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.

Problem computing reaction force (lagrange multipliers)

Please login with a confirmed email address before reporting spam

Hey Comsol users.

I have a problem you might can help me with.

I am dealing with a rocking pile embedded in a soil which I am loading with a time-dependent force. (F = 5000*sin(omega*t)). I use identity pair as pile/soil interface with a continuity condition. I have enabled weak constraint.

Problem is, when I integrate the lagrange multiplier (u_lm) for the surface area (the pile/soil interaction surfaces) I get a output (a force) which is approximately 1.9 times larger than the force I load with. What I do is the following, >derived values > integrate > surface integration), chose my surfaces and evaluates u_lm. I have also tried to add all three multipliers (u_lm+v_lm+w_lm) without success.

I have attached a figure showing the dicrepsencies between my input force and the force calculated by integrating u_lm (langrange multiplier)

Hope you can help.

Best regards,
Nilas


5 Replies Last Post 16.01.2012, 11:12 GMT-5
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 15.01.2012, 06:05 GMT-5
Hi

indeed a 1.9 factor is almost 2 which could indicate that your are doubling the items integrated over.
The LM integration is normally done as a "sum". COMSOL is programmed to detect this and arrange the integration accordingly, but sometimes you must help manually, there is a settings in the operator definition to force integration or node summation (the latter should apply to pure LM's)

--
Good luck
Ivar
Hi indeed a 1.9 factor is almost 2 which could indicate that your are doubling the items integrated over. The LM integration is normally done as a "sum". COMSOL is programmed to detect this and arrange the integration accordingly, but sometimes you must help manually, there is a settings in the operator definition to force integration or node summation (the latter should apply to pure LM's) -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 16.01.2012, 02:17 GMT-5
Hi Ivar,

You mention that I should try the operator definition to force integration. I suppose it is under the "integration settings'". I have tried to change that to Integration, but it did not change anything as far as I can see.

Mayby I misunderstood the thing you suggested?

Best regards,
Nilas
Hi Ivar, You mention that I should try the operator definition to force integration. I suppose it is under the "integration settings'". I have tried to change that to Integration, but it did not change anything as far as I can see. Mayby I misunderstood the thing you suggested? Best regards, Nilas

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 16.01.2012, 03:19 GMT-5
Hi

what about trying it out on a simple example and uploading the file here (no need to give out your full example). Even often by making a simple example you find the solution yourself, that happens often to me ;)

And reread carefully the Lagrange multiplier section in the doc (it's a bit scattered so use an indexer on the pdf doc). particularly about integration and summation to be sure you get it right

And I just checking I have been confused to about reaction forces and LMs and the summation / integration settings (i.e. see help Defining an Integration Model Coupling) reaction forces are summed, LM's are integrated (try it out)

The chapter in the KB and the help file "Calculating Accurate Fluxes" can also give some clues
--
Good luck
Ivar
Hi what about trying it out on a simple example and uploading the file here (no need to give out your full example). Even often by making a simple example you find the solution yourself, that happens often to me ;) And reread carefully the Lagrange multiplier section in the doc (it's a bit scattered so use an indexer on the pdf doc). particularly about integration and summation to be sure you get it right And I just checking I have been confused to about reaction forces and LMs and the summation / integration settings (i.e. see help Defining an Integration Model Coupling) reaction forces are summed, LM's are integrated (try it out) The chapter in the KB and the help file "Calculating Accurate Fluxes" can also give some clues -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 16.01.2012, 04:42 GMT-5
Hi Ivar,

I have here a more simplified version of my model (Unsolved). I found the node controlling the method on how to integrate. It does not change the picture. Mayby with my model at hand, you will be able to see what I am doing wrong.

Best regards,
Nilas
Hi Ivar, I have here a more simplified version of my model (Unsolved). I found the node controlling the method on how to integrate. It does not change the picture. Mayby with my model at hand, you will be able to see what I am doing wrong. Best regards, Nilas


Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 16.01.2012, 11:12 GMT-5
Hi

there are a few things, minor we have different habit ;) whcih makes the model more difficult to understand

I'm not sure you variable "Es" is correct, you are probably missing a factor 10. It's easier to use the material tabs for the material data.
Do not mix parameters and variables, put all your constant "Variables" in the Parameter list and keep the variables for items calculated from solver model results (as far as possible)
Avoid using point loads, user rather boundary loads, this makes less singularities and cleaner results (solver might fail too)

With such a nice 2D-axi model, you are quicker running (and debugging) in 2D-axi, but OK perhaps later you want to make it less symmetric

Try to avoid fixed points too, there too use boundaries, and I do not understand why you select point 33?

I would have started with lower boundary 7 fixed

More important: in "assembly mode" that strictly in this case you do not need, except if you want to later set up contact and friction ? You should select "Create Pairs" this will automatically create the required pairs for you, and if you select "create imprints" you will get a cleaner mesh assembly across the pairs

For periodic time sequence, use a time stepping of "intermediate and not the default "free" type, it skips too many solver steps

and to start with use only 10 points across a sinus period, to debug, you wait far less ,and you gain nothing. when your model is clean, then you can run once over night with a higher resolution

--
Good luck
Ivar
Hi there are a few things, minor we have different habit ;) whcih makes the model more difficult to understand I'm not sure you variable "Es" is correct, you are probably missing a factor 10. It's easier to use the material tabs for the material data. Do not mix parameters and variables, put all your constant "Variables" in the Parameter list and keep the variables for items calculated from solver model results (as far as possible) Avoid using point loads, user rather boundary loads, this makes less singularities and cleaner results (solver might fail too) With such a nice 2D-axi model, you are quicker running (and debugging) in 2D-axi, but OK perhaps later you want to make it less symmetric Try to avoid fixed points too, there too use boundaries, and I do not understand why you select point 33? I would have started with lower boundary 7 fixed More important: in "assembly mode" that strictly in this case you do not need, except if you want to later set up contact and friction ? You should select "Create Pairs" this will automatically create the required pairs for you, and if you select "create imprints" you will get a cleaner mesh assembly across the pairs For periodic time sequence, use a time stepping of "intermediate and not the default "free" type, it skips too many solver steps and to start with use only 10 points across a sinus period, to debug, you wait far less ,and you gain nothing. when your model is clean, then you can run once over night with a higher resolution -- Good luck Ivar

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