شبيه سازي جريان سيال با كمك روش شبكمه بولتزمن و روش چند شبكه اي و استفاده از پردازش موازي

چكيده انگليسي :

Fluid Flow Simulation Using Hybrid Lattice Boltzmann MultiGrid Method and Employing Parallel Programming Ehsan Fathi e fathi@me ac ir July 13th 2011 Department of Mechanical Engineering Isfahan University of Technology Isfahan 84156 83111 IranDegree M Sc Language FarsiSupervisor M Ashrafizaadeh Assistant Prof mahmud@cc iut ac ir AbstractThe standard Lattice Boltzmann LB method is an approach for fluid flow analysis This method uses theuniform mesh in the whole domain To model complicated case studies and to achieve more accuracy thewhole system s meshes should be made finer This causes an exponential growth in computation s cost witch its rate is related to the ratio of the lattice space in mesh systems Therefore advanced computationalmethods are being interested MultiGrid MG methods are a prime source of important advances inalgorithm efficiency finding a rapidly increasing number of users Unlike other known methods MG offersthe possibility of solving problems with N unknowns with O N work and storage for large classes ofproblems MG method can dominate the low convergence rate in complicated problems Damping lowfrequency error is the major time consumer part in iterative methods In iterative scheme in the primitiveiterations the high frequency part of error is diminished MG method damps the low frequency error of finegrid by restricting it to coarse grid Parallel processing is another method to decrease the solution time Thismethod can be applied with computational fluid dynamics simultaneously In this thesis a combination ofLB MG method and parallel processing is investigated This combination is used for simulating the fluidflow especially in this project for simulating fluid flow in a lid driven cavity and step flow in channel byusing parallel processing An introduction of MG method and LB method are presented The LB equationsare derived on the basis of density distributed function values which related to probability histogram ofparticle s situation In according to this fact that LB equations are non linear non linear MG equations areneeded Therefore the Governing equations of the MG and LB methods are derived In non linear solutionsthe appropriate initial guess should be employed The nested algorithm for this need is applied In theresults season for original LB modeling the simulation fluid flow in lid driven cavity is presented Thenthe results are validated with those in the literature The derivation of reformed non linear MG methodequations on the basis of LB equations is also presented In these equations in addition to the value offunction the residual is restricted Then the parallel processing algorithm is applied The effects of severalparameters on the precision and the speed of numerical simulation are investigated The quality of theresults are compared and discussed In this project by using the combination of LB MG and parallelprocessing method more than 150 times speed is achieved in compare with original LB method In theappendix an example of parallel programming using MPI is presented The innovation of this thesis is anexplicit combination of the LB and MG method by employing parallel processing architecture to increasethe computational performance Key Words Lattice Boltzmann MultiGrid Speed up Parallel Processing

فتحي، احسان

شبيه سازي جريان سيال با كمك روش شبكمه بولتزمن و روش چند شبكه اي و استفاده از پردازش موازي

افزايش سرعت