Deep Learning for Multifidelity Aerodynamic Distribution Modeling from Experimental and Simulation Data

Wind-tunnel experiment plays a critical role in the design and development phases of modern aircraft, which is always limited by prohibitive cost. In contrast, numerical simulation, as an important alternative research paradigm, mimics complex flow behaviors but less accurate compared to experiments. This leads recent emerging interest applying data fusion for aerodynamic prediction. particular, prediction aerodynamics with lower computational cost can be achieved fusing experimental (high-fidelity) (low-fidelity) data. Currently, existing works on models using mainly concern integrated data, like lift, drag, etc. this paper, multifidelity model based deep neural network (DNN) developed distribution over wing surface, where both are introduced loss function proper weighting factor balance overall accuracy. The proposed approach illustrated modeling surface pressure ONERA M6 transonic flow, including different scenarios condition varies or small number high-fidelity available. results demonstrate that trained decent low-fidelity few accurately predict wing. outperformance other DNNs single fidelity has been reported.