# V2.3.dev: Load assignment on "internal" faces gets lost after merging parts

**URL:** <https://prepomax.discourse.group/t/v2-3-dev-load-assignment-on-internal-faces-gets-lost-after-merging-parts/2710>\
**Category:** General Questions\
**Created:** [May 17, 2025, 7:25pm UTC](https://prepomax.discourse.group/t/v2-3-dev-load-assignment-on-internal-faces-gets-lost-after-merging-parts/2710 "2025-05-17T19:25:31Z")\
**Posts on this page:** 1\
**Showing post:** 9

<div class="post-metadata">

**Author:** ![OpenSourceFEA](https://avatars.discourse-cdn.com/v4/letter/o/4da419/32.png) [@OpenSourceFEA](https://prepomax.discourse.group/u/OpenSourceFEA)\
**Post date:** [August 18, 2025, 11:43am UTC](https://prepomax.discourse.group/t/v2-3-dev-load-assignment-on-internal-faces-gets-lost-after-merging-parts/2710/9 "2025-08-18T11:43:29Z")

</div>

Hi Matej. I am facing different issue.

My model:

I have a tubular structure where I apply the bolt pretension at a cross section just above the threads. Since it is a complex part, I use second order tet mesh with a boundary layer on both sides of the cross section.

Pre 2.3.0 I was able to select the surface and merge the part back into one and was able to run the simulation successfully. But the newer versions, I can only select the FE based surface like you suggested here. I set the pretension direction “auto-compute”.

Geometry based surface selection I get the pretension direction calculated by calculix

`*INFO: pre-tension normal for surface BOLTLOAD-1`  
`-4.5946230207038368E-018 1.6354628026558331E-016 -1.0000000000000000`

which is purely in z-direction (x and y values are very small and negligible for my calculation)

FE based surface selection I get the pretension direction calculated by calculix

`*INFO: pre-tension normal for surface BOLTLOAD-1`  
`0.54178583594546259 -0.84051657209651520 0.0000000000000000`

I am completely getting pretension in some other undesired direction.

So I did two things

1. Either set the pretension direction manually in z direction for FE surface selection
2. Or leave the split parts unmerged for geometry based surface selection

I am still getting some convergence issues while trying the workarounds which was not present in the older versions. The older version was more convenient, is it possible to bring it back?

---

_[View the full topic](https://prepomax.discourse.group/t/v2-3-dev-load-assignment-on-internal-faces-gets-lost-after-merging-parts/2710)._