Different Gen methods and different Number of LDPM particles

[Roozbeh]

I have tried using different LDPM Geometry Generation methods for a simple prism, and I obtained meshes with considerable difference in number of particles:

1) First method:
//----------------------------------------------------------------------
Material CONC LDPM {
MixDesign {
CementContent 420 kg/m3
WaterToCementRatio 0.45
AggregateToCementRatio 4.25
MinAggregate 4.75 mm
MaxAggregate 19 mm
FullerCoefficient 0.5512
}
}
//----------------------------------------------------------------------
HexSolidList MODL Geometry {
Material CONC
generate {
block {
dimensions 75 mm 75 mm 265 mm
elements 1 1 1
}
}
Make ParticleList PRTC {
seed 5067
plotSieveCurve
}
}
//----------------------------------------------------------------------
TetSolidList PRTC {
Write PartMeshDataFile Prism-Hex.mrs
}
//----------------------------------------------------------------------

Total number of particles is: 2317

2) Second method with the same material:
//----------------------------------------------------------------------
LdpmModelGenerator PRSM {
Material CONC
Seed 34565
GeneratePrism {
Dimensions 75 mm 75 mm 265 mm
//ConstantSpacedParticlesOnEdges
//MaximumIterations 400000 // default 10000
}
WriteLdpmMeshFile prismLdpmModelGen.mrs
}
//----------------------------------------------------------------------

Total number of particles is: 7462

3) Third method is:
//----------------------------------------------------------------------
HexSolidList Cube Geometry {
Generate {
Block {
dimensions 75 mm 75 mm 265 mm
Elements 20 20 60
}
}
}
LdpmModelGenerator MODL {
Material CONC
Seed 34565
GenerateFromFineHexMesh {
HexList Cube
//[ Examine ]
//[ Split ]
}
WriteLdpmMeshFile prismFineHex.mrs
}
//----------------------------------------------------------------------

Total number of particles is: 2865

I was wondering if this is something reasonable. Thanks.

[zhouxinwei]

Instead of posting pieces, please attach the input files you used so I can surely reproduce your results. Thanks.

[Roozbeh]

Three Gen files are attached. Thanks.

[peless1d]

I reconstructed the 3 input files before seeing the follow up postings, where Roozbeh posted the complete input files. For case 2 and 3 I get exactly the same mesh, with same number of nodes and same connectivity. For all three cases there are warnings (please read output files): for case 1, it was able to place 305 internal particles out of 1804, for cases 2 and 3 it was able to place 850 internal particles out of 1858. That's the main reason for the difference. The code is not able to insert all the particles that it wants to because the ratio between Max and Min aggregate is 19/4.75 = 4, this ratio creates a very large number of particles for a given volume and the smaller ones cannot be placed. In general we use a ratio of 2, and even for 2 not all particles can be placed. Case 1 and 3 are discontinued. Please use input commands as in case 2. Daniele

[zhouxinwei]

Just add one comment to Daniele's reply: for case 2 and case 3, the number of internal particles are the same, the difference you see in total number of mesh points is due to the external surface, try to plot them.

In method 2, MARS generates the external surface by calling Triangle, which accounts for randomness; in method 3, MARS uses the external surface from the hex mesh, which is regular. In case 3, you are using a much finer external mesh than case 2, that's why you see the difference in total number of points in the mesh.

[peless1d]

I just realized that when I set up the input for case 3, I actually used the commands for case 2. That's why I was getting exactly the same results. Xinwei, thanks for clarifying this issue. Daniele.

[Roozbeh]

Thanks Daniele and Xinwei for your comments. I have two questions:

1) Why methods 1 and 2 place different number of particles inside the specimen?

2) How fine should one make the surface mesh in 3rd method? What is the realistic one? Should changing resolution of surface mesh change the results?

Thanks.

[zhouxinwei]

Thanks Daniele and Xinwei for your comments. I have two questions:

1) Why methods 1 and 2 place different number of particles inside the specimen?

2) How fine should one make the surface mesh in 3rd method? What is the realistic one? Should changing resolution of surface mesh change the results?

Thanks.

1) the seeds are different. If the seeds are the same, we should get the same internal particles. It is recommended that we use LdpmModelGenerator.

2) the resolution of Hex surface mesh should be similar to the aggregate size specified in the mix design. The internal particles are decided by the mix design, not the hex surface mesh. However, the nodes of the hex surface mesh will be combined with the internal particles to generate the final LDPM mesh (by TetGen).

There was an issue where the maximum iteration number is not working when MARS is placing internal particles. This issue is fixed in version 2016.5.00.

[Roozbeh]

I tried with the latest mars version and this problem is solved and all methods work consistently.