Kriging Surrogate-based Optimization for Shape Design of Thin Electric Propeller

Abstract

This paper outlines an optimized propeller design for an unmanned aerial vehicle (UAV) employing a Kriging surrogate model-based optimization approach. The primary objective was to maximize propeller efficiency while adhering to the thrust-to-torque ratio constraint at a rotational speed of 6,500 rpm. The design variables encompassed the twist angle and the ratio of blade thickness to chord length across the twenty-section airfoil of the propeller. A comprehensive analysis was conducted using computational fluid dynamics to assess the aerodynamics of the propeller. The Kriging surrogate model serves as a valuable tool for approximating objective and constraint functions. The optimal Latin hypercube sampling technique was employed for design of experiment, generating a set of sampling points to construct a Kriging surrogate model. To tackle the optimization problem, seven metaheuristic optimizers were employed, including a genetic algorithm, particle swarm optimization, population-based incremental learning, differential evolution, teaching-learning based optimization, ant colony optimization, and an evolution strategy with covariance matrix adaptation. The obtained results revealed that Kriging surrogate model-based differential evolution optimization stood out as the most efficient method for addressing the propeller optimization problem. The propeller efficiency experienced improvement of approximately 0.6% compared to the maximum result obtained from the sampling points.