A team of scientists from the Center for Computational Physics at the Moscow Institute of Physics and Technology (MIPT), along with their colleagues, has developed the PANDA-NN algorithm, which can automatically calculate contact angles in liquid-liquid-solid systems. The related research results were published in the journal Colloids and Surfaces A: Physicochemical and Engineering Aspects.
The contact angle is a quantitative parameter that characterizes the interaction between media and determines surface phenomena such as capillary pressure and permeability. It is crucial in simulating fluid flow in porous media. Accurate values help create reliable fluid flow models and optimize the production of minerals such as oil and gas. Traditional contact angle modeling requires manual adjustment of important parameters such as density thresholds, which easily leads to errors and affects the reproducibility of results.
The PANDA-NN algorithm is an improved version. Unlike the previous PANDA version, it uses machine learning to automatically determine the shape of the two-phase interface in nanopores, such as water droplet shape, donut shape, etc. The neural network automatically identifies the interface type and selects the appropriate mathematical model for density profile analysis based on molecular modeling data. Ilya Kopanichuk, Senior Researcher at the MIPT Center for Computational Physics, explained that the new version no longer requires calculating the wetting angle based on the droplet density distribution; both the droplet shape and the approximate calculation equations are determined by the neural network.
PANDA-NN also includes an additional parameter — the thickness of the wetting layer. Considering this parameter can minimize errors when determining the density distribution function. The algorithm is stable and reproducible, with results consistent with experimental data and an accuracy of 5%, which is much higher than the previous PANDA method (13%) and the classical circle-fitting method (10%). However, for complex interfaces, the deviation from experimental values may exceed 5%, so the system composition must be selected carefully.
Its automation capability allows large-scale calculation of contact angles under different fluids, substrates, pore sizes, and temperatures, and is expected to be used to optimize oil and gas production. Ilya Kopanichuk pointed out that the algorithm does not require manual participation in the calculation of each system or hyperparameter calibration, and can automatically consider the thickness of the adsorption film. Alexey Semenchuk, Junior Researcher at the MIPT Center for Computational Physics, added that this algorithm can highly automate the determination of contact angles in nanoporous structures, opening new possibilities for simulating filtration and fluid displacement processes, including oil production problems and analysis of capillary effects in reservoir rocks.
Currently, the algorithm is only applicable to liquid-liquid-solid systems. For liquid-gas-solid systems, improvements are still needed regarding changes in droplet volume. In the future, the authors plan to further improve the algorithm and develop related applications. Scientists from the MIPT Center for Computational Physics, the Joint Institute for High Temperatures of the Russian Academy of Sciences, the National Research University Higher School of Economics, the Schmidt Institute of Physics of the Earth of the Russian Academy of Sciences, and AIRI participated in this work.
