# Difference between revisions of "SC²S Colloquium - September 21, 2016"

(→Evangelos Drossos: Parallel-in-Time Simulation for Fluid Dynamics Using Lattice Boltzmann Methods) |
|||

Line 10: | Line 10: | ||

== Evangelos Drossos: Parallel-in-Time Simulation for Fluid Dynamics Using Lattice Boltzmann Methods == | == Evangelos Drossos: Parallel-in-Time Simulation for Fluid Dynamics Using Lattice Boltzmann Methods == | ||

− | + | The modeling and simulation of flow phenomena, ranging from the molecular to the macroscopic scale, has proven to be a challenging task in the area of computational fluid dynamics. Of particular interest is the Lattice Boltzmann method on the mesoscopic scale and the Navier-Stokes equations on the macroscopic scale. Previous research has shown that fluid dynamics simulations can benefit from parallel computation based on a spatial decomposition of the underlying domain. However, for a fixed-problem size any further benefits to speedup and parallel efficiency rapidly diminish for a small number of grid cells per core. To counter this problem, we examine a parallel-in-time method for fluid dynamics simulations based on the parareal algorithm. In this context a predictor-corrector scheme is used. In this scheme, results that are calculated on a grid at lower resolution are used to predict initial values for a solver running in parallel on a higher-resolution grid. As part of this thesis | |

+ | , we first present theoretical considerations and simulation results for a coupled Lattice Boltzmann solver running on both grids. We subsequently develop a model for a parareal simulation using a predictor method on the macroscopic scale based on the Navier-Stokes equations and a fine solver based on the Lattice Boltzmann method. | ||

== Michael Zellner, Interaktive Strömungssimulation auf Android basierten Endgeräten mit entfernter Berechnung == | == Michael Zellner, Interaktive Strömungssimulation auf Android basierten Endgeräten mit entfernter Berechnung == |

## Latest revision as of 10:48, 15 September 2016

Date: |
September 21, 2016 |

Room: |
02.07.023 |

Time: |
3:00 pm, s.t. |

## Evangelos Drossos: Parallel-in-Time Simulation for Fluid Dynamics Using Lattice Boltzmann Methods

The modeling and simulation of flow phenomena, ranging from the molecular to the macroscopic scale, has proven to be a challenging task in the area of computational fluid dynamics. Of particular interest is the Lattice Boltzmann method on the mesoscopic scale and the Navier-Stokes equations on the macroscopic scale. Previous research has shown that fluid dynamics simulations can benefit from parallel computation based on a spatial decomposition of the underlying domain. However, for a fixed-problem size any further benefits to speedup and parallel efficiency rapidly diminish for a small number of grid cells per core. To counter this problem, we examine a parallel-in-time method for fluid dynamics simulations based on the parareal algorithm. In this context a predictor-corrector scheme is used. In this scheme, results that are calculated on a grid at lower resolution are used to predict initial values for a solver running in parallel on a higher-resolution grid. As part of this thesis , we first present theoretical considerations and simulation results for a coupled Lattice Boltzmann solver running on both grids. We subsequently develop a model for a parareal simulation using a predictor method on the macroscopic scale based on the Navier-Stokes equations and a fine solver based on the Lattice Boltzmann method.

## Michael Zellner, Interaktive Strömungssimulation auf Android basierten Endgeräten mit entfernter Berechnung

It is possible to calculate fluid dynamics simulation on modern smartphones. The LB2M android application realizes this by using the lattice Boltzmann method. As the memory and computing power of a smartphone are limited, only small scenarios can be calculated. As part of this guided research, a server was developed that calculates time steps and transmits solely the results to a smartphone. The smartphone itself is only used to display the results.