Integration of experimental analysis and machine learning to predict drop behavior on superhydrophobic surfaces

Abstract The design of water-repellent surfaces is of great importance as water repellency of surfaces against impacting water drops is a promising approach for most of applications as anti-icing and self-cleaning. To comprehensively investigate drop interactions with hydrophobic and superhydrophobic surfaces, we conducted a large suite of experimental tests to evaluate the morphology of impacting drops on these surfaces as a function of drop properties (drop diameter, density, viscosity, and surface tension), kinematic parameters (velocity), and surface features (contact angle, contact angle hysteresis, and surface roughness). Following analyzing the experimental results, we utilized a novel approach in this field by applying a predictive approach based on machine learning to predict the behavior of impacting drops on hydrophobic and superhydrophobic surfaces. Our developed model, based on a random-forest approach, predicted drop behavior at up to 98% accuracy. Aiming at finding those conditions favorable for producing a bouncing behavior upon drop impact, we predicted the outcome of an impinging drop for a wide range of Weber numbers, i.e., impact velocities, and numerous hypothetical surfaces. Our results offer some design criteria for creating superhydrophobic surfaces that favor bouncing upon drop impact on these surfaces.