Surrogate Models combined with a Support Vector Machine for the Optimized Design of a Crude Oil Distillation Unit using Genetic Algorithms

This paper introduces a novel optimization-based framework for the design of a crude oil distillation unit. The approach presented integrates surrogate models based on artificial neural networks (ANN) with feasibility constraints generated using a support vector machine (SVM) in order to optimise the column configuration and its operating conditions. The SVM filters infeasible design options from the solution space of the design problem, which reduces the computational effort and ultimately improves the quality of the final solution. Rigorous process simulations are used to build the surrogate model, while pinch analysis is employed to determine the maximum heat recovery and minimum utility costs. The objective is to minimise the total annualized cost, which is optimised by combining a genetic algorithm with the surrogate model. The approach is illustrated in an industrially relevant case study.