System identification for a MIMO process

This paper describes the development of various model for MIMO process using system identification method. Ethane and ethylene separation process is considered to obtain the dynamics of a continuous distillation column, different models are developed based on the simulation data obtained from daisy database. Various linear methods like Autoregressive exogenous (ARX) model, State space (SS) model and nonlinear methods like Nonlinear autoregressive exogenous (NARX) model, Hammerstein-Wiener models are employed. Result of each models were compared and tested. The best model for the given data set is established using model validation methods. For model validation Correlation Analysis, Mean Squared Error (MSE), Final Prediction Error (FPE) methods are discussed. Based on this methods, the Nonlinear Autoregressive Exogenous (NARX) produces the better results for the ethane ethylene distillation column.