Discussion Forum

Hi again,

No replis or suggestions, must I then assume that I'm the only one out here simulating optical elements (structural deformations of) and wanting to use Zernike analysis or ?

Ivar

I also find M-file, such as opto-mechanics. I'm interested in simulating of polymer patterning by optical force using mass transport mechanism.

Hi

sorry for that, but my "Zernike" analysis goal is only to decompose surface deformations, typicallyused in opto-mechanics.

I have leaft a simple model attached, with the principles, it uses structual and optimisation, I'm working on updating it and documenting it correctly, and then I could post it on the model exchange. Ideally I should make a GUI but I have no time, and I'm waiting for V4, as I expect to then have to redo everything.

So documentation is to come, the crude, simplest model with a few levels of the polynomial is attached

Ivar

Hi Ivar,
For quite sometimes I was looking for doing a similar analysis for a telescope mirror. I need to convert the FEA deformation data in COMSOL to appropriate Zernike polynomial to analyze the wavefront aberrations of a reflecting surface. The problem is not straightforward. At least that is the impression I got from reading couple of papers on the same issue. This capability, however, is built into a commercial software called SigFit (www.sigmadyne.com/). This software tool is capable of converting the mechanical deformation data from MSC/PATRAN, FEMAP, and ANSYS into desired optical data format for ray tracing software.

I guess it should be possible to achieve the same thing with COMSOL data as you have mentioned in your earlier post. Could you please throw some more light on this issue as it is vitally important to people like us who want to make use of COMSOL for such analysis. Looking forward to hearing from you.

With kind regards.
-Banyal

PS: You had uploaded a file called Shell.zip in the earlier post. Strange that I could not open it in Windows. Could you please upload it once more.

Hi

In fact I have finally managed quite well in Comsol, but havnt got time yet to explain it and set it up on the model exchange, you have a teaser in the "shell" file, you can easily build up the full ZERNIKE case. Perhaps I find some time over Xmas

Actually I managed to upload it and unzip it (WinZip non MS software) even I add it again here unzipped

My only concern is to get the Zernike normalisation right (as Zemax and other differ) as I'm currently working on an elliptical mirror with central hole (M5 of E-ELT), my partners and "main" use ANSYS + private software, so I have to catch up and doeverything in double, this is annoying, I have a good partner for ANSYS to NASTRAN translations http:\\www.aerofem.ch FEMUTIL, Mr. Vollan there has writen an excellent translator between several FEM tools, but he is not active in the optics for the Zernike, nor directly for COMSOL.

I have already identified Sigmadyne, through their publications, one of my problems is that the mirror designs are not mine so I do not dear give them out, have enough problems to get them myself (I need them as I'm building the opto-mechanics and the precision actuation for AO thereunder), and my budget is too short to buy new software, furthermore, we have cancelled ANSYS and NASTRAN for COMSOL cannot keep them all, and my future i clearly in multiphysics for which COMSOL is currently by far the best

Really I would appreciate to have a better I/F for ANSYS and NASTRAN to and from COMSOL. The current import is too light, does a good job on the geometry imported, but nothing on material data import, and really nothing for beams and shells, a pity.

I have been pushing COMSOL to improve their structural, in the sens make it more engineering oriented, as I believe that is where they can get many new clients while taking them from the older tools, but until they get functions such as gravity by "click", superelements, list of distributed loads, RBE's, bulk to rigid load, mass participation factors for eigenmodes, simpler mode reductions ... scripts (as the formulas you need to write now are getting quite complex, very easy to do typos and cumbersome to debug) the effort demanded for "oldies" on classical FEM tools is too important and they will give up. COMSOL is still a superb tool for physicist, not yet as good for engineers, but that will come.

What are your feeling, how do you "simplify" your complex multibody FEMs ?

Ivar

Hi Ivar,
I was out of station for sometime. My apology for the delayed response. I could unzip the Shell_1.mph but it did not open in my COMSOL 3.4 version. If I am correct, there is no downward compatibility between version 3.4 and 3.5 of COMSOL. Would it be possible for you to upload the 3.4 version of the same file again? I guess your problem of M5 mirror of European-Extremely large Telescope is far more interesting, challenging and complex than the thermal analysis of a telescope mirror that I am wrestling at present!

