Surrogate Models for Distillation Boundaries in Azeotropic Multicomponent Mixtures

The accurate approximation of distillation boundaries is of crucial importance for evaluating the feasibility and conceptual design of distillation-based processes for the separation of azeotropic multicomponent mixtures. The lack of an explicit mathematical model for distillation boundaries is a major limitation for the application of shortcut models in the synthesis and design of distillation-based processes. In order to overcome this limitation, this work proposes a generalized method to derive surrogate models for distillation boundaries in multicomponent systems that are applicable for flowsheet evaluation and optimization during the conceptual design of distillation processes. On the basis of an extensive screening, a Gaussian process model combined with the quasi-random Halton sequence as sampling method was found to provide the best surrogate models for the use in flowsheet optimization. The benefits of integrating these surrogate models into hybrid process models are demonstrated in two case studies that show a reduction of up to 95% in the computational cost of process optimization while maintaining the accuracy of purely mechanistic modeling approaches.