On Phase II SPC in Cases When Normality is Invalid

Conventional statistical process control (SPC) charts require the normality assumption on the process response distribution. In reality, this assumption is often invalid. In such cases, it has been well demonstrated in the literature that results from control charts using the normality assumption may not be reliable in the sense that their actual false alarm rates could be substantially larger or smaller than the assumed false alarm rate. In this paper, we explore one natural solution to the phase II SPC problem in cases when the normality assumption is invalid, which tries to define a transformation based on an IC dataset so that the transformed process response distribution is close to normal and thus the conventional SPC charts can be applied to the transformed phase II data. This approach is compared with several alternative approaches in the literature, and some practical guidelines are provided regarding the use of all relevant control charts.