International Journal of Industrial Engineering & Production Management

 There are multivariate processes in which two or more quality characteristics must be controlled simultaneously. In controlling such processes, two goals must be achieved. The first one is to identify an out of control situation and the second is to determine the quality features caused the out of control signal. In this paper, both goals are investigated. In addition to the current methods used to diagnose an out of control situation, for the purpose of making the Hotelling T ² more sensitive, the warning limits for T ² are also defined. In determining the quality characteristics caused the out of control situation, current methods are investigated and a new procedure is suggested. Not only the new approach does not have some of the deficiencies with the current methods, but also its application is much simpler in practice. The results of simulation using the warning limits for very small shifts in process mean vector indicate that in 81% of the time, shifts are being detected. While in similar conditions, when the regular T ² is used, only 19% of the time true signals are observed. In comparison with similar techniques, use of new procedure in detecting the quality characteristics responsible for an out of control situation identifies the shifted quality features in 76% of the time. While in existing methods at most in 64% of the time the shifted quality characteristics are detected.