关键词: 大数据, 制造过程数据, 半导体制造, 分布式处理, Hadoop

Abstract: To analyze,predict and control semiconductor manufacturing process with the help of large amounts of data and its inherent value,a big data processing platform was designed and implemented.The characteristics of big data in semiconductor manufacturing were summarized,and a 3-layer framework was built,whose key techniques in the framework were discussed.According to the utilization features of big data in semiconductor manufacturing,the common operations of original data records such as import,deduplicate,filter and standardize were classified into the general pre-processing layer of platform.The pre-processing data records were grouped by topic in the middle layer named data warehouse layer.According to specific requirement of upper application,the top layer named special data processing algorithm layer provided reliable and reusable data resource by extracting grouped data.A prototype system was built through the actual situation of semiconductor manufacturing corporation,and the real semiconductor manufacturing data was imported,stored,organized,pretreated and analyzed.Results showed that the proposed platform was a high-performance data analysis system.

Key words: big data, manufacturing process data, semiconductor manufacturing, distributed processing, Hadoop