Cerfacs Enter the world of high performance ...

@ARTICLE{AR-CFD-21-105, author = {Renard, F. and Wissocq, G. and Boussuge, J.-F. and Sagaut, P. }, title = {A linear stability analysis of compressible hybrid lattice Boltzmann methods}, year = {2021}, volume = {446}, pages = {110649}, doi = {10.1016/j.jcp.2021.110649}, journal = {Journal of Computational Physics}, abstract = {An original spectral study of the compressible hybrid lattice Boltzmann method (HLBM) on standard lattice is proposed. In this framework, the mass and momentum equations are addressed using the lattice Boltzmann method (LBM), while finite difference (FD) schemes solve an energy equation. Both systems are coupled with each other thanks to an ideal gas equation of state. This work aims at answering some questions regarding the numerical stability of such models, which strongly depends on the choice of numerical parameters. To this extent, several one- and two-dimensional HLBM classes based on different energy variables, formulations (primitive or conservative), collision terms and numerical schemes are scrutinized. Once appropriate corrective terms introduced, it is shown that all continuous HLBM classes recover the Navier-Stokes-Fourier behavior in the linear approximation. However, striking differences arise between HLBM classes when their discrete counterparts are analyzed. Multiple instability mechanisms arising at relatively high Mach number are pointed out and two exhaustive stabilization strategies are introduced: (1) decreasing the time step by changing the reference temperature and (2) introducing a controllable numerical dissipation σ via the collision operator. A complete parametric study reveals that only HLBM classes based on the primitive and conservative entropy equations are found usable for compressible applications. Finally, an innovative study of the macroscopic modal composition of the entropy classes is conducted. Through this study, two original phenomena, referred to as shear-to-entropy and entropy-to-shear transfers, are highlighted and confirmed on standard two-dimensional test cases.}, keywords = {HLBM, LSA, compressible}, url = {https://www.sciencedirect.com/science/article/pii/S0021999121005441}}

@ARTICLE{AR-CFD-21-115, author = {Fiore, M. and Daroukh, M. and Montagnac, M. }, title = {Broadband noise prediction of a counter rotating open rotor based on LES simulation with phase-lagged assumption}, year = {2021}, volume = {514}, pages = {116420}, doi = {10.1016/j.jsv.2021.116420}, journal = {Journal of Sound and Vibration}, abstract = {This paper presents the broadband noise analysis of a Counter Rotating Open Rotor (CROR) configuration. The numerical study is based on a hybrid approach: a Large Eddy Simulation (LES) code solves the near-flow field of the CROR configuration and the near-to-far-field propagation is then predicted by a Ffowcs Williams–Hawkings analogy either based on a solid or porous formulation. The LES solver uses a phase-lagged assumption with a proper orthogonal decomposition for the data storage. The numerical approach is validated against wind tunnel experimental data of a CROR configuration (AI-PX7) at three different operating points focused on the broadband noise. The analysis of the numerical simulations shows the predominant effect of the rear rotor suction side on the radiated broadband noise with an increasing contribution with span location. This source is related to the impingement of front rotor wakes and the development of leading edge vortices that induce large pressure fluctuations close to the leading edge on the suction side.}, keywords = {Counter-rotating open rotor Broadband noise prediction Ffowcs Williams–Hawkings analogy Large-Eddy simulation Phase-lagged assumption POD data storage}, url = {https://www.sciencedirect.com/science/article/abs/pii/S0022460X2100465X}}

@ARTICLE{AR-CMGC-21-101, author = {Ménard, R. and Skachko, S. and Pannekoucke, O. }, title = {Numerical discretization causing error variance loss and the need for inflation}, year = {2021}, doi = {doi.org/10.1002/qj.4139}, journal = {Quarterly Journal of the Royal Meteorological Society}, url = {https://rmets.onlinelibrary.wiley.com/doi/10.1002/qj.4139}}

@ARTICLE{AR-CMGC-21-87, author = {El Garroussi, S. and Ricci, S. and De Lozzo, M. and Goutal, N. and Lucor, D. }, title = {Tackling random fields non-linearities with unsupervised clustering of polynomial chaos expansion in latent space: application to global sensitivity analysis of river flooding}, year = {2021}, doi = {10.1007/s00477-021-02060-7}, journal = {Stochastic Environmental Research and Risk Assessment}, pdf = {https://cerfacs.fr/wp-content/uploads/2021/08/globc-AR-ELGaroussi-StockasticEnvironmentalResearchRiskAssessment-2021.pdf}, url = {https://link.springer.com/article/10.1007%2Fs00477-021-02060-7}}

@ARTICLE{AR-PA-21-107, author = {Xing, V. and Lapeyre, C. and Jaravel, T. and Poinsot, T. }, title = {Generalization Capability of Convolutional Neural Networks for Progress Variable Variance and Reaction Rate Subgrid-Scale Modeling}, year = {2021}, number = {16}, volume = {14}, pages = {5096}, doi = {10.3390/en14165096}, journal = {Energies}, abstract = {Deep learning has recently emerged as a successful approach to produce accurate subgrid-scale (SGS) models for Large Eddy Simulations (LES) in combustion. However, the ability of these models to generalize to configurations far from their training distribution is still mainly unexplored, thus impeding their application to practical configurations. In this work, a convolutional neural network (CNN) model for the progress-variable SGS variance field is trained on a canonical premixed turbulent flame and evaluated a priori on a significantly more complex slot burner jet flame. Despite the extensive differences between the two configurations, the CNN generalizes well and outperforms existing algebraic models. Conditions for this successful generalization are discussed, including the effect of the filter size and flame–turbulence interaction parameters. The CNN is then integrated into an analytical reaction rate closure relying on a single-step chemical source term formulation and a presumed beta PDF (probability density function) approach. The proposed closure is able to accurately recover filtered reaction rate values on both training and generalization flames.}, keywords = {large eddy simulation; turbulent combustion; deep learning; convolutional neural network; progress variable variance; generalization}, pdf = {https://cerfacs.fr/wp-content/uploads/2021/08/energies-14-05096-v2.pdf}}