A Neural-network-based Approach to Study the Energy-optimal Hovering Wing Kinematics of a Bionic Hawkmoth Model

Abstract

This paper presents the application of an artificial neural network to develop an approach to determine and study the energy-optimal wing kinematics of a hovering bionic hawkmoth model. A three-layered artificial neural network is used for the rapid prediction of the unsteady aerodynamic force acting on the wings and the required power. When this artificial network is integrated into genetic and simplex algorithms, the running time of the optimization process is reduced considerably. The validity of this new approach is confirmed in a comparison with a conventional method using an aerodynamic model based on an extended unsteady vortex-lattice method for a sinusoidal wing kinematics problem. When studying the obtained results, it is found that actual hawkmoths do not hover under an energyoptimal condition. Instead, by tilting the stroke plane and lowering the wing positions, they can compromise and expend some energy to enhance their maneuverability and the stability of their flight.