Prediction of Self-Diffusion in Binary Fluid Mixtures Using Artificial Neural Networks

Abstract

Artificial neural networks (ANNs) were developed to accurately predict the self-diffusion constants for individual components in binary fluid mixtures. The ANNs were tested on an experimental database of 4328 self-diffusion constants from 131 mixtures containing 75 unique compounds. The presence of strong hydrogen bonding molecules may lead to clustering or dimerization resulting in non-linear diffusive behavior. To address this, self- and binary association energies were calculated for each molecule and mixture to provide information on intermolecular interaction strength and were used as input features to the ANN. An accurate, generalized ANN model was developed with an overall average absolute deviation of 4.1%. Forward input feature selection reveals the importance of critical properties and self-association energies along with other fluid properties. Additional ANNs were developed with subsets of the full input feature set to further investigate the impact of various properties on model performance. The results from two specific mixtures are discussed in additional detail: one providing an example of strong hydrogen bonding and the other an example of extreme pressure changes, with the ANN models predicting self-diffusion well in both cases.

Machine learning predictions of diffusion in bulk and confined ionic liquids using simple descriptors

“…For this work we used a single hidden layer ANN with 6 nodes for the computed descriptors and 8 nodes for the SMILES-based descriptors. This architecture has been tested in previous studies 58,61,63 and was demonstrated to yield good results for this work. A hyperbolic tangent activation function (flag tansig) was used for the first hidden layer.…”

“…We have discussed the benefits of ANNs compared to other machine learning architectures like random forests for predicting diffusion previously. 58,61,83,84…”

“…60 The current work builds on our recent efforts applying ML methods to predict the diffusion properties of ideal and real fluids in bulk and confined environments. [56][57][58][61][62][63][64]…”

Machine Learning Predictions of Simulated Self-Diffusion Coefficients for Bulk and Confined Pure Liquids

“…Even molecular dynamics (MD) simulation methods are computationally expensive to thoroughly explore the full range of liquid and pore properties and their effect on liquid diffusion. This effort builds on our previous work developing ML models for self-diffusion in bulk and pore fluids, based on experimental or simulated data. − Importantly, we showed that artificial neural networks (ANNs) can accurately predict MD results for pure LJ fluid diffusion in pores of various geometries (planar, cylindrical, hexagonal) over a range of fluid–wall interactions …”

Influence of Ethanol Parametrization on Diffusion Coefficients Using OPLS-AA Force Field

“…Other approaches include Artificial Neural Networks (ANN) and Molecular Dynamics (MD) simulations. ANN have been employed for the prediction of self-diffusion coefficients ( ) in pure liquids [ 18 ] and in binary fluid mixtures [ 19 ], whereas MD simulations have been successfully used, e.g., for the estimation of , both in liquids [ 20 ] and in supercritical fluids [ 21 ], and of in liquids [ 22 , 23 ].…”