Optimization of Simulated Moving Bed Chromatographic Processes using Surrogate Models

In this paper, we investigate a surrogate-based optimization of Simulated Moving Bed (SMB) Chromatography with Langmuir adsorption isotherm using an iterative approach. Artificial neural networks are fitted in each iteration based on randomly distributed sampling points around the optimal solution of the previous iteration. Crucial (Hyper)parameters of this surrogate-based optimization are related to the sampling region, e.g. the size, the position, and the number of samples within. It is shown that for highly efficient chromatographic columns with a large number of theoretical stages, the surrogate-based optimization is much faster than the numerical optimization of the full-blown model.