Discussion Forum

attachd is 4.2a file.
hope someone can give me some hint

Hi

well for me, you are fist of all starting with all initial conditions = 0 but BC's that tells me your are far from "0". OK you have used a step function with a smooth inlet, but you have a rise time of 0.1 ms while your first time step is 10 ms or some 100 time larger. COMSOL cannot see that rise when starting and then the initial conditions are bad, so it tries, then backtrack , tries again etc, you see it from our solver log.

there are a few ways around:
1) add a few steps so that the flow builds up while you are asking for it with a few points per rise time, or adapt your rise time to the time stepping you have

2) my preferred in addition to 1), add a pressure drop à la Poiseuil or just a linear pressure drop from inlet to outlet along with Y (in your case) and add a parabolic velocity profile (by the way why none "no-slip" on your FSI interface ?)

3) solve the FCD case alone with a fixed structure and use that as initial flow conditions for the FSI

but there might be other points too, I have no time just now to try it out

--
Good luck
Ivar

HI, lvar,

Thank you for your insightful information.
it give me a new way to try build the model.

I will update you the latest information once i get back to my office.

Thank you for knowledge sharing.

Dear lvar and my friends,

Thanks for lvar for bring me back to using COMSOL, as it looks has a lot of secret recipe now :p
1. Refer to the rectangular 4.2, the wall is add in to formulate no slip condition, and i am able to formulate the same layer as fluid-solid boundary layer. The stationary solver ok, as you can see, result looks good, however, when i enable time dependent solver, it cause problem.Any idea?

2. Second model is little bit complicated. as i try to use laminar flow as initial condition, then i process to FSI module. Then it appears "wrong number of dof in linearization point" ,good to know it.

Anyway, we stick to 1st one.

Any idea?

will it still work if we release the fixed constraint ? I think that is more close to the blood vessel problem, right ?

Thanks for your discussion and I did not get any help from comsol support either....:(

Hi, Siyuan,

Thank you for keep the post alive.

What makes you think to remove the fixed constraint?

COMSOL support try their best already, we can discuss ourselves.

Any method you have try to fix the problem?

Hi

one thing is striking me just now, you have two fluid physics, as you are mixing spf and fsi, if you look at the dependent variables of the main node you see the dependent variables therein. And you have not turned off the mix of both in all solver sequences.

FSI comprises both fluids and solid already

--
Good luck
Ivar

Thanks lvar.

My mistake, sorry. great, now it slowly progress.
As my mentor always say, as engineer, you should has eagle eye and patience.
so bad, I miss two.

Update you all later. Now tune the solver setting.

by the way, I have question about it
2) my preferred in addition to 1), add a pressure drop à la Poiseuil or just a linear pressure drop from inlet to outlet along with Y (in your case) and add a parabolic velocity profile (by the way why none "no-slip" on your FSI interface ?)

Confusion:
1. in the predefined inlet, i select the laminar inflow, it is means a parabolic velocity profile is formed?as that condition means that flow pass a long tube to form a laminar flow as input.

2. add a pressure drop is ok, however, in boundary condition, that boundary i select can only has one condition, either i select it as velocity,or pressure, is there any way except choose inlet condition so i can add a linear pressure drop?

Hi, lvar,

Thank you for the help.
It is solved, however, is in laminar flow side, solid not reaction, any idea?

back to just now confusion,
1. i include laminar inflow and outlet as my inlet to formulate pressure drop, is it consider a pressure drop?
i still do not understand the parabolic velocity, as in this way, i can only formulate one condition on one boundary. o by the way, in FSI interface actually include the no slip wall condtion.

Hi

well if you vertical "tube" has an axis along "y" (in 2D adapt for x&z in 3D) then try an initial value of the pressure :

p = p0 * ( y - y0 ) / Length

where "p0" is the maximum pressure at the inlet. "Length" is the tube length from input to output, and "y0" is the coordinate height of the inlet location (I believe that is correct)
Note: the input pressure is the total pressure, before it included also the extra tube length you added to get the developed laminar flow, I believe it's still so, check carefully in the doc

