Discussion Forum

Posted: 2 decades ago 6 oct. 2009, 06:01 UTC−4

Hi You can find a very nice example in the SMEUG.pdf file EXAMPLE: Torsional moment on a cylinder. With all explanations how to apply it, I would suggest to try out the example first. As rotations are not defined by default in 3D structural, you need to work a little with coordinate transforms, but it works very well. Good luck Ivar

Posted: 2 decades ago 6 oct. 2009, 06:10 UTC−4

Thank you for your reply. I've calculated the transformation from the angle of rotation to the coordinate change. I've tested it in Excel firstly and in the Comsol I've got the same displpacements like in Excel. But in Comsol I've got the irrealistique stresses. I'll looking for this exemple. Thank you Ivar.

Posted: 2 decades ago 6 oct. 2009, 06:14 UTC−4

I am sorry I cannot find this file. How I can browse it?

Posted: 2 decades ago 6 oct. 2009, 07:15 UTC−4

Hi Sorry the SMEUG.PDF is the "Structural Mechanics Module, User's Guide", see page of V3.5a page 69++ or just scan for "Torsional". For me its in .../comsol35a/doc/sme/ I have indexed all my helpfiles and pdf manuals with MS indexer, it's then easyer to find items there, but there is also the help files with a search function

Posted: 2 decades ago 6 oct. 2009, 08:33 UTC−4

I have no yet this module, I've done my model in Multiphysics/Structure. Could you please attach me the pdf file to read the exemple or on my email or could we find this doc on the Comsol site? Thank you very much for your help.

Posted: 2 decades ago 6 oct. 2009, 10:29 UTC−4

Hi If you are going to do a lot of structural as I can see on your image, you should very seriously consider the purchase of the structural module, you will quickly gain time covering the price difference . I hope the explanation below works ok without, but as I understand it there is not much specific to the structural module. And I hope COMSOL finds it OK to export an excerb of their doc, if not, they know how to suppress it ;) By the way, some of these docs can also be found on the web with a Google search, albeit I find that less conform to normal (c) standards Good luck Ivar

Attachments:

• smeugp69-71.pdf

Posted: 2 decades ago 6 oct. 2009, 11:26 UTC−4

Thank you very much Ivar, Do not worry, I've also asked the support guys to send me this file. It'll be OK. I'll study the exemple. I'll keep you updated if my problem interest you. Thanks

Posted: 1 decade ago 23 janv. 2010, 05:47 UTC−5

hello Ivar, I had gone through the discussion. Iam having one doubt in similar to it. Iam doing a module in comsol on concept of winding and unwinding a elastic spiral spring w.r.t a cylindrical rod in middle of the spring. Now i want to rotate the rod spherically, so that the spring should wind and unwind. Here i fixed the two ends of the spring and tried normally in cartesian coordinates. Output obtained is just moving the rod in one direction . I require information in rotating rod spherically. pls do the needful help. Thanking you..

Posted: 1 decade ago 23 janv. 2010, 06:08 UTC−5

Hi If I understand you well, you are in 2D or 3D cartesian and want to define a "rotation" around an axis. Well that is for me rather similar to the case described in the doc, apart that its not a load but a prescribed deplacement you want. My first try qwould be something like define your axis and a radial vector r=sqrt((x-x0)^2+(y-y0)^2) where x0,y0 is the centre of the axis, and I assume the axis si paralle to"z", if not you would ahve to apply some rotation matrices to the formual o get the correct components, but the principle remains. Then you can define the angle theta_z=atan2((y-y0),(x-x0)) (or is it the opposite?, I always get it wrong, Murhy I suppose again, pls check). then you apply a prescribed dispalcement on the x,y components of the cylindrical edge (assumed in the x-y plane) of your mechanical axis and you enforce a displacement relative to theta_z. You can do this cleaner in the way described, with a secondary coordinate system fromn a workplane defined normal to your ratation axis and with the origine on the axis intersection. Goos luck Ivar

Posted: 1 decade ago 25 janv. 2010, 02:15 UTC−5

thank you Iam doing model in 3D space dimension under comsol multiphysics module -> structural mechanics -> stress/strain. I didnt get how to apply displacement in this module, so you tell me the process how to apply displacement on total cylindrical rod

