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 The information provided may be out of date.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.
Failed to find a solution: Singular MATRIX.
Posted 03.08.2012, 07:15 GMT-4 Fluid & Heat, Heat Transfer & Phase Change, Mesh, Studies & Solvers Version 4.2a 10 Replies
Please login with a confirmed email address before reporting spam
Hi,
I am Patricia and I am a new Comsol user.
I am modelling the cooling scenario of a low voltage AC drives.To simplify the calculations and to obtain easier results, I have carried out many approximations: air flows between two parallel and vertical plates ignoring the different elements between both plates, box at the top left which dissipates heat and a fan in the bottom of the device which produces the cooling, hence, there is forced convection.
I was modelled on the example of "Convection Cooling of Circuit Boards". I solved 2 days ago and I obtained the temperature and velocity graphs. However, today I wanted to solve again (I didn't do any modification) and I have the next error:
" Failed to find a solution.
Singular matrix.
There are 27 equations giving NaN/Inf in the matrix rows for the variable mod1.u.
at coordinates: (0.00973619,0.290431), (0.013152,0.296094), (0.0114327,0.298224), (0.017299,0.290132), (0.0218176,0.283247), ...
There are 27 equations giving NaN/Inf in the matrix rows for the variable mod1.v.
at coordinates: (0.00973619,0.290431), (0.013152,0.296094), (0.0114327,0.298224), (0.017299,0.290132), (0.0218176,0.283247), ...
There are 28 degrees of freedom giving NaN/Inf in the matrix columns for the variable mod1.T.
at coordinates: (0.00973619,0.290431), (0.013152,0.296094), (0.0114327,0.298224), (0.017299,0.290132), (0.0218176,0.283247), ...
There are 17 degrees of freedom giving NaN/Inf in the matrix columns for the variable mod1.u.
at coordinates: (0.013152,0.296094), (0.0114327,0.298224), (0.017299,0.290132), (0.0218176,0.283247), (0.0226893,0.295827), ...
There are 17 degrees of freedom giving NaN/Inf in the matrix columns for the variable mod1.v.
at coordinates: (0.013152,0.296094), (0.0114327,0.298224), (0.017299,0.290132), (0.0218176,0.283247), (0.0226893,0.295827), ...
Returned solution is not converged. "
I checked so many times everything and I cannot find the errors. Btw, how is it possible to have an error if I didn't modify anything??
Also, I modelled the 3D model of my device (it was modelled on the example of "Convection Cooling of Circuit Boards" as well) and also I have an error
(this time different):
" Failed to find a solution
Divergence of the linear iterations.
Returned solution is not converged
-Feature: Stationary Solver1 (sol1/s1)
-Error:Failed to find a solution. "
I have attached both simulations (the 2D model attached is with the temperature and velocity graphs done, but if you compute again, you will be able to observe the matrix error). I attached them because I think it will be easier to find the error.
If you can clarify my doubts I will be really grateful to you.
Thank you very much for your help.
Sincerely,
Patricia.
I am Patricia and I am a new Comsol user.
I am modelling the cooling scenario of a low voltage AC drives.To simplify the calculations and to obtain easier results, I have carried out many approximations: air flows between two parallel and vertical plates ignoring the different elements between both plates, box at the top left which dissipates heat and a fan in the bottom of the device which produces the cooling, hence, there is forced convection.
I was modelled on the example of "Convection Cooling of Circuit Boards". I solved 2 days ago and I obtained the temperature and velocity graphs. However, today I wanted to solve again (I didn't do any modification) and I have the next error:
" Failed to find a solution.
Singular matrix.
There are 27 equations giving NaN/Inf in the matrix rows for the variable mod1.u.
at coordinates: (0.00973619,0.290431), (0.013152,0.296094), (0.0114327,0.298224), (0.017299,0.290132), (0.0218176,0.283247), ...
There are 27 equations giving NaN/Inf in the matrix rows for the variable mod1.v.
at coordinates: (0.00973619,0.290431), (0.013152,0.296094), (0.0114327,0.298224), (0.017299,0.290132), (0.0218176,0.283247), ...
There are 28 degrees of freedom giving NaN/Inf in the matrix columns for the variable mod1.T.
at coordinates: (0.00973619,0.290431), (0.013152,0.296094), (0.0114327,0.298224), (0.017299,0.290132), (0.0218176,0.283247), ...
