Linear Buckling Analysis - buckling modes

Philipp Link
Philipp Link Altair Community Member
edited January 2021 in Community Q&A

Hello,

I am performing a Linear Buckling Analysis in Optistruct. On one end of my rod I have created my Constraint and on the other one my load, a force. As I change the magnitute of the force I am getting different results for the buckling modes. The eigenvalues change in such a way that I would get different critical forces and often not even the correct buckling modes. Could someone help me out?

My settings for the EIGRL card image are:

V1: 0.0

V2: 10.0

NORM: MAX

 

Thanks for your help!

Best Answer

  • Adriano Koga_20259
    Adriano Koga_20259 New Altair Community Member
    edited January 2021 Answer ✓

    Thank you, I just tried to compress the .hm file.

    For this model (F=4000N) I am getting 9 buckling modes. Changing the force to 4500N results in just 5 buckling modes. I dont get what I am doing wrong here.

    You have a EIGRL card which computes your eigenvalues. You have set it to calculate eigenvalues from 0.0 to 10.0 (lambda values). So OS will only output lambda values until 10.0. By changing your loads, lambda will change, thus there might be more or less eigenvalues in this range.

     

    There are some warnings over your eigenvalues in the .out file. Have you checked that?

    Not sure if this is because your FE joint which is too soft and allows this instability or anything else. But at least the relevant modes were kept from 4000 to 4500N and the Pcrit for them too: ~5300N and ~5450N for the relevant modes.

     Lanczos solver warnings for eigenvalue subcase 2.
     - At some point in the process, an inconsistent inertial count was found.
       This implies that some of the calculated frequencies may be inaccurate,
       especially the duplicated frequencies. To avoid this, specify both bounds
       and keep the bound away from the currently calculated frequencies.

         *** warning *** the rayleigh quotient
         x'kx / x' kd x for the      1-th eigenvector
         does not match the computed eigenvalue
              lambda (     1) :     6.05581675250293D-07
              x' k x          :     1.28328780495167D+00
              x' kd x         :     2.22371353195738D+06
              discrepancy     :     2.84894434905308D-08


         *** warning *** the rayleigh quotient
         x'kx / x' kd x for the      2-th eigenvector
         does not match the computed eigenvalue
              lambda (     2) :     7.40777973402242D-03
              x' k x          :     9.99999921470268D-01
              x' kd x         :     1.01852430119369D+00
              discrepancy     :     9.74404750706621D-01


         *** warning *** the rayleigh quotient
         x'kx / x' kd x for the      3-th eigenvector
         does not match the computed eigenvalue
              lambda (     3) :     1.32861005015557D+00
              x' k x          :     1.00000000002386D+00
              x' kd x         :     7.53592429999081D-01
              discrepancy     :     1.22896398312138D-03


         *** warning *** the rayleigh quotient
         x'kx / x' kd x for the      4-th eigenvector
         does not match the computed eigenvalue
              lambda (     4) :     1.36436666199060D+00
              x' k x          :     9.99999999808459D-01
              x' kd x         :     7.36645019653350D-01
              discrepancy     :     5.02849319336677D-03


         *** warning *** the rayleigh quotient
         x'kx / x' kd x for the      5-th eigenvector
         does not match the computed eigenvalue
              lambda (     5) :     1.59399368792098D+00
              x' k x          :     9.99999999424107D-01
              x' kd x         :     6.00406992646006D-02
              discrepancy     :     9.04295504298749D-01


         *** warning *** the rayleigh quotient
         x'kx / x' kd x for the      6-th eigenvector
         does not match the computed eigenvalue
              lambda (     6) :     5.42775118564185D+00
              x' k x          :     9.99999999943861D-01
              x' kd x         :     1.84051865055608D-01
              discrepancy     :     1.01227116874495D-03


         *** warning *** the rayleigh quotient
         x'kx / x' kd x for the      7-th eigenvector
         does not match the computed eigenvalue
              lambda (     7) :     5.87325675572367D+00
              x' k x          :     9.99999999994693D-01
              x' kd x         :     1.69363086784738D-01
              discrepancy     :     5.28710636604187D-03

     

Answers

  • Adriano Koga_20259
    Adriano Koga_20259 New Altair Community Member
    edited January 2021

    I'm sorry. It is not so clear to me what issue you're getting.

     

    The linear buckling analysis will give you the lambda eigenvalue which is Lambda = P_crit/P_applied.

    as we get values above 1.0, it means that your applied load is less than the critical.

     

    If you change the magnitude of your applid load, the eigenvalue will change as well, isn't? The critical load P_crit is the same.

