Add test_2d_milling_2d_dat_cutter_planar example - #1043
Conversation
crb116923
changed the title
Hello Professor Hu, I have corrected the error that occurred during the gcc build process. Thank you for your time.
|
@crb116923 Now, as your case can be compiled. We can have discussion based on the new case. Could the case run? |
|
1.0023.bmp First of all, this case study can run. The problem is that only it runs partly, and I don't know why. Thank you for your time and guidance. I need you detail the issue so that we can discuss. |
|
@crb116923 |
|
2026_04_03_report.pdf |
|
@crb116923 I have read your report. Did you have tried to tune some parameters, such as time step size (by cfl number) and dissipation factor of the solver? |
|
@crb116923 I notice that you are using very hard material. did you try softer material for this problem? |
|
Hello, Professor Hu, |
|
@crb116923 your cutter is not moving. please commit a valid setting up. |
Sorry. I found the place for motion. |
|
@crb116923 I have reproduced your results, there is a numerical instability. We can have a online discussion after you have collected the parameter study. |
|
Professor Hu, |
I will be free on Friday 3pm. |
|
Hello Professor Hu, Thank you for your time. 曹庭赫 邀请您参加腾讯会议 点击链接入会，或添加至会议列表： #腾讯会议：847-365-077 复制该信息，打开手机腾讯会议即可参与 |
| auto &plate_update_pos = | ||
| main_methods.addStateDynamics<fluid_dynamics::UpdateParticlePosition>(plate); | ||
|
|
||
| auto &plate_correction = |
There was a problem hiding this comment.
the bug is that we have forget the parameter for free surface correction parameter. Use about 0.5 will solve the numerical instabilty.
|
Professor Hu, here is the result. First, the parameters used in the case are a rotation speed of 12 revolutions per minute and jelly; this rotation speed is far below the real operating conditions. Three constraints were set. No debris was generated. Second, if replaced with an aluminum block, it cannot run under the same conditions. test_1.mp4 |
Could you detail the cases which is you said "can not run"? |
|
Second, if replaced with an aluminum block, it cannot run under the same conditions. test_2.mp4
|
|
could you identify the material property range which can run. |
|
Analysis.pdf |
@crb116923 please rescale your dimension system so that there is no absurd value like 2.7 m displacement for a drilling case. |
|
Analysis.pdf |
|
Hello Professor Hu, I didn't bother you last Friday because of the Labor Day holiday. 2.7m represents the case with two-sided constraints; 0.036m represents the case with three-sided constraints. If you need verification videos for all nine cases, I can provide them. |
|
Analysis.pdf First, I'll submit a version based on this. Second, I'd like to ask you, does scaling to 0.001 to 1000 include normal parameters like density and mass? Third, the pressure of studying and changing careers is exhausting me. As I mentioned before, please feel free to ask me any questions; I'll continue learning. |
Updated material properties and simulation parameters for the 2D milling test. Added bottom fixed region and adjusted velocities and time settings.
**Add nine materials, and use meter-based SI units for the dimensional system.**
|
An idea This idea is wrong because there is gunpowder. |
|
Dear Professor Hu, I have a new idea to share with you. |
|
@crb116923 I think that is too much for a master thesis, I think that you may just need a make the present model can generate meaningful results. |
|
ok,I am modifying the code now. |
|
@crb116923 actually, you can work on your project based on SPHinXsim repository, there is a milling case works already. |
|
Professor Hu, I noticed that the material currently used for the Taylor Bar case is Hardening Plastic Solid, while my case uses J2 material. Second, I tried modifying the code API, but it never worked. |
what do you mean never worked. Please specify the issue so that we can discuss it. Words like never worked have no help at all. |
I modified the code by cloning the case based on the Taylor version and re-uploaded the relevant files. The code now builds successfully, but it still reports an error when running the case. Therefore, although the compilation succeeds, the case cannot actually run. In other words, the code builds, but it still does not work at runtime.
I understand that this may not be a simple compilation error. It seems more likely to be a runtime issue related to SYCL/kernel/device-side execution, or possibly caused by some inconsistency in the case configuration, kernel registration/linking, or runtime environment. The runtime error message is shown below. How do you see this issue? Which parts would you suggest I check first? |
You need identify which line of code leads to the crash. |
|
Professor Hu, first, I can run the relaxation successfully on the CPU using Second, I am not sure how to correctly make the corresponding kernel included in the SYCL device binary. |
But the code you have referred runs on GPU without issue. So you need check where is the difference you have changed leads to the issue. |
|
Professor Hu, first of all, I did not modify the code in the relaxation stage. Second, the version currently installed on my computer is the latest library cloned from:
Third, with the latest version, my case can be built successfully, but it cannot run on the GPU. I feel a bit frustrated because I did not change the relaxation-stage code, but this runtime issue still appears there. Do you mean a test case from the repo is not able run on GPU. It is bit strange as they are all sycl cases have tested with in github atveach pull request. |
|
Professor Hu, First, I have already run a considerable amount of simulation data before. Would it be acceptable for me to start writing a first draft of the paper based on the existing results? Second, the latest version of my code currently has a runtime issue. Specifically, the GPU kernel path for |
Do you mean a test case from the repository not able to run on GPU or your code not? |
Sure. If you think in that way. |
yes，this test case from my repository can build but not run on GPU. |
Thank for your reply.I will try to write immediately |
|
MA_CRB.pdf This is the first draft of my paper. |
|
Professor Hu, First, I have now reinstalled the NVIDIA GPU driver, and the problem has been resolved. Second，When would you be available to meet in person between August 4 and August 11? |
Basically, I am free on Friday afternoons. we can meet in person then. |
|
Professor Hu, this week I followed the Taylor bar example and changed the material model from J2 to HardeningPlasticSolid. I tested three materials. The mild steel block and aluminum simulations could only run partially, while the jelly showed the rotational behavior below. jelly.mp4 |
HardeningPlasticSolid
|
@crb116923 please come to university on Friday afternoon so that we can have in person discussion. |
Professor Hu, see you on August 7th. Thank you for your time |
|
Dear Prof. Hu, I have some good news. The code can now be successfully built and runs correctly on Linux. I have tested two constitutive models (J2 and HardeningPlasticSolid) together with both the fluid integration algorithm and the solid integration algorithm. Based on the current numerical results, the fluid integration algorithm provides more stable and physically reasonable results for the present milling case, while the solid integration algorithm still exhibits some numerical issues for certain materials. I am currently investigating the underlying reasons. |
2 participants
Professor Hu,
I made a mistake with the upload path before. What was the result this time? Thank you for your guidance.