Search

This work proposes a novel multi-output neural network for the prediction of the aerodynamic coefficients of wings in three dimensions using inviscid compressible flow data. Contrary to existing neural networks that are designed to predict the aerodynamic coefficients directly, the proposed network considers as output the pressure at a number of selected points on the aerodynamic shape. The performance of the proposed neural network is compared against the existing neural networks. The numerical results, involving high dimensional problems with flow and geometric parameters, show the benefits of the proposed approach.

This is the peer reviewed version of the following article: Sevilla, R., Giacomini, M., Karkoulias, A., Huerta, A. A superconvergent hybridisable discontinuous Galerkin method for linear elasticity. "International journal for numerical methods in engineering", 12 Octubre 2018, vol. 116, núm. 2, p. 91-116, which has been published in final form at https://doi.org/10.1002/nme.5916. This article may be used for non-commercial purposes in accordance with Wiley Terms and Conditions for Self-Archiving. The first superconvergent hybridisable discontinuous Galerkin method for linear elastic problems capable of using the same degree of approximation for both the primal and mixed variables is presented. The key feature of the method is the strong imposition of the symmetry of the stress tensor by means of the well known and extensively used Voigt notation, circumventing the use of complex mathematical concepts to enforce the symmetry of the stress tensor either weakly or strongly. A novel procedure to construct element by element a superconvergent postprocessed displacement is proposed. Contrary to other hybridisable discontinuous Galerkin formulations, the methodology proposed here is able to produce a superconvergent displacement field for low-order approximations. The resulting method is robust and locking-free in the nearly incompressible limit. An extensive set of numerical examples is utilised to provide evidence of the optimality of the method and its superconvergent properties in two and three dimensions and for different element types.

This work proposes a high-order hybridised discontinuous Galerkin (HDG) formulation of the Harten-Lax-Van Leer (HLL) Riemann solver for compressible flows. A unified framework is introduced to present Lax-Friedrichs, Roe and HLL Riemann solvers via appropriate definitions of the HDG numerical fluxes. The resulting high-order HDG method with HLL Riemann solver is evaluated through a set of numerical simulations of inviscid compressible flows in different regimes, from subsonic isentropic flows to transonic and supersonic problems with shocks. The accuracy of the proposed method is comparable with the one of Lax-Friedrichs and Roe numerical fluxes in subsonic and transonic flows. The superior performance of HLL is highlighted in supersonic cases, where the method provides extra robustness, being able to produce positivity preserving approximations without the need of any user-defined entropy fix., Comment: An extended and updated version of the paper covering both inviscid and viscous compressible flows is available at arXiv:2009.06396

An improvement of the classical finite element method is proposed. It is able to exactly represent the geometry by means of the usual CAD description of the boundary with Non-Uniform Rational B-Splines (NURBS). Here, the two-dimensional case is presented. For elements not intersecting the boundary, a standard finite element interpolation and numerical integration is used. But elements intersecting the NURBS boundary need a specifically designed piecewise polynomial interpolation and numerical integration. A priori error estimates are also presented. Finally, some examples demonstrate the applicability and benefits of the proposed methodology. NEFEM is at least one order of magnitude more precise than the corresponding isoparametric finite element in every numerical example shown. This is the case for both continuous and discontinuous Galerkin formulations. Moreover, for a desired precision NEFEM is also more computational efficient, as shown in the numerical examples. The use of NEFEM is strongly recommended in the presence of curved boundaries and/or when the boundary of the domain has complex geometric details. The possibility of computing accurate solution with coarse meshes and high order interpolations, makes NEFEM a more efficient strategy than classical isoparametric FE.

This paper presents the extension of the recently proposed NURBS-enhanced finite element method (NEFEM) to 3D domains. NEFEM is able to exactly represent the geometry of the computational domain by means of its CAD boundary representation with non-uniform rational B-splines (NURBS) surfaces. Specific strategies for interpolation and numerical integration are presented for those elements affected by the NURBS boundary representation. For elements not intersecting the boundary, a standard finite element rationale is used, preserving the efficiency of the classical FEM. In 3D NEFEM special attention must be paid to geometric issues that are easily treated in the 2D implementation. Several numerical examples show the performance and benefits of NEFEM compared with standard isoparametric or cartesian finite elements. NEFEM is a powerful strategy to efficiently treat curved boundaries and it avoids excessive mesh refinement to capture small geometric features.