Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.

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.

Laminar flow plus plane strain model

Please login with a confirmed email address before reporting spam

Dear Dr. Ivar,

1) I am sorry I typed wrong, now I understand the difference between parameter and variable. You are right from the beginning.

2) Now I tried to add the plane strain model, is shows the error:
"Undefined value found.
- Detail: NaN_or_Inf_found_when_solving_linear_system_using_PARDISO'
What is this error mean?
Is it because my equations or boundary Conditions are not sufficient to make an unique solution, I should add the moving mesh? (But I want to let it works first then add another)

3) I tried to learn from the structure model tutorial in this website, unfortunately I don't have the MIES model, I read the pdf file and the user guide, still have lots of confusion with weak constrains. Do you have some suggestions to learn about weak constraint?

4) 'lm' should be applied to boundaries with Dirchlet condition between two physics, for my model, the inner membrane is set with body load u_lm and v_lm, for my understanding, the fluid and plane strain is related by
a.) the pressure of the fluid on the membrane
b.) the inner membrane velocity should be same as the fluid velocity at the inner membrane wall
I am not sure LM can cover this or not?

5) Also in the laminar flow, I set the wall (inner membrane) uw as u2t and v2t (honestly I thought should be u2 and v2, it shows orange when set as u2 v2, but I see in Version 3.5, it is set as u2t and v2t), what is the meaning of of u2t, derivative or? I searched but with little explanation.

So can you please check my model and give me some suggestions? Thank you as usual!

Thank you so much!




13 Replies Last Post 23 apr 2011, 01:53 GMT-4
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 18 apr 2011, 10:52 GMT-4
Hii

one general comment, I see you do not use the "material section and enter each material into different nodes. This is perfectly correct, but I find it easier to define different Materials, and the use fewer physics nodes, but this is up to you ;) Now it could be useful to leave the small flap with a specific node entry, but then I would turn ON the Large deformation tab = Include Geometric Non linearities, But I would START WITHOUT

I also missed last time that you have 2x the gravity load for the initial pressure and a sign inversion (could be my fault too), the inlet pressure should be greater than the outlet, else the flow is inversed, more work for the solver

First fundamental question, why add the Prescribed Y direction value on your outer surface ? why not let the only the fluid pressure work ? Apart yes you need to give some "spring" or pressure load also from the outside to compensate for the fluid pressure, but your function, I see in the analytical case B goes to 4E9[m] !! You have stretchy material there, I believe this is not fully correct. (in my understanding the main reason for non convergence here, but not only)

I would rather use "Fixed" boundary conditions on the 2 edges, except if you want to give them some specific motion later

And then for the boundary load, why not use "p" from the fluid, I do not see (but I might be wrong) why to use the weak variables ? I would leave COMSOL defaults and turn OFF the non-default Weak contributions. This is in my sense also valid for the "spf" too. At least in the beginning it's already rather complex like that. The weak constrains can give a better approximation of the boundary/flux loads, but again it complicates. One can always turn them on at the end and see if they really give an effect, remains how to prove it's really better ;)

I'm not sure the "change thickness" will work as expected, Because this is a 2D model so its by default (per meter in depth) you are not really in a "tube" for that 2D axi is required, but then again the axi-symmetric convention might not be correct w.r.t. your gravity load (only the direction along the axis is possible in 2D-axi, and this axis is represented vertical but that is only a convention)

And finally your solver has some problems to find a stable solution, could be from the dead volume that needs more time, you can increase the Solver Fully Coupled Maximum number of iterations to 50 or 75

Then u2t is the velocity of the change along x hnce u2t=du2/dt. It will only be defined for a transient solver case

--
Good luck
Ivar
Hii one general comment, I see you do not use the "material section and enter each material into different nodes. This is perfectly correct, but I find it easier to define different Materials, and the use fewer physics nodes, but this is up to you ;) Now it could be useful to leave the small flap with a specific node entry, but then I would turn ON the Large deformation tab = Include Geometric Non linearities, But I would START WITHOUT I also missed last time that you have 2x the gravity load for the initial pressure and a sign inversion (could be my fault too), the inlet pressure should be greater than the outlet, else the flow is inversed, more work for the solver First fundamental question, why add the Prescribed Y direction value on your outer surface ? why not let the only the fluid pressure work ? Apart yes you need to give some "spring" or pressure load also from the outside to compensate for the fluid pressure, but your function, I see in the analytical case B goes to 4E9[m] !! You have stretchy material there, I believe this is not fully correct. (in my understanding the main reason for non convergence here, but not only) I would rather use "Fixed" boundary conditions on the 2 edges, except if you want to give them some specific motion later And then for the boundary load, why not use "p" from the fluid, I do not see (but I might be wrong) why to use the weak variables ? I would leave COMSOL defaults and turn OFF the non-default Weak contributions. This is in my sense also valid for the "spf" too. At least in the beginning it's already rather complex like that. The weak constrains can give a better approximation of the boundary/flux loads, but again it complicates. One can always turn them on at the end and see if they really give an effect, remains how to prove it's really better ;) I'm not sure the "change thickness" will work as expected, Because this is a 2D model so its by default (per meter in depth) you are not really in a "tube" for that 2D axi is required, but then again the axi-symmetric convention might not be correct w.r.t. your gravity load (only the direction along the axis is possible in 2D-axi, and this axis is represented vertical but that is only a convention) And finally your solver has some problems to find a stable solution, could be from the dead volume that needs more time, you can increase the Solver Fully Coupled Maximum number of iterations to 50 or 75 Then u2t is the velocity of the change along x hnce u2t=du2/dt. It will only be defined for a transient solver case -- Good luck Ivar

