search

CMU-CBML / HybridOctree_Hex

Automatic hexahedral mesh generation tool
BSD 3-Clause "New" or "Revised" License
34 stars 6 forks source link

Unable to generate results #4

Open zyyue99 opened 4 months ago

zyyue99 commented 4 months ago

Hello, thank you very much for providing the code! Can I ask you some questions about how the code works? May I ask why when I run the input model provided, there is no result for a long time, and I am forced to terminate the program because the memory is full?

gt2001 commented 4 months ago

You can set numbers in C_THRES and H_THRES in initialization.h bigger to make the code faster. This will leads to coarser mesh, but it may leave holes in the mesh if the object geometry has thin layers.

Regarding the code speed, I am writing a new version, which is 1000 times faster than the current version. It is expected to come out in September.

zyyue99 commented 4 months ago

You can set numbers in C_THRES and H_THRES in initialization.h bigger to make the code faster. This will leads to coarser mesh, but it may leave holes in the mesh if the object geometry has thin layers.

Regarding the code speed, I am writing a new version, which is 1000 times faster than the current version. It is expected to come out in September.

Thank you very much for your reply! But when I try to increase numbers in C_THRES and H_THRES, there is still no result for a long time, and I will always be in the hexgen.ConstructOctree() phase. May I ask how the data of more than 200 seconds of mesh generation given in your paper is achieved? In the implementation, I only changed the "model.raw" in the main.cpp file to the specific input model name. Is this correct? Thank you very much.

gt2001 commented 4 months ago

It depends on many factors (e.g., project settings, computer performance). Sorry that I forgot the parameter I set for that model ---- but C_THRES and H_THRES should be large. Also, please remember to turn on all the optimization options in the project setting interface, including but not limited to O3 optimization, OpenMP parallel computing, and compiling with Release mode instead of Debug mode. If you still couldn't solve the speed problem, you can Watch the repo, a much faster and more robust version will come out in two months.

zyyue99 commented 4 months ago

Alright, I will continue to try adjusting the parameters. Thank you very much for your patient responses, and I look forward to a faster and more robust version in the future!

BimwerxNZ commented 1 month ago

Hi, thank you for sharing this. I've tried several fairly simple models (linear extrusions of steel profiles) and it never seems to finish. I have tried a few combinations of settings, but do not understand the C_THRES and H_THRES values. Could anyone shed some light on how to determine the optimum values for these?

gt2001 commented 1 month ago

Thank you for your report. I have recently received numerous reports indicating that the code is unable to execute normally. It appears that version 1.3 has encountered some issues. Would you be able to try using version 1.2 instead? Additionally, I would like to remind that the current code is unable to preserve sharp features (i.e., dihedral faces) that are created by extrusions. This functionality will be incorporated into the later version that I am currently working on.

BimwerxNZ commented 1 month ago

OK thank you, I'll try that for now :) Looking forward to trying your new version in future!

BimwerxNZ commented 1 month ago

I've tried v1.2, it runs through but only produces the modifiedTri.vtk and no other files. My .raw file attached if you wanted to test yourself - no obligation off course, just for information if it helps. model_tri.zip

gt2001 commented 1 month ago

finalMesh.zip This is the mesh of the model you uploaded. The minimum scaled Jacobian is 0.6, but the sharp features cannot be preserved. I am uncertain why it only outputs a triangular mesh file. It is possible that the memory usage exceeded the computer's limitations (I ran the code on a 32 GB memory system, and it always uses the whole memory). I will address this issue in future versions of the software.