Double‐Layer Learning Control With Iterative Disturbance Observer for Unknown Nonlinear Batch Manufacturing Systems Subject to Nonrepetitive Disturbance

ABSTRACT In this article, a double‐layer learning control (DLLC) scheme with iterative disturbance observer (IDO) is proposed for unknown nonlinear batch manufacturing systems subject to nonrepetitive disturbance, by only leveraging the historical batch data of system input and output. The proposed scheme has two layers, where the inner layer is a simple proportional‐type iterative learning control (ILC) structure to directly update the control input based on the previous batch data for output tracking, and the outer layer is an IDO based high‐order set‐point learning control (IDO‐HOSPLC) structure to further expedite the convergence speed of tracking error and suppress the adverse effect of nonrepetitive disturbances. The convergence and boundedness of the resulting double‐layer learning system is clarified by the contraction mapping and mathematical induction principles. An illustrative steam‐water heat exchanger system is used to validate the effectiveness and superiority of the proposed DLLC scheme.