Discussion Forum

Hi

I wouder if you are not mixing a little the the cases:

Stationary:
You have no "t" defined, no "rho*d^2_u_/dt^2" terms or second order omega terms, only stable (in time) BC's and you wait infinite long to reach a stationary stable situation. Check the equations, no .../dt terms

Transient / time dependent:
You define time varying BC and you look at the response to this time dependent change (you have some BC which depends directly on "t" or via an initial condition etc The equations add in i.e. the "rho*d^2_U_/dt^2" term in solid, check the equations you have a .../dt term. You must define time steps for t.

harmonic / frequency domain:
You apply a continuous frequency of sinus type of amplitude and some phase is defined in the BC (but no "t" dependet BC here), the equations add in the "rho*omega^2*ampl" harmonic response term in the equations. Check the equation, you have omega and/or omega^2 terms, but no .../dt terms. You must define a frequency or a frequency range for a frequency sweep, one frequency at the time

frequency - stationary
You apply a continuous frequency type sinus wave with amplitude, the frequency analysis calculates the sinus response for ONE period, deduces average rms loads cases and apply those to the stationary case. There are no .../dt terms in the equations, You must define a SINGLE frequency

frequency - transient:
You apply a continuous frequency type sinus wave with amplitude the frequency analysis calculates the sinus response for one period, deduces average rms loads cases and apply those to the transient case (with BC depending on the time "t". There are .../dt terms in the equations. You must efine ONE frequency and time steppsfor "t"

At least this is how I read the doc.
General hypothesis for frequency-stationry and frequency-transient : the physics effect is much slower than the frequency defined, so extracting a one period rms value and applying this as average load makes sens. Global effects will appear, the effects at the frequency and jst below are ignored

I hope my reading of the doc is correct, and that I'm clear ;)

--
Good luck
Ivar

Dear Ivar

Thank you for reply.

I want to clear some points

i apply voltage with amplitude 15Kv and freq.=1 khz

i want to see the effect of this voltage on the temperature of the cable termination.

when i tried to solve this problem in time dependent sol. with time range (0,1,1000). It took a very long time and it stopped the calculation because ( Out of memory !!!!)

So i tried to solve it in frequency transient to reduce the time of calculations, but it gives me error

i attached my model, please take a look and give me your advise

thank you

Hi

indeed I get also a strange error message, it could be an 4.0 or translation effect (I'm in 4.2a)

However the transient - stationary is OK and gives a 35 to 40 °C temperature increase depending on how you add some domains to a Heat transfer in Solid nodes.

But still I fear a very high sensitivity to the resisitivty when you mix 6E7 to 8E-16 S/m side by side

And I would advise fine mesh and boundary elements along boundaries for such large conduction changes
Check the current denity, but use a log10(jh.normJ+eps) scale (the eps is there just to see the last domain where the results is 0

Alyo I would set the terminal to boundary 2 and the gnd to boundary 8, not that it changes much on the stationary temperature, but on the current density yes (and you do no longer need the eps)

--
Good luck
Ivar

Dear Ivar

i tried to make frequeny - transient in my model

but i get error

i attach the file
please take a look and give me your advise

Dear ivar

i tried to solve my model in time dependent sol.

but i want to solve for 1000 sec

i put the time rang (0,10,1000)

i see that the programme calculate at 0.001 sec,0.002.,0.003 ........etc, so that the time of calculation is horrible!!!!!


what can i do

Hi

but why frequency - transient and not frequency stationary ? for 1000 seconds its almost a stationary response

If you use a terminal instead of the potential BC you will get lumped global definition for current and voltage hence power, this should give you the values for your average power injected, then you could try a standard time series

And anycase with your Frequency Transient in the time-Dependent Solver sub node - Time stepping tab use a "strict" and not a freetime stepping then it will go better, but it will also show you that your mesh is too coarse and that you have a singularitiy that makes numerical issues for your solver

--
Good luck
Ivar

Dear Ivar
Thank you for your reply

Actually i make mesh more fine and i put the time dependent solver 'strict'

but still give me error why ?

please help me , my model is attached

and my problem is to run at 1000 sec to get the temperature distribution


thank you

Hi

that is because you still have your DV=wv1(t[1/s]) in there and you do not theed the sinaus time dependent input just the amplitude of your sinus. That corresponds to t=2*pi/4/1[kHz], so the simplest is to define the "t" in the Global Parameter, or replace DV in the terminal by the amplitude 15[kV]

Then it works.

By the way you are looking at the temperature by Joule effects in the isolation, but you are not considering the heat generated by the 15kV creted current in the main Cu cable, that will add some degrees too.

--
Good luck
Ivar

Dear Ivar

i define " t " in global parameter t=2*pi/4/1[kHz]

when i run the simulation it gives me error .


when i removed variables (wv1(t[1/s]) and put DV = 15 [KV] in parameters. it gives me error as well

what can i do?

i attach my file and i am very sorry for disturbing you


thanks

Hi

I'm not sure what to propose for you, as in my 4.2a version it works, it could be linked to your early v4 version, COMSOL has been quite improved since.

Last thing is to say your 1k[Hz] 15kV amplitude signal, boils down to 15/sqrt(2)[kV] rms and use a classical stationary analysis with this voltage

Well NO that is not working because your heating is mostly capacitive coupling at 1kHz I suppose, as I get a very different result comparing the two cases frequency-stationary and normal stationary with rms amplitude

As I'm not used to the frequency stationary (havent really ued it extensively so far) I would have to dig into the doc to find the different variables to better undersand myself ;)


I have joined what I see, file FreqStat is your mesh in freq stationary, then I made the mesh much denser for the regions where the temperature is higest, and added boundry mesh on these 2 domains,"...Dens2", so your mesh isratherrough, ut that is easy to change once it runs ;)

If I use a RMS DV and standard stationary I get the last file, and its almost the same for all mesh density

--
Good luck
Ivar

and the last file highest density frequency- stationary

I was a bit too quick, the finer mesh is not that much different (I had left the 1/sqrt(2))

In the mean time, I have notice that the temperature sems to e essentally dependnet ont he dielectric constatn of your first screen, which gives me the impression you are looking at the capaciative induced losses

--
Good luck
Ivar

Dear Ivar
thank you very much for your great help


i want to ask if i want to run my model for 1000 sec in time dependent solve . what can i do?

when i put the rang (0,10,1000)

it takes very very long time, i see that calculations are made at 0.01s,0.02,0.03 so takes horrible time


how can i decrease the time of calculations?

however, i am intersted only in the max.temp we can reach at the end of 1000 sec


thank you

Dear Ivar

a second question please

i want to put the conductivity of material as a function of the electric field

the progream should claculate the electric field then clacule tle corresponding conductivity of the materia

sigma= 0.5 * exp (0.08*E)

i do not know how to write this expression in comsol , and how i define electric field in the expression
ou

thank y