Welcome to the forums. Please post in English or French.

You are not logged in.

#1 2015-07-21 16:18:34

Johannes_ACKVA
Member
From: Ingenieurbüro für Mechanik, DE
Registered: 2009-11-04
Posts: 738
Website

CALC_FATIGUE - with mean stress

hello

With POST_FATIGUE mean stress can be considered in a fatigue analysis for just one point.

CALC_FATIGUE has the advantage that it calculates not only for one point, but for the complete model. This is good for making a coloured plot of the whole model.

But the option  CORR_SIGM_MOYE ='GERBER' or 'GOODMAN' to take  into account the mean stress is included in CALC_FATIGUE only for dynamic fatigue calculation (FATIG_VIBR). I miss it for the much simpler uni-axial static fatigue calculation with CUMUL_DOMAGE

Questions:
(1) why does it not exist for CUMUL_DOMAGE? I suppose it should be very easy to include, and it is very essential to use it
(2) can anybody propose a workaround?


Regards
Johannes_ACKVA

______________________________________________________________________
CODE-ASTER-courses at Ingenieurbüro für Mechanik, Germany

*** CODE-ASTER INTRO + CONTACT + MATERIAL
    4 days, Mon 05 to Thu 08 of Oct 2015

*** CODE-ASTER  DYNAMIC ANALYSIS
    2 days, Mon 09 to Tue 10 of Nov 2015


Ingenieurbüro für Mechanik
D 91717 Wassertrüdingen / Germany

www.code-aster.de                                                Training & Support for NASTRAN and CODE-ASTER

Offline

#2 2015-07-31 21:28:52

fsshaheen
Member
Registered: 2015-06-04
Posts: 58

Re: CALC_FATIGUE - with mean stress

hello,
I am facing same problem as you.

but in reference document R7.04.01 under the heading 2.3.3  INFLUENCE AVERAGE CONSTRAINT there is written

The influence of the average constraint is taken into account only on request of the user (key word
CORR_HAIG).

but i do not know where to use it in command file.please check it out whether it is relevant or not.

Please let me know if you find something.

Last edited by fsshaheen (2015-07-31 21:36:00)

Offline

#3 2016-11-08 10:45:12

Ect
Member
From: Switzerland
Registered: 2008-05-15
Posts: 103

Re: CALC_FATIGUE - with mean stress

Hello,

Sorry to dig out this old topic but the subject is still relevant.

To get the mean stress correction it's possible to use POST_FATIGUE on each mesh node, but it's utterly slow. For my case the computation elapsed time is 900 sec, the postprocessing is 11400 sec, that's a show stopper.

What's funny when looking at the CALC_FATIGUE code in bibfor/prepost/fgdomg.F90 is that it looks like everything is almost already done, only the call to fgcorr is missing. So ripping the code from POST_FATIGUE code in bibfor/prepost/pofaun.F90 and transferring it to fgdomg.F90 is very straightforward. Then adding the CORR_SIGM_MOYE option to CALC_FATIGUE with CUMUL_DOMAGE enables its usage in the comm file.

You'll find the two files in the attached archive. You'll need to replace them in the 12.6.0 source:
catapy/commande/calc_fatigue.capy
bibfor/prepost/fgdomg.F90
and compile aster.

I've made a simple test case (also attached) to compare POST_FATIGUE and CALC_FATIGUE and here is the output for the damage:

No correction POST: 2.275e-06
No correction CALC: 1.806e-06

Goodman POST: 2.480e-06
Goodman CALC: 2.165e-06

Gerber POST: 2.324e-06
Gerber CALC: 1.891e-6

There are some differences between the two operators, I guess partly because of the use of SIEQ_NOEU for POST_FATIGUE and SIEQ_ELNO for CALC_FATIGUE. But it could also come from the implementation, that's where the code would need to be reviewed by competent developers.

Ah and CALC_FATIGUE with the correction takes around 70 sec for my case, so there's a real benefit to have it implemented.

