Discussion Forum

Posted: 1 decade ago 21 lug 2014, 10:47 GMT-4

Dear Tejas, you can try to use your steady-state solution as an initial value for your time-dependent study - at least for your flow field. This will help the solver to find a solution for the first time steps. Simply add a stationary step before the time dependent step. Another option is to ramp up the boundary condition expression in time, e.g. by multiplying them with some step-function like "steady_state_expression * step1(t)". This will allow you to avoid discontinuities between domain and boundary conditions at time step t=0. -- Best regards, Erik ********************* Erik Bornhöft Senior Technical Sales Engineer COMSOL Multiphysics GmbH Robert-Gernhardt-Platz 1 37073 Göttingen Deutschland Knowledge Base: http://www.comsol.de/support/knowledgebase/browse/900/ COMSOL Blog: http://www.comsol.de/blogs/

Posted: 8 years ago 16 nov 2016, 03:41 GMT-5

[QUOTE] Dear Tejas, you can try to use your steady-state solution as an initial value for your time-dependent study - at least for your flow field. This will help the solver to find a solution for the first time steps. Simply add a stationary step before the time dependent step. Another option is to ramp up the boundary condition expression in time, e.g. by multiplying them with some step-function like "steady_state_expression * step1(t)". This will allow you to avoid discontinuities between domain and boundary conditions at time step t=0. -- Best regards, Erik ********************* Erik Bornhöft Senior Technical Sales Engineer COMSOL Multiphysics GmbH Robert-Gernhardt-Platz 1 37073 Göttingen Deutschland Knowledge Base: http://www.comsol.de/support/knowledgebase/browse/900/ COMSOL Blog: http://www.comsol.de/blogs/ [/QUOTE] Dear Erik, I have met the same problem. If I use the stationary solution as the initial values for the unsteady solver, I would met the same problem as in https://www.comsol.com/community/forums/general/thread/44477/ and https://www.comsol.com/community/forums/general/thread/33203/ In addition, I used the periodic boundary conditions, it seems that the definition "steady_state_expression * step1(t)", as has been advised, has no place to put. Could you please give me some suggestions? Thank you in advance. Best Regards, Bill

Posted: 8 years ago 17 nov 2016, 05:38 GMT-5

Dear Bill, I am not sure which issues of the referenced threads are the same for you. I can give the following suggestions though: - If your own PDE implementation of Navier-Stokes-Eq. does not converge try the implemented "Laminar Flow Interface" instead which comes with predefined stabilizations, boundary conditions, mesh- and solver-settings - Unsteady flow from constant boundary conditions needs generally to be induced by some "disturbance". You can do some investigations with this model (see also its documentation): https://www.comsol.de/model/flow-past-a-cylinder-97 - You could ramp the pressure difference in your periodic conditions; I cannot tell if that makes any sense for your specific model though Apart from that our technical support team is happy to give more specific advice on our customers models: https://www.comsol.com/support [QUOTE] Dear Erik, I have met the same problem. If I use the stationary solution as the initial values for the unsteady solver, I would met the same problem as in https://www.comsol.com/community/forums/general/thread/44477/ and https://www.comsol.com/community/forums/general/thread/33203/ In addition, I used the periodic boundary conditions, it seems that the definition "steady_state_expression * step1(t)", as has been advised, has no place to put. Could you please give me some suggestions? Thank you in advance. Best Regards, Bill [/QUOTE] -- Best regards, Erik ********************* Erik Bornhöft Senior Technical Sales Engineer COMSOL Multiphysics GmbH Robert-Gernhardt-Platz 1 37073 Göttingen Deutschland Knowledge Base: http://www.comsol.de/support/knowledgebase/browse/900/ COMSOL Blog: http://www.comsol.de/blogs/

Posted: 8 years ago 17 nov 2016, 08:22 GMT-5

[QUOTE] Dear Bill, I am not sure which issues of the referenced threads are the same for you. I can give the following suggestions though: - If your own PDE implementation of Navier-Stokes-Eq. does not converge try the implemented "Laminar Flow Interface" instead which comes with predefined stabilizations, boundary conditions, mesh- and solver-settings - Unsteady flow from constant boundary conditions needs generally to be induced by some "disturbance". You can do some investigations with this model (see also its documentation): https://www.comsol.de/model/flow-past-a-cylinder-97 - You could ramp the pressure difference in your periodic conditions; I cannot tell if that makes any sense for your specific model though Apart from that our technical support team is happy to give more specific advice on our customers models: https://www.comsol.com/support -- Best regards, Erik ********************* Erik Bornhöft Senior Technical Sales Engineer COMSOL Multiphysics GmbH Robert-Gernhardt-Platz 1 37073 Göttingen Deutschland Knowledge Base: http://www.comsol.de/support/knowledgebase/browse/900/ COMSOL Blog: http://www.comsol.de/blogs/ [/QUOTE] Dear Erik, Thank you for your reply. The problems I have met are as follows. 1) Using PDE equations, if I adopted the stationary solutions as the initial values for the time-dependent sovler, the obtained transient results will just the same as the steady ones, even for a very large Re number; 2) I have tried the built-in laminary flow interface. But the results strongly depend on the initial values as shown in the attachment (the initial value for u is 1e-2, 1e-20 separately at Re=10); since I run a large range for Re and the geometry is quite complex, which make it impossible to calculate the initial values at the corresponding Re. 3) As for the "flow past a cylinder" model, it defines a "step1" to trigger the disturbance at the inlet boundary; while I used the periodic boundary conditions, thus the only place to use "step1" is at the initial values setting windows, which will mix the problem (2) about determining the initial values. Hope to hear from you soon. Thank you in advance. Best Regards, Bill

Attachments:

• 1e-2,10.jpg
• 1e-20，10.jpg

Posted: 8 years ago 18 nov 2016, 06:05 GMT-5

Dear Bill, I can see from your questions a discussion on your specific model would be very useful for you; e.g. checking for your (pressure) boundary- and (absolute tolerance) solver-settings. Especially when your project involves complex geometries there might be more to discuss than the starting conditions. So I highly recommend to get in touch with our support team: www.comsol.com/support -- Best regards, Erik ********************* Erik Bornhöft Senior Technical Sales Engineer COMSOL Multiphysics GmbH Robert-Gernhardt-Platz 1 37073 Göttingen Deutschland Knowledge Base: http://www.comsol.de/support/knowledgebase/browse/900/ COMSOL Blog: http://www.comsol.de/blogs/

Posted: 8 years ago 1 dic 2016, 03:22 GMT-5

Dear Erik, Thank you. I have tried to contact the support, but since I used the PDEs, it is quite difficult for them to find a solution. I am trying to solve the Navier-Stokes equations with the built-in laminary solver instead of using the PDE equations. I have run the case and the results from the transient solutions seem to be that obtained with a steady solution. The expected result can be seen in the reference(attachment). Is there something wrong about my settings? or should I add a perturbation to induce the unsteady flow, as in http://www.cfd-online.com/Forums/main/135769-time-dependent-solver-unsteady-navier-stokes.html Would you please have a check? Hope to hear from you soon. Thank you. Best Regards, Bill

Attachments:

• equations.jpg
• reference.jpg
• laminary-periodic-unsteady.mph