1. Predicting BEOL Key Qualities by Mahalanobis-Taguchi System – An Example of Taiwan’s Semiconductor
- Author
-
Hsiao Lin Teng and Shu Yu Lin
- Subjects
Mahalanobis distance ,Engineering ,business.industry ,Process (computing) ,General Medicine ,Process variable ,Work in process ,Machine learning ,computer.software_genre ,Set (abstract data type) ,Taguchi methods ,Artificial intelligence ,business ,Multidimensional systems ,Quality assurance ,computer - Abstract
Taguchi Gen'ichi introduced Mahalanobis-Taguchi System (MTS) which is in combination with the concepts of quality engineering and Mahalanobis Distance (MD). The MTS is proposed as diagnosis and forecasting method using multivariate measurement scale with its intention to help policy maker as basis for decision making. This study applies MTS approach in a manufacturing process to reduce a set of parameters, at the same time there will be a pattern, which can forecast and identify important parameters, constructed by MTS method. Through this pattern can minimize unimportant inspection in process and save unnecessary time and cost. The primary goal to structure a measuring scale which makes accurate forecasting in multidimensional system. The case study in this paper reviews the planarity of back-end process in 8-inch silicon wafers on the purpose to construct a pattern of reduced set of parameters. In this paper, using thirty-two current variables as reference space and furthermore reducing the variables to seven parameters in order to predict defective items. As a result, it has still good discriminant accuracy. If validation of the reduced-set parameters is reliable with its good discriminant accuracy, it means that the company in this case study can built defective items warning of the pattern parameters in back-end process because this approach of selecting parameters is feasible.
- Published
- 2014