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 18 apr 2011, 16:42 GMT-4
Hi

Sorry for the confusions, but I'm starting to understand your model, live as we are exchanging, and I'm discovering several issues.

I can only suggest that we put aside a few items, for later;

1) lm's and weak constraints to be tested once the full model is running
2) gravity along -x, I would suggest that we add this also later as it makes life more complex than required in the beginning, hence the initial condition should also ignore gravity. To not to forget one can place a "0*" in front of the section, or better add a Parameter "GravitySwitch = 0" for no gravity, then set it to 1 later and leave the equations filled in.
3) large non-linear deformation of your valve part
4) your detailed structural boundary motion (still not fully understanding what it is doing.

Then for the model with the fluid (weak constraints off) with the structural (but without ALE we see the coupling of the fluid via the pressure on the thin wall, but not the effect of the displacement of the wall on the fluid since this comes later with the ALE. This shows that leaving the thin wall only attached at the two extremities then the fluid pressure will force a motion in the meter range, so we need to add some pressure or distributed "spring" action to mimic the tissues surrounding the artery, and that holds the system in place. This is probably the function of the point 4 Prescribed Displacement in y, but I find the values somewhat strange.
I can propose to replace the external boundary Prescribed displacement by either:

1) a pressure boundary load, set to the inside fluid pressure around 1000 Pa, or
2) a distributed spring value, as a boundary load with total force Fy = -K*v2 with K=1 to 10 [N/um] and v2 the structural solid displacement along Y

More next time

--
Good luck
Ivar
Hi Sorry for the confusions, but I'm starting to understand your model, live as we are exchanging, and I'm discovering several issues. I can only suggest that we put aside a few items, for later; 1) lm's and weak constraints to be tested once the full model is running 2) gravity along -x, I would suggest that we add this also later as it makes life more complex than required in the beginning, hence the initial condition should also ignore gravity. To not to forget one can place a "0*" in front of the section, or better add a Parameter "GravitySwitch = 0" for no gravity, then set it to 1 later and leave the equations filled in. 3) large non-linear deformation of your valve part 4) your detailed structural boundary motion (still not fully understanding what it is doing. Then for the model with the fluid (weak constraints off) with the structural (but without ALE we see the coupling of the fluid via the pressure on the thin wall, but not the effect of the displacement of the wall on the fluid since this comes later with the ALE. This shows that leaving the thin wall only attached at the two extremities then the fluid pressure will force a motion in the meter range, so we need to add some pressure or distributed "spring" action to mimic the tissues surrounding the artery, and that holds the system in place. This is probably the function of the point 4 Prescribed Displacement in y, but I find the values somewhat strange. I can propose to replace the external boundary Prescribed displacement by either: 1) a pressure boundary load, set to the inside fluid pressure around 1000 Pa, or 2) a distributed spring value, as a boundary load with total force Fy = -K*v2 with K=1 to 10 [N/um] and v2 the structural solid displacement along Y More next time -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 18 apr 2011, 17:34 GMT-4
Dear Dr. Ivar,

Thank you for your nice comments!

Thanks a lot!

1) For the function B, I think I made a mistake on the units, I divide x by 1000 to change mm to m, then its limit is 1.5mm.

2) And I turned off all the weak constraint to simplified the model.

3) For the prescribed displacement in y (outside membrane), I set as "(delta_max/h1^4)*sin(2*pi*t[1/s]/3)*x*0.001*(x*0.001-h1)*(x*0.001-h1-0.09)*(x*0.001-h1-0.135)", is because in our vein it is also affected by muscle compression, which is also periodic distributed compression, then the membrane will do this kind of motion, I set this according to my senior lab-mates' model in version 3.5, my professor said she did this experiment before. So maybe I can just leave it for later. NOW? I just set as fluid pressure p.

4) For the gravity, I thought same as you, I ignored it for now. Also your idea is great, add the parameter "GravitySwitch" to remind me in the last.

