ML-Driven Surface Structure Optimization of Aluminum Alloy Superhydrophobic Surfaces for Enhanced Anti-Icing Performance

Surface icing poses a serious threat to the safe operation of aerospace, transportation, and power transmission systems, highlighting the necessity of studying surface anti-icing performance. Superhydrophobic surfaces have been widely used in anti-icing applications; however, the relationship between their surface characteristics and anti-icing performance is very complex. Traditional experimental methods are costly and time-consuming, and optimizing anti-icing performance through surface characteristic tuning faces challenges. This study proposes an ML-driven optimization method for surface anti-icing performance. We use the gray level co-occurrence matrix (GLCM) to characterize multiscale surface features and predict the anti-icing performance of aluminum-based superhydrophobic surfaces. The key features for controlling the anti-icing performance were determined through various feature importance analysis methods. A mathematical model for optimizing anti-icing performance was constructed by combining these key features with classical nucleation theory. This model quantifies the synergistic regulatory effects of key features on anti-icing performance, elucidates their impact on freezing delay time, and provides a theoretical basis for the rational design of superhydrophobic anti-icing surfaces.