Wafer defect patterns recognition based on OPTICS and multi-label classification

In the industry of integrated circuits, defect patterns shown on a wafer map contain crucial information for quality engineers to find the cause of defect to increase yield. This paper proposes a method for wafer defect pattern recognition which could recognize more than one defect patterns based on Ordering Point to Identify the Cluster Structure(OPTICS) and Support Vector Machine(SVM). The effectiveness of the proposed method has been verified from following three aspects from a real-world data set of wafer maps(WM-811K): salient defect pattern recognition accuracy up to 94.3% and the accuracy of some types has an obvious improvement, multi-patterns recognition accuracy(82.0%), and computation time has a significantly reduction.