As I had stated earlier, I am not yet greatly familiar with COMSOL or Zemax capabilities. But hope to fill these gaps soon in future. That is why I had started with a simple system of thermal analysis of the telescope mirror. (though it does not appear simple any more !). I do not quite know the strengths and weaknesses of each software, but I guess it will be easy for you to comment or clarify my conceptual doubts related to the problem.
a) Regarding the thermal deformation: What if we import the mirror geometry from Zemax to COMSOL and do the thermal and structural analysis? The deformed shape of the mirror can then be analyzed in Zemax to study the impact of thermal stress on the image quality. We may have to use Zemax in no-sequential mode to import external (3D) object –in this case, the mirror. Please advise if such an approach is practical or undesirable.

b) Zernike fitting of surface data: This question is related to fitting the Zernike polynomial to the deformed geometry of the mirror surface in COMSOL. How does one approach this problem? COMSOL computes x, y, z and total displacement data after the mirror is subjected to the thermal load. That data is available for subdomain, different boundaries and edges of the mirror geometry. For my analysis, I need to worry about the reflecting face of the mirror, which the COMSOL treats as one of boundaries. Which among x, y, z and total displacement data, do you think correctly represents the deformation of the mirror surface? Is it suffice to consider the surface deformation alone without taking into consideration the deformation of the bulk mirror?

Hi Ivar,
I was out of station for sometime. My apology for the delayed response. I could unzip the Shell_1.mph but it did not open in my COMSOL 3.4 version. If I am correct, there is no downward compatibility between version 3.4 and 3.5 of COMSOL. Would it be possible for you to upload the 3.4 version of the same file again? I guess your problem of M5 mirror of European-Extremely large Telescope is far more interesting, challenging and complex than the thermal analysis of a telescope mirror that I am wrestling at present!

As I had stated earlier, I am not yet greatly familiar with COMSOL or Zemax capabilities. But hope to fill these gaps soon in future. That is why I had started with a simple system of thermal analysis of the telescope mirror. (though it does not appear simple any more !). I do not quite know the strengths and weaknesses of each software, but I guess it will be easy for you to comment or clarify my conceptual doubts related to the problem.
a) Regarding the thermal deformation: What if we import the mirror geometry from Zemax to COMSOL and do the thermal and structural analysis? The deformed shape of the mirror can then be analyzed in Zemax to study the impact of thermal stress on the image quality. We may have to use Zemax in no-sequential mode to import external (3D) object –in this case, the mirror. Please advise if such an approach is practical or undesirable.

b) Zernike fitting of surface data: This question is related to fitting the Zernike polynomial to the deformed geometry of the mirror surface in COMSOL. How does one approach this problem? COMSOL computes x, y, z and total displacement data after the mirror is subjected to the thermal load. That data is available for subdomain, different boundaries and edges of the mirror geometry. For my analysis, I need to worry about the reflecting face of the mirror, which the COMSOL treats as one of boundaries. Which among x, y, z and total displacement data, do you think correctly represents the deformation of the mirror surface? Is it suffice to consider the surface deformation alone without taking into consideration the deformation of the bulk mirror?

Wish you a Merry Christmas and Happy new year.
Regards,
-ravi

Hi

Unfortunately, I do not have any V3.4 up running, so I cannot easily provide you with a version compatible for you just like that, but I have a few workarounds to propose:

a) if you are rather a "hacker" type you can always open the mph file in a text editor, you will find an xml part in there that, if you are used to read such structures it allows you to find the essential of the variables, not that easily the geometry, one cannot have everything.
This opens up another question (a side-thought) that starts to bother me: how to document and compare two models, this will be even more complex in V4, as I make easily 5-10 variants per hour of some of my models, then 1-2 months afterward, how to easily sort them by their pertinent contet, not just their file name ?

b) I have attached a few screen dumps as jpeg hereby that would be the quickest for you I believe.

