Lattice-Boltzmann Solutions with Local Grid Refinement for
Transcription
Lattice-Boltzmann Solutions with Local Grid Refinement for
Lattice-Boltzmann Solutions with Local Grid Refinement for Nasal Cavity Flows Andreas Lintermann Institute of Aerodynamics and Chair of Fluid Mechanics RWTH Aachen University Aachen Germany Sonntag, 25. September 2011 Introduction Anatomy of the nasal cavity: Examples of pathological cases: • deformation (rhino-anaplasty) • malformation What is the best surgical method for the individual patient? missing center turbinate missing inferior turbinate Flow simulations support surgical decision process Sonntag, 25. September 2011 hole in septum open channel to paranasal sinus Lattice-Boltzmann method (LBGK) • • • Gas-kinetic approach for the simulation of continuum flows Advantages: • efficient for the simulation of complex internal flows • easy automatic grid generation Solve LBGK-equation iteratively for PPDFs fi (�x + ξi δt, t + δt) = fi (�x, t) + ωδt · (fieq (�x, t) − fi (�x, t)) D3Q19 model: • Solve TLBGK-equation to obtain temperature distribution gi (�x + ξi δt, t + δt) = gi (�x, t) + Ωg · (gieq (�x, t) − gi (�x, t)) Sonntag, 25. September 2011 Local grid refinement (Hänel etc.) • • Overlay of cells from different levels are required Reconstruction required: • obtain macroscopic variables by bi- or tri-linear interpolation bx f bxc available PPDFs on both levels missing PPDFs from different level O. Filippova and D. Hänel. Boundary-Fitting and Local Grid Refinement for Lattice-BGK Models. International Journal of Modern Physics C, 9:1271–1279, 1998. A. Dupuis and B. Chopard. Theory and applications of an alternative lattice Boltzmann grid refinement algorithm. Phys. Rev. E, Volume 67(6):066707, Jun 2003. Sonntag, 25. September 2011 Nasal cavity flows • • Application: Flow in a nasal cavity Re = 710 based on 125 ml/sec per averaged nostril septum paranasal sinus right nostril left nostril lower turbinate throat right nostril throat Sonntag, 25. September 2011 left nostril throat Nasal cavity flows • no levels no cells reduction time/it. • Grid refinement: Gs 1 4.744 × 107 0.0% 0.255s Gm2 2 4.335 × 107 8.6% 0.324s Gm3 3 2.277 × 107 52.0% 0.186s Temperature distribution (inflow: 20°C, wall 37°C) heating capability: 13°C increase right nostril throat Sonntag, 25. September 2011 left nostril throat Scalability and future work ideal speedup 0.262 x 106 cells/core 1.4 1.2 speedup 1 0.8 0.6 0.4 0.2 0 1024 • • 2048 3072 number of cores 4096 Scale-up tests on BlueGene System JUGENE at FZ Jülich good scalabilty 9 Maximal number of cells so far: 1.074 × 10 future simulations will require a massive amount of cells Sonntag, 25. September 2011 Scalability and future work • • Scale-up tests on BlueGene System JUGENE at FZ Jülich good scalabilty 9 Maximal number of cells so far: 1.074 × 10 future simulations will require a massive amount of cells simulations of the complete respiratory system Sonntag, 25. September 2011 Conclusion • Analysis of the flow in the nasal cavity based on • • • streamline characteristics backflow and vortex characteristics Local grid refinement: 52.0% reduction for three levels more efficient computation • Heat convection / diffusion allows analysis of heating capability • Fast and easy mesh generation with a Cartesian grid generator • Good scalability and well suited for Massive Parallel Computing efficient tool to simulate flows in highly intricate geometries Sonntag, 25. September 2011 Thank you for your attention Sonntag, 25. September 2011