Reply To: SPH brittle modeling with UMAT

[igandiko]

Hello, I would like to share some comments I recevied from an SPH expert. I hope this helps. 

1. The main issue to consider here is tensile instability present in SPH calculations. This might lead to inaccurate approximations that can then cause elements to fail at the incorrect time. Another thing to keep in mind is the fact that there are no "fragments" in SPH. In other words, there is no way to identify the fracture surface, so particles will influence each other if they are close enough to be considered neighbors. Another way to think about it is that particles that fracture into distinct pieces at some point will be "glued" together if they get close to each other again.
2. Yes, please check examples of usermat that make use of the a failure flag.
3. If the particles belong to different parts, users can use the default particle-to-particle interaction, or use *DEFINE_SPH_TO_SPH_COUPLING to specify a penalty-like coefficient defining the interactions between particles of the different parts. This is assuming the SPH particle itself has not eroded.

Some general information on spall which may help: 

When a particle fails, the pressure is set to 0. That affects the neighboring particles but that particle is not deleted. There is a feature that advanced user an use which is to try the *SECTION_SPH_USER option. User can then define some customized changes in the variation of the smoothing length according to the state of pressure of the particle (for instance). Once a particle fails, user can reduce the smoothing length on few time step so that spallation can become visible. This feature need a library of lsdyna and need to change the subroutine hdot in the dyn21.f file.