Dear Youssef,

Maybe you must change boundary conditions, and I think it is better to apply simply support in z direction (fix the plate movement in the normal direction). because the frictional contact does not have any constraint on normal separation.

Hello Aidin, adding an additional boundary condition to constrain the rubber sheet vertically will not simulate the real behavior as the rubber sheet is free to move in that direction. Do you have any suggestions or reasons for getting such behavior? 

Dear Youssef

Hello Amir,

I think you didn't complete your post, please retype your post again.

Thanks!

Hello. Is the cylinder rolling along with the sheet? If the cylinder rotates along with the sheet than you shouldn't observe this kind of vibration. If the cylinder is still then this behavior could be attributed to the deformation of the sheet. Just like in an empty conveyor belt in which you have a roller which is stuck. Another cause of problems may be the time step. Try a smaller one even if this means longer solution duration. Hope this helps.

Thanks Dimitrios, the cylindrical body rotates driven by the rubber sheet. I am convinced what should be the effect of time step on the resulting motion, but I will give it another try.

Oooops... Sorry...

Dear Gerges

1. The time duration (not time step) is very small, thus one may not capture the whole behavior correctly.

2. I am not sure if you have used a model to capture stick slip condition in contact. If yes, fluctuations are reasonable to observe. 

3. Since the cylinder is rolling, a transient fluctuation may be reasonable due to initial wrinkle formation on the sheet.(in this case stick slip can not be the reason)

3. To damp the vibrations constitutively you may use hysteresis capturing material models, or use experimental data ( most of polymers show some extents of hysteresis).

4. You can also define a viscous contact to damp the vibrations ( I'm not sure this matches the actual physics of your problem or material properties)

5. Can you monitor the temporal variation in the normal gap in your interface. Insufficient boundary conditions might result in instabilities at the interface. ( please notice up to this point in your model, we cannot find out if the model is going to be unstable)

Good luck

Hello Amir,

Thanks for re-posting your answer once again. I find it very useful!

1. My study includes applying variations to the sheet linear speed, which constraints the time duration accordingly. As the target is studying the resisting reaction forces to the motion, so both components have to kept in contact throughout the study time. Extending the sheet length will increase the computation time significantly..

2. I am using a frictional contact between both components. I am not sure if this contact type captures stick-slip phenomenon. Please advise.

3. That is correct, the cylinder rolls on applying a velocity boundary condition on the rubber sheet.

3'. The rubber material is defined as a viscoelastic material through extensive experimental testing.

4. I tried to use different values for stabilization damping factor, but that had no effect on the results at all.

5.  The contact tool results show that there is a zero gap between both components throughout all the study time span.

Hello Gerges,

Do you have tension applied on the sheet?

Yes, I do. I applied equal tension forces on both sides.

Dear Gerges

In my opinion, the fluctuations are a transient behavior of the model, you may need to extend the time span to capture the steady behavior. My reasons are:

1. the time span is very small for viscoelastic material to show its temporal effects (time constant of the rubber might be in orders of 10-20 seconds)

2. Frictional contact should capture stick-slip (if you use a proper model) but since you mentioned the gap is always zero, it did not occur in your model. As a symptom, slip condition occurs usually with a slightly positive gap.

3. Moreover the time is such small that wrinkling may not occur that soon.

Good Luck

Hello Amir, I have tried to use relatively longer analysis duration (up to 300%), but it still shows very similar behavior. Do you recommend a certain duration range? I think extending the analysis time to 1 second will require a minimum rubber sheet length of 8000 mm @ belt speed 8 m/s which will be very expensive in terms of computation time. Any suggestions? 

Is it possible that you face some penetration of the surface and target elements used to model the contact between the cylinder and the sheet? Since there seems to be some kind of a periodic behavior on the graph you have attached, I would check if that periodical behavior is linked with the number of elements you use in the periphery of the cylinder and the length of the sheet.

Hi Gerges, 

I would suggest in order to identify the origin of this behavior, you may make a coarser/finer mesh, increase/decrease belt speed, change the material properties of belt and roller...

These would help you identify the potential origins.

Please note that reducing the speed enables you to run for a longer time span with less length in belt.

Good luck

@Dimitrios, yes I face penetration at the contact area (max. 0.9 mm all over the study time). Any interpretations?

I tired to use finer mesh and wider study time interval, but I have at least one or two peaks in all cases.

Your advice will be highly appreciated!

I have small experience with contact elements. However, what I do know is that they perform better for a certain value of displacement rather for a certain load. Your case is the latter. I really think that you should insist on your efforts on a finer mesh. For example a 2-D model with a certain portion of the total cylinder (a small arc of 10-20 degrees) and an adequate length of the sheet should save you alot of elements. As long as you have a load for a boundary condition and a transient problem, you are bound to face penetration. Again I am no expert, but this applies to my little knowledge. The values you get for some instances that you don't experience penetration are satisfactory? Could you retain some of them as correct?

Thank you Dimitrios. I almost followed the same procedure to minimize computation time and getting more fine results.

You 're welcome Gerges. I am interested to know how this worked out.

@ Dimitrios, the results I have so far ranges from 65% to 150% compared to the experimental results, where the average accuracy is around 105%. I didn't change the model much, where I used hex dominant mesh, tried to use larger time interval to capture a wider range of results and tried to simplify the model to reduce computation time.

Thanks for your help!

I have got results for 5 different FEA model approaches (each one was solved at 3 different load values). Currently, I am struggling to find a quantitative measure that can help me the FEA approach whose combined results (for all loads) match with experimental results the most. Can any of you guys help me with this?

 I tried to explain that in a separate question

https://www.researchgate.net/post/I_have_several_FEA_results_sets_I_want_to_estimate_which_one_is_the_most_accurate_one_compared_to_experimental_results_Can_I_use_R2_R-squared?_tpcectx=profile_questions

Your help will be highly appreciated!