Posted: 1 decade ago 25 janv. 2010, 07:37 UTC−5

Hi If I understand you it's the notation that confuess you? in sld mode a Boundary constraint = forced displacement of Rx=0.1[m] is a forced displacement of 10 cm on all elements selected along the x direction. Not selecting a direction such as y, and z means the elements are free to move "freely" without any forced position constraints. What happens within COMSOL when selecting a forced displacement ? simply COSOL puts up an internal constraints equation Rx_sld-u = 0 to be solved check the Equation System - Boundary Settings GroupN values For a more generalised approach you can select H R consitions (I use these very seldom), for details see the documentation. for loads (forces on a face = pressure) its the load tab. for rotational moments you need to be naturally in a cylindrical coordinate with the axis on the axis of rotation, as this is not the default you must construct one and apply forces in this environment (or write out the curl form directly in the equations). This is what is described in the structural doc about applying torque boundary conditions, and this is vey similar to the reaction moment formulas you can see in the Physics - Equations - Boundary settings - Variables tab: RMxsld RMy_sld ...) Hope this helps Ivar

Posted: 1 decade ago 27 janv. 2010, 07:50 UTC−5

Hi, Thank u for reply.Iam not getting in correct way to do the model so I attached my geometrical model here, that is in 3D space dimension . The two ends of elasticspring are fixed by using holes kept on it. In middle of the spring a cylindrical subdomain is added. Now the subdomain should rotate with the axis of rotation, and due to that there should be deformation in the spring(strain calculation due to angulardisplacement) i.e, spring winds and unwinds. This is my problem. I didnt understand how to do this process(analysing) in Comsol. I followed comsol multiphysics-structural mechanics-solid-stress/strain-static analysis. I request u to send an idea how to do this process.

Attachments:

• model.jpg

Posted: 1 decade ago 27 janv. 2010, 17:01 UTC−5

Hi Well basically a whatchmakers spiral and the momentum wheel. For me it applies well to the SMEUG.PDF example (the "Structural Mechanics Module, User's Guide", see page of V3.5a page 69++ or just scan for "Torsional". For me its in .../comsol35a/doc/sme/). I agree that the explanations are not crystal clear, and I still have a unit issue I'm not happy with (at least in V3.5a, when I follow the explanations literally) but the rotation is there when I model it, even if I cannot yet get the reaction moment to match my applied moment. One of the items is that the "cylindrical coordinate" check used to define the new user coordinate system generates a specific rotation matrix (dump your model as a M file and check the coordinate section) that does the trick, the "geom2" workplane is only there as a placeholder to access this option for the new corrdinate transform. If you want to force and fix the rotation axis, then I would suggest to add two points along the axis on the top/bottom surface of your shaft, and enforce a displacement Rx=0 Ry=0 Rz=0 on both points. It's slightly crude as a "point constraint" but works still quite well. Good luck Ivar

Posted: 1 decade ago 9 févr. 2010, 02:55 UTC−5

Hello, i want information about how to calculate strain after getting the output of my model which attached previously. In plot parameters i selected ------boundary and deformed type plots. thank you

Posted: 1 decade ago 9 févr. 2010, 02:57 UTC−5

Hello, i want information about how to calculate strain after getting the output of my model which attached previously. In plot parameters i selected ------boundary and deformed type plots. I require formula or information how to calculate thank you

Posted: 1 decade ago 9 févr. 2010, 03:03 UTC−5

Thank for info. i was trying to do it Now i want to know how to calculate strain at particular point in the spring i need info. to calculate strain (either formula) using the displacements or any other

Posted: 1 decade ago 9 févr. 2010, 05:45 UTC−5

hello, basically i dont how to analyse the simulation,i just know how to geometric model. My theoritical strain formula is According to mechanical theory of torsional springs strain , ?=(6*J*??) /(h*b^2*l) =(h*??)/(2*l) ??=angular displacement J=inertial moment of transverse section of spiral torsion spring i know values h,b,l (geometric parameters of spring) i dont know about J i hve to relate the obtained the simulation strain with this theoritical strain In comsol how can we know the strain and how the the simulation strain will follow as same concept as theoritical strain pls do needful help thank you