There are 17 degrees of freedom giving NaN/Inf in the matrix columns for the variable mod1.u.
at coordinates: (0.013152,0.296094), (0.0114327,0.298224), (0.017299,0.290132), (0.0218176,0.283247), (0.0226893,0.295827), ...
There are 17 degrees of freedom giving NaN/Inf in the matrix columns for the variable mod1.v.
at coordinates: (0.013152,0.296094), (0.0114327,0.298224), (0.017299,0.290132), (0.0218176,0.283247), (0.0226893,0.295827), ...
Returned solution is not converged. "
I checked so many times everything and I cannot find the errors. Btw, how is it possible to have an error if I didn't modify anything??
Also, I modelled the 3D model of my device (it was modelled on the example of "Convection Cooling of Circuit Boards" as well) and also I have an error
(this time different):
" Failed to find a solution
Divergence of the linear iterations.
Returned solution is not converged
-Feature: Stationary Solver1 (sol1/s1)
-Error:Failed to find a solution. "
I have attached both simulations (the 2D model attached is with the temperature and velocity graphs done, but if you compute again, you will be able to observe the matrix error). I attached them because I think it will be easier to find the error.
If you can clarify my doubts I will be really grateful to you.
Thank you very much for your help.
Sincerely,
Patricia.
Attachments:
10 Replies Last Post 08.08.2012, 01:36 GMT-4
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi Patricia,
You have used laminar flow assumption. Inlet velocity seems quite high for laminar flow (=> high Reynols number). Maybe you should use turbulence model.
Best regards
Tero
You have used laminar flow assumption. Inlet velocity seems quite high for laminar flow (=> high Reynols number). Maybe you should use turbulence model.
Best regards
Tero
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi Tero,
Thank you for your answer. I changed the velocity (umax) for another lower, but the same failure in the solution is still appeared in the 2D model. However, the mistake was solved in the 3D model. Maybe, the problem is because of the mesh?
Also, I would like to ask you another question. Would it be possible to calculate the convective heat transfer coefficient throught Comsol? How?
Thanks very much.
Sincerely,
Patricia.
Thank you for your answer. I changed the velocity (umax) for another lower, but the same failure in the solution is still appeared in the 2D model. However, the mistake was solved in the 3D model. Maybe, the problem is because of the mesh?
Also, I would like to ask you another question. Would it be possible to calculate the convective heat transfer coefficient throught Comsol? How?
Thanks very much.
Sincerely,
Patricia.
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi,
Here is my version of 2D model. I changed mode to turbulent flow and made finer mesh. Finer mesh helps sometimes but make sure that flow is laminar whole area.
I made some version of convective heat transfer coefficient under temperature section (turbulence/wall function). This is not valid for laminar flow.
Br
Tero
Here is my version of 2D model. I changed mode to turbulent flow and made finer mesh. Finer mesh helps sometimes but make sure that flow is laminar whole area.
I made some version of convective heat transfer coefficient under temperature section (turbulence/wall function). This is not valid for laminar flow.
Br
Tero
Attachments:
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi Tero,
Thank you very much for your answer, but I cannot open the model. I don't know if that happens because I have a different version of Comsol...my version is 4.2.a. I don't really know.
Could you attach it again,please?
According to what you said, it would not be possible calculate the convection heat transfer coefficient with laminar flow, would it?
Thank you very much.
Sincerely,
Patricia.
Thank you very much for your answer, but I cannot open the model. I don't know if that happens because I have a different version of Comsol...my version is 4.2.a. I don't really know.
Could you attach it again,please?
According to what you said, it would not be possible calculate the convection heat transfer coefficient with laminar flow, would it?
Thank you very much.
Sincerely,
Patricia.
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi,
Sorry, wrong version. Here is 4.2a version.
Convection heat transfer coefficient with laminar flow is possible but you have to found right expressions to define it.
Br
Tero
Sorry, wrong version. Here is 4.2a version.
Convection heat transfer coefficient with laminar flow is possible but you have to found right expressions to define it.
Br
Tero
Attachments:
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi Tero,
Thank you so much for your 2D version. It helped me to understand better some concepts.