5) then I will digest them and try it on my model, to see how far I can go!

pls: I did this model according my senior lab-mate's dissertation, I attached the model setting part, if you have time, you can take a look at it(basically chap4). ( I know I occupied a lot of time from you, so if you are busy just ignore it:)

Thanks again!

Dear Dr. Ivar, Thank you for your nice comments! Thanks a lot! 1) For the function B, I think I made a mistake on the units, I divide x by 1000 to change mm to m, then its limit is 1.5mm. 2) And I turned off all the weak constraint to simplified the model. 3) For the prescribed displacement in y (outside membrane), I set as "(delta_max/h1^4)*sin(2*pi*t[1/s]/3)*x*0.001*(x*0.001-h1)*(x*0.001-h1-0.09)*(x*0.001-h1-0.135)", is because in our vein it is also affected by muscle compression, which is also periodic distributed compression, then the membrane will do this kind of motion, I set this according to my senior lab-mates' model in version 3.5, my professor said she did this experiment before. So maybe I can just leave it for later. NOW? I just set as fluid pressure p. 4) For the gravity, I thought same as you, I ignored it for now. Also your idea is great, add the parameter "GravitySwitch" to remind me in the last. 5) then I will digest them and try it on my model, to see how far I can go! pls: I did this model according my senior lab-mate's dissertation, I attached the model setting part, if you have time, you can take a look at it(basically chap4). ( I know I occupied a lot of time from you, so if you are busy just ignore it:) Thanks again!


Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 18 apr 2011, 21:47 GMT-4
Dear Dr. Ivar,

1) First, I am so glad that you can understand my model further, l am just too bad to let others aware of what I am doing. I will try to do it better:)

2) Second, I did what you suggest to set the outside membrane a) pressure -1000pa, -1500pa or -1200, b) force Fy -K*v2, both of them run ok when compute the stationary solver. But as you said, the largest deformation is almost 1m which is ridiculous. So I set the outside pressure to be -1000---1600 to see of there is some value is reasonable, but it shows the same amplitude( more than 1m), then I think is it because the thickness is too thin? So I tried to add the thickness, even I set it as 5mm-50mm. it still the same amplitude of displacement, I am confusion why it this?



3) when I then run the Time-dependent solver, it shows error "Pointwise contraints for time derives are not supported." for all the settings that I have ever tried.

4) I think is it because the properties of my solid is not suitable for the conditions that I applied, it will affect the mesh?ACTUALLY, I really don't know how to check the mesh? Just by looking the many triangles? Should I adjust the geometry to smooth some edges? (I see somewhere before)

but the properties and geometry should be right( because my senior lab-mate already done it in ver3.2 ( maybe) and experiment 6years ago)

5) I searched, and due to this large deformation, should I check the "include geometric non-linearity" check-box , but because I don't have the MIMS model, so I didn't find this checkbox in my linear elastic material settings windows, or I missed it somewhere??

That is what I did tonight, encouraging myself to go further and thank you for all you help:)


Dear Dr. Ivar, 1) First, I am so glad that you can understand my model further, l am just too bad to let others aware of what I am doing. I will try to do it better:) 2) Second, I did what you suggest to set the outside membrane a) pressure -1000pa, -1500pa or -1200, b) force Fy -K*v2, both of them run ok when compute the stationary solver. But as you said, the largest deformation is almost 1m which is ridiculous. So I set the outside pressure to be -1000---1600 to see of there is some value is reasonable, but it shows the same amplitude( more than 1m), then I think is it because the thickness is too thin? So I tried to add the thickness, even I set it as 5mm-50mm. it still the same amplitude of displacement, I am confusion why it this? 3) when I then run the Time-dependent solver, it shows error "Pointwise contraints for time derives are not supported." for all the settings that I have ever tried. 4) I think is it because the properties of my solid is not suitable for the conditions that I applied, it will affect the mesh?ACTUALLY, I really don't know how to check the mesh? Just by looking the many triangles? Should I adjust the geometry to smooth some edges? (I see somewhere before) but the properties and geometry should be right( because my senior lab-mate already done it in ver3.2 ( maybe) and experiment 6years ago) 5) I searched, and due to this large deformation, should I check the "include geometric non-linearity" check-box , but because I don't have the MIMS model, so I didn't find this checkbox in my linear elastic material settings windows, or I missed it somewhere?? That is what I did tonight, encouraging myself to go further and thank you for all you help:)


Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 19 apr 2011, 02:13 GMT-4
Hi

A few things, there are so many questions here that I'm missing some, we can always come back to them later. It's interesting to build up a model like that interactively.

careful with the "thickness": this is the "depth" in the paper direction or "z", changing the thickness will not change anything else (apart from some force calculations), as the fluid pressure is calculated "per meter = its in fact a density and pressure is scaling with the surface. In 2D an edge is in fact a surface of 1[m] deep by default (along Z) so if you change the "thickness" then the pressure will also scale and the net displacement remains unchanged

