Discussion Forum

Hi

there are several comments I can make:

1) you should update to the latest patch, check with your system admin or your comsol rep
2) your model:
a) even if there are a few temperature references in the SPF physics, in fact "T" is not globally defined in your model, try to evauate "T" on a point in the derived values, you will get an error message. Easiest way around: add a parameter T = 700[K], this defines a static temperature value, that COMSOL might everwrite later, particularly if you add a HT or other physics
b) your mesh does not have any boundary layer, try to use the default "physics generated mesh" it is better
c) your initial conditions all = 0 (the default) are not ideal, try to add a reasonable velocity profile and a pressure drop i.e. à la Poiseuille (check the Forum try a search) this will help the solver to converge better
d) yo seem to have edited the solver sequence or the model several times. You should "delete" all your solver and let COMSOL regenerate a fresh default new one (this will also delete the plots so you need to redefine them) The reason is that you have probably modified te sovler a few times, then COMSOl leaves the solver as is, even if you change the physics and that these changes implies changes in the solver, COMSOl assumes that you know what new specialities are needed. Thereore deleting the solver sequence and regenerating it from scratch is a good way to get a fresh and clean solver sequence, based on the latest physics and main solver node settings
e) in fact for the "spf" you do not need the steel at all
f) you you want to use mode advanced fluid structure temperature exchange, then the best is to use the "nitf" physics which combines SPF and HT full for you, but be aware its tricky to get to solve as its rather setting and initial condition sensitive

--
Good luck
Ivar

Hi Ivar,

I did what you suggested to me, the segregated group 1 error dont occur more. I tried to put the poiseuille law to pressure drop , but initial condition error occur. I used the following equation:

DP = 8*mu*Q*L/pi*R^4

obs: Q = V/pi*R^2

So, to my model with values:

DP = (8*mat1.def.mu[Pa*s]*3.25[m]*(0.000001[m/s]/(pi*(0.009[m])^2)))/(pi*(0.009[m])^4)

my initial velocity field is 0.000001 [m/s] because the pump operation is 0.002 - 0.012 [m^3/s].

I tried a lot of values in the initial pressure and initial velocity field, but the solution dont converge. i dont know whats wrong in my model.

About the mesh... I tried to use the "physics generated mesh", but i needed to use an extra fine model to my geometry not deform. this result in a high computational cost and when i compute the solution the message of "out of memory" appear , so a way around: I created two meshs, one to domain 1 and other to domain 2. The computational cost is lower.

You say me to use the nitf physics, but this physics use the compressible formulation for the fluid and I want to use imcompressible formulation, so I think the only way is to use the spf and ht physics. Am I wrong?

I tried to attach the file of my model, but the size is to big.


Can you say me what s wrong with my initial conditions?

best regards


Gustavo

Hi

if you have a fully symmetric model, why not us e the 2D-axi, it's much cheaper computationally, and just as precise, this allows you also to play more with initial conditions and other parameters, at least until you are happy with your model, then only you can try it out in fully 3D, you gain a lot of time like that.

For the mesh, as you have a long thin pipe you can use a sweep mesh as all your gradients (velocity and heat when that comes) will be radial and not along the pipe axis you can live with elongated mesh elements (along the axial direction) without major loss of info. This is valid for laminar flow, much less for turbulent ;)

--
Good luck
Ivar

Hi Ivar..
I have attached my gemotry file.I could not solve for Darcy flow through the domains-hydrogen channel,as i specified in the selections list.I tried also,with laminar flow in the same domain.I couldnt solve for the velocity profile. I get error in multiphysics compilation or error in evaluating variable spf.U,if I use laminar.My problem - hydrogen channel has laminar flow and the fluid diffuses into the underlying next layer and reaches its next for reaction.I'm very much new to COMSOL but given a complex project.I'm struck up with the initial flow solving itself,and couldn't proceed next.Please help me to resolve the issue.I have attached the assembly geometry since the mph file faced uploading error and aslo have notifiedthe channel in the png file attached.Please bear with me and help for the problem

Hi

setting up such a model, even if it seems simple, and learning COMSOL the same time is far from trivial. I can only suggest to start simple, i.e. take a thin straight tube of correct section and representative length. Then get the laminar flow running, add more complication, and i.e. diffusion through a wall, turbulence, ...

But only one at the time.

Then check the parameters, can the model be simplified, does a long tube need to be simulated in detail through its section or can it be made 1D ? else you need a cluster to solve it, and more time to learn and validate that approach

And others out there might have further suggestions ;)

--
Good luck
Ivar

Hi

I am trying to stimulate an array of magnets and coils to find the induced emf in the coil. But i am getting the error message:
Failed to find consistent initial values.
Segregated group 1

System matrix is zero.
Last time step is not converged.

I am not able to attach the file due to file size being large, Can anyone help