LAUSR

Abstract In this work, we present a data-driven methodology to overcome actuator faults using a model-based feedback controller that optimizes process economics termed economic model predictive control (EMPC). Specifically, we utilize a moving horizon error detector that quantifies prediction errors and triggers updating the empirical model used for state predictions in the EMPC on-line using the most recent input/output data collected after the fault when significant prediction errors occur due to the loss of an actuator. The proposed approach is applied to a catalytic chemical reactor example where an actuator fault occurs, affecting the coolant temperature. The proposed scheme was able to reduce prediction errors caused by the actuator loss by replacing the model within the EMPC with a more accurate model, resulting in improved economic performance compared to not updating the model.