Accelerating flapping flight analysis: Reducing CFD dependency with a hybrid decision tree approach for swift velocity predictions

Abstract

Insect flight depends on flapping their wings, allowing their agile movement. The modulation of wing flapping enables insects to manoeuvre with flexibility. Given the inability to directly control or observe the nuances of insect wing flapping in biological experiments, numerical simulation emerges as a more feasible approach for investigating insect flight dynamics. Through computational fluid dynamics (CFD) analysis, it is possible to obtain highly accurate results and gain insights into the effects of various flapping behaviours on flight. However, the substantial time cost associated with individual simulations poses a challenge, making it difficult to explore the comprehensive range of parameter combinations and variations. In order to enhance the efficiency of research, this study introduces an algorithmic framework that utilises signal decomposition techniques and decision tree to reduce the computational time required for flight simulation. The approach simplifies data complexity, enabling rapid identification of specific flight manoeuvres of interest, followed by detailed examination with conventional method. It allows for predicting flight end-states with minimal simulation data while maintaining high accuracy and reducing dependency on CFD computation. These advancements benefit studies on insect flight postures and the design of micro air vehicles (MAVs), enriching both theoretical and practical aerodynamics.