Design of an Aeroelastically Scaled Model in a Compressed Air Wind Tunnel Facility Using Multifidelity Multi-Objective Bayesian Optimization

This paper presents the design of a geometrically nonlinear aeroelastically scaled model in a compressed air wind tunnel (CAWT) facility using a two-pronged approach that integrates the classical dimensional analysis and a systematic multi-disciplinary optimization procedure. The CAWT facility, recently constructed at Penn State, enables large Reynolds numbers to be tested using small models, which effectively removes the usual approximation of Reynolds number in aeroelastic wind tunnel tests. To develop the scaled model, the two-pronged approach first identifies the groups of similarity parameters of the aeroelastic model using classical dimensional analysis. Next, when some of the similarity parameters cannot be satisfied due to limitations of manufacture and test conditions, numerical optimization is performed to adjust the scaled model to maintain the similarity in the aeroelastic characteristics, such as the flutter boundary. A sample-efficient multifidelity multi-objective Bayesian optimization (M2BO) algorithm is proposed to tackle the highly nonlinear aeroelastic optimization problem. The developed methodologies are applied to the scaling of the Pazy wing model, which is designed for large deformation aeroelastic experiments. The results have demonstrated the efficacy of utilizing the CAWT facility for aeroelastic tests that enables a significantly wider range of model scales beyond that of a conventional wind tunnel, while maintaining the aerodynamic similarity. Furthermore, the two-pronged approach has been demonstrated to produce the design of a practically-viable aeroelastically scaled model in the presence of model imperfections. The initial success opens up a unique venue for the design, analysis, and testing of scaled nonlinear aeroelastic models with enhanced reproduction of operating conditions in the CAWT facility.