Fatigue analysis correlation

Dirk76
Dirk76 Altair Community Member
edited February 2021 in Community Q&A

Dear all,

 

I'm facing a problem regarding a fatigue analysis I'm doing in hyperlife. My goal is to correlate a simple model with hand calculations and FE model. I have a HEA500 solid beam modeled in Hypermesh with a length of 5 m. One end is fully constrained in all dof's, the other end has a force of 1 kN applied with RBE2 element. I performed a static analysis with Optistruct in order to have the static subcase for my fatigue case. I chose the max stressed element from the static analysis to evaluate damage and life (you can see it in the attached picture). The load history selected to perform fatigue analysis is the one you can find in the fatigue e-book, where the rainflow counting method algorithm is explained (I'll attach as well the .txt file in case). Now, I converted the force value from the load history into stress history of the above selected element (let's say, I applied 200 kN and the resulting signed von mises on the element was 336.4 MPa and so on for the remaining values). Performing a rainflow counting with hand calculations and obtaining the cycles I could create the rainflow matrix with amplitude and mean values of stress. From the matrix it was immediate to obtain life (linked to that particular amplitude and mean) and, finally, damage. Now the problem is that damage and life from FE model is different from the one obtained with hand calculations, especially the values of mean and amplitude on the selected element. To perform the analysis I used a stress-life approach (Uniaxial, MPa, Signed von, 0.5 for survival, Goodman mean stress correction, top layer selection), a Steel 1045,Annealed,BHN=225(932) as material and in the load map section I used the mentioned curve with a scale of 100, in order to visualize damage on my beam (the static subcase has a force of 1 kN). I was wondering if, maybe, hyperlife calculates amplitude and mean in a different way than theory (amplitude = abs |max-min/2|; mean = (max+min)/2). Can someone please give me an insight about this problem? I'll attach the .fem,.h3d and the .csv files you need in order to compare the two solutions. Thanks in advance for any response.

 

p.s. I'll give you the link of we transfer to download the .fem and .h3d file since they were too big to be attached here.

 

https://we.tl/t-aPUMno26zh

Tagged:

Answers

  • Joshua Pennington
    Joshua Pennington New Altair Community Member
    edited January 2021

    Dirk,

    Apologies for the long delayed response. Maybe you've resolved the issue by now. Here are my initial thoughts:

    1) In order to compare OptiStruct results with HyperLife you need to be sure to check the MATFAT option on the MAT1 card to Amplitude based. By default it is set to R, which is range based.

    image

     

    2) To calculate for the most damage you should use Abs Max Principle stress and not Von Mises approach. 

    3) Make sure to use the "Worst" layer

    4) All of the formulations HyperLife uses are well documented. I would recommend accessing it via the File drop down menu within HyperLife.

    For file sharing purposes would you please use my FTP: https://ftam1.altair.com/filedrop/jpennington@altair.com

    Regards

    JP

  • Dirk76
    Dirk76 Altair Community Member
    edited February 2021
    Joshua Pennington said:

    Dirk,

    Apologies for the long delayed response. Maybe you've resolved the issue by now. Here are my initial thoughts:

    1) In order to compare OptiStruct results with HyperLife you need to be sure to check the MATFAT option on the MAT1 card to Amplitude based. By default it is set to R, which is range based.

    image

     

    2) To calculate for the most damage you should use Abs Max Principle stress and not Von Mises approach. 

    3) Make sure to use the "Worst" layer

    4) All of the formulations HyperLife uses are well documented. I would recommend accessing it via the File drop down menu within HyperLife.

    For file sharing purposes would you please use my FTP: https://ftam1.altair.com/filedrop/jpennington@altair.com

    Regards

    JP

    Hi Joshua,

     

    thanks for the answer. Meanwhile I could find the solution of the problem. Indeed instead of using hyperlife, I did the same fatigue analysis on Hypermesh UI, obtaining the fatrf.3 file to check cycles and damage of each element. What I found out was that the damage calculated by Hypermesh on that particular element was wrong! The value, indeed, of n/N was different from the one with hand calculations. This was surprising since the procedure was correct but I don't know why on that particular element, Hypermesh calculated a wrong n/N. Maybe it's a bug ahaha. However thanks for your tips, I'll take them in mind for my next analysis. Have a nice day :)

Categories