Aerodynamic pressure prediction and non-Gaussian characteristics of airborne floating objects under turbulent flows: A hybrid modeling strategy integrating proper orthogonal decomposition and deep learning

Abstract

Airborne floating objects (AFOs) pose a serious threat to the safe operation of civil aviation and low-altitude aircraft. Accurately determining the flight trajectories of AFOs is an urgent task, with the core component being the precise prediction of wind pressure. This study focuses on AFOs in the form of ellipsoidal balloons and proposes a hybrid model framework for predicting the surface wind pressure on these AFOs. The proposed approach combines proper orthogonal decomposition (POD), a convolutional neural network (CNN), and a long short-term memory (LSTM) module. First, the wind pressure characteristics of the balloons are investigated through wind tunnel tests under two different turbulence intensities. The wind pressure on the balloon surface exhibits non-Gaussian characteristics under both types of turbulence, being distributed along the leading edge in the windward region and toward the leeward zone. Typical measurement points are then selected for model training and wind pressure prediction. A comparison between the experimental and predicted results demonstrates that the POD–CNN–LSTM model proposed in this paper effectively captures the spatiotemporal characteristic structures underlying complex turbulent wind pressure fluctuations. Our research findings provide a reliable, data-driven method for predicting wind pressure on AFOs.