Neural hammerstein based MPC (NHMPC) for MTBE reactive distillation

Reactive distillation of MTBE has strong interaction among the variables and is highly nonlinear process, hence difficult to be controlled. In this work, Neural Hammerstein Based MPC (NHMPC) is proposed to tackle the nonlinearity and the variables´ interaction issue while controlling tray temperature of MTBE reactive distillation. The proposed NHMPC uses Neural-Hammerstein (N-H) model which is one of the advanced nonlinear block oriented model that is feasible to model highly nonlinear process with low computational time. The work has been simulated using Simulink (Matlab) that integrated with Aspen dynamic model. The sequential quadratic programming (SQP) is used as optimizer in the NHMPC proposed. The results show that NHMPC was able to control the tray temperatures in MTBE reactive distillation process very well.