Data-driven EMPC with contractive constraint for multi-mode flexible production

This work proposes a decision-making strategy leveraging the data-driven economic model predictive control (EMPC) for multi-mode flexible production. The flexible production process is first modeled as an unknown nonlinear system and subsequently transformed into an equivalent linear representation using the dynamic linearization method. A data-driven state observer and parameter estimator are jointly designed to estimate the system states and model parameters. Then, a composite optimization objective involving energy consumption and control performance is integrated into the EMPC framework. Notably, a contraction-stability constraint is constructed leveraging an auxiliary controller, which enforces the data-driven EMPC to drive the system states into the stable range. Finally, through a thread-tapping process, results indicate that the energy consumption of EMPC is lower than that of model-free adaptive predictive control in both noisy and non-noisy scenarios. A comparative and analytical study has also been carried out on the selection of the optimization horizon, aiming to identify the appropriate optimization horizon of the production process.