What you must do is to say you have surrounding tissues, that is acting on the external surface and compresses it, so you can add an external pressure on this boundary, first about the value of the internal fluid pressure i.e. 1000[Pa]. But that is static, if the pressure of the fluid increases, your artery wall will displace more and more, you have also a "spring" effect, this means that the pressure from the tissue will increase with the displacement (in your case v the vertical direction). therefore I suggest that you start to add a boundary pressure value on this external boundary of "1000[Pa] -K*v2" (or v if the structural displacement dependent variable is "u") K is then a spring constant that you can adapt, typically to some 5[N/um/m^2] in a parameter and than change it to get a displacement that you feel is reasonable. This formula is not fully correct, as the pressure is acting perpendicular to the surface, and your bumb there has a normal going away from v as normal so to be fully correct, and by defining a Parameter P0=1000[Pa] (as this value appears many places, it's better to link them via a Parameter,
pressure = P0-K*(solid.nx*u2+solid.ny*v2)

Where (solid.nx*u2+solid.ny*v2) is the scalar product between the displacement vector (u,v) and the boundary normal vector (solid.nx,solid.ny)

You need to turn "on" the displacement and use u2,v2 as variables to see the structural motion

By the way, keep the Wall 1 of spf to "no-slip", the vertical motion of the wall will be introduced via the ALE later

in the initial pressure spf, use a *Gravity Switch and not a "+" typos comes quickly
For me the inlet 2 should not exist, the velocity in inlet 1 is rather U_amp*(1-s)^2, and you need to add the gravity Switch for Volume force in SPF. You can add the initial velocity but as you add it to a volume the "s" variable is not defined, you must then write u_amp*((0.004-y[1/m])/0.004)^2*(y<0.004) thisequations will ensure a starting velocity and set the velocity in the chamber for y>0.004[mm] to "0"

Solid Boundary load use +P0 or +1000[Pa] a normal is outwards and pressure acts perpendicular against the normal (by convention)

If you try you will get a wavy result, wee need some damping ;) But that can come with the link to the fluid via the ALE

I have tested this in Static, it's OK, but it is too wavy and unstable for the transient solver

I see too that your prescribed load condition now looks better, so its probably also possible to use it.

Final comment, your valve is thin, and the material propeties you are using is making it move quite far, to ease the convergence of the solver, you could increase the stiffness by some 100 or 1000* to begin width, then lower it gently to see the effect (just write 15e6*1000 in the Young modulus and change the *1000 at will)

--
Good luck
Ivar
Hi A few things, there are so many questions here that I'm missing some, we can always come back to them later. It's interesting to build up a model like that interactively. careful with the "thickness": this is the "depth" in the paper direction or "z", changing the thickness will not change anything else (apart from some force calculations), as the fluid pressure is calculated "per meter = its in fact a density and pressure is scaling with the surface. In 2D an edge is in fact a surface of 1[m] deep by default (along Z) so if you change the "thickness" then the pressure will also scale and the net displacement remains unchanged What you must do is to say you have surrounding tissues, that is acting on the external surface and compresses it, so you can add an external pressure on this boundary, first about the value of the internal fluid pressure i.e. 1000[Pa]. But that is static, if the pressure of the fluid increases, your artery wall will displace more and more, you have also a "spring" effect, this means that the pressure from the tissue will increase with the displacement (in your case v the vertical direction). therefore I suggest that you start to add a boundary pressure value on this external boundary of "1000[Pa] -K*v2" (or v if the structural displacement dependent variable is "u") K is then a spring constant that you can adapt, typically to some 5[N/um/m^2] in a parameter and than change it to get a displacement that you feel is reasonable. This formula is not fully correct, as the pressure is acting perpendicular to the surface, and your bumb there has a normal going away from v as normal so to be fully correct, and by defining a Parameter P0=1000[Pa] (as this value appears many places, it's better to link them via a Parameter, pressure = P0-K*(solid.nx*u2+solid.ny*v2) Where (solid.nx*u2+solid.ny*v2) is the scalar product between the displacement vector (u,v) and the boundary normal vector (solid.nx,solid.ny) You need to turn "on" the displacement and use u2,v2 as variables to see the structural motion By the way, keep the Wall 1 of spf to "no-slip", the vertical motion of the wall will be introduced via the ALE later in the initial pressure spf, use a *Gravity Switch and not a "+" typos comes quickly For me the inlet 2 should not exist, the velocity in inlet 1 is rather U_amp*(1-s)^2, and you need to add the gravity Switch for Volume force in SPF. You can add the initial velocity but as you add it to a volume the "s" variable is not defined, you must then write u_amp*((0.004-y[1/m])/0.004)^2*(y0.004[mm] to "0" Solid Boundary load use +P0 or +1000[Pa] a normal is outwards and pressure acts perpendicular against the normal (by convention) If you try you will get a wavy result, wee need some damping ;) But that can come with the link to the fluid via the ALE I have tested this in Static, it's OK, but it is too wavy and unstable for the transient solver I see too that your prescribed load condition now looks better, so its probably also possible to use it. Final comment, your valve is thin, and the material propeties you are using is making it move quite far, to ease the convergence of the solver, you could increase the stiffness by some 100 or 1000* to begin width, then lower it gently to see the effect (just write 15e6*1000 in the Young modulus and change the *1000 at will) -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 20 apr 2011, 00:37 GMT-4
Dear Dr. Ivar,

