Computer-Aided Solvent Design Integrated with a Machine Learning-based Atom Contribution Method
Abstract Solvents are widely applied in chemical industries. The use of efficient model-based solvent selection techniques is an option worth considering for rapid identification of solvents with desired properties. In this paper, a Machine Learning-based Atom Contribution (MLAC) method is developed for fast predictions of molecular surface charge density profiles (ρ(σ)) using the three-dimensional atomic descriptors. Integrating the developed MLAC method and Computer-Aided Molecular Design (CAMD) techniques, an optimization-based MLAC-CAMD framework for solvent design is established by formulating and solving a Mixed-Integer Non-Linear Programming (MINLP) model, where model complexities are managed with a decomposition-based solution strategy. Finally, a case study involving crystallization is presented to highlight the effectiveness of the MLAC-CAMD framework.
Paper
Full text
Computer-Aided Solvent Design Integrated with a Machine Learning-based Atom Contribution Method
Semantic Scholar · Chemistry · 2021
Abstract
Abstract Solvents are widely applied in chemical industries. The use of efficient model-based solvent selection techniques is an option worth considering for rapid identification of solvents with desired properties. In this paper, a Machine Learning-based Atom Contribution (MLAC) method is developed for fast predictions of molecular surface charge density profiles (ρ(σ)) using the three-dimensional atomic descriptors. Integrating the developed MLAC method and Computer-Aided Molecular Design (CAMD) techniques, an optimization-based MLAC-CAMD framework for solvent design is established by formulating and solving a Mixed-Integer Non-Linear Programming (MINLP) model, where model complexities are managed with a decomposition-based solution strategy. Finally, a case study involving crystallization is presented to highlight the effectiveness of the MLAC-CAMD framework.