     

    Now if you're saying that you got different mode shapes, this would be strange.

    You might get different mode shapes depending on your boundary conditions. Have you defined any BC in the load end? or is it totally free? If it is totally free, it is expected that some instabilities in other directions might occur, but probably very close to each other (for a sym section).

  • Philipp Link
    Philipp Link Altair Community Member
    edited January 2021

    Yes, the lambda eigenvalues change, which is correct, but a slight change in the magnitute of the force also results in a changed P_crit. Furhtermore the mode I am looking for, is missing sometimes. For example with a force of 4000N everything seems correct and fine, but as I change the force to 4500N, I am getting different buckling modes.

     

    The file is larger than the 5MB limit here, is there another possibility to share it with you?

     

    Thanks

  • Adriano Koga_20259
    Adriano Koga_20259 New Altair Community Member
    edited January 2021

    Yes, the lambda eigenvalues change, which is correct, but a slight change in the magnitute of the force also results in a changed P_crit. Furhtermore the mode I am looking for, is missing sometimes. For example with a force of 4000N everything seems correct and fine, but as I change the force to 4500N, I am getting different buckling modes.

     

    The file is larger than the 5MB limit here, is there another possibility to share it with you?

     

    Thanks

    usually, compressing the .fem file gives you a smaller file But algo, including some images might be helpful.

  • Philipp Link
    Philipp Link Altair Community Member
    edited January 2021

    Thank you, I just tried to compress the .hm file.

    For this model (F=4000N) I am getting 9 buckling modes. Changing the force to 4500N results in just 5 buckling modes. I dont get what I am doing wrong here.

  • Adriano Koga_20259
    Adriano Koga_20259 New Altair Community Member
    edited January 2021 Answer ✓

    Thank you, I just tried to compress the .hm file.

    For this model (F=4000N) I am getting 9 buckling modes. Changing the force to 4500N results in just 5 buckling modes. I dont get what I am doing wrong here.

    You have a EIGRL card which computes your eigenvalues. You have set it to calculate eigenvalues from 0.0 to 10.0 (lambda values). So OS will only output lambda values until 10.0. By changing your loads, lambda will change, thus there might be more or less eigenvalues in this range.

     

    There are some warnings over your eigenvalues in the .out file. Have you checked that?

    Not sure if this is because your FE joint which is too soft and allows this instability or anything else. But at least the relevant modes were kept from 4000 to 4500N and the Pcrit for them too: ~5300N and ~5450N for the relevant modes.

     Lanczos solver warnings for eigenvalue subcase 2.
     - At some point in the process, an inconsistent inertial count was found.
       This implies that some of the calculated frequencies may be inaccurate,
       especially the duplicated frequencies. To avoid this, specify both bounds
       and keep the bound away from the currently calculated frequencies.

         *** warning *** the rayleigh quotient
         x'kx / x' kd x for the      1-th eigenvector
         does not match the computed eigenvalue
              lambda (     1) :     6.05581675250293D-07
              x' k x          :     1.28328780495167D+00
              x' kd x         :     2.22371353195738D+06
              discrepancy     :     2.84894434905308D-08


         *** warning *** the rayleigh quotient
         x'kx / x' kd x for the      2-th eigenvector
         does not match the computed eigenvalue
              lambda (     2) :     7.40777973402242D-03
              x' k x          :     9.99999921470268D-01
              x' kd x         :     1.01852430119369D+00
              discrepancy     :     9.74404750706621D-01


         *** warning *** the rayleigh quotient
         x'kx / x' kd x for the      3-th eigenvector
         does not match the computed eigenvalue
              lambda (     3) :     1.32861005015557D+00
              x' k x          :     1.00000000002386D+00
              x' kd x         :     7.53592429999081D-01
              discrepancy     :     1.22896398312138D-03


         *** warning *** the rayleigh quotient
         x'kx / x' kd x for the      4-th eigenvector
         does not match the computed eigenvalue
              lambda (     4) :     1.36436666199060D+00
              x' k x          :     9.99999999808459D-01
              x' kd x         :     7.36645019653350D-01
              discrepancy     :     5.02849319336677D-03


         *** warning *** the rayleigh quotient
         x'kx / x' kd x for the      5-th eigenvector
         does not match the computed eigenvalue
              lambda (     5) :     1.59399368792098D+00
              x' k x          :     9.99999999424107D-01
              x' kd x         :     6.00406992646006D-02
              discrepancy     :     9.04295504298749D-01


