Interfacial behavior prediction and regulation on inclined superhydrophobic surfaces with transfer learning-assisted method

Abstract

Accurately predicting droplet impact behavior on inclined superhydrophobic surfaces (SHSs) has been a significant challenge in achieving effective interfacial regulation application. To address the precision and efficiency for interfacial behavior prediction, this study proposes a transfer learning-assisted (TL-assisted) method with broad applicability for predicting droplet impact and slip phenomenon on inclined SHSs with different micro/nano-structures. The proposed method leverages knowledge learned from a source SHS to assist the interfacial behavior prediction for target SHSs, even when historical measurements are limited, without model performance degradation caused by data distribution discrepancies and insufficient target samples. And the droplet dynamic behavior on target SHSs can be predicted by fine-tuning a limited number of labeled samples, eliminating the need for additional experimental data extraction. The underlying mechanism of the interfacial behavior, co-regulated by droplet surface tension and microstructural contraction forces, is quantified through Pearson correlation analysis and permutation feature importance (PFI) analysis. Based on these insights, the synergistic TL-assisted method accurately optimizes parameters for two instances of self-cleaning regulation (SCR) and droplet-based electrical generator (DEG) electrification output, respectively. Consequently, SCR was controlled efficiently by rationalizing settings contact location (Di) and traveled distance (Li). DEG electrification output was enhanced when recognizing impacting distance (Loptimal) for maximum spreading diameter (Ds).

Keywords

• Superhydrophobic
• Machine learning
• Interfacial behavior prediction
• Self-cleaning
• Droplet-based electrical generator