Semiconductor Engineering Data Analysis System Based On Data Mining

Wafer manufacture is the most time-consuming and complicated phase, with the heaviest investment, of semiconductor manufacture. The production cycle of one wafer passes about hundreds of process step and have the dozens return character of production. This results in that the yield of the finished wafer has great uncertainty.Today, with the increasingly developed computer technology, to improve the yield of production, the data has been collected as much as possible during the wafer manufacture process. As there’s a bound of data and dimensions, which disperse in the data bases of sub systems, data lacks integrity and consistency. On the other hand, it’s hard to get the result, which has guiding significance, from the data by the traditional statistical methods. This causes the low analysis efficiency. The opportunity of problem finding and solving is missed.Meanwhile, data mining technology has been widely used in a lot of engineering area. The target is to find the useful and unknown relation and knowledge for the data owner. This improves the analysis efficiency and accuracy to some extent. So it’s imperative to use data mining technology in semiconductor manufacture process.For this reason, the thesis analyzes and summarizes the application of data mining technology in semiconductor engineering. Design and implement one semiconductor engineering data analysis platform base on data mining. This platform is built on the data mart of semiconductor engineering. The target of data mart is to prepare the data for the effectively use of data mining technology. The platform consists of three application systems. Monitor and Alarm system, which improves the automatic manufacture efficiency through monitoring the data automatically and trigger the necessary alarm; Interactive Data Analysis System, in which system, engineer can use ANOVA analysis, correlation analysis and some other algorithm of data mining to analyze data and find the root cause of yield loss; Data Spec Auto Calculation System, which helps to improve the setting of data spec scientifically.