| Summary: | 碩士 === 銘傳大學 === 管理科學研究所 === 92 === This research is to find out the major factors which influence the yield, then forecast yield. Furthermore, we want to find out the rule to stipulate specification from the yield by three methods: back propagation neural network, regression analysis and decision tree. The cost structure of industry that IC designs of Taiwan, 70% of the cost is used in the chip to make, so reduce the manufacturing cost to turn into a very important problem of IC designing industry. But, the commissioner knowing how to make one is very scarce, general small-scale industry who designs IC person can only is it give parameter value to come according to experience or historical data; therefore this research is tried to find out the relation between input parameter and outputting yield, dig out valuable information and probe into , and then go to control the parameter value of the products, in order to improve the yield, control quality, achieve the goal of lowering costs finally. This research is according to the 92nd year of the Republic of China collected wafer accept test data, construct three kinds of yield predict model by factor analysis combine back propagation neural network, regression analysis and decision tree. So this research is the first application to use back propagation neural network, regression analysis and decision tree into the yield predict model of wafer accept test parameter of IC designs and compare the three models. These simulate methods could save the time and cost of manufacture and test, those are this research contribution.
|
|---|
