Development of a suitable model is the most challenging problem for the distillation columns due to their nonlinear and complex behavior. First Order Plus Time Delay (FOPTD) models using transfer functions and Nonlinear Autoregressive with Exogenous Input (NLARX) models described by sigmoid function were used in this study to identify the dynamics of the industrial debutanizer column. The results demonstrated that linear models were not able to approximate the behavior of system whereas the NLARX models performed well in capturing the nonlinear dynamics of debutanizer column. The prediction capability of the developed NLARX model was found to be 91.26 and 86.56 for top and bottom composition respectively. © 2019 IEEE.