Altair RISE

A program to recognize and reward our most engaged community members

Nominate Yourself Now!

PSD analysis

Dp12

Hi,

I am currently performing random vibrartion analysis, where I have sweep up PSD data for muffler.

I have some query mention below:

1)From testing, I got PSD data (G^2/Hz vs Freq) . Do I have to change it into { (mm/sec2)^2 vs freq}.

2)when I am giving the input in (G^2/Hz vs Freq) and FRF excitation as 1 , then PSD stress is very very less e-7.

3)Plz suggest the appropriate methodology step by step for providing - input (in which form) and post-processing method to interpret results (which stress is useful).

Thanks in Advance

Find more posts tagged with

English

OptiStruct

Accepted answers

All comments

Rahul Rajan_21763

Please refer to pdf file shared in below post.

https://community.altair.com/community?id=community_question&sys_id=66864c3a1b2bd0908017dc61ec4bcbf5

Dp12

Thanks for the information.

At the time of postprocessing, which output is more useful to determine the life cycle. PSD or RMS??

Rahul Rajan_21763

Since all the results are complex you cant use standard calculation for vonmises stress because it is phase dependent. So at any loading frequency at any element you have 360 degree sweep of the vonmises stress. So you have to calculate the maximum value of that 360 degree sweep and there is method called segalman method to do that in HyperView.

Set the global output request 'STRESS(H3D,BOTH,CORNER,RMS) = YES'

Venkatesh Balulad_21941

Hi Rahul,

Can you please explain me how to get the von-mises stress using segalman method.

Thanks

Venkatesh

Rahul Rajan_21763

Define PSDM output request for STRESS

Venkatesh Balulad_21941

Hi Rahul,

I am simulating random vibration in optistruct. I have input psd in (m/s^2)/Hz, model content in mm, output required is stress in MPa. I submitted 1 run and i am getting high stress(RMS stress over frequencies) which are not realistic. Unit excitation is 9810 (SPCD) and I have submitted another run with unit excitation 1 (SPCD), here i am getting very low stress which is not realistic. Can you suggest me the right procedure. I guess i have mistake in unit conversion. If anyone having dummy procedure file please share me.

and one more my stresses should not be higher than endurance limit (criteria). Stresses whatever i am checking is the right one to compare with endurance limit ??

Note: Earlier i used to do in abaqus. recently we purchased altair software and trying to simulate in optistruct. Please help me. Your suggestion is very valuable to us.

Please call me if possible. here is my number 9019717355

Thanks you in advance

Aux_Converter_RV_4.fem

Rahul Rajan_21763

Please refer to the best practices for random response analysis from Resource Library.

Random Vibration Best Practices.pdf

Rex_22369

On ‎10‎/‎3‎/‎2019 at 5:03 PM, Rahul R said:

Please refer to the best practices for random response analysis from Resource Library.

Hi Rahul,

Where I can find all these interesting presentations you attached above?

I am really interested with these!

Let me know, thanks!

Rahul Rajan_21763

You can refer to below knowledge articles with keyword as random.

https://community.altair.com/community?id=kb_search&spa=1&query=random%20response

Rex_22369

15 hours ago, Rahul R said:

You can refer to below knowledge articles with keyword as random.

https://community.altair.com/community?id=kb_search&spa=1&query=random%20response

oops, it's really difficult to download the presentations, I mean the download speed, is there a way to download all the Optistruct presentations in a more efficiency way?

Tommy_21858

Hello everyone,

now I am facing with the problem of setting up an random response analysis. Therefore I have some questions…

1. The input data from testing is in acceleration [mm/s²] vs. time and my fe-model is in units of [mm], [tonne]. I have created in HyperGraph a curve for my frequency resonse analyse by using FFT --> Hamming. But how can I create the PSD-signal for my RRA in HyperGraph?

2. Further I want to achieve my results for stress in [MPa] and displacements in [mm]. So the unit of my PSD should be in [(mm/s²)²/Hz], I am right?

3. I have read that the TABRND in my RRA must have the LOG - LOG. Can someone please explain this?

Best regards Tommy

Rex_22369

Hello everyone,

now I am facing with the problem of setting up an random response analysis. Therefore I have some questions…

1. The input data from testing is in acceleration [mm/s²] vs. time and my fe-model is in units of [mm], [tonne]. I have created in HyperGraph a curve for my frequency resonse analyse by using FFT --> Hamming. But how can I create the PSD-signal for my RRA in HyperGraph?