Posted: 1 decade ago 9 févr. 2010, 05:45 UTC−5

hello, basically i dont how to analyse the simulation,i just know how to geometric model. My theoritical strain formula is According to mechanical theory of torsional springs strain , ?=(6*J*??) /(h*b^2*l) =(h*??)/(2*l) ??=angular displacement J=inertial moment of transverse section of spiral torsion spring i know values h,b,l (geometric parameters of spring) i dont know about J i hve to relate the obtained the simulation strain with this theoritical strain In comsol how can we know the strain and how the the simulation strain will follow as same concept as theoritical strain pls do needful help thank you

Posted: 1 decade ago 9 févr. 2010, 05:45 UTC−5

hello, basically i dont how to analyse the simulation,i just know how to geometric model. My theoritical strain formula is According to mechanical theory of torsional springs strain , ?=(6*J*??) /(h*b^2*l) =(h*??)/(2*l) ??=angular displacement J=inertial moment of transverse section of spiral torsion spring i know values h,b,l (geometric parameters of spring) i dont know about J i hve to relate the obtained the simulation strain with this theoritical strain In comsol how can we know the strain and how the the simulation strain will follow as same concept as theoritical strain pls do needful help thank you

Posted: 1 decade ago 10 févr. 2010, 04:37 UTC−5

strain=(6*J*angulardisplacement)/(h*b^2*l) angular displacement is in radians

Posted: 1 decade ago 17 févr. 2010, 01:19 UTC−5

Hello Ivar, In the example torsional cylinder which u gave i tried,its working. I have a doubt in that ,i.e., how to find the angle of rotation at the particular point in that rotating cylinder Thank you

Posted: 1 decade ago 18 févr. 2010, 02:12 UTC−5

Hi Sorry coming in late, but I am still very busy and have had little time to go on the forum. Nevertheless, I'll try to give some replies hereafter, as I understand from your questions: - strain, stress calculations,: COMSOl have them internally, the trick is to get read them = postproces and to be sure you have them in the deisred coordinate refeence (by default COSOL is in cartesian x,y,z, The strain tensor (6 independent items for isotropic material) are in the Subdomain postprocessing with the names: ex,ey,ez, normal and exy, eyz, exz shear elements. Now there are other ways to work with strain (as well as stress) vlues, this depends on your domain, I can only suhggest to look into the classical books on elasticity. Depending what you are looking for you might need to integrate these values over your domain. Now angles of rotation, COMSOL does not inlcude angles in 3D smsld, I have discusseed that a couple of times with the people of support and the developpers when I meet tham at the conferences, one of the issues is that the conventional angle value is only really valid for "small" angles, and the non-linear ode of COMSOL allos you to simulate large angles and then these "small" angles values are really no longer correct. In your case with your spiral spring you are somwhere in the middle towards the "large angle" approach so pls be aware. If you read through chapter 2 of the excelent short book: http://www.amazon.com/Elasticity-Tensor-Dyadic-Engineering-Approaches/dp/0486669580/ref=sr_1_2?ie=UTF8&s=books&qid=1266476488&sr=8-2 see also the thread "breathing eigenmode of a 3D solid sphere" you will see that you can define the "small" angles as thu = 0.5*(wy-vz) thv = 0.5*(uz-wx) thw = 0.5*(vx-uy) It's the antisymmetric, or curl part of the general displacement formulation shown hereafter in 2D: du = (du/dx)*dx + 1/2*(du/dy + dv/dx)*dy + 1/2*(du/dy - dv/dx)*dy or du = ex*dx+1/2*exy*dy-thz*dy You can recognise the normal and shear strain and the "rigid body" rotation thz I hope this will help you on your way Good luck Ivar

Posted: 1 decade ago 18 févr. 2010, 03:41 UTC−5

Thank you Ivar for your reply and for your help in this. Julia

Posted: 1 decade ago 23 févr. 2010, 00:13 UTC−5

Thank you , i will go through it n discuss to u if any doubts udaya lakshmi

Posted: 1 decade ago 23 févr. 2010, 04:04 UTC−5

Hi there are a few threads on the forum about torque load an torsional moments, and in the structural documentation smeug.pdf there is a good example, I needed some tries before I catched the "trick" but it works fine for me try "torsional moment" in the search above Good luck Ivar

