Automatic Synthesis of Distillation Processes for the Separation of Homogeneous Azeotropic Multicomponent Systems

Abstract

The synthesis of distillation processes for azeotropic systems is a complex task for which commercial simulation software mainly offers a graphical analysis of residue curve maps, limited to ternary (sub)­systems, or an iterative evaluation based on tedious simulation studies. The current work features a novel algorithmic approach that aims at an automatic synthesis of distillation-based separation processes for azeotropic mixtures. Unlike other methods available in the literature, the feasibility of separation in a simple column is evaluated for both limiting operating conditions, total reflux and reversible distillation. Furthermore, the method integrates a pinch-based shortcut method to verify split feasibility and estimate the minimum energy demand. Thus, feasible process variants are generated solely based on a thermodynamic model of the multicomponent mixture, incorporating suitable recycle streams, and pressure variation, thus enabling a computationally efficient and reliable process synthesis. The capabilities and efficiency of the proposed method are illustrated for different multicomponent azeotropic mixtures.

Optimization-Based Azeotropic Distillation System Synthesis Using Geometric Insights

“…The synthesis of distillation systems for separating azeotropes has received renewed interest because it is necessary to recover polar biochemicals in biorefineries. Various sequential-based approaches and metaheuristics have been developed − in the past; however, mathematical optimization-based approaches are still fairly limited. Feng et al proposed a geometric-based approach for synthesizing azeotropic distillation systems based on partitioning the composition space.…”

AI-Based Geometry Recognition for Evaluating the Feasibility of Intensified Reaction and Separation Systems

“…39−43 However, the feasibility evaluation of an RD column still demands expert knowledge, and it is time-consuming due to the complexities of RD columns that arise from multiple interactions between vapor−liquid equilibrium and chemical kinetics. 44,45 Particularly, the extraction and interpretation of graphical information from ternary diagrams are necessary, which consists of determining the occurrence of the chemical reactions and ensuring the feasibility of crossing the distillation boundary and the desired purity of the products. This study newly proposed an automated method for extracting and interpreting topological information from a ternary diagram and reaction equilibrium curve using the image segmentation method with k-means clustering, an unsupervised machine learning algorithm, to effectively evaluate the feasibility of RD columns for ternary systems.…”

“…A feasibility evaluation algorithm for a complex batch reactive distillation column was introduced in a subsequent work . Various frameworks for the feasibility evaluation of RD columns have been continuously reported in the literature. − However, the feasibility evaluation of an RD column still demands expert knowledge, and it is time-consuming due to the complexities of RD columns that arise from multiple interactions between vapor–liquid equilibrium and chemical kinetics. , Particularly, the extraction and interpretation of graphical information from ternary diagrams are necessary, which consists of determining the occurrence of the chemical reactions and ensuring the feasibility of crossing the distillation boundary and the desired purity of the products.…”

Data‐driven integrated design of solvents and extractive distillation processes

“…), primarily due to the complex integration of mass and energy balances as well as multiple vapor–liquid equilibria within the distillation column. To achieve high‐performing distillation‐involved processes, previous work proposed different methods for solvent selection 49–51 and process optimization 52–54 . In terms of CAMPD for distillation processes, the use of rigorous process models is prohibitive due to the significant dependence of the optimization process on a high‐quality initial guess.…”

“…To achieve high-performing distillation-involved processes, previous work proposed different methods for solvent selection [49][50][51] and process optimization. [52][53][54] In terms of CAMPD for distillation processes, the use of rigorous process models is prohibitive due to the significant dependence of the optimization process on a high-quality initial guess. Therefore, CAMPD for distillation-involved processes is challenging and usually solved using simplified process models and advanced optimization strategies for higher efficiency.…”