Describe the bug
The stress computation shows a wrong result if there is a particle lying on an element edge/boundary. This is perhaps one of the major causes where the stress and the strain in the simulation sometimes show a singularity error. This happens in all constitutive model.
The error is resulted due to the incapability of preserving the rigid-body motion zero energy modes when a particle is lying exactly on element edge. See the following math for a single particle moving in rigid-body motion with v=+1 m/s:
The following simulation is produced for an elastic particle with E=1000, v=-1, dt=1:
To Reproduce
Steps to reproduce the behavior:
Compile 'cmake -DCMAKE_CXX_COMPILER=g++ .. && make -j8'
Run on one node with './mpm/build/mpm -f \path\to\input\file -i input-file.json'
Run any simulation where a particle is lying exactly on edge.
Expected behavior
Rigid body motion should have no stress. The corrected version:
Runtime environment (please complete the following information):
Any OS
Branch develop
Additional context
The easiest way to fix this is to implement a nudge of std::numeric_limits<double>::epsilon() while computing reference location.
Describe the bug The stress computation shows a wrong result if there is a particle lying on an element edge/boundary. This is perhaps one of the major causes where the stress and the strain in the simulation sometimes show a singularity error. This happens in all constitutive model.
The error is resulted due to the incapability of preserving the rigid-body motion zero energy modes when a particle is lying exactly on element edge. See the following math for a single particle moving in rigid-body motion with v=+1 m/s:
The following simulation is produced for an elastic particle with E=1000, v=-1, dt=1:
To Reproduce Steps to reproduce the behavior:
Expected behavior Rigid body motion should have no stress. The corrected version:
Runtime environment (please complete the following information):
Additional context The easiest way to fix this is to implement a nudge of
std::numeric_limits<double>::epsilon()while computing reference location.