Posted: 1 decade ago 21 avr. 2010, 07:26 UTC−4

hello Iam trying to simulate bragg gratings in 2D in perpendicular modes -> hybrid mode waves After assigning conditions in subdomain settings(abt refractive index) when i tried to solve , iam getting an error of"singular matrix" Boundary conditions is considered as default. I would like to know how to resolve the problem So kindly do the needful Regards udaya

Posted: 1 decade ago 21 avr. 2010, 07:42 UTC−4

Hi "singular matrix" means, as it states, lack of enough equations to solve the model, so either you are lacking a one or a few BC's, or there is a variable missing (such as a temperature T if you use non-linear materials from the database and you do not specifically solve the model with any given T, that one is simple: just add a constant T=20[degC] or whatever and its solved). But it could be trickier. Hope this helps on the way Ivar

Posted: 1 decade ago 23 avr. 2010, 02:16 UTC−4

Hello Ivar Iam trying to design fiber bragg grating sensor in RF module -> Electromagnetic waves -> harmonic propagation. FBG principle is , when applied a broadband waveform to FBG, at gratings some part of input gets reflected and the other is transmitted in the fiber.The reflection is occured after satisfying condition bragg wavelength = 2*neff*grating period neff = effective refractive index I finished model by considering two cylinders as core and cladding and in middle i kept gratings( of 5 number). coming to simulation part i assigned conditions in physics as 1. In subdomain settings i assigned refractive index for core, cladding, gratings 2. In boundary settings i tried to assign port for core as input and selected wave excitation at this port and in port block in boundary settings i kept the default TE wave propagation In same way i assigned port2 for core output and for port3 for gratings without any wave excitation. 3. Initialized mesh, it got 37585 elements 4. In solver settings, stationary analysis is selected So after solving it is showing an error of "singular matrix" So i need help in . to overcome the error. I would like to know , any other conditions to be added for simulation process to satisfy the above principle and get the reflection wave form ncore = 1.45 ncladd - 1.4355 neff= 1.44995 grating period = 0.5e-6m these are parameters i considered. So if possible kindly do the needful Regards Udaya

Posted: 1 decade ago 30 avr. 2010, 00:45 UTC−4

Hello Ivar, i have a doubt in solver settings. when i solve my model result is obtaining. to solve model again with same values iam doing the following procedure file -> reset model After initializing mesh i kept for solving , this time iam not getting the same result as previous when i try third time again the result is differing, for no. of times i tried but result is not same iam not understanding why its happening iam doing static analysis in comsol multiphysics -> structural mechanics -> solid stress/strain so i would like to know the reason. kindly do the needful regards udaya

Posted: 1 decade ago 5 mars 2014, 10:44 UTC−5

Hi Udaya Lakshmi, Can I please know how u designed 2-d bragg grating in comsol?

Posted: 1 decade ago 10 juin 2014, 15:45 UTC−4

Dear Ivar I was trying to reproduce the "Torsional moment on a cylinder" example in version 4 As I understood the moment is expressed in terms of stress distribution at boundary surfaces as M=int(Taz_smsld*y-Tay_smsld*z) Where the value of M is set to 100 and tau is defined as tau=M-100 What I dont understand is what is the need to add "tau*r" as boundary load? . If M is set to 100 then tau goes 0 ! Regards, H.

Posted: 1 decade ago 1 juil. 2014, 09:05 UTC−4

[QUOTE] Hi Udaya Lakshmi, Can I please know how u designed 2-d bragg grating in comsol? [/QUOTE] Hi V Neeharika, I'm also interested in modeling the FBG in 2D from along time and tried to do that but still not quite sure. I think Udaya Lakshmi have done it with 3D since he used cylindricals to build the FBG-sensor. For me in 2D is much conveinent since it will not take loads from the memory size. My concerns also about the number of grating pitches to get a strong reflection and FBG. And I wonder if we need to apply the PML or Scattering condition around the FBG model ? please, if any have any information or some tips about that, can kindly share it ?! Kind regards Medya