Predictive control considering model uncertainty for variation reduction in multistage assembly processes

Active control for dimensional variation reduction in multistage assembly processes (MAPs) is a challenging issue for quality assurance. It is desirable to implement a system-level control strategy to minimize the end-of-line product variance, which is propagated from upstream manufacturing stages. Research has been conducted to realize such objective, based on the variation propagation models derived from the nominal parameters of product and process design. However, due to the uncertainties induced by the significant changes of process parameters, such designated model will be different from that of the actual process, and will not precisely represent the actual physics of the process. This model discrepancy may lead to the performance deterioration of the controllers. This paper proposed a feed-forward MAP control strategy that explicitly takes into account the uncertainties of model coefficients. The case study demonstrates that, when the model uncertainties are significant, the controller derived from the proposed approach outperforms that derived without considering the model uncertainty.