Chromatographic separation plays a vital role in various areas, as this technique can deliver high-quality products both in lab- and industrial-scale processes. Economical and also ecological benefits can be expected when optimizing such processes with mathematical methods. However, even small perturbations in the operating conditions can result in significantly altered results, which may lead to the disposal of the whole product.
In this paper, we propose an algorithmic approach for the robust optimization of chromatographic processes. In our approach we combine techniques from optimization with partial differential equation (PDE) models with techniques form distributionally robust optimization in an alternate manner. Specifically, we split the nominal optimization of the separation process from the robustification of the fractionation times, i.e., the time window where the eluting mass is collected. We thus avoid solving computationally intractable robust mixed-integer optimal control problems constrained by PDEs.
We apply our proposed methodology to a realistic case study where a single component is to be separated from two enclosing impurities. We show that our approach indeed generates optimized controls and fractionation times that lead to a robust separation, requiring only a small number of iterations. As this algorithmic approach is quite general, it can be used for various experimental settings, thus making it a versatile method.