2) When I site Zemax or other optical ray tracing softwares, it's because these have different ways to normalise the Zernike polynomials, so if you want to compare your results then you must understand this and correct wehn required, I do not have any direct access anylonger to any of these optical ray tracing tools, so I have neither no means to compare. Neither do I know how to import a geometry file into Zemax, but you can generate a deformed geometry in COMSOL as a meshed surface, and then with some format tweaking I'm sure its importable into Zemax or CodeV, that is how most people do, but this requires specific translation software, and not many document and distribute this, if even someone do, apart a few who are selling their services, that has the advantage that in principle these people know what their are doing.

3) The Zernike fitting, in my approach, is to consider first only full circular shaped area that you can renormalise to a radius of 1, and then use the deformation (COMSOL u,v,w) vectors, my example is even simpler, as my "mirror" is in the x-y plane and I'm mainly interested in w, but with the normal vectors n2x n2y n2z in COMSOL you can do more too (select Physics - Properies - Create Deformed Frame ON) solve your structural case and to visualise: "Plot Parameters - Arrow plot - Plot arrows as Boundaries , select Geoemtry and Mesh - deformed mesh, normal vector". Note: the end defomed points are (xi,yi,zi)+(ui,vi,wi) for all "i" nodes) so the deformation or strain is (u,v,w), and not (x,y,z). if you want the rms deformation of a surface/mirror in Z (assuming it is stable in the X-Y plane) is simply int(w^2,dx,dy)/Area over the X-Y boundary. So only the u,v,w for the deformed reflecting surface is of interest for the Zernike analysis.

Back to the Zernike in COMSOL, I use the optmisation toolbox, but it can be done, with some more formulas directly in plain COMSOL:
On the deformed surface I define a variabble "zer2" = the square of the difference of "w" and the Zernike polynomial, to be fitted to"0", in the scalar settings I define the variables = Zernike polynomials with names I have written out as a set of Scalar and Global Equations. My example file is just a thin shell cylinder, partly defomed by edge constraints and by gravity + pressure on the surface nothing fancy.

Hope this helps, try a google search for "Zernike" to get help for the notation a good refernce is take a look what is done by Pr. J.C Wyatt, University of Arizona and his people/students/ph-D s

Good luck
Ivar

Ivar, thanks a lot for your help,
What you suggested should be more than sufficient for me to get started with my analysis. I will seek your opinion again after making some progress. Once again thank you very much.
Regards,
-Ravi

PS: I really appreciate your strict adherence to the COMSOL notations and internal variables while replying to all types of queries in this discussion forum. That reflects your command on COMSOL and also makes you admirably precise and to the point -a quality worth emulating for new comers like myself :)

Hi there
Thanks for your kind words, I must admit that I apply what I have learned over time, I'm usually consulting on complex system pojects, and have noticed that >80% of the "problems" arriving are interface mismatch or mis-understandings, that is why I try to adhere to the wording of the tool, or science of the users, there are enough challenges anyhow, outside I/F mismatch ;)
And just now I'm taking a few days off from my work, so I have time to see what is going on here

Have a nice day
Ivar

Dear Ivar,
I hope you read still this old discussion... I found it in Forum some days ago and the topics was interesting for me, so that I search for "Zernike" to more information.

I try to complete the model according to the old report but I had some error in automatic sequence generation. Could you please say me what it is wrong in the file?. Thanks a lot!!!

You are very nice in all discussions!!!
Have a nice weekend.
Adriana.

Hi

indeed I get an error message about "z00_bc" a variable I cannot find in the model but still seems to be generated somewhere, could be a version issue with the newer V4 (by the way you still have the patch V4.0.0.993 to apply ;)

there is one typo (probably my fault from 3.5) the p0 should be written as (1/250)[atm] with the paranthesis around to get the correct units in "Pa2 and not "1/Pa".

Furthermore, in V4 "g_const" has been defined for 9.81[m/s^2] so no need to use the old tricj G0=1[lbf/lb] even if it's still correct. There are other physical constants that have appeared in V4 too, very handy.

So baiscally I need some more time o redo this in V4, but time is something I'm lacking, I'ma lways too busy, but will try to get a working V4 model running as I need it myself too ;)

Good luck
Ivar