Data-Driven Complex Icing Scaling via Fused Super-Resolution Network

Accurate icing scaling is crucial for the icing wind tunnel tests required for aircraft airworthiness certification. Unlike traditional methods that rely on physical similarity principles, this study proposes a data-driven approach for complex icing scaling. Using the Fused Super-resolution Network (FSRN), a comprehensive ice shape database (million level) is generated, enabling efficient matching of target ice shapes with similar shapes and their corresponding icing parameters. Furthermore, we define an Ice Shape Complexity Index, based on the variance of the equivalent ice height, as a new metric for quantifying the complexity of ice shapes under various icing conditions. This innovative icing scaling method achieves high-precision full-dimensional scaling, with a pixel similarity score exceeding 97% for various complexities of ice shapes. Moreover, the proposed method extends to generalized scaling, addressing scenarios where parameters such as ice height and ice horn vary within certain ranges, enabling the identification of ice shapes with similar characteristics within the database.