Robust control settings for repeated manufacturing processes

The product quality in manufacturing processes is dependent on various process parameters (called control settings) usually set by an human overseer at the beginning of each process run. However quality is often adversely influenced by several additional “noise” parameters which cannot be controlled and hence can take random values for each process run. The problem of computing the optimal sequence of control settings for an infinite series of such manufacturing runs, such that the product quality is maintained for all runs and for all possible values of the noise parameter, is addressed in this article. The dependence of product quality on control and noise parameters is modeled using the response surface methodology. Subsequently, the concept of mixed strategy equilibrium is borrowed from the theory of games to compute a probability distribution which optimally determines the control settings for each process run. These choices of control settings maximally improves product quality over all possible values of the noise parameters for all runs of the process.