Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
How do you implement FEM model V&V with COMSOL ?
Posted May 29, 2015, 2:46 a.m. EDT 0 Replies
Please login with a confirmed email address before reporting spam
I followed an interesting webinar yesterday
www.nafems.org/events/nafems/2015/vvpredictive/
and I bough the book of Oberkampf. The issues discussed are essential for all modelling, but I find little on how best to apply V&V to multi-physics FEM, as we do commonly with COMSOL Multiphysics.
How do you do it ?
My way:
- analyse my model to get out the physics needed, the type of solver use (steady-state, time dependent frequency domain, pre-stressed ...) the interfaces and flow of physical quantities between the physics
- update: check the physics hypothesis versus the model dimensions are they valid ? this is less critical for COMSOL (except for ACDC/RF/Optics) than for older FEM as COMSOL is far more general, but should be done carefully, a 100m^3 oil rig looks the same as a mm^3 MEMS device on our screens but the domain to boundary ratio changes drastically and the physics hypothesis might not be valid for one or the other of our scale dimensions.
- build the models: one physics at the time !! (often by modifying some Application Library models)
- check the different sub models for simple cases against analytical solutions, and better, when available simple measured cases.
The latter is the best _verification_ approach I can do, for the rest I must rely on COSMOL's developers, and from all the other users i.e. the Forum
- I do often also check the mesh sensitivity of the results already on these simpler models
- build and run my complete model,
- check the mesh sensitivity (when possible as often these become very heavy),
- check results against a few analytical calculations and the sub models
- identify measurement check points to ask for when the model is build and is undergoing tests
- wait for test results from the device to _calibrate_ the model and to hopefully _validate_ it by complementary measurements.
Basically for one full multi-physics model, I can have made a few dozen simpler sub-models, this is now very easy and fast with COSMOL, and is the most efficient way I have found to progress with confidence in this multi-dimensional world of complex systems modelling
Comments appreciated :)
--
Have fun COMSOLing
Ivar
Hello Ivar KJELBERG
Your Discussion has gone 30 days without a reply. If you still need help with COMSOL and have an on-subscription license, please visit our Support Center for help.
If you do not hold an on-subscription license, you may find an answer in another Discussion or in the Knowledge Base.