Lines of Code
[root]/CCA/Components/MPM/Matlab/Multilevel
| Author | Changes | Lines of Code | Lines per Change |
|---|---|---|---|
| harman | 67 (100.0%) | 1896 (100.0%) | 28.2 |
shapeFunctions.m
reduce debugging output
findNodesAndWeightGradients_linear()
When a particle is in an extra cell you need to determine if the particle is to the left
or right of a node to properly set the gradient.
particleFunctions.m
reduce debugging output
initializationFunctions.m
findCFI_nodes(): fineLevel-> fl, coarseLevel->cl
initialize_Lx(): Initialize Lx to NaN.
Do not add extra Cells to Level 1 when using Linear interpolation
ML_AMRMPM.m
- output intermediate variables to files for comparison with composite grid formulation.
- The internalForce for the compaction problem is 10X bigger than it should be.
- use ps.NP(level) instead of ps.NP in loops!!
- compute intForceG: use the interpolation dx set in the particle set, dx = ps.interpolation_dx(l).
- project state back to the particle: Only perform this step in non-extracell cells.
- cleaner method for computing the tipLocation and tipDisplacement
- plotResults(), now plotting extForceG.
- bulletproofing: using a cleaner way of finding rouge particles that have left the domain.
- relocateParticles: Only perform this step in non-extracell cells.
The code is fully of print statements and needs to be cleaned. At this revision
number ML_AMRMPM gives similar results to the composite grid formulation AMRMPM
for the compaction problem.
0 lines of code changed in 8 files:
shapeFunctions.m
Added ExtraCell_positionToNode() -returns the CFI node closest to xp
findNodesAndWeights_linear()
- now using ExtraCell_positionToNode if the compDomain is in the extra cells.
findNodesAndWeightGradients()
- now using ExtraCell_positionToNode if the compDomain is in the extra cells.
ParticleFunction.m & initializationFunctions.m
name change fineCFI_nodes -> CFI_nodes
coarseCFI_nodes -> CFI_nodes
ML_AMRMPM.m
interpolateParticles to grid: only iterate to the CFI node when the compDomain
is in the extra cells. We only care about that node.
ComputeInternalForce: only iterate to the CFI node when the compDomain
is in the extra cells. We only care about that node.
Runs until you compute the deformation gradient.
75 lines of code changed in 4 files:
shapeFunctons.m
positionToNode() - use Matlab's built in function 'find' instead of my kludge to fine the nodes.
initializationFunctions() - initialize node positions to Nan.
17 lines of code changed in 2 files:
More thrashing:
shapeFunctions.m
postionsToNode() -return the left and right node of the cell that the particle is in.
findNodesAndWeights() -turn off bulletproofing when looping over the extra cell particles
particleFunctions()
createEC_partcleStruct(): keep track of the number of particles in a extra cell
particle set. Set the CompDomain flag to "ExtraCells"
Set the dx used for interplolation, which is not level.dx.
ML_AMRMPM.M
-interpolateParticlesto grid: loop over the particles in the particleSet.
relocateParticles(): update the number of particles (NP) in the particleSet.
122 lines of code changed in 3 files:
The cell spacing (dx) is now being passed into all shapeFunctions.
34 lines of code changed in 2 files:
variable name changes
nRegions -> nPatches
Regions -> Patches
23 lines of code changed in 1 file:
Still thrashing around.
particleFunctions.m
Added getCopy() - which returns the requested array data from the particle set.
createParticleStruct() - now using variable length input argument lists
createEC_particleStruct() - added ";" to eliminate some of the output noise.
ML_AMRMPM.m
-The particle arrays (xp, Fp, velP, massP....etc) have been placed in a particle struct.
-Looping over particle sets inside the "interpolate particles to grid", and "computeInternalForce"
steps.
- No longer calling adjustCFI_Nodes. Keep code around for comparisons.
- plotResults() - passing in particleSet structure.
134 lines of code changed in 2 files:
particleFunctions.m
- completed findExtraCellParticles()
- added createEC_particleStruct()
function that creates a particle struct that contains only extra cell particle data.
84 lines of code changed in 2 files:
More thrashing around:
ParticleFunction
createGhostParticles -> createExtraCellParticles
Added code to createExtraCellParticles()
Added function findParticlesInRegion()
GridFunctions
hasFinerCell() added argument withExtraCells or withoutExtraCells
added function outputLevelInfo() - prints out level and patch information
initializationFunctions.m
using newly created functions.
