An experimental investigation of batch distillation column control

The development of an inferential soft sensor for a pilot-plant distillation column separating an ethanol-water mixture using neural network (NN) models has been investigated in this work. Inferential sensors are increasingly used in the process industries to infer the value of the main quality variable while utilizing much easier to measure secondary variables of the process. The lags between the input variables and the output variables vary due to changes in operating conditions. Previous studies have introduced different methods to estimate lags for input and output variables, but all of them have assumed these lags to be constant regardless of the changes in the operating conditions. In this work, an inferential sensor that can predict the composition of ethanol at the top product using time lags for the input variables and varied first-order time constant lags with the output variable has been developed. The developed inferential sensor is based on a neural network (NN) model. Principal Component Analysis (PCA) and Projection to Latent Structures (PLS) methods are used in this work to remove the outliers from the input variables set and to determine the most correlated values of the input variables and their lags with the output variable Xa (ethanol composition of distillate product) respectively. The model adaptively selects the correct first-order time constant lags of an output variable according to the instantaneous operating condition (the composition of ethanol is increased or decreased) and assigns a best value for each case. The experimental data resulting from the operation of pilot-scale batch distillation column of ethanol-water system has been used to build these NN models first and then to validate their performance. The proposed NN model structure with time lags for input variables and varied first-order time constant lags for output variable gave higher accuracy compared with the NN model without any time lag for input and output variables. This new developed NN based soft sensor has been used in an inferential proportional-integral (PI) control scheme to control the ethanol composition of the distillate. The initial inferential control results of using one tuning parameter set during the whole operation showed imperfect control results. So, using updated tuning parameter sets (gain scheduling/adaptive tuning) within this inferential PI control scheme based on the ethanol mole fraction region is necessary to improve the control performance. The results of this new developed PI control scheme showed a good control performance compared with the initial control results of this inferential controller using one set of tuning parameters. Then, this new developed NN based soft sensor has also been used in an advanced control scheme (model predictive control or MPC scheme). Two DeltaV MPC control schemes (MPC11 and MPC22) have been developed in this work. The control results of DeltaV MPC22 control scheme showed better control performance compared with other control schemes (inferential PI and MPC11 control schemes). This is due to the capability of this control scheme (MPC22) to handle the interactions between different variables (multivariable effect) especially for the distillation process. Also, it provided a faster response with very small undershoot or overshoot.