6744

6286

كارشناسي ارشد

تبديل انرژي

اصفهان: دانشگاه صنعتي اصفهان، دانشكده مكانيك

1390

[هشت]، 85ص : مصور، نمودار

ص.ع. به فارسي و انگليسي

محمود اشرفي زاده

احمدرضا پيشه ور

17/3/91

ابراهيم شيراني، احمد صداقت

1396/10/06

كتابنامه

مكانيك

مهندسي مكانيك

ID6286

به فارسي و انگليسي: قابل رويت در نسخه ديجيتالي

Parallel Solution of Euler Equation on GPUs Mohammad Saber Zamanpour Zahraee Ms zamanpourzahraee@me iut ac ir Date of Submission 2012 02 18 Department of Mechanical Engineering Isfahan University of Technology Isfahan 84156 83111 Iran Degree M Sc Language FarsiSupervisor M Ashrafizaadeh Mahmud@cc iut ac irAbstractCompressible flow problems are frequently encountered in mechanical engineering In fluid flow simulation at highReynolds number due to reduction of the viscous effects compressibility can be neglected By dropping the viscousterm out of the Navier stokes equation the Euler equation is obtained To investigate supersonic and hypersonicflows the Euler equation can be used Simulation of compressible flow is usually very time consuming hence itscomputational cost is high Parallel processing with Graphical Processing Unit GPU has secondly been widelyused to reduce such costs In this work numerical solution of the governing equations for inviscid compressible flowusing HLL HLLC and WENO5 methods have been described and the accuracy of them for one dimensionalproblems for which the analytical solution of Riemann is available has been investigated to validate results Theexplosion problem in two dimensions has been solved The present solution has been compared with the Godunovmethod The solver of the Riemann problems in the present work is almost the same as the exact solver however anadditional source term has been included to implement the boundary condition for the 2D problem The immersedboundary method has been used In this method a ghost fluid is used for the extension of the boundary conditionwith high accuracy Using this method one can use the rectangular Cartesian grid method instead of considering aboundary fitted grid This will simplify the implementation costs To validate the immersed boundary method thetriple points resulted from the trailing works will be located by an interpolation method and the result will becompared with experimental and numerical results reported by others The computed codes for the simulation of thecompressible fluid flow have been written using the CUDA programming language and may have been executed inparallel on the of graphical processing units The GPU that has been use for this work was GerfoceGTX580 with512 cores and 1536 MB accessory memory and 192 Gbyte sec speed of data transfer The speed up ratio is directlyrelated to the number of cores and it will be effectively increased as the number of computational cores is increased The speed up ratio for the HLL code for the one dimensional problem was 30X and for the two dimensional case itwas more than 80X For the WENO5 code we achieved a 36X speed up ratio for the 1D case and we reached 174Xspeed up ratio for the 2D problem The above mentioned speeds up ratio are very considerable and for compressiblesolvers have not been reported previously In this research the effect of increasing the size of the block on the speedup ratio has been studied and we showed that best size of the block for the used graphical card is 512x1 Usingimmersed boundary method the speed up ratio was decreased however with increasing the domain size the speed upratio will approach to the speed of ratio of the other method which does not use the immersed boundary technique Keywords Compressible flow CUDA HLL HLLC WENO Immersed Boundary

محمود اشرفي زاده

احمدرضا پيشه ور

ابراهيم شيراني، احمد صداقت