Influence of Feed Composition on the Optimum of Batch Distillation Determined by Surrogate Model-Based Optimisation

Hégely, László and Láng, Péter (2026) Influence of Feed Composition on the Optimum of Batch Distillation Determined by Surrogate Model-Based Optimisation. In: 13th International Conference on Distillation & Absorption, 2026. augusztus 30.-szeptember 2., Salzburg.

Text DA26_paper_BD-SM6 published.pdf - Published Version Restricted to Registered users only until 1 August 2027. Download (441kB) | Request a copy

Abstract

A surrogate model-based optimisation (SMBO) method is proposed to determine how the feed composition (xF) influences the optimum of batch distillation (BD) processes. Profit of a single batch of a BD process (OF), where methanol is recovered from a five-component mixture, is maximised. Optimisation variables are the reflux ratios of the steps and the stopping criteria of the fore-cuts. OF is estimated by surrogate models (SMs) fitted to simulation results (mass of product, mass of first fore-cut, process duration) by ALAMO, which uses a machine-learning approach to generate algebraic models. The influence of xF is taken into account by including the mass fraction of three components as input variables of the surrogate models along with optimisation variables. Using Latin hypercube sampling, a large number of simulations are performed in the space of xF and optimisation variables by using a professional flow-sheet simulator. OF calculated by the SMs is optimised for 31 different charge compositions by sequential quadratic programming (SQP), and variation of the value of optimisation variables over the composition space is evaluated. The method enables the estimation of the optimal value of operational parameters for a wide range of xFwithout any further simulation.

Actions (login required)

View Item