Developing Prediction Model for Wafer Acceptance Test — For Capacitance

• Main Authors: Hsien-Wen Huang, 黃賢文
• Other Authors: 鄭春生
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/77012654885112229259

碩士 === 元智大學 === 工業工程與管理學系 === 97 === Wafer acceptance test (WAT) results are the basis of shipping wafers to foundry customers. The fundamental parameters of WAT, such as capacitance, voltage, resistance …etc., are employed to verify IC’s function. The purpose of WAT is to response the wafer production status by testing the electrical parameters and to avoid low yield. This study attempts to develop a Back-Propagation Network (BPN) prediction model for WAT. Real equipment data, wafer process measurement data and WAT capacitance data are employed to verify our proposed prediction model. Two kinds of input variables, completeness and simplified by stepwise, are considered in this research. The traditional regression analysis is used as a benchmark for comparison with BPN. The mean absolute percent error (MAPE) is used as the primary performance measure in this research. A comparative study shows that the MAPE values of four prediction models are less than 2 %. The stepwise variables selection of BPN has the best overall performance.