Low fidelity data driven machine learning based optimisation method for box-wing configuration

Abstract

Wing design optimization traditionally involves computationally expensive high-fidelity simulations, limiting the exploration of design spaces. In this study, we propose a methodology that combines low-fidelity numerical models with machine learning algorithms to efficiently navigate the complex parameter space of box-wing configurations. Through the utilisation of a surrogate model trained on a limited dataset derived from low-fidelity simulations, our method strives to predict results within an acceptable range, significantly curtailing computational costs and time. The effectiveness of this methodology is demonstrated through a series of case studies, involving the Onera M6 and NASA CRM wing as test cases and Bionica box-wing optimization as an application case. The initial application of the proposed methodology to the box-wing case successfully achieved an almost 9.82 % increase in overall aerodynamic efficiency. Its competitive performance compared to conventional optimization methods, along with its substantial reduction in computational time and resource requirements, is evident. This efficient methodology holds promise for enhancing the design optimization process for aviation start-ups by efficiently exploring complex design spaces with reduced computational burden.