Iterative learning predictive fault-tolerant control for nonlinear batch processes via dynamic output feedback approach

Abstract

For batch processes with unmeasurable states, previous iterative learning control methods compute the control gains offline, which leads to constant control law gains. Due to the lack of sufficient historical batch control input data in the initial operating cycles, the constant control law gains make the controller require more operating cycles to track the set value stably. In addition, the traditional iterative learning control method that integrate fault-tolerant control enhances system robustness by imposing upper and lower bounds on faults, which restricts the range of tolerable faults. To address these issues, an iterative learning predictive fault-tolerant control method based on dynamic output feedback is proposed. The robust positive definite invariant and terminal constraint sets are utilized to derive the stability conditions under faults, which suppress the effects of faults by ensuring that the system state is always within the safe region. By computing control gains in real time, the method can handle system variations during operation and reduce the impact of missing initial batch data, thus accelerating convergence across batches. Finally, we perform a simulation verification using a nonlinear batch reactor, and the results demonstrate the validity of the proposed method.