Statistical analysis of high-Rayleigh
Transcription
Statistical analysis of high-Rayleigh
Statistical analysis of high-Rayleigh-number turbulent convection data (SBDA003) Ambrish Pandey1,3, Janet D. Scheel2 and Jörg Schumacher1 1Institute of Thermodynamics and Fluid Mechanics,Technische Universität Ilmenau, Germany 2Department of Physics, Occidental College, Los Angeles, USA 3Department of Physics, Indian Institute of Technology, Kanpur, India Motivation 2. Extreme thermal dissipation events 1. Analysis of the large-scale circulation in the closed convection cell by means of Proper Orthogonal Decomposition Full cell 2. Analysis of the evolution of extreme events of thermal dissipation and kinetic energy dissipation Bulk volume Level 0 Track 4th-order moment of thermal dissipation rate Original production run 3. Analysis of energy spectra in the bulk of convection cell and comparison with Kolmogorov spectra in isotropic turbulence of ref. [2] (0) tout ⇡ 0.72Tf Rayleigh-Bénard (RB) Simulation Model @~u 1 + (~u · r)~u = rp + ⌫r2 ~u + g↵(T @t ⇢0 r · ~u = 0 @T + (~u · r)T = r2 T @t Level 1 Formation of the temperature front between two colliding plumes 1st rerun with finer data output T0 )~ez (1) tout = Level 2 Detailed analysis of the production terms in the gradient balance equations 2nd rerun with finest output Spectral element method (nek5000) with polynomials of order N=13 [1]. Production jobs on up to 131,072 BG/Q cores for small-scale structure studies. (2) tout = • Total amount of simulation data (9 parameter sets): approx. 45 Tbyte 1 (0) t 5 out 1 (1) t 10 out Extreme Dissipation Event • One simulation snapshot for the biggest simulation for RB convection in a liquid metal flow = 130 GByte Thermal Plume Collision • Data in subvolumes of the original unstructured element mesh have been spectrally interpolated onto a uniform Cartesian or cylindrical grids. • Support by JSC personel for improvement of parallel IO. • Analysis of different temperature and velocity gradient production terms shows that energy dissipation lags behind thermal dissipation 1. Proper Orthogonal Decomposition of data • Extreme event is caused by cessastion of large-scale flow References • Expansion of velocity and temperature data into a basis of empirical orthogonal modes which are sorted with respect to their kinetic energy (see left panel) and/or scalar variance • Subtraction of large-scale velocity and temperature fields from original data • Analysis of derivative moments of the remaining fields (see right panel) does not change the scaling results which were reported in [3]. [1] J. D. Scheel, M.S. Emran and J. Schumacher, Resolving the fine-scale structures in turbulent Rayleigh-Bénard convection, New. J. Phys.15, 113063 (2013). [2] A. Kumar, A. G. Chatterjee and M. K. Verma, Energy spectrum of buoyancydriven turbulence, Phys. Rev. E 90, 023016 (2014). [3] J. Schumacher, J. D. Scheel, D. Krasnov, D. A. Donzis, V. Yakhot and K. R. Sreenivasan, Small-scale universality in fluid turbulence, Proc. Natl. Acad. Sci. USA 111, 10961-10965 (2014).