AI for wetting hydrodynamics
Wetting hydrodynamics studies the dynamics of liquids in contact with surfaces and entails multiphysics processes that operate at disparate scales (see Figure). Depending on the level of detail at which the dynamics is resolved, this can often pose formidable large-scale computing challenges. In many instances, researchers resort to low-fidelity models, which typically rely on empirical approximations to capture unresolved physics, thus limiting their applicability to specific parameter regimes.
However, this limited predictive capability is not desirable in applications of practical importance, where the need to control, with precision, how a liquid behaves when deposited on a surface is key. These applications span a broad spectrum of areas, including novel technologies in microfabrication, the biomedical, smart materials, pharmaceutical and printing industries, as well as energy conversion and water harvesting in environmental applications, to name a few. Thus, the inability of industry to leverage the predictive capabilities of computationally intensive and fully resolved models for the optimal design of these applications, leads to approaches which are largely driven by intuition and empirical observations.
Figure 1: Capturing the dynamics of wetting relies critically on simulations that encompass various mechanisms that operate at disparate lengthscales. At the macroscale, the dynamics is dominated by hydrodynamic and body forces. At mesocopic distances away from the contact line, where the solid, liquid and gas phases meet, the continuum description of hydrodynamics is still applicable and we have a balance of viscous and capillary forces. At atomistic scales, out-of-equilibrium physics manifest themselves, which largely determine the wetting properties of the system.
The recent developments in computing architectures and the advances of AI methodologies present immense opportunities and challenges, in order to deliver impact across diverse fields. Given that the uptake of these technologies in wetting hydrodynamics is currently at a nascent stage and their potential benefits remain largely unexplored, in this use case of RAISE, surrogate modelling will be undertaken, leveraging data-driven multi-fidelity modelling using machine learning methodologies to enable an accurate inference of wetting dynamics. In this manner, by learning the non-linear mappings between droplet trajectories and surface features from simulation data, the present use case aims to:
-
enhance our insights about the morphology of surfaces and how droplets evolve on them;
-
help accelerate parametric studies aimed towards designing surfaces for controllable droplet transport.
​
Ultimately, it is envisioned that this framework will be applicable in situations where the computing demands are so high that we cannot afford to exclusively rely on the outputs of high-fidelity, full-resolved simulations, and it will be able to meet the requirements in other applications across diverse domains which face similar challenges in terms of computational and modelling complexity.
​
Computation-based Science and Technology Research Center, The Cyprus Institute
20 Konstantinou Kavafi Street, 2121, Nicosia (Cyprus)
Email: coordination.castorc [@] cyi.ac.cy
https://castorc.cyi.ac.cy/