Similarity matching of wafer bin maps for manufacturing intelligence to empower Industry 3.5 for semiconductor manufacturing

Abstract Yield improvement is increasingly important as advanced fabrication technologies are complicated and interrelated for semiconductor manufacturing. Wafer bin maps (WBM) present specific failure patterns which provide crucial information to track the process excursions to empower intelligent manufacturing for wafer fabrication. In practice, WBM identification is still subjective relied on domain knowledge and human-eye. As the semiconductor industry continuously migrates for advanced nano technologies, many rare defect patterns are also generated by different pattern, pattern size, noise degree, pattern density, pattern shift, and wafer rotation. Existing studies regarding WBM focus on classification and lack of capability to detect a rare pattern. In order to overcome the shortage of WBM classification, the similar WBMs provide useful information of WBM identification. Following Industry 3.5 as a hybrid strategy between Industry 3.0 and to-be Industry 4.0, this study aims to develop a novel approach to measure the similarity of defect patterns of WBMs to enhance decision quality for fault detection and defect diagnosis effectively and efficiently. In particular, the proposed approach applied a mountain clustering algorithm to enhance the defect features depending on clustering density. Then, Weighted Modified Hausdorff Distance (WMHD) is employed to measure the similarity level. Furthermore, a decision support system embedded the developed algorithms is constructed. An empirical study of WBM clustering was conducted in a fab for validation. The results have shown practical viability of the proposed approach.