Thank you again:)

I really inspired by your idea about setting the external boundary as
pressure=P0-K*(solid.nx*u2+solid.ny*v2)
which is so smart, I think it can work and substitute the prescribed displacement that I should set later. (Also I tried this prescribed displacement in stationary solver, it works ok, nut also failed in the timedependent solver)

1) First about the thickness, I was wondering should I change the "thickness"in the plane(not out of plane), I mean how it is drawn, to make it thicker? But still I don't know how to check it, is it related by the mesh, which means that we can see this is reasonable or not by checking the mesh?

2) I did what you suggested, also the valve's Young modulus*1000 can shows reasonable displacement in the stationary solver, which is also reasonable in the bump, but when I run the time-dependent, it shows error
"Nonlinear solver did not converge. Time : 0 Last time step is not converged."


'Nonlinear solver', now, the displacement is not so large, so I am not sure what is exactly the large deformation (geometric non-linearity) mean? (relative to the object's original shape, like deflect 10% or 50% of the original shape?)
Even when I change the Young modulus as *10000, it still did not converge.

So should I ? I really lost , next step should be what.

Thank you again for your time and great ideas, to help me on my comsol trip!




Dear Dr. Ivar, Thank you again:) I really inspired by your idea about setting the external boundary as pressure=P0-K*(solid.nx*u2+solid.ny*v2) which is so smart, I think it can work and substitute the prescribed displacement that I should set later. (Also I tried this prescribed displacement in stationary solver, it works ok, nut also failed in the timedependent solver) 1) First about the thickness, I was wondering should I change the "thickness"in the plane(not out of plane), I mean how it is drawn, to make it thicker? But still I don't know how to check it, is it related by the mesh, which means that we can see this is reasonable or not by checking the mesh? 2) I did what you suggested, also the valve's Young modulus*1000 can shows reasonable displacement in the stationary solver, which is also reasonable in the bump, but when I run the time-dependent, it shows error "Nonlinear solver did not converge. Time : 0 Last time step is not converged." 'Nonlinear solver', now, the displacement is not so large, so I am not sure what is exactly the large deformation (geometric non-linearity) mean? (relative to the object's original shape, like deflect 10% or 50% of the original shape?) Even when I change the Young modulus as *10000, it still did not converge. So should I ? I really lost , next step should be what. Thank you again for your time and great ideas, to help me on my comsol trip!


Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 20 apr 2011, 01:24 GMT-4
Hi

indeed I have played a little its not obvious how to get it to converge. I had also a quick look at the article, interesting. Still my main remark is this is normally a cylindrical symmetry case, should have been studied in 2D axi, but we van always repeat that later (you need to copy our geometry, to rotate it by 90° counter-clockwise and to mirror it, but that is quickly done in COMSOL, in 2D axi the geometry MUST be in the R>=0 quadrants, with the axis of symmetry at R=0.

OK thickness (for me mostly out of plane, note in 2D keep it at default 1[m] and thing per meter depth, the units of all results should also express this [1/m] dependence) if you talk about the venous membrane thickness (in your Y direction) you can indeed increase its dimension, but I would then rather add another layer and call it "muscle", By the way the book of An Introuction to Modelling of Transport Processes, application to Biomedical Systems, of Ashim Datta and Vineet Rakesh Cambridge Texts in BioMedical Engineering ISBN 978-0-521-11924-5 is very interesting and contain a lot in nice material data for your type of simulations.

I suspect the treansient solver is having problems because we have perhaps too large time steps, this morning I just saved a 5.5Gb model file, as I needed 200 iterations to apply my load by small steps, some 9h calculations on my WorkHorse of PC. But the results look finally nice ;) Conclusions, for me, sometimes one need to be patient

My suggestion: set back the stiffness of the material to the default values, and lets try to add ALE, and simulate i.e. just the star of the rising of the pulse, perhaps 0.3 sec but with perhaps 30 steps. I'm slightly overbooked today, will see what I can try out later tonight

