The Study of IC Process Fault Detection

• Main Authors: Guen-Wei Chen, 陳冠瑋
• Other Authors: Yao-Jen Chang
• Format: Others
• Language: zh-TW
• Published: 2002
• Online Access: http://ndltd.ncl.edu.tw/handle/73157627741160982386

碩士 === 大葉大學 === 自動化工程研究所 === 90 === The evolution of semiconductor manufacturing process on the enlargement of wafer size together with the shrink of feature size results in the difficulty of process control. In addition，faulty processes relatively increase. Tight process control then becomes an essential requirement in the fabs. Up to the present，SPC has been used as a tool for quality control. However，many process parameters exhibit correlated relationship and inevitable steady drift. Using SPC control charts sometimes leads to false alarms and erroneous judgments. Therefore，the major motivation of this research is to learn the characteristics of these process variations by using radial basis function (RBF) neural networks. Equipment malfunction and/or the faults can thus be detected and the false alarms can be avoided. Furthermore，the maintenance can be performed based on our provided diagnosis function in order to promote the overall equipment effectiveness. Radial basis function neural networks have the capability of parallel computation. The neural networks are trained by the input-output data so that the internal weights of networks can be obtained. The constructed non-linear models have characteristics of curve-fitting and mapping relations. RBF networks can provide generalizations with minimum structures. Therefore，they are applicable to the complicated systems, especially for the purposes of fault detection and classification.