Design of a Neuro‐Based Computing Paradigm for Simulation of Industrial Olefin Plants

Abstract

Abstract A neuro‐based computing technique is used for simulation of olefin plants at industrial scale. Artificial neural networks are applied to estimate the flow rate of the main products of the olefin unit from available information in terms of flow rate of feed streams and operating condition of furnaces. The structure of the smart model is determined through a trial‐and‐error procedure taking the real plant information over four successive years. The proposed paradigm estimates the tonnage of the product streams by an absolute average relative deviation in the range of 0.9 % for methane to 3.14 % for propylene. Results confirmed that this smart simulation not only presents accurate predictions, but is easy to use, straightforward, and can be simply employed for optimization and control of the unit.