Design of DOE-Based Automatic Process Controller With Consideration of Model and Observation Uncertainties

Robust parameter design (RPD) has been widely used as a cost-effective tool in quality control to reduce variability, in which the controllable factors are set to minimize the variability of response variables due to noise factors, assuming their distributions are known. It is essentially an offline tool without considering that some noise factors can be measured online. Recently, the concept of design of experiment (DOE)-based automatic process control (APC) has been proposed for online process control based on regression models obtained from DOE and with consideration of the online measurement of noise factors. The existing literature investigates the DOE-based APC with assumption that both regression models and the online noise measurement are precisely known, which limits the applicability of the technique. This paper develops the DOE-based APC scheme that considers both the observation and the modeling uncertainties. The controller is implemented under two APC strategies, i.e., cautious control strategy and certainty equivalence control strategy. The comparison among online APC and robust design approaches demonstrates that automatic controller with consideration of both uncertainties can achieve better process performance than conventional design, and is more stable than normal DOE-based APC controllers. The proposed approach is illustrated using an industrial process.