[COMSOL_Users] output of stationary solver as input to time-dependent solver

[jeremy_rutman]

I need to implement a flux condition on an internal boundary -

Flux= k*(c1-B*c2)
where c1 and c2 are concentrations on left and right sides of
boundary,
k and B are constants, and flux into left side = - flux into right

It looks like I need to split this two regions with a nonlocal
coupling variable. This would all be fine except now I get a

Warning:
Time 0:
Failed to find consistent initial values.
Last time step is not converged.

In femsolver at 239

In femtime at 61
In nonlocal3 at 141

whenever I use a grid finer than the initial default.

The comsol help files suggest using the output of a stationary solver
as input to the time-dependent solver. Any clues how this is done?
I have been using femnlin to solve he stationary problem and femtime
to solve the time-dependent problem.
Thanks
Jeremy

[dibyadeep]

I too had a similar problem in which I had one-way coupling between
the time dependent solution and the stationary solution. What I did
was to solve the stationary solution first using the solver manager,
as a time dependent problem. This I did by substituting the
coefficient "da" of del_u/del_t to 0. Hence Comsol solved for the
stationary solution at different points of time. Then I stored the
resulting solution, and used it to to solve the stationary solver
again by using the solver manager.

This doesn't seem to me the most elegant of methods, since I am
essentially solving a stationary solution using a time dependent
problem, for lack of any better methods. But I am just a new user...so
if you have found anyway to solve such coupled solutions containing
solver types...I would appreciate very much if you could send your
solution to this group.

Sincerely

Dibyadeep

Post by jeremy_rutman
I need to implement a flux condition on an internal boundary -
Flux= k*(c1-B*c2)
where c1 and c2 are concentrations on left and right sides of
boundary,
k and B are constants, and flux into left side = - flux into right
It looks like I need to split this two regions with a nonlocal
coupling variable. This would all be fine except now I get a
Failed to find consistent initial values.
Last time step is not converged.

In femsolver at 239

In femtime at 61
In nonlocal3 at 141
whenever I use a grid finer than the initial default.
The comsol help files suggest using the output of a stationary solver
as input to the time-dependent solver. Any clues how this is done?
I have been using femnlin to solve he stationary problem and femtime
to solve the time-dependent problem.
Thanks
Jeremy

[Hamidreza ADIB RAMEZANI]

Hi,

Do you use the predefined comsol code or you use a PDE one!
It is different. If you use comsol solver for predefined codes, it
well natural that it works properly. Otherwise, take care with
coupling parameters. If you use coupled variables, your Jacoubian
matrix can become very sensible in function of your coupling. Please
read comsol user's guide for more detail information.

H. Ramezani

Post by dibyadeep
I too had a similar problem in which I had one-way coupling between
the time dependent solution and the stationary solution. What I did
was to solve the stationary solution first using the solver manager,
as a time dependent problem. This I did by substituting the
coefficient "da" of del_u/del_t to 0. Hence Comsol solved for the
stationary solution at different points of time. Then I stored the
resulting solution, and used it to to solve the stationary solver
again by using the solver manager.
This doesn't seem to me the most elegant of methods, since I am
essentially solving a stationary solution using a time dependent
problem, for lack of any better methods. But I am just a new

user...so

Post by dibyadeep
if you have found anyway to solve such coupled solutions containing
solver types...I would appreciate very much if you could send your
solution to this group.
Sincerely
Dibyadeep

Post by jeremy_rutman
I need to implement a flux condition on an internal boundary -
Flux= k*(c1-B*c2)
where c1 and c2 are concentrations on left and right sides of
boundary,
k and B are constants, and flux into left side = - flux into right
It looks like I need to split this two regions with a nonlocal
coupling variable. This would all be fine except now I get a
Failed to find consistent initial values.
Last time step is not converged.

In femsolver at 239

In femtime at 61
In nonlocal3 at 141
whenever I use a grid finer than the initial default.
The comsol help files suggest using the output of a stationary solver
as input to the time-dependent solver. Any clues how this is done?
I have been using femnlin to solve he stationary problem and

femtime

Post by dibyadeep

Post by jeremy_rutman
to solve the time-dependent problem.
Thanks
Jeremy

[Dibyadeep Paul]

Hii,

thanks for your reply...actually i think this method is better because, if
you just solve the stationary and the time dependent solutions
simultaneously you are unnecessarily wasting a lot of computer time. In my
case actually I solved only for one time step for the time-independent pdes,
and then used the solution for the time dependent problem. Otherwise since
my problem is quite demanding it would take a lot more time to solve.

sincerely

dibyadeep

On Thu, Apr 3, 2008 at 9:39 PM, Hamidreza ADIB RAMEZANI <

Post by Hamidreza ADIB RAMEZANI
Hi,
Do you use the predefined comsol code or you use a PDE one!
It is different. If you use comsol solver for predefined codes, it
well natural that it works properly. Otherwise, take care with
coupling parameters. If you use coupled variables, your Jacoubian
matrix can become very sensible in function of your coupling. Please
read comsol user's guide for more detail information.
H. Ramezani

Post by dibyadeep
I too had a similar problem in which I had one-way coupling between
the time dependent solution and the stationary solution. What I did
was to solve the stationary solution first using the solver manager,
as a time dependent problem. This I did by substituting the
coefficient "da" of del_u/del_t to 0. Hence Comsol solved for the
stationary solution at different points of time. Then I stored the
resulting solution, and used it to to solve the stationary solver
again by using the solver manager.
This doesn't seem to me the most elegant of methods, since I am
essentially solving a stationary solution using a time dependent
problem, for lack of any better methods. But I am just a new

user...so

Post by dibyadeep
if you have found anyway to solve such coupled solutions containing
solver types...I would appreciate very much if you could send your
solution to this group.
Sincerely
Dibyadeep

"jeremy_rutman"

Post by dibyadeep

Post by jeremy_rutman
I need to implement a flux condition on an internal boundary -
Flux= k*(c1-B*c2)
where c1 and c2 are concentrations on left and right sides of
boundary,
k and B are constants, and flux into left side = - flux into right
It looks like I need to split this two regions with a nonlocal
coupling variable. This would all be fine except now I get a
Failed to find consistent initial values.
Last time step is not converged.

In femsolver at 239

In femtime at 61
In nonlocal3 at 141
whenever I use a grid finer than the initial default.
The comsol help files suggest using the output of a stationary

solver

Post by dibyadeep

Post by jeremy_rutman
as input to the time-dependent solver. Any clues how this is

done?

Post by dibyadeep

Post by jeremy_rutman
I have been using femnlin to solve he stationary problem and

femtime

Post by dibyadeep

Post by jeremy_rutman
to solve the time-dependent problem.
Thanks
Jeremy