2. Further I want to achieve my results for stress in [MPa] and displacements in [mm]. So the unit of my PSD should be in [(mm/s²)²/Hz], I am right?

3. I have read that the TABRND in my RRA must have the LOG - LOG. Can someone please explain this?

Best regards Tommy

@Tommy

Hello, I try to answer you based on my knowledge, if I am wrong, please correct me!

1. I did not use the HyperGraph to do this, so I do not know, but I know there is a free software can do this convert, usually I will use Scilab or Smath to do this work, but you can use the vibrationdata made by Tom Irvine

2. Usually, the PSD used in the vibration test is log-log format, then we should use the same in the simulation, the log-log makes the PSD looks like linear, but they are not a straight line in the real, if you use linear-linear, the vibration energy will be reduced.

Hope it helps, keep in touch!

Hung Tran

Hello everyone,

now I am facing with the problem of setting up an random response analysis. Therefore I have some questions…

1. The input data from testing is in acceleration [mm/s²] vs. time and my fe-model is in units of [mm], [tonne]. I have created in HyperGraph a curve for my frequency resonse analyse by using FFT --> Hamming. But how can I create the PSD-signal for my RRA in HyperGraph?

2. Further I want to achieve my results for stress in [MPa] and displacements in [mm]. So the unit of my PSD should be in [(mm/s²)²/Hz], I am right?

3. I have read that the TABRND in my RRA must have the LOG - LOG. Can someone please explain this?

Best regards Tommy

Hi Tommy,

For question 2 : I think that you should set PSD in G^2/Hz. Then your output will be MPa and mm.

I'm interested in Random vibration as well.

The altair documents is not too detail, it make me confuse.

Can anyone talk with me, which stress is to calculate fatigue life, PSD or RMS stress ?

Thanks and best regards

Adriano_Koga

For fatigue you usually compare RMS stress value for assessment.

RMS value is the 1-sigma stress value, occuring ~68% of your test/simulation time. 2-sigma is 2*rms value, combining 95%..and 3-sigma is 3*rms summing up ~98%.

You can use miner's rule for computing fatigue.

If your model has highest participation factors of 10Hz and your structure has to withstand 10000s, you can say that you will have a total of 10*10000 cycles, from which 68% will be at max 1sigma, plus 27% at max 2sigma and the remaining at 3sigma.

Per miner's you can computa damage for each of these ranges.

There are other methods for computing fatigue in frequency..i believe dirlik is one of the most used.

Hung Tran

For fatigue you usually compare RMS stress value for assessment.

RMS value is the 1-sigma stress value, occuring ~68% of your test/simulation time. 2-sigma is 2*rms value, combining 95%..and 3-sigma is 3*rms summing up ~98%.

You can use miner's rule for computing fatigue.

If your model has highest participation factors of 10Hz and your structure has to withstand 10000s, you can say that you will have a total of 10*10000 cycles, from which 68% will be at max 1sigma, plus 27% at max 2sigma and the remaining at 3sigma.

Per miner's you can computa damage for each of these ranges.

There are other methods for computing fatigue in frequency..i believe dirlik is one of the most used.

Hi mr Adriano,

Thank you very much.

Could you please explain more detail about this sentence

' highest participation factors of 10Hz and your structure has to withstand 10000s, you can say that you will have a total of 10*10000 cycles '

For example, if my model is being excited from 1 to 20 Hz during 10000s, and it is resonance at 10Hz , so i can calculate total cycles as : 10hz*10000s = 100.000 cycles

Is it right ? and why ?

Thanks

Adriano_Koga

kind of. You can request PARAM,EFFMASS so that you have an Output ranking the mass participation for each mode per direction (xyz).

As a simplification of cycle counting for your psd analysis, you can assume that if your most participating mode (ranked by effmass) is shown as 10hz(i.e), you will have 10000s * 10hz = total 100.000 cycles.

So, from these 100.000, ~68% will stress levels up to 1sigma (rms), ~34% up to 2sigma and the other portion up to 3sigma.

So if you simply use miner's rule, your damage will be calculated using these cycle counting.

Adriano_Koga

Hopefully this helps you:

https://blog.altair.com/random-response-fatigue-analysis/

Hung Tran

Hopefully this helps you:

https://blog.altair.com/random-response-fatigue-analysis/

Dear Mr. Adriano,

Thank you very much, it make me more clearly about Random vibration, but i still want to learn more about it /emoticons/default_biggrin.png' srcset='/emoticons/biggrin@2x.png 2x' title=':D' width='20' />

I will keep studying, then when i have a question, i will post here, please help me !!

Thank you !