For the inlet velocity in 2D, if you so not use the predefined laminar flow build-up, you can define an inlet velocity as

v0 * 4 * s * ( 1 - s )

where v0 is the maximum velocity (mean velocity = 2/3*v0) and s is the edge length that by default in COMSOL 2D goes from 0 to 1 "along the arrow" of each 2D edge (more complex in 3D). But this is only valid on the input edge, if you want to apply it to the full tube length: again for a vertical tube with flow parallel to "y" axis, in 2D)

v0 * 4 * ( x - XL ) * ( XR - x ) / ( XR - XL )^2

where v0 is the maximum velocity, XL, XR respectively the left and right coordinates of your inlet cross section, or something close to the average tube section for a non straight tube (again I hope I typed this right pls check carefully as I haven't checked myself, and I'm not behind my PC ;)

thislast formula is valid for the full height ("for all y") of your 2D domain, with the risk you get negative flow if the total width is larger, somewhere, than (XR-XL)
--
Good luck
Ivar

Dear lvar,

Thank you for knowledge sharing, what you say i will made in my next model and open the new discussion.;p
by the way, the model get converge, it can converge smoothly, in my computer just few minutes..thanks lvar.

now try some real application,
I intends to simulate the large deformation, so i formulate the inlet as laminar inflow with pressure with 20mmHg, while not outlet, quote from support:
"if you do not have outlet, your wall must move, to let model converge"

i apply boundary load on the wall, -fsi.T.stressr.where the stress come from laminar flow.as blood vessel case.
so we can expect the wall actually will expand in the middle, since i apply point fix constraint on top and bottom.

Surprisingly, the model get converge.but do not expand too much. You can play the animation.

Q
1. Why it do not deform much even with fluid load?
2. If you play the animation, while the fluid actually do not move with the wall, they are like separate each other, anyway i can improve?

Thank you

Hi Chiwei,

According to my reading, the case we are trying to do is a bit different from the blood vessel model, which DID NOT really use the FSI module (you can read it from the model manual). The reason is that they consider the deformation of the blood vessel is small so it will not strongly affect the flow field. However, we are using the FSI and our tube deformation is big.

Specifically, I think we do not need to apply the fluid stress to the tube as boundary load as long as we set the fluid-solid interface correctly (similar to Ivan mentioned). Actually if you remove the boundary load and leave the fluid-solid interface along the model still works.

I am now trying to find out if the coupling of the two fields make sense in my simulation result. Please keep us updated and thanks for your posts.

Hi, Siyuan,

Thank you for supporting this discussion.
I agree with what lvar and you say, FSI boundary layer should not have anything.
I would encourage you using FSI module rather than separate module to make your result easily interpreted, as i found out actually they are nice to use, as all in one.If you notice, in FSI module, there is one FSI boundary interface, while blood vessel one do not have.

I made a balloon fill in with fluid,start from simple stretch. It is not function, i will try in weekend see can fix the problem or not. I think the problem come from the fix constraint, as if we want something to be expand, we should not fix any line, but if we do not fix any line, the reaction force cannot be measured.It make me feel weird. you may run the model and you can see the rupture of balloon.

The moving wall also is a problem for me, as i am wondering how i can let the external wall moving.

I will update you all what is the conclusion if i fix the problem.

lvar ,you and others are free to join this forum, so we can reduce question which send to comsol support.

Hi

A few comments on latest model:

You have too many "prescribed displacements, no need to say something along the "r" direction of the axis of symmetry, as this defines dr=0 by default

Your solid is only held by 2 points, that is "bad" practice, I would propose that you block in z&r the two vertical inlet and outlet regions of the structural wall (vertical lines)

You cannot have a pressure inlet, no outlet and enforce a "0" pressure inside the fluid that is "physically incorrect

Because of the dependent variable naming convention I do NOT believe your FSi initial condition is being copied over from the SPF, and further with no outflow the SPF will not work

then for the laminar inflow entrance pressure, read carefully the doc:

"If Entrance pressure is selected, enter the Entrance pressure p_entr (SI unit: Pa) at the entrance of the fictitious channel outside of the model."

You need to take into account the 1m inflow tube w.r.t. pressure drop values, as you have no outlet.

Again the no outlet get me to ask the question, then why choosing the fluid at all ? just model the structural with a variable pressure value on the inside wall, you get the same "physical" effect.

It's often worth to sit down and sketch out a given problem, define the physics and the dependent variables + interactions on a piece of paper, or even better, with a second or two other persons to get a discussion going, then one find often simpler ways to get to the same result

--
Good luck
Ivar

Dear lvar,

Finally i can persuade myself it is not COMSOL problem and it is my physic problem after read your reply.
It convince me that my model has physic problem.

Thank you for llfe saving.
Great to have your feedback.

p/s: now sitting in office reformulate the problem, will open new discussion thread next time.

hi
I want to couple mixture model with structural. But in mixture model I can't define the loading or displacement induced by structural model. can you help me?
regards