عنوان مقاله [English]
In this paper, a new methodology is proposed to detect the magnitudes of shift in the mean of some correlated quality characteristics in multivariate-multistage production processes. The proposed methodology is able to consider not only the cascade property of the production stages, but also, the inherent correlation structure of the variables in one stage. It is also able to visualize the movements of the process mean during monitoring time. The proposed method consists of two parts; a self-organizing map (SOM) neural network to detect the mean shifts and a fuzzy inference system to judge the exact magnitudes of the mean shifts. In-control average run lengths, out-of-control average run lengths, and average run lengths to detect the correct class of the proposed method, are investigated by a simulation study of a 2-variate 3-stage process. The results of the simulation are encouraging.