--
Good luck
Ivar
Hi indeed I have played a little its not obvious how to get it to converge. I had also a quick look at the article, interesting. Still my main remark is this is normally a cylindrical symmetry case, should have been studied in 2D axi, but we van always repeat that later (you need to copy our geometry, to rotate it by 90° counter-clockwise and to mirror it, but that is quickly done in COMSOL, in 2D axi the geometry MUST be in the R>=0 quadrants, with the axis of symmetry at R=0. OK thickness (for me mostly out of plane, note in 2D keep it at default 1[m] and thing per meter depth, the units of all results should also express this [1/m] dependence) if you talk about the venous membrane thickness (in your Y direction) you can indeed increase its dimension, but I would then rather add another layer and call it "muscle", By the way the book of An Introuction to Modelling of Transport Processes, application to Biomedical Systems, of Ashim Datta and Vineet Rakesh Cambridge Texts in BioMedical Engineering ISBN 978-0-521-11924-5 is very interesting and contain a lot in nice material data for your type of simulations. I suspect the treansient solver is having problems because we have perhaps too large time steps, this morning I just saved a 5.5Gb model file, as I needed 200 iterations to apply my load by small steps, some 9h calculations on my WorkHorse of PC. But the results look finally nice ;) Conclusions, for me, sometimes one need to be patient My suggestion: set back the stiffness of the material to the default values, and lets try to add ALE, and simulate i.e. just the star of the rising of the pulse, perhaps 0.3 sec but with perhaps 30 steps. I'm slightly overbooked today, will see what I can try out later tonight -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 20 apr 2011, 01:42 GMT-4
Dear Dr. Ivar,

Thank you so much for your time and comments, even when you are "overbooked", these days , receiving your comments and back to comsoling seems attractive to me, I am really touched by all those suggestions...

9h calculation, oh my god, how complicated it will be:) I will try to do it tomorrow and see what I can do with this....

We do have time difference due to different timezones:)
Have a nice day/evening:)
Dear Dr. Ivar, Thank you so much for your time and comments, even when you are "overbooked", these days , receiving your comments and back to comsoling seems attractive to me, I am really touched by all those suggestions... 9h calculation, oh my god, how complicated it will be:) I will try to do it tomorrow and see what I can do with this.... We do have time difference due to different timezones:) Have a nice day/evening:)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 20 apr 2011, 23:53 GMT-4
Dear Ivar,

I tried to add the moving mesh (ALE), and see how it works.

1) I keep the Young modulus still as 15e6*1000, because when I run the stationary solver, the displacement is too unnormal (as large as several m ), if keep it as 15e6*1000, it shows 0.0954 mm or 0.1053 mm( use prescribed displacement as"(delta_max/h1^4)*sin(2*pi*t[1/s]/3)*x*0.001*(x*0.001-h1)*(x*0.001-h1-0.09)"), which is much reasonable.

2) I set the wall as u2t and v2t, because I think now I add the moving mesh, so the effect of the displacement on the fluid should be taken into consideration.

3) when I run the time-dependent solver, it shows error "Pointwise constraints for time derivatives are not supported" , even I just compute 0.2s. What is this error mean? Pointwise constraint?

4) I tried to turn on the weak constraint ( which is not the right time, at least for now I think) , even in the stationary solver, it cannot converge.

5) Then I am thinking, maybe I should find a way to set the weak constraint reasonably, then maybe things can getting better.

Also, due to in the stationary solver, the displacement seems reasonable, so I think I don't need to change the geometry, also this model has been like did in the experiment, so I think I still need to find the key mistake.

can you help me to check it a little bit, and give me some further comments?

Thank you as usual! Best wishes to you and your comsoling model:) Really wish one day, I can have a chance to take a look at the 5G model.:)


Dear Ivar, I tried to add the moving mesh (ALE), and see how it works. 1) I keep the Young modulus still as 15e6*1000, because when I run the stationary solver, the displacement is too unnormal (as large as several m ), if keep it as 15e6*1000, it shows 0.0954 mm or 0.1053 mm( use prescribed displacement as"(delta_max/h1^4)*sin(2*pi*t[1/s]/3)*x*0.001*(x*0.001-h1)*(x*0.001-h1-0.09)"), which is much reasonable. 2) I set the wall as u2t and v2t, because I think now I add the moving mesh, so the effect of the displacement on the fluid should be taken into consideration. 3) when I run the time-dependent solver, it shows error "Pointwise constraints for time derivatives are not supported" , even I just compute 0.2s. What is this error mean? Pointwise constraint? 4) I tried to turn on the weak constraint ( which is not the right time, at least for now I think) , even in the stationary solver, it cannot converge. 5) Then I am thinking, maybe I should find a way to set the weak constraint reasonably, then maybe things can getting better. Also, due to in the stationary solver, the displacement seems reasonable, so I think I don't need to change the geometry, also this model has been like did in the experiment, so I think I still need to find the key mistake. can you help me to check it a little bit, and give me some further comments? Thank you as usual! Best wishes to you and your comsoling model:) Really wish one day, I can have a chance to take a look at the 5G model.:)


Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 21 apr 2011, 12:08 GMT-4
Hi

A few comments to keep your model coherent:

