Discussion Forum

Hi

if you have the points as a list in a file use rather the Interpolation function, you can load the file directly, the in the BC you call the function i.e. int1(t[1/s]) (if you have not filled in the Units in your definition, and if the time base is seconds, or just int1(t) if you have defined the units in the function.

If your function is a space dependent and not time dependent, you load it in the same way (3-4 columns) then you call it with a int1(x,y) or int1(X,Y) or, if you chewck the "use space coordinates you can call it simply int1 without any arguments (add a third argument Z if you are in 3D)

These functions can be defined in the Definition node to be globally avaialbel, or in the material sub node if its purely a material interpolation function

--
Good luck
Ivar

When i am using Interpol Function , do i require smoothing or not ? because the graph i get looks like the jpg file that i have attached

Hi

indeed your line looks rahter spiky, why not try some smoothing, but with the cubic spline it might even oscillate more, in which case some smoothing could be applied, probably outside COMSOL on the raw data.

--
Good luck
Ivar

Hi Ivar !

i am happy about few moderators who are really really active in solving ppl's problem. I am sure you rank in the top few :)

Even though we thought that the lines looked spiky the solution converged . But the temperatures profile in transience did nt follow my exact path .. it rather followed kind of a smooth curve.. but experimentally i have a curve which is spiky as well based on power input .

Input : Spiky/ piecewise linear power input
Output : (model) Smooth temperature transition
Output: (Experimental) Spiky/ piecewise temperature transition

now to match the experiment data .. should i consider using an approximation called efficiency of heaters varies with temperate . ?

i have attached the required data and graphs .. ! please suggest ! <------------ (This statement is for all !!)

Hi

that depends on where and how you measure your temperature, there is the heat diffusivity to consider, even if you hange the power abruptly, the heat needs time to travel through any material, this has an impotant "damping " effect and will smoothen a T value, particularly if its obtained by integrationg over a given boundary or domain

--
Good luck
Ivar

the temperatues are taken from the exact same location .. i.e , the experimental and from the model . No averaging has been done by me ... only problem can be due to thermal diffusivity.
while it can smoothen out in calculation/ model , why cant it smoothen it out in practical as well ?

I have another doubt as well.

I need to incorporate natural convection into the model. as you know, comsol is poor in doing it for a highly complicated 3D Geometry with so many degrees of freedom to establish the natural convection part .. ! is theer any way i can incorporate and get a good approximation ?

Hi

perhaps your experimental sensor is seeing more external temperature than the internal ones you try to estimate, or the material properties are not the same, there can be many things ;)
worth a second check with some simple hand calculations to see if your model seems OK

--
Good luck
Ivar

what about the second part of my question ? any possible way to solve the natural convection of air in the central cavity ?

Hi

I believe its not really a COMSOL issue, but rather a size of your WS issue (for convection and gravity boance development of motion in any body)

Have you checekd the model librbary "cold water glass" or the light bulb examples ?

Try using NITF as Search key

--
Good luck
Ivar