.
Aiming at the characteristics of wafer map data angle and dimension diversity and quantity imbalance during wafer production processes,a wafer pattern defect recognition method was proposed based on generative adversarial networks.The two-stage wafer defect data pre-processing method was proposed to obtain standard wafer defect data,where Radon transform was designed to solve the multi-angle characteristics of the wafer map,and a resampling mechanism was used to realize the scaling of various data dimensions.The proposed wafer defect classification method used ageneration mechanism to balance the number of samples of each defect type based on a generative adversarial networks,which could improve the defect pattern recognition accuracy.The experimental results show that this method may greatly improve the accuracy of small class samples,and the overall recognition rate is much better than the support vector machine and Adaboost algorithm.
引文
[1]汪俊亮,秦威,张洁.基于数据挖掘的晶圆制造交货期预测方法[J].中国机械工程,2016,27(1):105-108.WANG Junliang,QIN Wei,ZHANG Jie.Data Mining for Orders’LT Forecasting in Wafer Fabrication[J].China Mechanical Engineering,2016,27(1):105-108.
[2] HANSEN M H,NAIR V J,FRIEDMAN D J.Monitoring Wafer Map Data from Integrated Circuit Fabrication Process for Spatially Clustered Defects[J].Technometrics,1997,39(3):241-253.
[3] KAEMPF U.The Binomial Test:a Simple Tool to Identify Process Problems[J].IEEE Transactions on Semiconductor Manufacturing,1995,8(2):160-166.
[4] FRIEDMAN D J,HANSEN M H,NAIR V N,et al.Model-free Estimation of Defect Clustering in Integrated Circuit Fabrication[J].IEEE Transactions on Semiconductor Manufacturing,1997,10(3):344-359.
[5] HARALICK R M,SHAPIROS L G.Computer and Robot Vision[M].Boston:Addison-Wesley,1993:28-48.
[6] TAN Y S,ZHOU X M.Research on the Statistical Characteristics of the Unipixel Gray Value of Natural Images[J].Computer Engineering&Science,2004,26(3):42-46.
[7] GONZALEZ R C,WOODS R E.Digital Image Processing[M].3rd ed.Harlow:Prentice-Hall,2008.
[8]WU M J,JANG J S R,CHEN J L.Wafer Map Failure Pattern Recognition and Similarity Ranking for Large-scale Data Sets[J].IEEE Transactions on Semiconductor Manufacturing,2015,28(1):1-12.
[9]邡鑫,史峥.晶圆缺陷检测与分类的卷积神经网络[J/OL].计算机工程:1-5[2018-05-31].http://kns.cnki.net/kcms/detail/31.1289.TP.20170907.1450.006.html.FANG Xin,SHI Jun.Convolutional Neural Networks for Wafer Defect Detection and Classification[J/OL].Computer Engineering:1-5[2018-05-31].http://kns.cnki.net/kcms/detail/31.1289.TP.20170907.1450.006.html.
[10] GOODFELLOW I J,POUGET-ABADIE J,MIRZA M,et al.Generative Adversarial Networks[J].Advances in Neural Information Processing Systems,2014,3:2672-2680.
[11] WANG C H.Recognition of Semiconductor Defect Patterns Using Spectral Clustering[J].Expert Systems with Applications,2008,34(3):1914-1923.
[12] KEYS R.Cubin Convolution Interpolation for Digital Image Processing[J].IEEE Transactions on Acoustics,Speech,and Signal Processing,1981,29,(6):1153-1159.
[13] SPRINGENBERG J T.Unsupervised and Semi-Supervised Learning with Categorical Generative Adversarial Networks[C]//International Conference of Learning Representation.San Juan,2015:1-20.
[2] HANSEN M H,NAIR V J,FRIEDMAN D J.Monitoring Wafer Map Data from Integrated Circuit Fabrication Process for Spatially Clustered Defects[J].Technometrics,1997,39(3):241-253.
[3] KAEMPF U.The Binomial Test:a Simple Tool to Identify Process Problems[J].IEEE Transactions on Semiconductor Manufacturing,1995,8(2):160-166.
[4] FRIEDMAN D J,HANSEN M H,NAIR V N,et al.Model-free Estimation of Defect Clustering in Integrated Circuit Fabrication[J].IEEE Transactions on Semiconductor Manufacturing,1997,10(3):344-359.
[5] HARALICK R M,SHAPIROS L G.Computer and Robot Vision[M].Boston:Addison-Wesley,1993:28-48.
[6] TAN Y S,ZHOU X M.Research on the Statistical Characteristics of the Unipixel Gray Value of Natural Images[J].Computer Engineering&Science,2004,26(3):42-46.
[7] GONZALEZ R C,WOODS R E.Digital Image Processing[M].3rd ed.Harlow:Prentice-Hall,2008.
[8]WU M J,JANG J S R,CHEN J L.Wafer Map Failure Pattern Recognition and Similarity Ranking for Large-scale Data Sets[J].IEEE Transactions on Semiconductor Manufacturing,2015,28(1):1-12.
[9]邡鑫,史峥.晶圆缺陷检测与分类的卷积神经网络[J/OL].计算机工程:1-5[2018-05-31].http://kns.cnki.net/kcms/detail/31.1289.TP.20170907.1450.006.html.FANG Xin,SHI Jun.Convolutional Neural Networks for Wafer Defect Detection and Classification[J/OL].Computer Engineering:1-5[2018-05-31].http://kns.cnki.net/kcms/detail/31.1289.TP.20170907.1450.006.html.
[10] GOODFELLOW I J,POUGET-ABADIE J,MIRZA M,et al.Generative Adversarial Networks[J].Advances in Neural Information Processing Systems,2014,3:2672-2680.
[11] WANG C H.Recognition of Semiconductor Defect Patterns Using Spectral Clustering[J].Expert Systems with Applications,2008,34(3):1914-1923.
[12] KEYS R.Cubin Convolution Interpolation for Digital Image Processing[J].IEEE Transactions on Acoustics,Speech,and Signal Processing,1981,29,(6):1153-1159.
[13] SPRINGENBERG J T.Unsupervised and Semi-Supervised Learning with Categorical Generative Adversarial Networks[C]//International Conference of Learning Representation.San Juan,2015:1-20.