In the NFS outlet replace the 1000[Pa] by P0 as this is now defined like that so you do not forget it next time you change P0

For the inlet (could also be used for the Initial conditions fluid) X velocity you define
u_amp*((0.004[m]-y/1000)/0.004[m])^2*(y<0.004[m])

You should normally NOT divide y by 1000 as "y "is, by default, expressed in SI or [m]
you should rather write, partly my fault the shape of the parabola was not the right one (always test your function on an analytical function plot, that is what they are here for ;)

u_amp*(1-(Y[1/m]/0.004)^2)*(Y<0.004)

An important point: When We add the ALE we are splitting up the frames: spatial, material and mesh, then our smaller case "x" is now referring to the spatial (deformed frame) while our velocity initial conditions should refer to the Material frame now expressed in UPPER CASE X,Y,Z, so wee need to replace "y" by "Y" and the same for x (and z when applicable in 3D), also for the pressure variation

Try a "Solver - Dependent variables - Compute to selected" and plot the velocity profile over the inlet boundary to check initial conditions.

Keep in mind that once ready you should try a case with the "Structural - Linear elastic material Model" check the "Include geometric non linearity", particularly for the flap. But as this is slowing down the solver, I propose to leave it to later

Do not forget the reset t=0[s] in the parameter when you do transient solver cases, as else you prepare the system with a velocity different from "0" (if I remember right)

For the rest my first analysis of the model seems OK ; still have to run it :)

For your questions:

OK the *1000 is OK the flap is normally moving I suppose some mm, but then one should ideally have the large displacement geometric non-linearities "on"

Then your v2t. I'm not sure it's needed, this is normally for motion w.r.t. the mesh, your relation is, for me coming in through the motion of the ALE mesh. This gives also the next error message of point derivatives, the end of the edges inherit properties from the edge, some of these edges are fixed hence derivatives =0 others have the issue derivatives from left edge, or right edge intersecting ? therefore the error (or warning) message. Try without wall velocity, use "no-slip" i.e. the fluid should follow the wall motion but be at "0" velocity along the wall boundary.

I'm not sure we really need the weak constraints here, the way we couple it, indeed I would leave it for later

So when I run the solver (and turn on the plot while solving) I see nice surface waves develop and no convergence. I see that this is due to my "spring load" as I do not introduce any damping, therefore the "spring" interacts with the fluid flow and we get different oscillations depending on the spring constant I use.

To get this "tricky" model to solve I believe we need to continue by fixing the external boundary. At least until we get the solver to solve it nicely. Anyhow the flap is free. I'm not sure you r colleague managed to get it to solve at the first run ;)
--
Good luck
Ivar
Hi A few comments to keep your model coherent: In the NFS outlet replace the 1000[Pa] by P0 as this is now defined like that so you do not forget it next time you change P0 For the inlet (could also be used for the Initial conditions fluid) X velocity you define u_amp*((0.004[m]-y/1000)/0.004[m])^2*(y

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 21 apr 2011, 23:14 GMT-4
Hi, Ivar,

Thank you so much for your valuable knowledge, it works better than the books and papers that I read:)

1) Your suggestion about set the P0 and using the Analytical Function to check the function is so right, now I know, I'd better plot first and see whether it seems reasonable or not before I apply them.

2) About the splitting frames: spatial, material and mesh, I am not so clear about them, but I will read the user's guide and see if I can understand more.

3) I do want to try with the "Structural- Linear elastic material model----Include geometric non linearity", but I can not find the check box, I only find "damping" when right-click the Linear elastic material model? Can you tell me where is this option?

4) I think your idea about setting the wall , at least for now to be non-slip is right. I will keep it until run the model well.

5) I set the damping under the Linear elastic material model, but it seems still cannot converge. Also as you said, we can keep fixing the external boundary, maybe I misunderstand it, because if we fix it, how can we check the displacement of the membrane and its effect on the fluid?

6) also a weld thing happened, when I turn on the Plotting while solving, my Comsol in my new sony laptop(I think IBM would be better) just shut down again and again, hehe, I got 4G memory, maybe I should get a Desktop later.

7) I will consider your valuable suggestions, and see what I can do next:)

Thank you for all your help to me, I really feel touched!

