摘要

There are always more than one quality responses or characteristics in product or process design and engineers often must choose optimum parameter combination for these responses simultaneously. Taguchi method, as a cost-effective method for off-line quality control, especially the quadratic loss function shows loss of quality directly, has been widely employed by quality engineering to deal with single-response problem, but few of them pay attention to multi-response problem. TOPSIS (Technique for order preference by similarity to ideal solution) method is a popular used multiple attribute decision making (MADM) methodology. In conventional TOPSIS method, responses are usually transformed into loss of quality, and then normalized by the largest quality loss. However, for nominal-the-best (NTB) problem, it's not a reasonable methodology. Based on conventional TOPSIS, the proposed method makes an improvement to adapt to NTB type response, it also takes the standard deviation of responses into consideration simultaneously for the purpose of robustness. The improved TOPSIS method is illustrated by an example from literatures, and the result indicates its reasonableness and effectiveness.