Parallel implementation of localized radial basis function interpolation for computational aeroelastic predictions

Mesh deformation and data interpolation using radial basis functions (RBF) in combination with data reduction greedy algorithm has proven to be an efficient method, both in providing high quality deformed meshes and speed up computations. In the present work an in-house hybrid unstructured Reynoldsaveraged Navier-Stokes solver (HUNS3D) has been extended to include dynamic mesh motion and aeroelastic behavior prediction. For present computational aeroelastic simulations, RBF interpolation serves as single subroutine and carries out the required data interpolation for both the surface loads and deformations. For mesh motion and displacement interpolation the already developed RBF interpolation methods works reasonably well. But for interpolation of aerodynamic loads the current procedures become expensive in terms of computational time and are greatly influenced by the parameters used in the interpolation. In this paper a more efficient and robust method is presented that localizes the interpolation. This method resembles in concept to the pointwise form of partition of unity method but somewhat differs in its implementation. It is efficient in terms of computational time and can be readily parallelized. Also it reduces the influence of the interpolation parameters on the coupling behavior. The proposed method has been tested by performing static aeroelastic computations in transonic flow over the AGARD 445.6 wing, HIRENASD wing/body configuration and a flexible wing with spar-rib-skin construction. The method has shown its effectiveness in aeroelastic behavior prediction for different aerodynamic configurations.