Cerfacs Enter the world of high performance ...

@ARTICLE{AR-PA-21-23, author = {Bauer, M. and Kostler, H. and Ruede, U. }, title = {lbmpy: Automatic code generation for efficient parallel lattice Boltzmann methods}, year = {2021}, volume = {49}, pages = {101269}, doi = {10.1016/j.jocs.2020.101269}, journal = {Journal of Computational Science}, pdf = {https://doi.org/10.1016/j.jocs.2020.101269}}

@ARTICLE{AR-CFD-21-53, author = {Renard, F. and Feng, Y. and Boussuge, J.-F. and Sagaut, P. }, title = {Improved compressible hybrid lattice Boltzmann method on standard lattice for subsonic and supersonic flows}, year = {2021}, number = {April 2021}, volume = {219}, pages = {104867 }, issn = {0045-7930}, doi = {10.1016/j.compfluid.2021.104867}, abstract = {A D2Q9 Hybrid Lattice Boltzmann Method (HLBM) is proposed for the simulation of both compressible subsonic and supersonic flows. This HLBM is an extension of the model of Feng et al. [1], which has been found, via different test cases, to be unstable for supersonic regimes. To circumvent this limitation, we propose:: (1) a new discretization of the lattice closure correction term that makes possible the simulation of supersonic flows, (2) a corrected viscous stress tensor that takes into account polyatomic gases, and (3) a novel discretization of the viscous heat production term fitting with the regularized formalism. The result is a hybrid method that resolves the mass and momentum equations with an LBM algorithm, and resolves the entropy-based energy equation with a finite volume method. This approach fully recovers the physics of the Navier–Stokes–Fourier equations with the ideal gas equation of state, and is valid from subsonic to supersonic regimes. It is then successfully assessed with both smooth flows and flows involving shocks. The proposed model is shown to be an efficient, accurate, and robust alternative to classic Navier–Stokes methods for the simulation of compressible flows}, keywords = {LBM, Compressible High speed flow, Shock waves, Aerodynamic noise}, url = {https://www.sciencedirect.com/science/article/abs/pii/S0045793021000335?via%3Dihub}}

@ARTICLE{AR-PA-21-45, author = {Mohanamuraly, P. and Müller, J.D. }, title = {An Adjoint‐assisted Multilevel Multifidelity Method for Uncertainty Quantification And Its Application To Turbomachinery Manufacturing Variability}, year = {2021}, number = {9}, volume = {122}, pages = {2179-2204}, doi = {10.1002/nme.6617}, journal = {International Journal for Numerical Methods in Engineering}, keywords = {Multilevel Multifidelity Monte Carlo, Uncertainty Quantification, Goalbased PCA, Adjoint Sensitivity, Manufacturing Variations}, pdf = {https://doi.org/10.1002/nme.6617}}

@ARTICLE{AR-CFD-21-38, author = {Agostinelli, P.W. and Rochette, B. and Laera, D. and Dombard, J. and Cuenot, B. and Gicquel, L.Y.M. }, title = {Static mesh adaptation for reliable large eddy simulation of turbulent reacting flows}, year = {2021}, number = {3}, volume = {33}, pages = {035141}, doi = {10.1063/5.0040719}, journal = {Physics of Fluids}, abstract = {The design challenge of reliable lean combustors needed to decrease pollutant emissions has clearly progressed with the common use of experiments as well as Large Eddy Simulation (LES) because of its ability to predict the interactions between turbulent flows, sprays, acoustics and flames. However, the accuracy of such numerical predictions depends very often on the user’s experience to choose the most appropriate flow modeling and, more importantly, the proper spatial discretization for a given computational domain. The present work focuses on the last issue and proposes a static mesh refinement strategy based on flow physical quantities. To do so, a combination of sensors based on the dissipation and production of kinetic energy coupled to the flame-position probability is proposed to detect the regions of interest where flow physics happens and grid adaptation is recommended for good LES predictions. Thanks to such measures a local mesh resolution can be achieved in these zones improving the LES overall accuracy while, eventually, coarsening everywhere else in the domain to reduce the computational cost. The proposed mesh refinement strategy is detailed and validated on two reacting-flow problems: a fully premixed bluff-body stabilized flame, i.e. the VOLVO test case, and a partially premixed swirled flame, i.e. the PRECCINSTA burner, which is closer to industrial configurations. For both cases, comparisons of the results with experimental data underline the fact that the predictions of the flame stabilization, and hence the computed velocity and temperature fields, are strongly influenced by the mesh quality and significant improvement can be obtained by applying the proposed strategy}, url = {https://aip.scitation.org/doi/pdf/10.1063/5.0040719}}

@ARTICLE{AR-CMGC-21-40, author = {Mouradi, R.-S. and Goeury, C. and Thual, O. and Zaoui, F. and Tassi, P. }, title = {Physically interpretable machine learning algorithm on multidimensional non-linear fields}, year = {2021}, volume = {428}, pages = {1-34}, doi = {10.1016/j.jcp.2020.110074}, journal = {Journal of Computational Physics}, pdf = {https://cerfacs.fr/wp-content/uploads/2021/03/GlobC-Article-Mouradi-Physically-interpretable-machine-learning-algorithm-on-multidimensional-2021-Elsevier.pdf}}