Reliability Based Design Optimization of a novel double-layer staggered octocopter with improved surrogate model

Abstract Surrogate models are commonly used in the aircraft design process to save costs, but the predictions of simplified surrogate models often exist errors. Therefore, improving the accuracy of the surrogate model and conducting uncertainty analysis specifically for the surrogate model are necessary. In this study, a multi-objective optimization was performed on a novel configuration octocopter drone that demonstrated significantly higher aerodynamic efficiency compared to conventional configurations. The aim was to reduce the potential design failures resulting from the use of surrogate models by conducting uncertainty analysis. Under the guidance of experimental design, sample points were selected for computational fluid dynamics(CFD) simulation and a surrogate model was constructed. The errors of the surrogate model were fitted using a normal distribution, and uncertainty analysis and constraints were applied to the surrogate model using the Reliability Index Approach (RIA). The NSGA-III algorithm was employed for optimization. Based on the approximate Pareto front obtained from optimization, additional sample points were selected using orthogonal experimental design to supplement the dataset. The surrogate model was then improved, and the changes in accuracy were analyzed by comparing the probability distribution functions of the errors before and after adding the sample points. The final Pareto front was obtained. Through CFD simulation verification, the surrogate model achieved a prediction accuracy of 99.6%. The optimized unmanned aerial vehicle(UAV) exhibited a 113.341% improvement in thrust-to-weight ratio(y) and a 26.5% increase in aero ratio(a). The research demonstrates that the improved surrogate model has a high prediction accuracy and the methodology based on the improved surrogate model and RIA uncertainty analysis plays an important role in the optimization of novel rotorcraft.