Bagging support vector data description model for batch process monitoring

Abstract To improve the monitoring performance of the support vector data description model (SVDD), an ensemble form of SVDD is developed, which is termed as bagging SVDD in this paper. While different kinds of ensemble learning approaches have been developed in the past years, bagging is probably the most traditional and simplest one. By randomly selecting subsets from the original dataset, bagging constructs an individual SVDD model for each of these subsets. For practical utilization, the results of different individual SVDD models are ensembled/combined together. In this paper, two kinds of combination strategies are proposed, named as voting-based strategy and Bayesian-based strategy. Compared to a single SVDD model, the monitoring performance can be improved by the bagging SVDD method in most cases. The feasibility and effectiveness of the proposed method are demonstrated by an industrial semiconductor etch process.

[1]  Furong Gao,et al.  Improved two-dimensional dynamic batch process monitoring with support vector data description , 2011 .

[2]  John F. MacGregor,et al.  Multi-way partial least squares in monitoring batch processes , 1995 .

[3]  Zhiqiang Ge,et al.  Multimode process monitoring based on Bayesian method , 2009 .

[4]  Tao Chen,et al.  Multivariate statistical monitoring of two-dimensional dynamic batch processes utilizing non-Gaussian information , 2010 .

[5]  Robert P. W. Duin,et al.  Support Vector Data Description , 2004, Machine Learning.

[6]  Furong Gao,et al.  Batch process monitoring based on support vector data description method , 2011 .

[7]  J. Macgregor,et al.  Monitoring batch processes using multiway principal component analysis , 1994 .

[8]  Furong Gao,et al.  Review of Recent Research on Data-Based Process Monitoring , 2013 .

[9]  Jin Wang,et al.  Large-Scale Semiconductor Process Fault Detection Using a Fast Pattern Recognition-Based Method , 2010, IEEE Transactions on Semiconductor Manufacturing.

[10]  Zhiqiang Ge,et al.  Performance-driven ensemble learning ICA model for improved non-Gaussian process monitoring , 2013 .

[11]  Zhiqiang Ge,et al.  Improved kernel PCA-based monitoring approach for nonlinear processes , 2009 .

[12]  Lifeng Xi,et al.  A neural network ensemble-based model for on-line monitoring and diagnosis of out-of-control signals in multivariate manufacturing processes , 2009, Expert Syst. Appl..

[13]  Zhiqiang Ge,et al.  Nonlinear process monitoring based on linear subspace and Bayesian inference , 2010 .

[14]  Zhiqiang Ge,et al.  Semiconductor Manufacturing Process Monitoring Based on Adaptive Substatistical PCA , 2010, IEEE Transactions on Semiconductor Manufacturing.

[15]  Fuli Wang,et al.  On-line batch process monitoring using batch dynamic kernel principal component analysis , 2010 .

[16]  ChangKyoo Yoo,et al.  Fault detection of batch processes using multiway kernel principal component analysis , 2004, Comput. Chem. Eng..

[17]  Jie Yu,et al.  Nonlinear Bioprocess Monitoring Using Multiway Kernel Localized Fisher Discriminant Analysis , 2011 .

[18]  Julian Morris,et al.  Dynamic model-based batch process monitoring , 2008 .

[19]  Bin Wu,et al.  A neural network ensemble model for on-line monitoring of process mean and variance shifts in correlated processes , 2010, Expert Syst. Appl..

[20]  Barry M. Wise,et al.  A comparison of principal component analysis, multiway principal component analysis, trilinear decomposition and parallel factor analysis for fault detection in a semiconductor etch process , 1999 .