Abstract
Due to increases in the complexity of processes involved in semiconductor manufacturing, increasingly high inspection costs associated with defective wafers has become a critical concern of modern manufacturers. More importantly, because of high-dimensional wafer bin maps (WBMs), it is difficult to capture the variations of each dimension via traditional pattern recognition or classification methods. By contrast, this work proposes a novel two-phase morphology-based similarity search consisting of: 1) training sample generation based on the morphological method and 2) SVM categorization for test datasets according to the variant degrees of similarities. The morphology-based samples contain five kinds of features, including original morphology definitions in addition to our proposed features. The second phase, using SVM for similarity searches, extends the usage of pattern recognition to real applications in large sample dimensions. The preliminary results demonstrate the usefulness of our approach in the context of yield improvements in semiconductor manufacturing. © 2012 IEEE.