Model Predictive Control Method with Variable Structure to Input-output Faults

Input actuator faults and output measurement device faults always lead to poor control performance of model predictive control (MPC). By analysis of the steady-state relationship between input and output, we propose an improved MPC method with variable structure, which removes the faulty input or output variables from the controller. Some high-priority output set points need to be recalculated to meet the control requirements, due to the reduction of input degree caused by the variable structure. The output fault variable structure control used the dynamic matrix control (DMC) algorithm, which ensures integrated input steady-state targets and prevents the system from being influenced by output sensor faults. Simulation results of Shell heavy oil fractionator benchmarks validated the effectiveness of the proposed method.