Data-driven approach for gust evolution and velocity field prediction

As a representative unsteady flow phenomenon, wind gusts exhibit complex spatiotemporal evolution that severely impacts aircraft aerodynamic performance and flight safety. In this study, a random forest (RF) algorithm is employed to predict phase-averaged distorted gust profiles, and the RF model is further combined with wavelet decomposition to predict instantaneous gust profiles. For higher-dimensional gust velocity fields, a low-dimensional and interpretable representation method is proposed based on a nonlinear autoencoder (AE), using only two latent features. Building on these two models, an AE–RF hybrid prediction model is developed to enable accurate and efficient prediction of high-dimensional velocity fields. Results demonstrate that the RF algorithm effectively establishes a regression model linking pitching amplitude, reduced frequency, and temporal parameters with gust profiles. It accurately predicts phase-averaged gust profiles from sparse experimental data. Wavelet decomposition reliably separates the trend and noise components of instantaneous gust profiles. Results also indicate that the AE model can reconstruct gust velocity fields from low-dimensional latent features, with feature dimensionality shown to affect reconstruction accuracy. The two-dimensional (2D) features provide an interpretable representation of the spatiotemporal evolution characteristics of gust fields. After optimization, the AE–RF hybrid model improves the structural similarity index of predicted phase-averaged gust velocity fields from 0.731 to 0.773, with the probability density concentrated within the 0.700–0.900 range. Moreover, the low-dimensional feature representation of instantaneous velocity fields is analyzed, and the AE–RF hybrid model effectively predicts the large-scale evolution trends of instantaneous gust fields. Overall, the proposed gust prediction model offers a novel framework for rapid reconstruction and analysis of complex gust flow fields, thereby providing a theoretical support for aircraft gust response prediction and alleviation design.