For Abaqus is more complex to calculate contact between Lagrangian shell elements and Eulerian material than with Lagragian solid elements.

Contact between Lagrangian surfaces and Eulerian material shows penetration along corners of shell elements without filled radius. For your problem I would create a solid mesh to avoid this problem. I have attached you an INP with this solution and a video with the first tenths of a second of the ODB.

I have modeled the INP with Abaqus 2016, so it could be that you cannot open the file, but as I have told you the change in your model is simple: replace shell elements by solid elements.

The other solution is creating a shell mesh with fillet radius, but this method does not work well with surface shell elements (SFM3D4R) used in your model. If you use shell elements S4R and use a mass scaling only to the structure (not to the Eulerian elements), you will obtain good results. See attached video with this other configuration.

Hello José Calaf Chica, thank you for your suggestions.

I thought I could fasten computational time with rigid shell element.

I'll get back to you, after I apply your suggestion.

I did solve the problem however although it is not an ideal solution, by making the water at step 0 floating at the air (to avoid contact between water and other parts at the beginning). I still don't understand, why the simulation went well that way...

1. By the way, is the global mesh size of eulerian and lagrangian elements important? Could it be that if the eulerian mesh smaller is than lagrangian mesh causing the leaking? I am not sure every time I have to decide mesh size for eulerian and lagrangian elements. Which one should be smaller or does it even influence the result?

2. What is the purpose of applying the "mass scaling only to the strucuture"? I have once read about it in manual. As long as I understand it correctly, this is a way to make the simulation faster.

Thanks again for your replies, José Calaf Chica :)

Hello Jose, how could you make a shell mesh with fillet radius? I could see in your 2nd video, that the parts (other than the water) have thickness (or shell element with thickness) and fillet radius.How did you do that? I have tried assigning thickness larger than 1 at section manager in property module, but it doesn't change the mesh.

Thank you for your reply :)

Hello,

I think the problem can be related to more than one reason:

i. mesh size should be of the same order of magnitude for both the Lagrangian and Eulerian meshes;

ii. average mesh size of the Eulerian part should be less than the thickness of the shell elements that you're using in the Lagrangian part;

iii. you should look at the boundary conditions that you have defined for the Eulerian mesh, you can find how to prevent or not the outflow at boundary surfaces here, https://www.sharcnet.ca/Software/Abaqus/6.14.2/v6.14/books/usb/default.htm.

Hope you solve the problem.

first reason is not right " i. mesh size should be of the same order of magnitude for both the Lagrangian and Eulerian meshes "

because some thickness are too real small like 0.0025 m , sothat much bigger mesh in eulerian part