ML_AMRMPM.m
using renamed functions
115 lines of code changed in 5 files:
ML_AMRMPM.m
- drawNodes() color the extra nodes blue
InitializationFunctions.m
- each level is keeping track of which nodes are extra cell nodes
45 lines of code changed in 2 files:
Each level now contains info about the interior
node hi/lo and min/max
26 lines of code changed in 1 file:
Major thrashing:
No longer syncronizing the nodes at the CFI, this won't work in 2D, moving to a new
strategy. Now adding extra cells to the fine level CFI and adding ghostParticles
from the adjacent level.
particleFunctions.m
- added create Particle Struct
- added disolveParticleStruct
- added createGhostParticles() - empty
- added deleteGhostParticles() - empty
initializationFunctions.m
- added addExtraCells()
103 lines of code changed in 3 files:
Added sfigure()
Now using sfigure() instead of figure(). Sfgure() prevents the window focus from being
stolen by matlab.
Wrapped the code that dumps out movie frames with the conditional
if(dumpFrames){
}
Unfortunately, the function getFrame() steals the window focus.
51 lines of code changed in 2 files:
Rewrote adjustCFI_nodes to handle a variable number of input arguments
Calling adjustCFI_nodes after computing the internal force on the nodes
Added the functions ML_to_1D() and ML_Grid_to_1D(). These functions
convert multi-level arrays into 1D arrays that are sorted by position.
These are used for plotting.
- Now plotting extForceG
** Quasi-static compaction problem appears to be giving the right answers for
2 levels, refinement ratio of 2 on the finest level. More testing is needed.
142 lines of code changed in 1 file:
gridFunctions.m
-mapNodeToCoarser()
now using a tolerance comparison instead of (a == b)
InitializationFunctions.m
-initialize_lx()
Removed hard coded values for the Lx values at the fine/coarse level CFIs.
Now computing them.
ML_AMRMPM.m
- reordered initialization steps
- updated compaction initialization
- added loop over levels around ExternalForceAccl()
ExternalForceAccl()
now returning bodyForce from function.
- changed E, bar_max & domain to start testing the quasi-compaction problem on
multiple levels.
60 lines of code changed in 3 files:
gridFunctions.m
-added mapNodeToCoarser() -- function that returns the a coarse level node index
of a node with position X.
initializationFunctions.m
-added findCFI_nodes() -- function that determines the fine level CFI node indices and
the underlying coarse level node indicies.
- initialize_grid() - removed hard coded CFI node indicies.
ML_AMRMPM.m
- adjustCFI_nodes()
removed all the hard coding and replaced with more general code.
88 lines of code changed in 3 files:
shapeFunctions.m
-findNodesAndWeights_linear()
add & subtract fuzz (1e-15) to the shape functions whenever the particle is
coincidental with a node. If you don't then the simulation will go unstable.
initializationFunctions.m
- initialize_NodePos()
simplified algorithm to compute node positions
-initialize_Lx()
l -> left, r -> right
hard coded Lx values at the CFI.
-initialize_grid()
define dx for each level
compute the high and low node index for each patch
hard coded the CFI nodes for each level
ML_AMRMPM.m
- after interpolate particles to grid call adjustCFI_Nodes()
added adjustCFI_Nodes()
At the CFI nodes, q_new = q(Level1) + q(Level2) +....
q(allLevels) = q_new.
Currently the CFI node indicies are hard coded.
Currently able to advect a block of particles on a 2 level grid, rr = 2.
Linear shape functions.
After 80 timesteps the sum(positional error) = 2.131628E-14
148 lines of code changed in 3 files:
initialize as many variables as possible to NaN. This should catch
when a variable is being used inappropriately.
- Adjusting plotting routine to handle arrays with NaNs
fixed error calculation in the "advectBlock" section
50 lines of code changed in 2 files:
advectBlock Initialization
- initial position is just a single array.
computeDeformationGradient()
- declare vol & dF. You must declare the size of the arrays if they're not passed in.
ComputeStress(()
- must declare size of array.
relocateParticles()
- Look for gaps in the arrays and eliminate any gaps. A gap is a NaN followed by
a non-zero number.
Can now move particles from coarse Level -> Fine Level and vice versa.
62 lines of code changed in 1 file:
isOutsideLevel
using <= & >= as opposed to < & >
1 lines of code changed in 1 file:
(9 more)