Learning Forum

[Autonewbie]

Hi,

I am running a random vibration based on modal results for an assembly. Is there other method that I can assess the non-linear contact stress in Random Vibration as I understand it is linear contact in modal and random vibration?

[peteroznewman]

You must use Full Transient Structural (not using Modal as a Pre-analysis) in order to see nonlinear contact such as Frictional or Frictionless contact working as you want, by opening and closing and sliding around.  Note that No Separation is a linear contact so you can get the sliding around part working in Random Vibration.

However, the input to Full Transient is a load-time history, while the input to Random Vibration is a PSD load spectrum, so you have to have or synthesize a load-time history to use Full Transient.

[Autonewbie]

 I also interested of Random Vibration but Random Vibration is linear analysis, right?

[peteroznewman]

Yes, Random Vibration is linear.

[Autonewbie]

What happened was the area of interest has been shifted to the area of fixation (bolt hole). If bolt is modeled in with contact but non-linear is not supported ... what would be the best solution to account for the nonlinear contact in random vibration?

[peteroznewman]

Please explain using images, labels and pointers.

[Autonewbie]

Basically, I would like to know the nonlinear contact stress under Random Vibration...

[peteroznewman]

A nonlinear contact is Frictional, Frictionless or Rough contact.  To use that in a simulation, you must use Transient Structural.

Transient Structural is a time-history simulation where you can use nonlinear contact.

Random Vibration is a frequency-based analysis which requires Linear Contact.

You can synthesize a time-history load for use in a Transient Structural model from a Random Vibration PSD specification. I can help you with that if you want.

[Autonewbie]

I would be love to hear that. Thanks in advanced! 

What should I provide you? PSD profile and a model?

[peteroznewman]

I only need the the PSD profile to synthesize a time history. You can reply with that as lines of text in the post.  Attach the model .wbpz file for my interest.

You should do the MSUP Random Vibration model first anyway to see how many orders of magnitude the solution time is compared with the Transient Structural. We're talking minutes vs. tens of hours, and that is assuming you don't run out of disk space during the solution.

[Autonewbie]

I share the PSD below. For the model, I need times to prepare a similar for you.

 

I already have the MSUP random vibration but that leads to another concern as many modes are required if meeting the participation factor or maximum frequency of the PSD profile. I ran the model up to lower frequency (800 Hz if refer to the PSD profile below) as damage is higher in lower frequency. Correct me if I am not right. Thanks!

 

 

[peteroznewman]

I have matlab and use a GUI package called vibrationdata that has a script to synthesize an acceleration time history from white noise to meet a PSD specification.  The script requires input in terms of G^2/Hz, but the package also includes a unit conversion.

The sampling of a time history should have 20 samples/cycle, so to properly reproduce a 1500 Hz signal, the sampling would ideally be at 30 kHz. Attached is a signal sampled at 26 kHz. It is 2.5 seconds long and has 65,536 data points.

Here is how well it meets the PSD specification

[Autonewbie]

Hi Peter, 

It is so cool! I guess next step is to use this time domain profile in running transient simulation, is that right?

[peteroznewman]

Yes, insert a Transient Structural model into Workbench and open Mechanical. Copy all the data in the text file and paste it into an Acceleration load when the units are set to meters and seconds.

[Autonewbie]

Hi Peter,

Is it what you have done is converting the PSD input to displacement over time at every frequency for transient analysis? How do we calculate that?

[peteroznewman]

I have matlab and use a GUI package called vibrationdata that has a script to synthesize an acceleration time history from white noise to meet a PSD specification in terms of G^2/Hz.

The GUI package with that script is free and open source. Here is a link to download the package. You can look at all the calls to various functions and see how it is done.

[MTBXC]

 Hi, Peter,

Have you met with similar software? E.g. to scilab, or other free software?

[GP_FEM_CFD]

Hi Peter! I have the same situation like @Autonewbie. The topic here is really useful but that MATLAB tool has too many opportunities and it is not that well explained how exactly can be transform PSD profile to acceleration signal which I want to use later in Transient Structural. Coul you give me somehow tips how to use it  Thanks in advance!n