Have a nice day:):):):):):)
Hi, Ivar, Thank you so much for your valuable knowledge, it works better than the books and papers that I read:) 1) Your suggestion about set the P0 and using the Analytical Function to check the function is so right, now I know, I'd better plot first and see whether it seems reasonable or not before I apply them. 2) About the splitting frames: spatial, material and mesh, I am not so clear about them, but I will read the user's guide and see if I can understand more. 3) I do want to try with the "Structural- Linear elastic material model----Include geometric non linearity", but I can not find the check box, I only find "damping" when right-click the Linear elastic material model? Can you tell me where is this option? 4) I think your idea about setting the wall , at least for now to be non-slip is right. I will keep it until run the model well. 5) I set the damping under the Linear elastic material model, but it seems still cannot converge. Also as you said, we can keep fixing the external boundary, maybe I misunderstand it, because if we fix it, how can we check the displacement of the membrane and its effect on the fluid? 6) also a weld thing happened, when I turn on the Plotting while solving, my Comsol in my new sony laptop(I think IBM would be better) just shut down again and again, hehe, I got 4G memory, maybe I should get a Desktop later. 7) I will consider your valuable suggestions, and see what I can do next:) Thank you for all your help to me, I really feel touched! Have a nice day:):):):):):)

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 22 apr 2011, 01:21 GMT-4
Hello again

1) For me too, as I got caught also ;)

2) Understanding the frames is essential in COMSOL, in fact I should correct again, now in V4, with structural you have already 2 frames x,y,z the spatial, and X,Y,Z the material as you have an option default on in the new version: "Spatial frame is driven by the deformations". This is worth to study, its linked to Eulerian and Lagrangian references

3) check the main structural physics "Linear Elastic Material Model", at the far bottom of the GUI

4) for me the wall velocity is i.e. for two disks surrounded by a fluid, and the fluid/structure is "slipping". The ALE is taking into account the linking of the deformation of the structure to the fluid volume change, while the fluid pressure is linking the fluid to the structure.

5) we need a large damping here, it links to another issue I have when solving vibrating structures in a fluid, on can easily add the the "added mass" of the fluid onto the structure, but then the damping is mainly the viscous damping of the fluid, then we enter easily into acoustics (and I do unfortunately not have that module) so I need to enter further manual equations, another story ;)

6) ah this is probably linked to the graphics and the way that is sharing the RAM on the laptops. Today 4Gb on a 32 bit OS is a max you can address, after you need to go to a 64 bit OS too, it's probably not brand related, could be a driver issue.
It's true that most CFD models are run on larger machines. My workhorse has 48 Gb (and it's a year old now) it heats up my office to above 30°C and makes the noise of a helicopter when it really runs, so I would have to find the time to move it down in the basement for this summer ;)

--
Good luck
Ivar
Hello again 1) For me too, as I got caught also ;) 2) Understanding the frames is essential in COMSOL, in fact I should correct again, now in V4, with structural you have already 2 frames x,y,z the spatial, and X,Y,Z the material as you have an option default on in the new version: "Spatial frame is driven by the deformations". This is worth to study, its linked to Eulerian and Lagrangian references 3) check the main structural physics "Linear Elastic Material Model", at the far bottom of the GUI 4) for me the wall velocity is i.e. for two disks surrounded by a fluid, and the fluid/structure is "slipping". The ALE is taking into account the linking of the deformation of the structure to the fluid volume change, while the fluid pressure is linking the fluid to the structure. 5) we need a large damping here, it links to another issue I have when solving vibrating structures in a fluid, on can easily add the the "added mass" of the fluid onto the structure, but then the damping is mainly the viscous damping of the fluid, then we enter easily into acoustics (and I do unfortunately not have that module) so I need to enter further manual equations, another story ;) 6) ah this is probably linked to the graphics and the way that is sharing the RAM on the laptops. Today 4Gb on a 32 bit OS is a max you can address, after you need to go to a 64 bit OS too, it's probably not brand related, could be a driver issue. It's true that most CFD models are run on larger machines. My workhorse has 48 Gb (and it's a year old now) it heats up my office to above 30°C and makes the noise of a helicopter when it really runs, so I would have to find the time to move it down in the basement for this summer ;) -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 23 apr 2011, 01:53 GMT-4
Hi, Dear Ivar,

Nice to receive your reply:) A picture of a large computer and an excellent , nice COMSOL engineer come into my mind, haha, 30 degree, maybe only winter is ok, also I think it is better to move them to the basement also for your health:)

I am sorry, I still cannot find the check box "Include geometric nonlinearity ", is it in the MEMS model, for my model I still cannot find them under the Linear Elastic Material Model:)

Also final is coming with 3 exams and 3 projects, I think I have to come back to comsol at the beginning of May. Really thank you for all your help and encourage during these period:) Wish we can discuss more about this model soon:)

Take care:)))):)))))::)))
Hi, Dear Ivar, Nice to receive your reply:) A picture of a large computer and an excellent , nice COMSOL engineer come into my mind, haha, 30 degree, maybe only winter is ok, also I think it is better to move them to the basement also for your health:) I am sorry, I still cannot find the check box "Include geometric nonlinearity ", is it in the MEMS model, for my model I still cannot find them under the Linear Elastic Material Model:) Also final is coming with 3 exams and 3 projects, I think I have to come back to comsol at the beginning of May. Really thank you for all your help and encourage during these period:) Wish we can discuss more about this model soon:) Take care:)))):)))))::)))

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.