MOOSE case with strange parallel output

[刘骁]

I’m Liuux from Github who proposed an issue “Something Wrong with Displaced Outputs”. Under that issue you wrote that the strange output for displaced mesh is caused by nodes which are on two boundaries. Then you gave me an input file that adds a new nodeset to solve that problem. I’ve tested it and its result looks good. However, those days I’ve tested another 3-D problem with displaced mesh, strange results came out again and I don’t think it’s caused by nodes on two boundaries. the input files, mesh files, and results . I put serial and parallel results separately into two folders, because the serial computation outputs are good while the parallel computation outputs are strange ,and the defective elements occur on the middle of the block, with none nodes on two boundaries. So I think something is wrong with parallel output. In addition, the solving process is perfect no matter we use serial or parallel schema, Only the output block goes wrong. To prove this, we can set ‘use_displaced = false’ in the output block and load the displacement to the initial mesh using Paraview, then the results will be same both for serial and parallel computation. I don’t know how to solve this problem. Could you please have a look at the cases and give me some help? 

 

 

 

 

[Cody Permann]

Many developers on this list are very busy. You've put together a lot here but to make this task more palatable, could you post a link to your repository instead? Generally it's much easier to clone a working repository than it is to download random source files and place them in the right locations to get started. Do you have an image of your defective case?

On Sun, Jun 25, 2017 at 9:27 AM 刘骁 <liuxi...@163.com> wrote:

I’m Liuux from Github who proposed an issue “Something Wrong with Displaced Outputs”. Under that issue you wrote that the strange output for displaced mesh is caused by nodes which are on two boundaries. Then you gave me an input file that adds a new nodeset to solve that problem. I’ve tested it and its result looks good. However, those days I’ve tested another 3-D problem with displaced mesh, strange results came out again and I don’t think it’s caused by nodes on two boundaries. the input files, mesh files, and results . I put serial and parallel results separately into two folders, because the serial computation outputs are good while the parallel computation outputs are strange ,and the defective elements occur on the middle of the block, with none nodes on two boundaries. So I think something is wrong with parallel output. In addition, the solving process is perfect no matter we use serial or parallel schema, Only the output block goes wrong. To prove this, we can set ‘use_displaced = false’ in the output block and load the displacement to the initial mesh using Paraview, then the results will be same both for serial and parallel computation. I don’t know how to solve this problem. Could you please have a look at the cases and give me some help? 

 

 

 

 

--
You received this message because you are subscribed to the Google Groups "moose-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to moose-users...@googlegroups.com.
Visit this group at https://groups.google.com/group/moose-users.
To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/4a5fcfb6.4da7.15cdfdd289f.Coremail.liuxiao8146%40163.com.
For more options, visit https://groups.google.com/d/optout.

[liuxi...@163.com]

The cases I uploaded can be run using moose_test-opt , very convenient.And I put the .i file and mesh into the same folder so you just need place the folder in any location and run the iput file using moose_test-opt directly,much earier than clone a whole repository .In addition, I put the output files into the same folder so you can also open them using paraview or ensight without running the case.Here are pictures show the difference between parallel and serial results for the same case,the first one is serial result,the second one is parallel result:

在 2017年6月26日星期一 UTC+8下午10:35:02，Cody Permann写道：

[Slaughter, Andrew E]

I have an initial reaction, which might be relevant. If you have variables "disp_x, disp_y, etc." Paraview and Peacock will automatically apply these to the mesh. Thus, if you also use "use_displaced" it will output the displaced mesh and also apply the displacements, which of course is meaningless.  In general, there isn't much use for the "use_displaced" flag in the Outputs block.

To unsubscribe from this group and stop receiving emails from it, send an email to moose-users+unsubscribe@googlegroups.com.

Visit this group at https://groups.google.com/group/moose-users.

To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/ec98ed43-d1c8-4188-aac7-51e9a72c165f%40googlegroups.com.

[liuxi...@163.com]

I don't think so. If the strange results are caused by Paraview or Peacock, the parallel and serial results should be the same.The  "use_displaced" flag in the Outputs block is 
necessary if we want to know the varialble values at any point on the displaced mesh.
在 2017年7月5日星期三 UTC+8下午10:20:34，andrew.slaughter写道：

[Derek Gaston]

"use_displaced" in the Output block is not usually used/needed.  Both Paraview and Peacock have the ability to "programmatically" displace the mesh using the displacement fields.  That is the "normal" way to do things.

I will repeat (even stronger): "use_displaced" in the Output block is almost never used / needed.  It is for VERY specific circumstances.

Derek

To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/269990b8-c759-498a-a4ba-90c3ac5a9037%40googlegroups.com.