I have responded to the first message (which got posted with some delay, as I have to approve first-time posts of new mailing list members as an anti-spam measure).
On 10/26/2012 11:15 AM, Bjarke Dalslet wrote:
Hello sfepy users.
I am doing thermal simulations of (very big) electric resistors in sfepy. Nothing fancy, but as the resistors get very hot, i would like to use a thermal conductivity that depends on the temperature.
At the moment i am using the laplace term: int( c \nabla q \nabla p) where c is the thermal conductivity, q is the test (temperature field) and p is the temperature field.
I have come up with two possible approaches to make c dependent on q:
1: restrict c(q) to a linear function c(q) = c_0 + \alpha q. I think this would allow me to do this: int( c(q) \nabla q \nabla p) = int( c_0 \nabla q \nabla p) + int( \alpha q \nabla q \nabla p) That would allow fast steady state solutions, although the second term is not yet implementet (and judging from the laplace source, it seems like some work to do it).
2: Iteratively seek the solution, i.e. adjusting c(x,y,z) after each timestep. This would be fine when looking at transients, but for steady state it would be much slower than solution 1.
I am new to sfepy and also to weak form FEM, so I ask anybody with experience in this: Are the above approaches silly, and does a better one exist?
Kind regards Bjarke Dalslet