Discussion ClosedThis 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 consistent initial values?

Please login with a confirmed email address before reporting spam

Hello,
I am having difficulty getting comsol to converge on time dependent solutions. I get this error pretty consistently:

Error:
Failed to find consistent initial values.
Last time step is not converged.

I am not sure what is meant by 'consistent initial values'. If someone could shed some light on this I would greatly appreciate it.

-Michael Petrarca

8 Replies Last Post 2009年8月22日 GMT+2 03:01

Please login with a confirmed email address before reporting spam

Posted: 2 decades ago 2009年7月13日 GMT+2 22:49
To be more specific, I am doing variations on the cylinder flow model presented as an example for comsol. I have changed the geometries of the cylinder and the fluid properties. The cylinder is .4 meters in diameter and the fluid is air at room temperature (density of 1.225 kg/m^3 and dynamic viscosity of 1.81e-7). I received this error for reynolds numbers of both 1e2 and 1e7.
To be more specific, I am doing variations on the cylinder flow model presented as an example for comsol. I have changed the geometries of the cylinder and the fluid properties. The cylinder is .4 meters in diameter and the fluid is air at room temperature (density of 1.225 kg/m^3 and dynamic viscosity of 1.81e-7). I received this error for reynolds numbers of both 1e2 and 1e7.

David Wenger Certified Consultant

Please login with a confirmed email address before reporting spam

Posted: 2 decades ago 2009年7月14日 GMT+2 17:26
In these cases, some boundaries or properties are not specified properly. I usually try to "Get the Initial Values". Have a look at the results. Does look like your model specification?

David Wenger
www.comsol.de/company/consultants/wenger/
In these cases, some boundaries or properties are not specified properly. I usually try to "Get the Initial Values". Have a look at the results. Does look like your model specification? David Wenger www.comsol.de/company/consultants/wenger/

Please login with a confirmed email address before reporting spam

Posted: 2 decades ago 2009年7月16日 GMT+2 01:50
My boundary conditions are all set properly, as are my subdomain settings. I found that by setting the mesh to be coarse and the reynolds number to be extremely low, it solves the problem in seconds...however once i reach a certain reynolds number it decides its missing some initial condition? I am at a loss here.
My boundary conditions are all set properly, as are my subdomain settings. I found that by setting the mesh to be coarse and the reynolds number to be extremely low, it solves the problem in seconds...however once i reach a certain reynolds number it decides its missing some initial condition? I am at a loss here.

Please login with a confirmed email address before reporting spam

Posted: 2 decades ago 2009年7月16日 GMT+2 08:27
Dear Michael,
maybe it is a matter of "absolute tolerances" for the time dependent solver.
Try to estimate the magnitude of the physical quantities you are solving for and set an absolute tolerance for each of them which is at least one order of magnitude less than the expected magnitude of that quantity.
Usually pressure and velocities require different tolerances since their order of magnitude are quite different.
I hope this helps.

Alessandro
Dear Michael, maybe it is a matter of "absolute tolerances" for the time dependent solver. Try to estimate the magnitude of the physical quantities you are solving for and set an absolute tolerance for each of them which is at least one order of magnitude less than the expected magnitude of that quantity. Usually pressure and velocities require different tolerances since their order of magnitude are quite different. I hope this helps. Alessandro

Please login with a confirmed email address before reporting spam

Posted: 2 decades ago 2009年7月21日 GMT+2 00:13
The attached simple example does the same thing. When solver parameter is set to stationary mode, it converges very quickly. When the domains are set to transient, it comes to a quick halt with an inability to find a good initial start.

I've tried helping it along: run in stationary mode, store the solution, then tell solver manager to use the stored solution as the initial guess.

Nope.
The attached simple example does the same thing. When solver parameter is set to stationary mode, it converges very quickly. When the domains are set to transient, it comes to a quick halt with an inability to find a good initial start. I've tried helping it along: run in stationary mode, store the solution, then tell solver manager to use the stored solution as the initial guess. Nope.


Please login with a confirmed email address before reporting spam

Posted: 2 decades ago 2009年7月21日 GMT+2 00:35

In these cases, some boundaries or properties are not specified properly. I usually try to "Get the Initial Values". Have a look at the results. Does look like your model specification?

Wenger
www.comsol.de/company/consultants/wenger/


Hi David,

Where is "Get the Initial Values"? Where can I find it?

Thanks
[QUOTE] In these cases, some boundaries or properties are not specified properly. I usually try to "Get the Initial Values". Have a look at the results. Does look like your model specification? Wenger www.comsol.de/company/consultants/wenger/ [/QUOTE] Hi David, Where is "Get the Initial Values"? Where can I find it? Thanks

Please login with a confirmed email address before reporting spam

Posted: 2 decades ago 2009年7月21日 GMT+2 00:59
It is under the Solve Menu, just above "Solver Parameters".

I've been playing around some more with the file I uploaded. If I run it in Stationary mode, then go to Solver Settings, and click "Store Solution", then go back to Solver parameters and switch everything over to Transient with segrated solution. Then click on "Get Initial Value", then COMSOL does exactly what I want, namely it plots as the initial value exactly the same solution as it has converged to last time.

I can then go to Solver Manger and click "use stored solution as initial value", but it still doesn't converge.

Depending upon the solver parameter setting tolerances, time-step, etc, it still sometimes complains "Can't find consistent initial values", more commonly it just complains about not converging.

Regards, John




It is under the Solve Menu, just above "Solver Parameters". I've been playing around some more with the file I uploaded. If I run it in Stationary mode, then go to Solver Settings, and click "Store Solution", then go back to Solver parameters and switch everything over to Transient with segrated solution. Then click on "Get Initial Value", then COMSOL does exactly what I want, namely it plots as the initial value exactly the same solution as it has converged to last time. I can then go to Solver Manger and click "use stored solution as initial value", but it still doesn't converge. Depending upon the solver parameter setting tolerances, time-step, etc, it still sometimes complains "Can't find consistent initial values", more commonly it just complains about not converging. Regards, John

Please login with a confirmed email address before reporting spam

Posted: 2 decades ago 2009年8月22日 GMT+2 03:01
In a similar turbulent natural convection case, how do you initialize the turbulent quantities (log of kinetic energy and dissipation rate) in a transient simulation starting from rest? Also, technically the flow will be laminar until a crtitical velocity is reached; is it ok to model it as turbulent throughout?
In a similar turbulent natural convection case, how do you initialize the turbulent quantities (log of kinetic energy and dissipation rate) in a transient simulation starting from rest? Also, technically the flow will be laminar until a crtitical velocity is reached; is it ok to model it as turbulent throughout?

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

Baidu
map