There appears to be an error in the Solver. I'm using solver library version 5.5.0.
In a simple axisymmetric model of a hollow cylinder with the inner and outer temperature fixed, the heat flux is low by 6.28X (ie 2*pi).
Thermal resistance of a cylinder radially is ln(outer radius / inner radius)/(2*pi*(heat cond.)*width), in units of K/W, but elmer's result is missing the 2*pi.
By changing the model to Cartesian, the result matches theory.
Am I using an older solver? Is this a bug, perhaps? I ran a 3d model and it matches theory fine.
Error in Axisymmetric Thermal Model
-
- Site Admin
- Posts: 4832
- Joined: 22 Aug 2009, 11:57
- Antispam: Yes
- Location: Espoo, Finland
- Contact:
Re: Error in Axisymmetric Thermal Model
Hi
Could you define how you computed the flux. Perhaps using the the 'Temperature Loads'?
I guess there the problem is that most axisymmetric equations in their discrete form (Ax=b) are multiplied by just 'r' instead of '2*pi*r' since that's just a constant weight. Now the only thing it affects is then in the computational of the matrix residual (r=Ax-b) that is used to compute the Loads. So without testing I would say you're right. The problem is now how to make the generic matrix feature (computation of loads) aware of the varying formulations in a bunch of solvers. In matrix level it is rather dirty to use geometric information again, so the consistant remedy would be to multiply things in each equation by 2*pi which otherwise is just waste of resources. So there is something to think about...
-Peter
Could you define how you computed the flux. Perhaps using the the 'Temperature Loads'?
I guess there the problem is that most axisymmetric equations in their discrete form (Ax=b) are multiplied by just 'r' instead of '2*pi*r' since that's just a constant weight. Now the only thing it affects is then in the computational of the matrix residual (r=Ax-b) that is used to compute the Loads. So without testing I would say you're right. The problem is now how to make the generic matrix feature (computation of loads) aware of the varying formulations in a bunch of solvers. In matrix level it is rather dirty to use geometric information again, so the consistant remedy would be to multiply things in each equation by 2*pi which otherwise is just waste of resources. So there is something to think about...
-Peter
Re: Error in Axisymmetric Thermal Model
Yes, the heat flow (Watts) was calculated using a sum of the Temperature Loads on the boundaries.