Hi
I have simulated the SHPB test based on the attached image conditions in Abaqus (3D and one quarter model). The analysis is dynamic/explicit. The bars are linearly elastic. The cylindrical specimen test is made of epoxy and its plastic part is presented in xy data in property module. After analysis, the resultant stress-strain, stress-time and strain-time curve in the particular node of the specimen (the node has maximum mises stress), has an illogical appearance and is very different from the corresponding experimental results (output results of simulation are logarithmic strain LE33 and the mises stress at node). For example, the maximum stress in the experimental curve is 209 MPa, while in simulation, the maximum mises stress is not greater than 160 MPa. Also, the simulation output curve has broken lines and does not have a stress-strain curvature form. What is the problem? Is my simulation wrong?
Thank
There is no way to say from the information you provided.
Your abaqus plot seems to be stress vs. time, not stress vs. strain.
You may observe elastic oscillations, it may be that the may von Mises stress is at a time point where you did not look (it seems that you only plot at a few time points), there may be a problem with hourglassing, your mesh being too coarse or about a million other things.
I would recommend to check other variables - speed of the bars, kinetic energy or whatever else is available. I would also recommend to resolve the time much more closely.
Increase the number of of output interval.. make it very big so you get smooth strain history output... >> Go to file output >> Evenly space interval >> put 1000 or more.
Make sure you also refined the mesh for the specimen, and also the bar ends near the specimen...
Actually, finding stress field in FEM, needs in differentiation, the saw-like plot may indicate (along with insufficient output data, as was already pointed out), numerical errors. Unfortunately, in explicit schemes quadratic elements that give much better results for strain and stress, are usually absent, so herein my recommendations on how to make better stress or strain in explicit analysis:
1) increase loading variation time (up to 20 x (stable (Courant) time increment), that will considerably reduce non-physical oscillations
2) apply filtering in Abaqus there are several filters available, best of which are Butterworth filters, try also median filter (in Abaqus it is called Smooth)
Good luck
- Badges
- Science topic