Search

The appearence of unphysical velocities in highly distorted meshes is a common problem in many simulations. In collocated meshes, this problem arises from the interpolation of the pressure gradient from faces to cells. Using an algebraic form for the classical incompressible Navier-Stokes equations, this problem is adressed. Starting from the work of F. X. Trias et. al. [FX.Trias et al. JCP 258: 246-267, 2014], a new approach for studying the Poisson equation obtained using the Fractional Step Method is found, such as a new interpolator is proposed in order to found a stable solution, which avoid the appearence of these unpleasant velocities. The stability provided by the interpolator is formally proved for cartesian meshes and its rotations, using fully-explicit time discretizations. The construction of the Poisson equation is supported on mimicking the symmetry properties of the differential operators and the Fractional Step Method. Then it is reinterpreted using a recursive application of the Fractional Step Method in order to study the system as an stationary iterative solver. Furthermore, a numerical analysis for unstructured mesh is also provided., This work has been financially supported by the Ministerio de Economía y Competitividad, Spain (project ref. ENE2017-88697-R). D. Santos acknowledges a FI AGAUR-Generalitat de Catalunya fellowship (2020FI B 00839), and N. Valle also acknowledges a FI AGAUR-Generalitat de Catalunya fellowship (2017FI B 00616). The authors thankfully acknowledge these institutions., Peer Reviewed, Postprint (published version)

In this work, we study the accuracy of CFD-based data-driven models, which predict comfort-related flow parameters in a ventilated cavity with a heated floor. We compare the computational cost and accuracy of three different models, namely artificial neural network, support vector regression, and gradient boosting regression. The tested scenarios include short and long cavities with different inlet velocities. Among the studied frameworks, the artificial neural network provides the most accurate predictions for most of the tested flow configurations. However, test configurations with jet separation and a secondary vortex are more difficult to predict correctly; thus more high-fidelity data is required in order to construct a more robust and reliable model., This work is supported by the Ministerio de Economía y Competitividad, Spain [ENE2017-88697-R]. N. Morozova is supported by the by the Ministerio de Economía y Competitividad, Spain [FPU16/06333 predoctoral contract]. Part of the calculations was performed on the MareNostrum 4 supercomputer at the Barcelona Supercomputing Center [RES project IM-2021-1-0015]. The authors thankfully acknowledge these institutions., Postprint (published version)

Wall-bounded flows at high Reynolds number are an important research field given that they are present in many important industrial applications. Nonetheless, the use of accurate numerical methodologies such as Large Eddy Simulation (LES) for routine industrial applications, is still unfeasible due to their heavy computational cost. Wall-Modeled LES is intended to circumvent the massive costs of accurately resolving the boundary layer while benefiting from the temporal and spatial resolution of an LES computation. In this work, a Two-Layer Wall Model with a general formulation suitable for non-equilibrium flows and complex geometries is presented. The Two-Layer Models (TLM) suffer from two recurrent problems, the ”log-layer mismatch” (LLM) and the resolved Reynolds stresses inflow (RRSI). Until now, complex and expensive techniques have been proposed to overcome these problems separately. In the present work, a timeaveraging filter (TAF) is considered to deal with both issues at once, with a single and low-computational-cost technique. It is the first time that the TAF technique is used to block the RRSI. In this regard, a numerical experiment is initially performed to assess the TAF ability in doing so. Afterward, the WM is tested in regular operating conditions with a Ret ˜ 3000 pipe flow case. Good results are obtained in all the performed tests, showing the capability of the TAF of suppressing the LLM while avoiding the RRSI at the same time. This makes the new TLM methodology a very effcient technique compared to other implementations proposed so far., Peer Reviewed, Postprint (published version)

Massively-parallel devices of various architectures are being adopted by the newest supercomputers to overcome the actual power constraint in the context of the exascale challenge. This progress leads to an increasing hybridisation of HPC systems and makes the design of computing applications a rather complex problem. Therefore, the software e¿ciency and portability are of crucial importance. In this context of accelerated innovation, we developed the HPC2 (Heterogeneous Portable Code for HPC). It is a portable, algebra-based framework for heterogeneous computing with many potential applications in the ¿elds of computational physics and mathematics, such as modelling of incompressible turbulent ¿ows. In its application to CFD, the algorithm of the timeintegration phase relies on a reduced set of only three algebraic operations: the sparse matrix-vector product, the linear combination of vectors and the dot product. This algebraic approach combined with a multilevel MPI+OpenMP+OpenCL parallelisation naturally provides modularity and portability. In this work, we present the strategies for the e¿cient heterogeneous execution of large-scale simulations on hybrid supercomputers that are part of the HPC2 core. The performance is studied in detail for the major computing kernel, the sparse matrix-vector product, using a sparse matrix derived from a simulation on a hybrid unstructured mesh and up to 32 nodes of a hybrid CPU+GPU supercomputer., Postprint (published version)

The monograph presents recent developments on the solution of high speed compressible flow problems using finite element and finite point (meshless) methods with adaptive refinement, techniques. This work treats two of the main problems of today's problems related to computational fluid dynamics. One problem dealt with is the assured accuracry of a numerical model, at least for the problems which are presented. The other problem which is addressed is the vast amount of time spent in generating (or regenerating) a mes in the design process., Postprint (published version)