You are on page 1of 13

Parallel Computing for solving ANSYS-CFX problems

MAE 405/505 Lab By Jay Udayan Patel Graduate Student Assistant University at Buffalo

The main time we spend while working on ANSYS-CFX is on Meshing and Solver CFX-Solver itself will take 50-80% of the total time spend on whole problem, depending on its size(i.e. No of elements) Whatever we are doing in Design Modeler, Mesh and CFX-Pre will convert into a definition file(.def) at the end We are suppose to solve this file to get the result file(.res)

How can we reduce this time? One way is to use parallel computing What is parallel computing? Lets take an example Imagine a man is given a task to paint a wall and let say he took 4hrs to complete the job Now, if he gets help from other 3 people, they can get the job done in like 1hr(Ideal case)

The same concept is in Parallel computing Instead of solving (let say NS Eqn) equations on single processor we use multi-processor to solve our equations The method which ANSYS uses to do so is called MPI(Message Passing Interface)

Coming back to Furnas 1019

Lets take a similar example of CFD

Each Processor will operate separately All Slaves Processors will return their results to the Master/Main Processor at the end of each iteration So at the end of entire run, Master Processor will create the main result file(.res) This file is exactly same as one obtained by solving in serial run

The two option we have in ANSYS for running .def file in parallel are MPI local parallel and MPI distributed parallel

Today we will just see the effect of Local Parallel v/s Serial and some Post processing shortcuts

So lets begin..

You might also like