Multifidelity Gaussian Processes for Predicting Shear Viscosity over Wide Ranges of Liquid State Points Based on Molecular Dynamics Simulations

Abstract

A linear multifidelity model based on Gaussian process (GP) regression is proposed that uses shear viscosities from a few molecular dynamics simulations as well as a few experimental shear viscosities to enable a highquality prediction of this transport property over a large range of thermodynamic state points. Transport properties, such as viscosity, determined from molecular simulations are sensitive to the underlying force field, which is often parametrized with emphasis on vapor−liquid coexisting properties, so that mean absolute deviations of 20% or more are frequently observed. A linear multifidelity model based on GP regression allows for compensating for such quantitative deviations based on very limited experimental data. The requirement for the molecular simulation is to describe the univariate relationship between the dimensionless shear viscosity and the reduced residual entropy qualitatively over a wide range of reduced residual entropy. This allows for a high-fidelity prediction of the shear viscosity for regions not covered by the experimental training data. The reliability of this new approach is showcased for 14 substances with more than 6000 data points, whereby only about 5 data points per species were used to train the multifidelity model. The few data points for training (selected in a narrow range of temperatures and ambient pressure, if possible) were used to mimic situations where substances are characterized by few experimental measurements, only.

Integrating Entropy Scaling into a Deep Neural Network Architecture to Predict Viscosities

“…Hence deviations between model prediction and experimental references are of systematic quantitative nature. This qualifies the outputs of the models proposed in this work as low fidelity inputs of a multi-fidelity model as proposed in previous work [Fleck et al, 2024].…”

“…Our previous multi-fidelity work has shown that the entropy space is very beneficial for machine learning applications as it translates a complex two-dimensional problem with phase boundaries into a continuous low-dimensional space [Fleck et al, 2024]. Some side results are worth mentioning with regard to this work.…”

“…Agreement between prediction and experiment is good for alkanes but decreases significantly for more complex molecules with polar or associating groups such as alcohols. In previous work a multi-fidelity approach that requires very little experimental data was proposed [Fleck et al, 2024]. However, molecular simulations are required to obtain low fidelity data, which take up a large amount of time and resources limiting the explorative capabilities.…”

Transferable Anisotropic Mie Potential Force Field for Alkanediols

“…Numerical values and corresponding standard deviations are given in Supporting Information Table S5. PC-SAFT parameters and multifidelity predictions based on previous work are shown in Table S6 and Figure S16.…”

“…When used in MD simulations, the hydrogen mass needs to be assigned to the position of the hydrogen atom. Transport coefficients such as shear viscosity or thermal conductivity were not considered during force field optimization and some deviations to experimental data are observed. , …”