[aspect-devel] Averaging material properties

Bruno Turcksin bruno.turcksin at gmail.com
Mon Apr 20 09:11:44 PDT 2015


this is pretty cool. There is actually a similar problem in reactor 
physics and a lengthy literature on how to do the homogenization on a 
cell correctly. To (grossly) simplify the idea, if you can compute the 
distribution of the viscosity on some typical cases, you can use these 
distributions for your coarse calculation. This will fix the huge 
difference in results that you see.



On 04/20/2015 10:21 AM, Wolfgang Bangerth wrote:
> All,
> it's not quite ready yet, and it should have gone out earlier, but I 
> still wanted to share this before we meet online in an hour:
> Cedric Thieulot spurred me to implement a scheme by which one can 
> average material properties on cells. There are a number of 
> possibilities for this, and I implemented all I could think of -- with 
> dramatic results: much smaller over/undershots of the pressure at 
> places where the viscosity jumps, and 10x faster solves even for 
> realistic models.
> I'm attaching the corresponding section of the manual that describes 
> all of this.
> @Cedric: Does this cover roughly what you were looking for?
> @Sarah: The realistic model is yours -- it now runs in 17 instead of 
> 145 hours :-)
> Best
>  Wolfgang
> _______________________________________________
> Aspect-devel mailing list
> Aspect-devel at geodynamics.org
> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.geodynamics.org/pipermail/aspect-devel/attachments/20150420/9d64094e/attachment.html>

More information about the Aspect-devel mailing list