Hybrid Semi‐parametric Modeling in Separation Processes: A Review

Abstract

This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes. Separations of mixtures play a critical role in chemical industries. Over the last century, the knowledge in the area of chemical thermodynamics and modeling of separation processes has been substantially expanded. Since the models are still not completely accurate, hybrid models can be used as an alternative that allows to retain existing knowledge and augment it using data. This paper explores some of the weaknesses in the current knowledge in separations design, simulation, optimization, and operation, and presents many examples where data-driven and hybrid models have been used to facilitate these tasks.

Machine Learning in Chemical Product Engineering: The State of the Art and a Guide for Newcomers

“…At the same time, the implementation of semi-supervised learning methods (2%) and reinforcement learning methods (<1%) is thus far marginal in these application domains. These findings are consistent with the reported observations of other reviews, which also emphasized the major use of supervised learning methods compared to unsupervised learning methods in CPE or chemical engineering problems [2,21,26,70]. The interest in semi-supervised methods appears to be quite recent and displays an increasing trend, showing that this category of ML methods may become more significant for problems in the domain.…”

“…Several reviews of the applications of hybrid and combinatorial models are available in petroleum and energy systems engineering, multiscale material and process design and in separation processes [52,69,70]. The authors in [52] presented hybrid models as an alternative of data-driven models and first principle models in terms of knowledge of process, computational burden, data demand and extrapolation capabilities.…”

Challenges and Opportunities to Enhance Flexibility in Design and Operation of Chemical Processes

“…Recent work uses, e.g., artificial neural networks to set up a surrogate model enabling multi-scale optimization exemplified with a membrane process [91] or thin film growth processes [92]. Other solutions proposed to deal with multiscale problems are model reduction methods or surrogate modeling approaches to reduce computational effort for complex nonlinear systems to allow for efficient control and scheduling [93][94][95][96]. Some aspects concerning the abovementioned challenges are expected to be treated using machine learning methods in chemical engineering in the future [97].…”

AI versus Classic Methods in Modelling Isotopic Separation Processes: Efficiency Comparison

“…By comparison, the usage of neural networks in the case of the mathematical model proposed in the paper is more justified since they are used to learn nonlinear functions. In [55], some examples of using neural networks for the direct learning of the separation processes behavior are given. In the case of the approached 18 O separation process, due to its complexity, the learning of its behavior using only one neural network, even a high complexity one, is almost impossible (it is possible only by interconnecting more neural networks).…”