Abstract: The accurate prediction of water droplet collection coefficients is essential for icing analysis and the design of anti- and de-icing systems. Traditional high-fidelity numerical simulation methods, however, are often hindered by their computational complexity and time-intensive nature. Deep learning-based rapid prediction methods for water droplet collection coefficients present a promising avenue to address these challenges. In this study, we propose a fast prediction approach leveraging Proper Orthogonal Decomposition (POD) and Kolmogorov–Arnold Networks (KAN) to accurately predict water droplet collection coefficients on 3D spherical surfaces. Using the POD method, intrinsic modes and their corresponding fitting coefficients are extracted. A KAN-based deep learning model is then developed, mapping working condition parameters to fitting coefficients for efficient prediction. To further enhance model generalization, droplet inertia parameters are incorporated. Experimental results demonstrate that the proposed POD-KAN model is well-suited for predicting water droplet collection coefficients on 3D spheres, delivering rapid predictions with high accuracy, achieving an average absolute error of 3.386 × 10^–4.