Thanks for your feedback.
Ect

Last edited by Ect (2016-11-08 10:49:32)


Attachments:
fatigue.tar.gz, Size: 901.73 KiB, Downloads: 475

Offline

#4 2016-11-09 13:43:56

Johannes_ACKVA
Member
From: Ingenieurbüro für Mechanik, DE
Registered: 2009-11-04
Posts: 738
Website

Re: CALC_FATIGUE - with mean stress

Dear Ect,

thank you for your contribution, that's really interesting for me. Sadly I have currently no time to look deeper into it.

There are some differences between the two operators, ..

This cries for the reason to be understood

.., I guess partly because of the use of SIEQ_NOEU for POST_FATIGUE and SIEQ_ELNO for CALC_FATIGUE.

This could easily be checked with a simpler mesh: a column of HEXA8 elements under uniaxial loading. Also refining extremly a mesh should make converge the SIEQ_NOEU and SIEQ_ELNO values.

But it could also come from the implementation, that's where the code would need to be reviewed by competent developers.

ok, this applies if SIEQ_NOEU and SIEQ_ELNO cannot be converged and if you code has no bug.

Could you propose your code in the Code-Aster bitbucket for implementation in the EDF-version?

Regards Johannes_ACKVA

__________________________________________________________________
next CODE-ASTER-courses at Ingenieurbüro für Mechanik, Germany

*** CODE-ASTER INTRO + CONTACT + MATERIAL
     16-20 January 2017

*** CODE-ASTER  DYNAMIC ANALYSIS
     16-17  February  2017


Ingenieurbüro für Mechanik
D 91717 Wassertrüdingen / Germany

www.code-aster.de                                                Training & Support for NASTRAN and CODE-ASTER

Offline

#5 2016-11-09 16:40:47

Ect
Member
From: Switzerland
Registered: 2008-05-15
Posts: 103

Re: CALC_FATIGUE - with mean stress

Dear Johannes,

I've tried to refine the mesh as you suggested and got very close values for SIEQ both on ELNO or NOEU, but still different results for POST_FATIGUE and CALC_FATIGUE damage. And this with the original Aster 12.6, without CORR_SIGM_MOYE.

Looking deeper into it it happens that the number of cycles considered by the two operators are different:

POST_FATIGUE finds 7 peaks and 4 cycles:

 VALEURS DE LA FONCTION CHARGEMENT:
        0.100000E+01           0.509665E+03
        0.200000E+01           0.560631E+03
        0.300000E+01           0.458698E+03
        0.400000E+01           0.560631E+03
        0.500000E+01           0.458698E+03
        0.600000E+01           0.509665E+03
        0.700000E+01           0.000000E+00

 PICS EXTRAITS DE LA FONCTION CHARGEMENT
 APRES AVOIR PRIS EN COMPTE DE KT
 SEUIL =       0.000000E+00          NB DE PICS =     7

CYCLE        VALE_MIN     VALE_MAX     DOMMAGE      DOMM_CUMU   
           1  0.00000E+00  5.60631E+02  6.94749E-07 -           
           2  4.58698E+02  5.60631E+02  5.26852E-07 -           
           3  4.58698E+02  5.09665E+02  5.13075E-07 -           
           4  3.64046E+02  5.09665E+02  5.39263E-07 -           
-            -            -            -             2.27394E-06

CALC_FATIGUE also finds 7 peaks, but 3 cycles (disregard the different stress values):

 PICS APRES LE COMPTAGE RAINFLOW
 NOMBRE DE CYCLES =                     3
      CHARGEMENT_MAX     CHARGEMENT_MIN
       0.618392E+03       0.505957E+03
       0.562175E+03       0.505957E+03
       0.618392E+03       0.000000E+00
   nombre de valeurs        =  7
     562.174982, 618.392480, ...
   nombre de pics extraits   =  7
     618.392480, 505.957484, ...
   nombre de cycles détectés =  3
   1  /  618.392480   505.957484
   2  /  562.174982   505.957484
   3  /  618.392480   0.000000
   dommage en ce point/composante  =  0.000002

pofaun.F90 (POST_FATIGUE) makes a call to fgpeak to extract the stress peak values from the history function, whereas fgvdmg.F90 (CALC_FATIGUE) uses a very different piece of code that I don't fully understand yet.
This is actually not the source of the difference since both operators find the same number of peaks, so it should reside in the rainflow counting. But both operators call the same fgpic2 and fgrain routines... need to investigate further.

