Dear Simun,

If I understood you question, your problem is concerning the energy during the crash remains constant after the crash. If I'm wrong, please correct me.

1- When we work with 3D models, it's very important (especially with compos mat) create the plies and if possible each layer with the both constitutive materials, i.e., matrix (one parasolid file) and the fiber/cloth/roving/unidir your intends to use and prior to import to ABAQUS or anything else FEA/CFD package, you must create an assembly of the both materials and then, you should create the composite plate or other geometry do you need.

2- The right laws of particle collision should be observed to set the adequate boundary conditions, independently of the algorithm you'll use. any mistake in this step conduct the FEA to spurious results.

3- It's obviously, but not impossible to put values and associated units in a mixed mode, example: use mm and m to set any measure, or kN and N to express force.

So, firstly I'll consider to review all these steps, than I follow more specific issues regarding the Procedures of input solver method.

Dear Simun,

At first check element type you used for composite laminate and artificial strain energy variation. If elements include reduced integration formulation such as S4R, hourglass control will cause increasing artificial strain energy that increase total energy. Increasing artificial strain energy because of unrealistic strains make results unacceptable. However if artificial strain energy be less than 10% of total energy, its effect on results could be ignored.

Therefore if artificial strain energy value is greater than 10% of total energy, use S4 elements instead of S4R, especially for impact zone.

Using S4 elements increase run time. Therefore its better to use it for limited area.

Another way for solving this problem is decreasing incrementation time. When you define Dynamic, Explicit procedure, use all default values. Scale factor default value is 1. Just decrease its value up to 0.5 ( start with 0.9 and decrease it 0.1 step by step untile ETOTAL, get constant).

Hope this help.

Regards,

Khodaei

First of all thank you both for your quick answers!

Meysam Khodaei ,

thank you for the suggestion but I always check ALLAE and in this case is under 3% and going to full integration elements would be too expensive because this case that I showed is just a test case. My real case have over 1 million elements.

Altibano Ortenzi ,

my concern is that my ETOTAL is not constant as you can see from the ETOTAL picture but it's increasing in the few first steps because my elements are damaged and my ALLDMD is increasing. My units are consistent so this is not a case but concerning boundaries, I think that all of them all stetted up correctly.

My problem is that I don't believe ETOTAL is calculated correctly because my kinetic energy from the cylinder should be spent on internal energy (internal combines strain, artificial, viscous and damage ALLDMD energy) and remaining kinetic energy of the cylinder.

For the first proposal, sorry but I didn't fully understood you. Are you suggesting to model matrix and fiber separately because I'm using Lamina properties?

Thanks again!

Yes. It takes more time to develop the parasolid assembly, however, the results are more precise than an homogeneous solution. We compared the solutions and we know through the tests and simulations, that the assembled models reach 18-32% greater precision, and it doesn't matter the algorithm used.

Yes, you should use more compute time and more powerful computer, however, any laptop with a 500 GB SSD dedicated graphic board with 4GB and 32 GB of RAM works fine. Unless your question is regarding to a homework and you MUST use these routines to solve a problem. In this case, follow the solution presented by Meisan!