Fuzzy Hybrid Modeling of a Reactive Distillation Column for Ethyl Acetate Process

The complex interactions between vapour-liquid equilibrium, mass and energy transports and reaction kinetics in reactive distillation (RD) process have constituted serious computational challenges to its modelling and control. This is because large numbers of nonlinear differential algebraic equations are needed to represent the process when using first principle modelling approach. This work developed simplified model RD with the essential nonlinear characteristics relevant to control design purpose but with reduced computational demand. A Continuous RD flow sheet for ethyl acetate process was designed using a commercial simulator (ASPEN) to generate relevant dynamic data for model development. A hybrid Fuzzy Hammerstein (FH) model consisting nonlinear fuzzy model and linear state space model was then developed. The developed model was compared with linear autoregressive input exogenous (ARX) and nonlinear autoregressive input exogenous (NARX). The developed FH model has the order of [8, 8, 8, 8, 8] for dynamic linear part and 300 parameters consisting of 192 premise parameters and 108 consequent parameters for static nonlinear part. The output variables were temperature on trays 9 and 29 (T9 and T29) of the column. The values of root mean square error (RMSE) for the three models RMSE (ARX: 0.951; NARX: 0.75 and FH: 0.011) and RMSE (ARX: 0.961; NARX: 0.79 and FH: 0.015) obtained for T9 and T29 respectively show that FH model outperforms the other two models. Correlation tests showed that FH model mostly described the dynamic nature of the process as it passed both whiteness and independent test criteria at p < 0.01.