Residual-based fault detection isolation and recovery of a greenhouse

This paper is concerned with fault detection, isolation and recovery (FDIR) of the greenhouse whose temperature is regulated by the model predictive controller (MPC). The residual generation approach is adopted for fault detection and isolation. The new considerations in the proposed FDIR approach are the residual generation for actuator faults, regulation failure detection as the indication of inappropriate regulation by the controller, below threshold actuator fault detection strategy, and recovery operation updating model used by MPC once the FDIR isolates the actuator fault. The proposed control strategy FDIR with MPC was compared with fixed model information MPC for simulated scenarios of the actuator fault. It has been shown that FDIR successfully detects, isolates and updates model information with low computation burden for non-delayed control evaluation.