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.
very low value of velocity during natural convection( e-23)
Posted Mar 12, 2012, 10:28 p.m. EDT 2 Replies
Please login with a confirmed email address before reporting spam
Hi,
I am working for natural convection in a heating cavity which has opening at the top and I have placed it in a air enclosure.My simulation works well I have introduced buyouncy effect also.But the velocity profile is very low like 1.5 e-23 something.Even cavity bottom is at high temperature around 973K.
I am really confused whether my model is correct or it has some error.As most of the natural convection problem i have seen the air velcity is not so low.
Please if some one could guide me I ll be highly thankful
Thank you in Advance
Tassawar
I am working for natural convection in a heating cavity which has opening at the top and I have placed it in a air enclosure.My simulation works well I have introduced buyouncy effect also.But the velocity profile is very low like 1.5 e-23 something.Even cavity bottom is at high temperature around 973K.
I am really confused whether my model is correct or it has some error.As most of the natural convection problem i have seen the air velcity is not so low.
Please if some one could guide me I ll be highly thankful
Thank you in Advance
Tassawar
2 Replies Last Post Mar 13, 2012, 6:41 p.m. EDT
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
this could well be a scaling issue, do not forget that binary numbers, with complex formulas as we use in physics mainly has only 6-8 usefull digits so for the same simulation without rescaling one should not expect ratios of the same variable much above 1:1'000'000 (even in double floating numbers)
I f one need more there are ways to stop and restart with a rescaling, or using log() scales but this requires to rewrite most of the equations ;)
I'm not saying you cannot simulate with number < 1E-6 it's just the ratio between smallest and largest values of the same variable. If you are in the 1E-18 to 1E-24 range, you need to rescale your variable by 1E21 or there around. Normally COMSOL does this automatically at solver start (check the log, or ask for a detailed log) but sometimes one must do it also by hand (in the solver dependent variables node). Check the doc on "scaling". This is not a special COMSOL feature (apart that it's easier to access to the scaling than in many other programmes) it's linked to the binary numbers representation and matrix inversion algorithms, pure math number theory ;)
--
Good luck
Ivar
this could well be a scaling issue, do not forget that binary numbers, with complex formulas as we use in physics mainly has only 6-8 usefull digits so for the same simulation without rescaling one should not expect ratios of the same variable much above 1:1'000'000 (even in double floating numbers)
I f one need more there are ways to stop and restart with a rescaling, or using log() scales but this requires to rewrite most of the equations ;)
I'm not saying you cannot simulate with number < 1E-6 it's just the ratio between smallest and largest values of the same variable. If you are in the 1E-18 to 1E-24 range, you need to rescale your variable by 1E21 or there around. Normally COMSOL does this automatically at solver start (check the log, or ask for a detailed log) but sometimes one must do it also by hand (in the solver dependent variables node). Check the doc on "scaling". This is not a special COMSOL feature (apart that it's easier to access to the scaling than in many other programmes) it's linked to the binary numbers representation and matrix inversion algorithms, pure math number theory ;)
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Thank you so much :)
Regards,
Tassawar
Regards,
Tassawar