Could you describe the model setup a little bit more in detail? From your geometry description I assume you excite Lamb waves. With the thickness and the frequency you defined, you will see higher order Lamb wave modes.

Hi Hamidreza,

As Christian suggested at such a high frequency, and 2mm thickness you are trying to simulate guided stress waves aka lamb waves.

At 5 MHz except the fundamental modes (A0, S0) also exist higher asntisymmetric and symmetric modes that you are triggering through your applied load. But I dont think this is the problem. Probably your system hasn't converged yet. At 5MHz A0 has a wavelength of approximately 1 to 1.5 mm and with conventional FEs you need at best 15 nodes per wavelength, with a fast calculation you need 135 wavelengths per side, that is 2025 nodes per side. And by assuming you have 3 FE through the thickness to achieve an adequate solution you reach up to 12301875 nodes for your full construction. The aformentioned number times 3 its your total system to achieve a correct solution. To sum up you need up to 36M DOFs for your system to converge. Do you have the computentional power for this enormus system?

I have one question, why 5MHz, what are you researching?

Dear Mr. Willberg,

At first, I want to thank you for your attention. I added another image to the question. What kind of description are you looking for ?

I want to simulate longitudinal waves but maybe I'm in the wrong direction. What should I consider through this work ?

Thanks in advanced,

Hamidreza

Dear Mr. Rekatsinas,

Thank you for your kind attention. The thickness of this model is 5mm and other dimensions are 2 cm. If I lower the frequency, it may confluence with back-wall echo. The size of seeds are 7E-005 and the result didn't considerably change from 8E-005. The time increments are 5E-009 maximum. The number of elements are about 3.7M with 6 DOF. I couldn't decrease seed size anymore because of my system limitations. Do you have any suggestions to get this work done ? I'm trying to simulate some kind of defects in 3D parts.

Best regards,

Hamidreza

You can solve the symmetric problem by applying symmetry boundary conditions at the left and bottom edges.

https://www.clear.rice.edu/mech403/HelpFiles/CW_sym_anti-symmetry_BC.pdf

with this way you will have 1/4 of the total problem.

Just so you know, the other way packets are higher wave modes. But the convergence and dispersion problem might still exist.

Good luck!

I'm not quite sure how to do this. You mean I model a quarter of the hole model and define XSYMM to the boundaries ?

Another question. Should I decrease the frequency ? and Do I have to add at least 2025 seeds to every edge ? This number is not accepted by ABAQUS.

Best lucks,

Hamidreza

Yes I mean the to apply X and Y SYMM to the boundaries by modelling the 1/4 of the model.

I cant tell you to reduce the excitation frequency because maybe its crucial to your research. I am just saying that the FEs need approximately 15 nodes per wavelength to achieve a converged solution.

My advice would to start simple.

That is model a simple plate and try lower frequencies, 50-100 kHz to see how the guided waves work. You need the dispersion characteristics of the cross section of your structure. By dispersion characteristics I mean the Group Velocity and wavelength dispersion curves.

If it is not much trouble please provide me information of how you apply the load. Location, pressure, concentrated force etc..

" I want to simulate longitudinal waves "

For a thin plate you won't have such wave. In your case you will have symmetric mode. Due to your frequency and thickness you will excite at least two more mode, called S1 and S0.

Article An effective damage identification approach in thick steel b...

In the end it does not matter. If you want to study a real problem with this frequency you have to take these effects into account. To reduce the numerical effort you can use the symmetry of your problem as Christoforos suggested.

Another possibility is to start with a 2D problem (plane stress and plane strain). The key features are equal to the 3D problem with lower numerical effort.

Hi Javidrad,

I agree with previous answers about the high order Lamb waves. Before we have an in-depth discussion, can you please clarify the following concerns?

1. Where are the locations of your actuating (exciting) and sensing?

2. How do you apply the excitation? Transversely or out-of-plane? with piezoelectric elements or simply applying displacement boundary?

Thanks,

Guoyi

Hi Javidrad,

In ordrer to identify each contribution, longitudial direct and edge wave or shear edge wave in solid, you can use the ray methods. Otherwise, I recommend you to represent the normal and the tangential stress separately.

Nadir.

Dear sir,

Thanks for the all answers and kind attention. I afraid we are not in the same page. I provide a document to clarify the problem.

I'm looking forward to taking your solution and comments.

I appreciate your time and kindness.

p.s. I sent the file password to your private message.

Best regards,

Hamidreza

Hi Javidrad,

You document and animation are very useful for me to understand your issue. I think the additional waves, except for reflections between two boundaries, are high order modes. And some nonlinearities are due to the waves reflected from the edges interacting with your longitudinal waves.

Some suggestions: 1) model the structure with a large in-plane (u1-u2) size to eliminate the effects of edges; 2) through the simulation & animation, calibrate the velocities so that you can identify the reflected waves with the known thickness; 3) it is a good idea to use a relatively lower frequency firstly, might be able to avoid high order modes.

Hope it helps,

Guoyi

Hi dear Guoyi,

Thank you very much for your time and kind feadbacks. I am going to change thickness to 2 cm and 1 MHz frequency and model full part at first. I don't understand your statement "high order modes". Is it possible that I get natural frequency of the part ? As you can see at the bottom of the report, the magnitude of second peaks (which are extra) are higher than load magnitude. I think it amplify by natural frequencies. And another thing. Do you think time increment (1E-009) is small enough for this calculation ?

I really appreciate your help. You are really helping me move forward.

Best lucks,

Hamidreza

By "higher order modes" Guoyi means modes that are not fundamental and appear at higher frequencies i.e. 1-5 MHz. S1, S2, A1, A2 etc.

Maybe your purpose is not to study guided waves, but with your choosen thickness and frequencies you currently are.

To study bulk waves you need the thickness to be in the same range with the other 2 dimensions.

Hi Javidrad,

I think Dr. Rekatsinas has answered the question about the "higher order modes", and I completely agree with him.

For this: "As you can see at the bottom of the report, the magnitude of second peaks (which are extra) are higher than load magnitude. I think it amplify by natural frequencies. ". I think it probably are the superposition of wave reflected from boundaries; you can guarantee it through a larger plane size as mentioned before. I think if the nature frequency matters, it would affect all the waveforms.

For the time-step, I think a convergence study will be the most convincible way, even I think you time step is fine....

Thanks,

Guoyi

Thank you Mr. Guoyi and Mr. Christoforos. I'm working on some strategies to solve the problem. I will let you know any progress.

Thanks for all supports and suggestions,

Hamidreza