Hello there,
I'm working on a simulation of viscoelastic flows in a square duct. Since our measurements are performed in the center plane of the channel, we can basically consider that a 2-dimensional flow situation, an expansion to three dimensions might happen later, but is currently not necessary.
As a start, I tried to implement the Oldroyd-B model, as it is basically already given in the Model library, to check for consistency with literature results (e. g. Rhees, Zimmerman, Craven and Oliveira).
The control parameter in the model is the Weissenberg number (rescaled intake velocity) Wi. For the range of Wi = 0 .. 1.1, so basically in the regime where viscoelastic effects play only a minor role, this works fine. However, if Wi becomes greater than 1.1, COMSOL gives me an error message about failing to reach convergence.
The literature models present simulations of Weissenberg numbers up to around 3 to 5, so basically there should be no problem, even if the relative tolerance is chosen as low as 10^(-7), in contrast to my 10^(-3). I'm using a MUMPS direct solver in COMSOL 4.3. All other parameters are as given in the library model.
Any takes on this, or has anyone encountered the problem as well?
Regards
Philipp Bohr
I'm working on a simulation of viscoelastic flows in a square duct. Since our measurements are performed in the center plane of the channel, we can basically consider that a 2-dimensional flow situation, an expansion to three dimensions might happen later, but is currently not necessary.
As a start, I tried to implement the Oldroyd-B model, as it is basically already given in the Model library, to check for consistency with literature results (e. g. Rhees, Zimmerman, Craven and Oliveira).
The control parameter in the model is the Weissenberg number (rescaled intake velocity) Wi. For the range of Wi = 0 .. 1.1, so basically in the regime where viscoelastic effects play only a minor role, this works fine. However, if Wi becomes greater than 1.1, COMSOL gives me an error message about failing to reach convergence.
The literature models present simulations of Weissenberg numbers up to around 3 to 5, so basically there should be no problem, even if the relative tolerance is chosen as low as 10^(-7), in contrast to my 10^(-3). I'm using a MUMPS direct solver in COMSOL 4.3. All other parameters are as given in the library model.
Any takes on this, or has anyone encountered the problem as well?
Regards
Philipp Bohr