Wafer Map Defect Recognition Based on Improved 2DPCA Method

Abstract

The effective recognition of the wafer map defect patterns is of vital importance to ensure the high product quality the wafer manufacturing process. The present Two- Dimensional Principal Component Analysis (2DPCA) based wafer map defect recognition methods only consider the overall information of defect maps, and ignore the hidden between-class information. In order to solve this problem, this paper proposes an improved 2DPCA called Weighted-Class 2DPCA (WC2DPCA) method to recognize the defect pattern of the wafer maps. The WC2DPCA method considers the information of the classes in the feature subspace of 2DPCA, and designs the weights for representing the relationship between the classes in the feature subspace. The wafer maps projections on the feature subspace not only have common information in the same patterns of wafer maps defects, but also have the discriminant information between different wafer maps defect patterns. Experimental results on the benchmark WM-811K data set demonstrate the effectiveness of the proposed WC2DPCA method in wafer map defect recognition.