bearing vibration analysis in explicit dynamics

  • Last Post 11 March 2020
kortagu posted this 29 December 2018

hey guys, since last year i am trying to complete rolling element bearing vibration analysis in explicit dynamics, i have some questions for experts;

- bearing load is not valid for explicit dynamics, therefore how can i apply bearing load ? ( i apply force on inner ring inner face in down direction but i think it is not proper way)

- because of geometry i have some very tiny gaps like e-16 mm unit, and in explicit dynamics i have no options like adjust to touch or normal langrange contact type. What should i do for it ?

Here is a video about it 

kind regards.


Attached Files

  • Liked by
  • minhasnavdeep95
  • ajaynsv
Order By: Standard | Newest | Votes
peteroznewman posted this 30 December 2018

Hey Ufuk,

I liked your video, very professional looking.

One thing I would suggest is editing out the part about Hertzian Contact Stress at 35 s because the mesh is too coarse to see the subsurface nature of Hertzian contact stress.  Compare the video with the very refined mesh I used in this post.

I haven't used a force on a rotating body in Explicit before. Does it keep pointing down as the rotation occurs? If so, that is okay. You would have to add a shaft to improve the fidelity.

Don't worry about the tiny gaps. Real roller bearings have some clearance too.

I don't understand why you are showing the vibration of the node on the top of the outer race.

At the end you show the neighbor node next to the defect. That seems much more relevant. You should spend some time explaining what f(bpfo) is and why it is 161 Hz by using the shaft angular velocity, inner race diameter, ball diameter, number of balls, outer race diameter.


kortagu posted this 30 December 2018

Thanks sir, yes you re right, i have an old laptop (4 processors) and as i remember defectless analysis took about 10 hours even mesh was course.

I dont know whether it keeps pointing down or not but how can i see that works you have an idea about it ? Should i look moment or smt ?

Actually i worry about contact mechanics in cad i created balls with point contact and they seem well but in ansys after mesh you know circular bodies need refine mesh to get that circular shape.

Actually my aim was to observe bpfo which means OUTER RING DEFECT FREQUENCY at spectrum graph (FFT) but i failed (i dont know why) and i decided to signify stresses and waveform graph to complete task. Under real conditions, defect frequency is measured from top center of outer ring or housing of rolling element bearing.

In a nutshell i havent completed yet and i need to see that 161 Hz frequency on spectrum graph.

Do you have any recommend for me ? Now i m using half model and apply symmetry to reduce solving time, finer mesh etc. 

161 Hz means;

fbpfo = Z.fs/2 (1-(db/dm).Cos α )

with fs is the shaft frequency, db and dm is ball diameter and pitch diameter respectively. Since in this study assumed that the ball bearing is subjected to a pure radial load, hence the contact angle will be 00.

fs = 3000 cycle/min = 50 cycle/s = 50 Hz

 fbpfo = 8.50/2 (1-(7,5/38,5)) = 161 Hz

According to above equation, in this thesis BPFO is 161 Hz.


peteroznewman posted this 30 December 2018

Thanks for explaining fbpfo.  What is Z?  I expected to see the number of balls in the equation since if you add a ball, the frequency goes up, right?

Using symmetry is a good idea.

When you took the Ay signal, did you subtract the mean before you tried to compute the FFT?  If you attach a zip file with the Ay time-history data, I will take a look at it.

A simple model to prove if a force on a cylindrical face always points down or if it rotates with the face would be a disc on a flat block with an initial angular velocity. If the disc rotates 180 degrees and stays in contact, you know the force always pointed down. If the disc lifts off the block, then it is rotating with the face.


kortagu posted this 31 December 2018

Yes sir, Z is number of balls and it increases the frequency right. I did a simple yesterday and obtained acc-time results but i havent tried to take fft of it yet, i ll add that simple excel table here.

Yes, i understand your idea then, do i need a contact with disc and box ?     this one is a sample clip of simulation     this one is excel file

By the way, thanks a lot sir for your effort, im so appreciated.



peteroznewman posted this 31 December 2018

Hello Ufuk,

I looked at the data in the excel file sampled at 25 kHz, which is an adequate sampling rate.

There is a lot of noise in the data.

If you do an FFT of that data out to 10 kHz, the high magnitude is in the high frequencies.

You can filter the data with a 1000 Hz cutoff filter and get this time history plot

Then do an FFT on the filtered data

It still doesn't come close to the 161 Hz frequency you were looking for.

I recommend using the Ay value from the neighborhood of the defect. That seemed like a nice clean signal.


kortagu posted this 31 December 2018

Hello sir, thanks a lot for your effort again, Did you use matlab to get those results ?

i am wondering that sampling rate frequency (25 kHz)  affect anything ? Should i change parameters ?  