Regarding pushing the code to bitbucket this is a procedure I'm not familiar with, but I'll give it a try if it's worthwhile.

Regards,
Ect

Last edited by Ect (2016-11-09 17:35:16)

Offline

#6 2016-11-09 22:57:38

Ect
Member
From: Switzerland
Registered: 2008-05-15
Posts: 103

Re: CALC_FATIGUE - with mean stress

Ok the differences were coming from the COMPTAGE='RCCM' in the POST_FATIGUE operator, giving a different peak order... the issue alway lies between the chair and the keyboard.

With the RAINFLOW option the damage values are the following:

No correction POST:  1.73714E-06
No correction CALC: 1.73715E-06

Goodman POST: 1.90966E-06
Goodman CALC: 1.90970E-06

Gerber POST: 1.77618E-06
Gerber CALC: 1.77617E-06

It actually looks like it's working !

Regards,
Ect

Offline

#7 2016-11-10 11:45:36

Johannes_ACKVA
Member
From: Ingenieurbüro für Mechanik, DE
Registered: 2009-11-04
Posts: 738
Website

Re: CALC_FATIGUE - with mean stress

congratulations!
Again, can you propose it in the bitbucket:
thank you!


Regards Johannes_ACKVA

__________________________________________________________________
next CODE-ASTER-courses at Ingenieurbüro für Mechanik, Germany

*** CODE-ASTER INTRO + CONTACT + MATERIAL
     16-20 January 2017

*** CODE-ASTER  DYNAMIC ANALYSIS
     16-17  February  2017


Ingenieurbüro für Mechanik
D 91717 Wassertrüdingen / Germany

www.code-aster.de                                                Training & Support for NASTRAN and CODE-ASTER

Offline

#8 2019-01-14 07:07:45

Ceduardolino
Member
Registered: 2017-03-19
Posts: 20

Re: CALC_FATIGUE - with mean stress

Good morning, sorry to reopen the topic.
I would like to know how the CALC_FATIGUE damage was printed in Paravis. As in the attached .png image. I am using the codes as an example and I could not perform this postprocessing.

Thank you!!!

Offline

#9 2021-04-12 13:49:30

sameer21101970
Banned
Registered: 2019-09-06
Posts: 354

Re: CALC_FATIGUE - with mean stress

though the post is old, but very much important.

the changes as suggested in CA14.6 programme is not yet made.

Though i would like to know,  the below ffigures shows damage in   some unit meter or mm for particular selected node.
No correction POST: 2.275e-06
No correction CALC: 1.806e-06

Goodman POST: 2.480e-06
Goodman CALC: 2.165e-06

Gerber POST: 2.324e-06
Gerber CALC: 1.891e-6

I am intrested to find Gerber Factor of Safety & Goodman Factor of Safety Values. IS it possible with CA14.6 . Can anybody guide.

Last edited by sameer21101970 (2021-04-13 08:32:07)

Offline

#10 2021-04-13 14:26:57

miib
Member
Registered: 2021-03-23
Posts: 102

Re: CALC_FATIGUE - with mean stress

where do you found the sigma (epsilon)'s values of HISTORIE ? :-)

Last edited by miib (2021-04-13 14:29:02)

Offline

#11 2021-04-13 14:31:11

sameer21101970
Banned
Registered: 2019-09-06
Posts: 354

Re: CALC_FATIGUE - with mean stress

download tar.gz as given above, it has steps giiven

Offline

#12 2021-04-14 11:45:19

miib
Member
Registered: 2021-03-23
Posts: 102

Re: CALC_FATIGUE - with mean stress

ha!!!

often i have the problem which the function generate by RECU_FONCTION isn't monotonous and the software give me back fatal error.

I bypassed this problem which this solution :

1) doing a first study without fatigue computation only for  obtain VMIS_SG sigma's values. This is possible by paravis annotating the simga values time by time.

2) doing a second study with fatigue computation and introducing a function  by points (time;VMIS_SG) obtained by first study and introducing this funcion in HISTORIE.

is it right ???

Also, i observed which introducing the values of Kt the three results ( domage by wholer, goodman, gerber) is often different.

:-)

Offline