Surrogate Based Design Optimization of Aerostat Envelope
Darshit M. Mehta, Rajkumar S. Pant, Sunil Lakshmipathy, Sawan Suman · 2013
To improve performance, decrease costs and enhance safety in aerospace systems, typical design processes involve time consuming and computationally expensive high fidelity models. In this context, use of surrogate-based approach can play a vital role in analysis and optimization. Surrogates are models constructed through regression or interpolation using results from high fidelity models at certain test points. They provide fast approximations of objective functions at new points thereby making sensitivity and optimization studies much more feasible. This paper presents an overview of surrogate-based analysis and design (SBAO) and its application to aerostat envelope optimization. We generated aerostat shapes using parameters created by Optimal Latin Hypercube Sampling and passed them through a CFD solver. We then constructed various surrogate models and found Kriging to be a suitable surrogate model because of its accuracy, as proved by the cross validation tests. Using Kriging, the envelope is optimized for drag using the Efficient Global Optimization algorithm.