Hi Everyone,
i'm working with particle finite element method. The idea is very simple, after every finite element calculation, i project the data (e.g. the plastic deformation gradient and the hardening variable) from gauss points to the nodes, detect the boundary of the nodes and remesh the nodes as shown in Fig1. In the next time step, i interpolate the data from nodes to gauss points.
For the finite element calculation, i have implemented 6 nodes element with segment to segment strategy for elasto-plastic problem. At the beginning of some time steps it works fine. But after some time steps, the finite element calculation of feap got problem, because on an element (the pink one in the lower right corner of mesh.png) the determinate of the deformation gradient is negative, even though the mesh quality is not bad, see mesh.png.
So i try to record how the mesh deforms during the Newton iteration at this time step. Following is a part of the input file.
loop,equi,99
af20
CONT,CHEC
tang
form
solve,line,0.8
next,equi
stre,node,1,999999999
I implement "af20" to record the mesh for every iteration. The result is in NewtonMethod.gif. You can see after 4 iterations the calculation interrupt. The most strange thing is that the mesh deforms too strong which is not expected.
I wonder why it happens. Is it because of the projection? Because every time step i do a projection to transfer the data from gauss points to the nodes, which let us lose some accurate. After some time steps, the data which stores in the nodes is so not accurate, that causes what happens.
Is my idea right? Can anyone give me some advice why it happens? Thanks a lot.
