Loading...
Parallel Simulation of Generalized Newtonian Flows Using Smoothed Particle Hydrodynamics
Roustaei, Ali | 2009
1466
Viewed
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 40018 (08)
- University: Sharif University of Technology
- Department: Mechanical Engineering
- Advisor(s): Taghizade Manzari, Mehrdad
- Abstract:
- Smoothed Particle Hydrodynamics(SPH) is the oldest Lagrangian method for solving fluid equations. Fluid is approximated with particles that represent a control mass and carry physical properties su as mass, temprature and . . . Simpler formulation relative to other methods like finite volume and finite element, no need for surface traing and ease of adding new physics are benefits of this method. In this method it is necessary to find neighboring particles that are closer than smoothing lenght (h) for ea particle. is is a time consuming operation causing SPH to be a computationaly expensive method. First simulation of Power-law fluids is considered. It has been observed that using of usual viscos force computing for newtonian fluids leads to divergence of solution for Poiseuille flow. Based on a previously idea in SPH for combining of finite differece & SPH derivative computation, a new method for computing of viscos force is proposed. With new method the solution does not diverges. Also coet and 2D flow between to cylinders has been solved with new proposed method. It has been observed that the error of solution does not decrease with increasing the number of particles. e reasons of that has been discussed. A new explicit formulation of SPH is osen. Considering formolation and target maine whi is a Beowulf, a parallel method based on domain decomposition has is proposed to aeive highest performance. Processors use ghost particles for message passing and alghorithms contains 3 communication phases. SePeHr code and MPICH2 are used for implementation purpose. Speedup curves of the algorithm are presented for different number of particles and processors and it has been seen that the proposed parallel algorithm has a good ability to increase the computation speed. In relatively large number of processor, communication time grows considerably and reduces speedup. A communication hiding strategy has been suggested to solve this problem and it has been shown that this strategy is capable of increasing the parallel efficiency considerably
- Keywords:
- Parallel Processing ; Non-Newtonian Flow ; Smoothed Particle Hydrodynamics (SPH) ; Domain Decomposition ; Beowulf Cluster
- محتواي پايان نامه
- view