Importing previous results as a starting point for future simulations

mcintoshjamie · October 28, 2009, 19:22

I'm conducting a research project for my Masters degree to do with profiled endwall design for a linear cascade. The project uses a genetic algorithm that creates perturbations across the end wall and then sends these designs to be meshed in gambit and then analysed in fluent.

Currently the system takes 2.5 hours per design with 30 simulations per generation. I'm trying to optimise the system so that the calculation time is decreased using parallel processing on a myrinet system and also by using a 'surrogate model'.

Essentially what I'm trying to do is use a planar end wall simulation as the starting point for calculations on each new design as the hope is that this will significantly reduce the number of iterations required to converge on a solution.

I've had a look on the internet and the help files in fluent and couldn't find anything specific about this. I would have thought it's possible but can't think of a way to make it happen. Any help is appreciated.

Thanks
Jamie McIntosh

mcintoshjamie · November 15, 2009, 07:38

Solved:

1. File-->Read-->Case (as normal)
2. File-->Interpolate*
3. Read...

Iterate towards solution as before...

*Requires interpolate data file from previous results:
1. File-->Interpolate
2. Write...

Jamie

October 28, 2009, 19:22	Importing previous results as a starting point for future simulations	#1
mcintoshjamie New Member Jamie Join Date: Oct 2009 Posts: 5 Rep Power: 17	I'm conducting a research project for my Masters degree to do with profiled endwall design for a linear cascade. The project uses a genetic algorithm that creates perturbations across the end wall and then sends these designs to be meshed in gambit and then analysed in fluent. Currently the system takes 2.5 hours per design with 30 simulations per generation. I'm trying to optimise the system so that the calculation time is decreased using parallel processing on a myrinet system and also by using a 'surrogate model'. Essentially what I'm trying to do is use a planar end wall simulation as the starting point for calculations on each new design as the hope is that this will significantly reduce the number of iterations required to converge on a solution. I've had a look on the internet and the help files in fluent and couldn't find anything specific about this. I would have thought it's possible but can't think of a way to make it happen. Any help is appreciated. Thanks Jamie McIntosh

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Using results form previous run as inlet condition	lauren	CFX	3	June 9, 2008 14:32
[Gmsh] Gmsh and samplesurface	touf	OpenFOAM Meshing & Mesh Conversion	2	December 10, 2007 03:27
Exporting Fluent results and importing into StarCD	Jim Loebig	FLUENT	0	July 20, 2007 17:26
bad results when compared with other simulations	Le Stanc	CFX	12	November 8, 2006 02:47
CFX4.3 -build analysis form	Chie Min	CFX	5	July 13, 2001 00:19

November 15, 2009, 07:38		#2
mcintoshjamie New Member Jamie Join Date: Oct 2009 Posts: 5 Rep Power: 17	Solved: 1. File-->Read-->Case (as normal) 2. File-->Interpolate* 3. Read... Iterate towards solution as before... *Requires interpolate data file from previous results: 1. File-->Interpolate 2. Write... Jamie