Optimization-Based Control Strategy with Wavelet Network Input–Output Linearizing Constraint for an Ill-Conditioned High-Purity Distillation Column

A new nonlinear optimization control strategy is developed for multivariable control of an ill-conditioned, high-purity distillation column. A high-gain directional effect resulting from the ill-conditioned nature of the system causes difficulty in controllability and requires a higher performance control system. The developed optimal controller applies a minimization of energy consumption as the optimal objective function to treat the ill-conditioning effect, while wavelet neural network input/output linearizing constraints force the outputs to reach the desired set points. In this paper, ethylene dichloride purification is used as a case study. The process dynamics are evaluated based on relevant thermodynamic properties in Aspen Plus Dynamics and are controlled by the proposed controller in the MATLAB/Simulink platform. Control performances are investigated in this cosimulation environment for set point tracking and regulatory problems. The simulation results demonstrate that robust tracking is attained, while compensation of the input disturbances is effectively improved compared with a model predictive controller.