Merge nodes after mesh creation

hi,

currently a new feature of advanced mesher (Gmsh) in extrusion method is limited to single part. A compound part with multi-axes extrusion is not yet supported or implement. However, there’s workaround to get similar result by little effort only. I do small trial by iteration in “Scale factor” options of extrusion method.

below several figures of example, some crucial steps required in the last steps is merge nodes. Unfortunately, this feature is not yet available in PrePoMax and i have done using external programs i.e CGX by import/export of INP files.

Hopefully this merge or connect nodes feature will be available in the next or future versions, an option to connecting of coincident nodes by constraint using *Equation seems can be useful also.

regards,

Merging of nodes was suggested before, and I will definitely add it before adding any new features. This will happen right after I fix the features I am working on now. This will be added as a special step after the meshing is done. In this case, after the node merging is done, it cannot be easily undone. One problem will be what happens when one of the parts is remeshed.

So, the idea of *Equation as an alternative option is quite nice. I could create a constraint of type Surface-Surface Equation connecting all degrees of freedom. In this way remeshing will not affect the connection.

1 Like

Undone of merged nodes is not usual. Normally you always rebuild your mesh from the part meshes and merge it again. If you remesh only one part then you merge only one part again.

In the following example I made a mesh of only the half blade 1 connection. Then mirrored the mesh to get a full blade 1 connection. And then copy+rotate the mesh to blade 2 and blade 3. And after that you can merge the nodes between the parts. That saves a lot of mesh time and you get identical meshes.

For more information see https://fatigue.pro/2023/09/wind-turbine-hub-calculation-in-ansys/

So with a merge option it is important to get also:
a mirror mesh option
a copy+rotate mesh option
a copy+move mesh option
Otherwise you have still to do that outside of PrePoMax.

Abaqus has the Merge/Cut Instances tool, it can operate on mesh and/or geometry:

Yes, I also think that undo is not necessary, but PrePoMax now combines options to work on a mesh created from geometry and “native” mesh without geometry. If two parts are merged by nodes and one of the parts is remeshed, PrePoMax first deletes its mesh - deletes a part in the FE Model tab and then creates a new mesh and imports it again. In this procedure some problems might occur.

It is possible to use copy-rotate and copy-move using the Model → Part → Transform. Adding mirror and merging nodes is missing for now.

I see Abaqus creates a new part. That is good idea but in PrePoMax it will not fit nicely in the workflow. After remeshing one of the parts and merging it, a new part will be created again with a new Id and all mouse selections will be invalid. So keeping both parts is probably better for PrePoMax.

Abaqus indeed creates a new part and adds it to the assembly while suppressing or deleting (depending on the user selection) the original instances (not parts, those are always kept). It maps sets and surfaces as well as section assignments to the new part/instance.

many thanks for interesting at implementation also.

it could be okay when ‘undone’ is not possible, part and mesh setting previously set is available and mesh is easy to regenerates. Some problem may occur due to not exactly the same position of coincident nodes, most approach used is averaging of these two nodes coordinate.

merging (new single node) and connecting by constraint (two nodes by *Equation) has each advantages, constraint can eliminate result plot averaging of different material connected ideally (composite). Hopefully, both method are implemented and available, probably in Tree’s menu when user doing merge parts in FE model.

2023-11-22 16_58_22-PrePoMax v1.5.5 dev   C__Users_user_Documents_extmult.pmx

indeed, constraint using *Equation is useful for such condition. Some disadvantage is in non-conformal mesh, since required precalculate of weight factor or ratio based on distance to master nodes. Maybe for the future version of PrePoMax it will be implementing.

Using compound parts and the setting Split compound mesh, one can get coincident nodes for all parts in the compound. Then, using the *Equation connection, one could get non-averaged fields for different materials.

1 Like

Yes, you are right. Copy-rotate and copy-move are already in PrePoMax. And it works as expected. :+1: Maybe it make sense to add a number of repetitions. So it can be copied more than once.

1 Like

an opposite action to ‘unconnect’ a meshed part can also be useful, specifically for structured mesh.

before (three parts)

2024-02-17 22_17_27-

after (five parts)
2024-02-17 22_16_29-

You can do this by creating an element set for the selected parts and convert them to part. Then use “copy and translate” and delete the previous part.

well, there’s workaround by duplicated several times and placed at the same exact location. Each of them need further step to remove any unwanted element. These approaches have a bit different and uncommon in FE preprocessor, need more steps and working carefully also.

2024-02-17 23_28_33-

In example, simple cube need to be separated at top and bottom region, the ‘unconnect’ methods only need one step by creating part based on element set (convert). However, using duplicate and remove methods can have two steps, element numbers also can risk to overlaps when user inconsistent in selection.

*edited
but i can achieve after a little bit confused and carefully selection, may this approach is temporary methods to separating meshed part.

only want to confirm, this feature currently available in latest versions (2.18) made it extrusion based approach in meshing become more convenient than using tied contact or tie constraint, Option in distant tolerance also, next or future probably using *Equation for took advancement in sliding surface or non-average results. Many thanks.

*edited
previously, some edge division mismatch even mesh refinement has been defined. It’s required to apply in surface instead.

2 Likes

Closing as completed in v2.2.0.