Actuator fault-tolerant iterative learning control of nonlinear multi-phase batch processes

• A novel actuator fault-tolerant iterative learning control employing neural networks for multi-phase batch processes. • Convergence analysis of the proposed control scheme. • Injection molding batch process experimental study. • Comprehensive analysis of control system performance in the presence of repetitive and non-repetitive disturbances as well as actuator faults. The design of a neural-network-based fault-tolerant iterative learning control for nonlinear, multi-phase batch processes is addressed. Recurrent neural networks are used to model each phase of a batch process. The learning controller is developed using iterative procedures to solve an optimal control problem. Sufficient conditions are provided for the convergence of the tracking error and the control signal. The efficiency of the developed control scheme is illustrated by a non-trivial example of tracking control for a multi-phase injection molding process affected by repetitive and non-repetitive disturbances and/or actuator faults.