But, I have one question which I would like to ask you. In my 2D version I showed the forced convection through an inlet-velocity profile, uy. This velocity profile is approximated to parabolic curve in the vertical, y axe, direction due to I supposed the similitude to a developed laminar profile. The equation is:
uy = 4x’ (1-x’) Umax
Now, in the last 2D version I have turbulent flow, hence, will I need to change the velocity profile? If I need to change it, how would I be able to know the new profile?
Thank you.
Sincerely,
Patricia.
Thank you so much for your 2D version. It helped me to understand better some concepts.
But, I have one question which I would like to ask you. In my 2D version I showed the forced convection through an inlet-velocity profile, uy. This velocity profile is approximated to parabolic curve in the vertical, y axe, direction due to I supposed the similitude to a developed laminar profile. The equation is:
uy = 4x’ (1-x’) Umax
Now, in the last 2D version I have turbulent flow, hence, will I need to change the velocity profile? If I need to change it, how would I be able to know the new profile?
Thank you.
Sincerely,
Patricia.
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi Patricia
Fully developed turbulent flow (as least in pipes and between plates) has different type of form than parabolic. Flow looks like more "flatten" parabolic curve. One way to get real form is to make inlet channel long enough to get flow fully developed to right shape.
I hope this will help you.
Br
Tero
Fully developed turbulent flow (as least in pipes and between plates) has different type of form than parabolic. Flow looks like more "flatten" parabolic curve. One way to get real form is to make inlet channel long enough to get flow fully developed to right shape.
I hope this will help you.
Br
Tero
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi Tero,
Thank you for your reply, now it is clearer.
As you said one solution would be to make inlet channel long enough to get flow fully developed to right shape. But, my model has the dimensions set, hence, I will not be able to modify the lenght of the inlet channel.
The Reynolds number is not really big (this could be considered between laminar and turbulent flow) due to the maximum velocity is 2,2 m/s. So, I think that there would not be a big error if I use the parabolic velocity profile in the turbulent simulation. Please, let me know if I am wrong.
Thank you so much.
Sincerely,
Patricia.
Thank you for your reply, now it is clearer.
As you said one solution would be to make inlet channel long enough to get flow fully developed to right shape. But, my model has the dimensions set, hence, I will not be able to modify the lenght of the inlet channel.
The Reynolds number is not really big (this could be considered between laminar and turbulent flow) due to the maximum velocity is 2,2 m/s. So, I think that there would not be a big error if I use the parabolic velocity profile in the turbulent simulation. Please, let me know if I am wrong.
Thank you so much.
Sincerely,
Patricia.
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi Tero,
I forgot to ask you one thing more.
The device has a fan, so I considered forced convection and at the same time I took into account the buoyancy force due to heating of the fluid what produces difference between local density, ρ, and the inlet density, ρ0. I expressed that like: (ρ-ρ0) ∙ g
Anyway, checking other Comsol models from the model library, I am not sure if I need to consider the buoyancy force if I consider forced convection because in the "circuit board forced 3D natural" is taken into account the buoyancy force, but in the "circuit board forced 3D forced" not.
Thank you so much.
Sincerely,
Patricia.
I forgot to ask you one thing more.
The device has a fan, so I considered forced convection and at the same time I took into account the buoyancy force due to heating of the fluid what produces difference between local density, ρ, and the inlet density, ρ0. I expressed that like: (ρ-ρ0) ∙ g
Anyway, checking other Comsol models from the model library, I am not sure if I need to consider the buoyancy force if I consider forced convection because in the "circuit board forced 3D natural" is taken into account the buoyancy force, but in the "circuit board forced 3D forced" not.
Thank you so much.
Sincerely,
Patricia.
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi Patricia,
I don't see big problem to use parabolic velocity profile with turbulent flow.
Usually if you have forced convection, natural convection can be forget. But this depend lot of by case so I cannot give straight guidelines.... High temperature differences, places were forced convection cannot reach, etc could need natural convection.
Br
Tero
I don't see big problem to use parabolic velocity profile with turbulent flow.
Usually if you have forced convection, natural convection can be forget. But this depend lot of by case so I cannot give straight guidelines.... High temperature differences, places were forced convection cannot reach, etc could need natural convection.
Br
Tero