Research On Clustering Method On Workpiece For Semiconductor Wafer Fabrication

Since twenty-first Century, our country vigorously supports the manufacturing industry, equipment manufacturing industry, and the semiconductor manufacturing industry, as the continuous development of science and technology, semiconductor manufacturing equipment upgrading fast and the product demand increases unceasingly, the product quality requirements continue to rise. Semiconductor manufacturing system has become a hotspot by scientists all around the world.Semiconductor manufacturing system is with multiple, complex process, various constraints, uncertainty, multiple target feature. There are important characteristic indexes of semiconductor wafer fabrication, they represent the attributes, and the indexes of the workpiece. How to dig out useful information in these indicators in order to prove the semiconductor wafer fabrication is a focus by scientist all over the world. In many data mining algorithm, clustering algorithm is a common algorithm respectively, as the background feature of workpiece, it classify index into a scheduling example. In this paper, as the background of the semiconductor wafer fabrication, we focus on the study of fuzzy clustering algorithm and its application in the semiconductor wafer fabrication. In this paper, the main research contents are as follows:(1) As the background of semiconductor wafer fabrication, the FCM clustering algorithm is provided, the FCM algorithm is a theoretical foundation of this paper. Through the simulation experiment, this method has a good clustering ability, and we put it on the semiconductor wafer fabrication clustering by using the characteristic indexes of the workpiece and make a cluster analysis.(2) Through the research of FCM algorithm, we found that the initial clustering center is determined randomly, then the SUB-FCM algorithm is put forward, had better effect in the determination of the initial cluster center. By the simulation experiment, we found that this method is better in the accuracy and speed of FCM algorithm, and has a very good application in semiconductor wafer fabrication workpiece clustering.(3) Through the study of FCM algorithm and the characteristics of semiconductor wafer fabrication, we find most of semiconductor wafer fabrication is uncertainty, so there are some points out of structure in the semiconductor wafer fabrication. The abnormal points will interfere ordinary FCM cluster, in the case the Type-2FCM algorithm is proposed, against the abnormal structure points, it has a strong robustness. By the simulation experiment and the characteristic index clustering analysis, it has been verified.