Working with large models
Working with large models
Although bigger models will typically need more RAM, you should note that it is the volume of model data and results data created during analysis and design that generates the demand for RAM, so by controlling the volume of data you are able to influence the speed of solution.
Your modelling and design choices can affect the volume of data produced; some of the more significant of these choices are discussed below:
Don't mesh concrete slabs in 3D Analysis
Meshing is not necessary in 3D Analysis for traditional Beam and Slab models as these can use FE decomposition instead.
It is also not necessarily required for Flat Slab models, however, it does become required if you have a transfer slab. It is not critical unless the slab is part of the lateral resistance system.
By default the Mesh 2-way slabs in 3D analysis option that controls this is not checked at any level; you should only check it at a given level when you have good reason to do so.
Using coarser shell meshing in large models
You should review and consider adjusting the mesh parameters as the defaults can often be conservative.
- Using coarser mesh parameters has no impact on sway or modal frequencies.
- If you are not concentrating on slab design you can use very coarse meshes.
Ultimately it is your responsibility to be comfortable with the level of mesh refinement applied to the model. However we would definitely recommend using a courser mesh during design development and then perhaps consider refining a bit more at final design stage.
Using coarser semi-rigid diaphragm meshing in large models
If you have employed semi-rigid diaphragms and are experiencing performance issues you should review and consider adjusting the semi-rigid mesh parameters.
It has been observed that using refined semi-rigid meshes generally has negligible impact on results and only has the effect of slowing down analysis and increasing memory requirements. It is suggested that the default mesh parameters may be more refined that is actually necessary.
Limit the number of load cases and combinations
You can control the number of combinations created when running the Combination Generator.
In particular you should consider limiting the number of wind load cases and combinations.
- Don't add wind loading during the initial design development.
- Don't activate pattern load cases and combinations until you need to, probably only at the final design stage.
Alternative design approach for large models
For large models, rather than running Design All you may be able to save time by running Analyse All instead and then run a selective design, such as:
- Design by level
- Design by frame
- Design by group
- Design by sub-structure
- Design by member
Effective use of Auto Design
Although for the first design run you might choose to use select bars starting from Minima, on subsequent runs it is generally more efficient to use select bars starting from Current; this will run a check on the current steel provision and if inadequate, it will automatically re-select new steel bars to pass the design.
Check Design can also be very effective - you can turn off the Autodesign and then manually deal with any fails.
Design members for FE chasedown analysis results
This is set in the design options for concrete and by default it is checked on for beams, columns and walls. However in a traditional Beam and Slab model it may not actually be necessary; it is generally not critical unless you have some unusual transfer level challenge. In large models you should therefore consider unchecking it.
Re-design columns (or beams) using previous analysis results
If you change the size of a member you can try out its design without being forced to re-analyse.
Basically you can make any edit you want that does not change the number of stacks in a column, (or spans in a beam); although the analysis results will be marked as out of date, you can still do a design for the changed member based on the old analysis results.
So using the list below as examples design can still be done in all the cases noted:
- Changing column (or beam) size but retaining shape - designs ok (but see 3 and 4).
- Changing column (or beam) shape - designs ok (but see 3 and 4).
- Making column (or beam) smaller so that previously attached members no longer attach - this changes number of stacks - design beyond scope.
- Making column (or beam) larger so that previously un-attached members now attach - this changes number of stacks (spans) - design beyond scope.
- Adding / editing / deleting beams that attach to a column - design remains possible up to the point that it affects number of stacks - OK
- Adding / editing / deleting flat slabs that attach to a column - design remains possible up to the point that it affects number of stacks - OK
- Adding / deleting stacks or levels (or editing level properties) - Adjusting Levels designs ok because number of stacks is the same. If you add or remove stacks then design is beyond scope.
Tekla Structural Designer has a number of features for organising the model than can each be used to increase efficiency:
- Grouping - one design is applied to all members in the group.
- Sub-structures - allow you to focus on specific areas of interest.
- Duplicate levels - generally save modelling time and reduce the volume of data.
When using duplicate levels, you can achieve further efficiency by designing slabs for a fine mesh at one level only, and then check the slabs at duplicates of the level using a courser mesh result.
Our general advice for duplicate levels is:
- For preliminary design set a coarse mesh for entire structure
- For final design where there are a lot of duplicate levels
- possibly refine the mesh used for the entire structure a little
- but for each set of duplicate levels, select one and adjust the relevant sub-model parameters to get a finer mesh.
Do not model every little architectural detail - especially not things like small holes in slabs and walls.
Take control (get it right at the beginning!)