Multi-Objective Function of Ethylene Cracking Furnace using Bayesian Optimization based Artificial Neural Network

Ethylene is the most significant undeveloped resources for the advanced chemical production, contributing to 75% of petrochemical products. An Ethylene Cracking Furnace (ECF) is subjected to carbon disposition issues which hinder the thermal conduction of reaction tube. An energy-efficiency diagnosis of ECF is of greatest significant for enhancing energy usage and production operation. Hence, this research proposes the Bayesian Optimization-based Artificial Neural Network (BO-ANN) approach for the multi-objective function for ECF. BO uses a likelihood model to guide the search, allowing an effective exploration of the hyperparameter space and quickly converging to optimal solutions. This leads to minimum iterations and computational resources required to identify the best ANN model. Moreover, the examination formulas for the parameters are improved through the development of natural approach and the iterative search approach. The experimental represented that the proposed BO-ANN approach attains the better convergence degree of 0.302 and speed of 0.422 respectively as compared to ANN approach.