Abstract
Process engineers at an automobile assembly plant wanted to develop a model for on-line process control in an assembly line. During an experiment to estimate the model, an uncontrollable variable was discovered to have a significant effect, but its values in the experiment were not representative of values in actual production. Through this case study, we present design criteria and assess sampling strategies for augmenting a factorial experiment to incorporate an additional factor that was found to be an uncontrollable independent normally distributed random variable. Our results demonstrate the potential improvement in on-line process control when experiments are augmented using these design criteria.