about MPI

[xibaiw]

Hello!
For the same input file, there has something wrong with the MPI calculation or other questions.


The code works without problems in serial Julia execution, but fails with MPI. Is this a known issue, or could it be a bug?

[JTHesse]

Hi @xibaiw this seems to be a bug. I'll take a look if you can provide the slab.yaml file.

[xibaiw]

Hello Mr. Hesse, im already sent my input file to your email!

[JTHesse]

Perfect, I'll look into this.

[xibaiw]

sure，I will send my package to you within a few days！ 来自Coremail ----- Original Message ----- From: "Jan-Timo Hesse" ***@***.***> To: "PeriHub/PeriLab.jl" ***@***.***> Cc: "JI zhe ying" ***@***.***>, Mention ***@***.***> Sent: Mon, 28 Apr 2025 06:32:10 -0700 Subject: Re: [PeriHub/PeriLab.jl] about MPI (Discussion #260) Hi @xibaiw this seems to be a bug. I'll take a look if you can provide the slab.yaml file. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: ***@***.***>

[JTHesse]

I checked your files and yes there seems to be an issue with the synchronization of the MPI cores for your input mesh.
In order to make it work, just reduce the horizon. Currently, the number of neighbors is way too high for an efficient simulation.
A horizon of 0.05 works for me.

[xibaiw]

okay, I‘ll try it with a smaller horizon. thanks a lot!

[xibaiw]

By the way, I wonder about the derivation of interblock damage parameters IB 1_2 and IB 2_1. Could you clarify the method? @JTHesse

[JTHesse]

As shown here, bonds are bidirectional, so a bond could be broken in one direction and intact in the other direction. Depending on what you want to simulate there could be a scenario where this is relevant. In most cases the critical value of IB1_2 and IB2_1 should be the same.

[xibaiw]

In PeriLab, how are the mechanical parameters (e.g., micro-elastic modulus) of interlayer bonds in laminated composites determined? Does the solver automatically adjust these parameters based on adjacent material layers, or are they defined via homogenization approaches (e.g., volume-averaging methods) to account for multi-material interactions?

[CWillberg]

"In PeriLab, how are the mechanical parameters (e.g., micro-elastic modulus) of interlayer bonds in laminated composites determined? "
Each material point has a defined material. That means point 1 is has material 1 and is connected to point 2 with material 2. The bond 1-2 has material 1 and bond 2-1 has material 2. This leads to an unsymmetric stiffness matrix (if you build it).

You have to understand that bond 1-2 and 2-1 only the points and the geometry. One bond can be evaluated with bond based formulation the other with PD Solid. One can be broken the other not. That's a direct consequence of the PD theory.