Learning Forum

[Severus_Zhang]

I perform the transient analysis of a structure in ANSYS by Newmark method. At the same time, conduct the Newmark integration method in matlab based on the same structure. And then, export the acceleration time history of the same degree of freedom from ansys and matlab. I find that the figure of acceleration time history from matlab can't match with the figure of acceleration time history from ansys. And, the value at the peak of matlab is larger than the value of ansys. Then, the average value of the acceleration values at two adjacent moments of matlab is taken as the acceleration value at that moment to get a new acceleration time history, which is matched well with the results of ansys. So, why? why the acceleration time history out put from ansys is averaged? Can you give me some explanations and tell me how to solve this problem? Thanks.

[Erik Kostson]

Hi
I would suggest that you go through the theory on transient analysis (NMK, or HHT) in our theory reference chapter 15.2 in the APDL theory reference.

All the best

Erik

[Severus_Zhang]

Thanks for your reply. I have gone through the theory in chapter 15.2. But it doesn't give the explanation of "the average acceleration". I want to ensure whether the acceleration time history is the average value between the adjacent moments or not. And if so, why you give the average value instead of the initial transient value? Thanks again.

[Erik Kostson]

Hi

Please read that section more carefully - one will see that the average acceleration for NMK method is explained there.
(one can search for average in the doc :))
All the best and good luck

Erik

[Severus_Zhang]

Thanks. But I think you don't understand what I mean. Generally, the results generated from ansys should matched well with the results from matlab(accurate results), or not much different, as both of the two software use the same integration method. But the truth is that the displacement and the velocity are matched well with each other by ansys and matlab. However, the acceleration from the two software can not match with each other, like the following figure.

But when I take the mean value of the resluts of matlab, which means that take mean value of the initial first value and the initial second value as the new first value, take mean value of the the initial second value and the initial third value as the new second value, and so on. And I generated a new acceleration time history which is used to compared with the results from ansys, as shown in the following figure.
It can be observed from the figure that the new acceleration time history is matched well with the acceleration time history generated from ansys. So the conclusion is that the acceleration time history generated from ansys is not the initial transient value, but the mean value like I introduced before. So why the ansys give the mean value instead of the initial transient value? Thanks.

[Erik Kostson]

Hi

If you look in the manual like I told you will find that:

'The acceleration of a typical DOF is given by Equation 15ÔÇô12 for time . By default, the acceleration vectoris the average acceleration between timeand time , since the Newmark assumptions (Equation 15ÔÇô8 and Equation 15ÔÇô9) assume the average acceleration represents the true acceleration.'

So everything is explained in the theory manual, and beyond that we can not provide any specifics - everything publicly available is in the manual.

Thank you

Erik

[Severus_Zhang]

Thanks. Actually, the Newmark assumptions assume the variation tendency of acceleration between time to time is the average acceleration. That doesn't mean the average acceleration represents the true acceleration. As you can see from the figure I provide to you, the average acceleration is less than the true acceleration. If the engineers design the structure with the average acceleration, it's not conservative. That's my suggestion. Thanks for your reply. Best wishes.

[yevgeniya.filanova]

Hello! I think the correct answer would be that ANSYS uses smoothing for the acceleration, since the Newmark method often provides noisy results for the acceleration. However, smoothing can be turned off via setting AVSMOOTH option to 1 in the TINTP command. 

P.S. The "average acceleration assumption" is not important for the question.

[Greeshma S Vijayan]

i would like to know how you exported the acceleration time history to the matlab. Did you write a separate code? .

[Erik Kostson]

Hi

This post is very old and not related to the origial question - please open up a new one.

Closing here.

Erik