An intelligent multi-response off-line quality control for semiconductor manufacturing

A Taguchi-based off-line quality control method is a cost-effective quality-improvement technique that uses experimental-design methods for efficient characterization of a product or process, combined with a statistical analysis of its variability, with the ultimate purpose of its minimization, so that more stable and higher quality products can be obtained. The paper presents a generalized performance statistic which is a summation of the expected quality losses of all responses. Since the units and orders of these responses may be quite different and not reasonable for process design, a normalization procedure called the performance characteristic transformation method (PCTM) is proposed to overcome these difficulties. In addition, a corresponded generalized signal-to-noise (S/N) ratio may be constructed based on the above concepts and would consider the importance of each response simultaneously. Therefore, the optimal multi-response process can be obtained by the traditional Taguchi procedure with respect to the new proposed S/N ratio. An experiment of the plasma-etching process is conducted to verify the performance of the new multiresponse process optimization technique.<>