A dynamic prognosis algorithm in distributed fault tolerant model predictive control

This paper presents a dynamic prognosis algorithm in distributed fault tolerant model predictive control (DFTMPC). The dynamic prognosis, which means predicting the trajectories of process variables under distributed model predictive control, is performed when a fault is diagnosed and several candidate reconfigured controls are proposed. Then, the dynamic prognosis is utilized to check whether the candidate reconfigured controls are able to drive the system to the new operating conditions and to evaluate the performance during the transition period. Thus, the most suitable candidate reconfigured controller is selected and its feasibility is ensured without using a Lyapunov function that is difficult to obtain for large-scale systems. On the other hand, the on-line computation burden of the prognosis algorithm is moderate under the assumption that the sets of active constraints in non-faulty subsystems remain the same as they are at the nominal operating conditions. Thus, the dynamic prognosis for DMPC is aimed to improve the applicability of the existing fault tolerant methods to large-scale systems.