         *** warning *** the rayleigh quotient
         x'kx / x' kd x for the      6-th eigenvector
         does not match the computed eigenvalue
              lambda (     6) :     5.42775118564185D+00
              x' k x          :     9.99999999943861D-01
              x' kd x         :     1.84051865055608D-01
              discrepancy     :     1.01227116874495D-03


         *** warning *** the rayleigh quotient
         x'kx / x' kd x for the      7-th eigenvector
         does not match the computed eigenvalue
              lambda (     7) :     5.87325675572367D+00
              x' k x          :     9.99999999994693D-01
              x' kd x         :     1.69363086784738D-01
              discrepancy     :     5.28710636604187D-03

     

  • Philipp Link
    Philipp Link Altair Community Member
    edited January 2021

    You have a EIGRL card which computes your eigenvalues. You have set it to calculate eigenvalues from 0.0 to 10.0 (lambda values). So OS will only output lambda values until 10.0. By changing your loads, lambda will change, thus there might be more or less eigenvalues in this range.

     

    There are some warnings over your eigenvalues in the .out file. Have you checked that?

    Not sure if this is because your FE joint which is too soft and allows this instability or anything else. But at least the relevant modes were kept from 4000 to 4500N and the Pcrit for them too: ~5300N and ~5450N for the relevant modes.

     Lanczos solver warnings for eigenvalue subcase 2.
     - At some point in the process, an inconsistent inertial count was found.
       This implies that some of the calculated frequencies may be inaccurate,
       especially the duplicated frequencies. To avoid this, specify both bounds
       and keep the bound away from the currently calculated frequencies.

         *** warning *** the rayleigh quotient
         x'kx / x' kd x for the      1-th eigenvector
         does not match the computed eigenvalue
              lambda (     1) :     6.05581675250293D-07
              x' k x          :     1.28328780495167D+00
              x' kd x         :     2.22371353195738D+06
              discrepancy     :     2.84894434905308D-08


         *** warning *** the rayleigh quotient
         x'kx / x' kd x for the      2-th eigenvector
         does not match the computed eigenvalue
              lambda (     2) :     7.40777973402242D-03
              x' k x          :     9.99999921470268D-01
              x' kd x         :     1.01852430119369D+00
              discrepancy     :     9.74404750706621D-01


         *** warning *** the rayleigh quotient
         x'kx / x' kd x for the      3-th eigenvector
         does not match the computed eigenvalue
              lambda (     3) :     1.32861005015557D+00
              x' k x          :     1.00000000002386D+00
              x' kd x         :     7.53592429999081D-01
              discrepancy     :     1.22896398312138D-03


         *** warning *** the rayleigh quotient
         x'kx / x' kd x for the      4-th eigenvector
         does not match the computed eigenvalue
              lambda (     4) :     1.36436666199060D+00
              x' k x          :     9.99999999808459D-01
              x' kd x         :     7.36645019653350D-01
              discrepancy     :     5.02849319336677D-03


         *** warning *** the rayleigh quotient
         x'kx / x' kd x for the      5-th eigenvector
         does not match the computed eigenvalue
              lambda (     5) :     1.59399368792098D+00
              x' k x          :     9.99999999424107D-01
              x' kd x         :     6.00406992646006D-02
              discrepancy     :     9.04295504298749D-01


         *** warning *** the rayleigh quotient
         x'kx / x' kd x for the      6-th eigenvector
         does not match the computed eigenvalue
              lambda (     6) :     5.42775118564185D+00
              x' k x          :     9.99999999943861D-01
              x' kd x         :     1.84051865055608D-01
              discrepancy     :     1.01227116874495D-03


         *** warning *** the rayleigh quotient
         x'kx / x' kd x for the      7-th eigenvector
         does not match the computed eigenvalue
              lambda (     7) :     5.87325675572367D+00
              x' k x          :     9.99999999994693D-01
              x' kd x         :     1.69363086784738D-01
              discrepancy     :     5.28710636604187D-03

     

    Thanks a lot! That really helped me!

    I will have a look at the bounds.



    What do you mean with the FE joint is to soft?

  • Adriano Koga_20259
    Adriano Koga_20259 New Altair Community Member
    edited January 2021

    Thanks a lot! That really helped me!

    I will have a look at the bounds.



    What do you mean with the FE joint is to soft?

    i saw that you have some joint with some flexibility in your model. This might be "causing" the lower eigenvalues.

  • Philipp Link
    Philipp Link Altair Community Member
    edited January 2021

    i saw that you have some joint with some flexibility in your model. This might be "causing" the lower eigenvalues.

    Will have a look at that!

    Thanks again.