Polynomial ridge flowfield estimation

Computational fluid dynamics plays a key role in the design process across many industries. Recently, there has been increasing interest data-driven methods order to exploit large volume of data generated by such computations. This paper introduces idea using spatially correlated polynomial ridge functions for rapid flowfield estimation. Dimension reducing are obtained numerous points within training flowfields. The can then be used predict flow variables new, previously unseen, Their dimension nature alleviates problems associated with visualizing high-dimensional datasets, enabling improved understanding spaces and potentially providing valuable physical insights. proposed framework is computationally efficient; consisting either readily parallelizable tasks or linear algebra operations. To further reduce computational cost, need only computed at small number subsampled locations. physics encoded covariance matrices from flowfields quantities, conditional upon those predicted sampled points. demonstrate efficacy framework, incompressible around an ensemble airfoils as test case. functions’ predictive accuracy found competitive state-of-the-art convolutional neural network. local also reused obtain surrogate models integral avoiding long-term storage data. Finally, use varying boundary conditions demonstrated on transonic wing.