Im sorry but i couldnt understand that 10 kHz and other filtering processes.

Is there a source to learn those calculations which you recommend ?

Lastly, i applied 5000 rpm rotational velocity but it was constant, should i start it from 0 and gradually increase to absorbe shocks ? 

kortagu posted this 31 December 2018

Hello again sir, i obtained acceleration results from neighbour node in +y direction and this time i applied velocity gradually, i attached files, i hope i dont waste your time, could you check them in your free time ?

Thanks again,

Best wishes.


peteroznewman posted this 01 January 2019

Hello Ufuk,

This signal has much lower noise, possibly because the velocity was applied gradually.

Unfortunately, a time signal of the stress is not useful if the rotational velocity is not constant.

Run another simulation. In the first 12 ms, ramp the angular velocity up to 5000 rpm then simulate another 36 ms (three rotations) at a constant speed. I will take the 48 ms recording, discard the first 24 ms and analyze the last 24 ms with an FFT analysis.

Yes, I am using matlab to filter and plot FFT results. The sampling rate is fine.


kortagu posted this 01 January 2019

Hello sir, i will complete simulations and let you know asap, thank you

Kind regards,


kortagu posted this 02 January 2019

Hello sir, simulation has completed now and i obtained acceleration results from top and neighbour point. I hope this time we can see bpfo.In addition in these simulations i constraint all dof's of outer ring except translation in vertical direction to get high amplitudes but it might be problem i think if we fail with these i'm going to try outer ring completely fixed model. Also in these simulations i didn't apply radial load.  (neighbour point) ( top point)

Best regards,


kortagu posted this 05 January 2019

 Dear Peter, probably you are busy but you checked my solution, im looking forward your answer sir, thank you



peteroznewman posted this 05 January 2019

Dear Ufuk,

Thanks for the reminder, sorry for the delay.

I downloaded the neighbour point. Here is the data:

I only see 7.5 ms of data. I suggested the following:

In the first 12 ms, ramp the angular velocity up to 5000 rpm then simulate another 36 ms (three rotations) at a constant speed. I will take the 48 ms recording, discard the first 24 ms and analyze the last 24 ms with an FFT analysis.

Is there a problem to follow this guidance?  Maybe you uploaded the wrong file?

Kind regards,

kortagu posted this 05 January 2019

Oh i am so sory sir, i had uploaded old results, i checked it now, sorry again.

I uploaded new results which you guided first 12 ms ramped to 5000 rpm after than 36 ms constant 5000 rpm. ( NEIGHBOUR POINT)    (TOP POINT) 

By the way these result are acceleration-time results not stress actually maybe you missed it but its not problem, not affect results.


Kind regards,


peteroznewman posted this 06 January 2019

I opened the Neighbour point. The sample rate is now only 10.42 kHz.  What happened to the 25 kHz that you had before?  When you ask for 6 times longer end time, you have to ask for 6 times more output also.

Here is the time history of the last 24 ms.

In any case, here is the FFT of the last 24 ms of the signal.

Explicit Dynamics is notorious for noisy output. I don't know what you need to do to get a useful signal. Maybe add damping and simulate a much longer time scale (with a higher sampling rate).



kortagu posted this 06 January 2019

You re absolutely right sir, i had to increase sampling frequency but i skipped it. Now i am going to run simulation again with higher sampling frequency also i am going to add damping, i hope it works, i amm gratefull for your effort, have a nice day sir.

Kind regards,


peteroznewman posted this 06 January 2019

I also recommend a lot more rotations, say 10 rotations, which would take 120 ms at a constant velocity, plus the 12 ms to ramp up from zero to full speed, for an end time of 132 ms.  I understand that is going to be tens of hours of simulation time. That is the nature of explicit dynamics solvers.


kortagu posted this 08 January 2019

Hello sir, im going to carry on 132 ms simulation, by the way i tried to apply very small damping coefficient like 0,001 but it failed because even this small number affects rolling elements rotation, they didnt move in harmony then i decided to neglect damping inevitably. I will let you know when im finish simulation.

Kind regards,


ajaynsv posted this 29 January 2020

Hi Kortagu,

I am also working on a similar type of problem. Could you please help me in solving my problem.

I want to know the settings you have used to get the data from mounting probe on outer race.

rjhunt posted this 05 February 2020

can anyone one teach me how to setup simulation as in above video.

FaezWahid95 posted this 27 February 2020

 ufuk kortagu , can you share with me how your simulation set up was done ? really appreciate your handy help .

ajaynsv posted this 11 March 2020

Hi ufuck,

Can you share the analysis settings of your simulation? 

Really appreciate your help