Bullet-Proof Modeling

[zissou]

A few years ago, a gentleman gave a presentation at SE Summit about "Bullet-Proof Modeling". The concept is to eleminate the vertical "parent/child" model tree and make every feature, etc independant. This uses a plane and an associated sketch to drive every feature. Also some custom variables to get key dims that cross features... for example; a height var to drive the height of the extrusion and the plan distance, etc.

Has anyone tried this method - or does anyone use it on a regular basis?

I'm looking to spur a little discussion about it's merits or lack of merits here.

Thanks.

 

[swertel]

In a parametric 3D modeler, you'll never get away from the parent/child relationship. The design intent takes precedence.

But, the fewer parent/child reletionships the more robust your model will be. The definition of robust in this case, is that when a feature high in the tree gets modified, lower features are less likely to fail.

Good modeling practice dictates that the designer create as few parent/child relationships as possible while maintaining design intent. That's how you find your balance.

--Scott

For some pleasure reading, try FAQ731-376

 

[donyoung]

Hi,

this approach is done by non history modellers like 'IronCad'
but here is an 'excerpt' from another source:

[cite]
The downside to a non-history based modeler is the inability to design any intellegence
into the features such that they behave in an expected way based on other features and
their order. For instance, if I have a block and I want to shell it with one face removed,
either system can do that easily. If I want to go back after the fact and add a hole
through the bottom up to the face removed with a boss around it, the history based modeler
can accomplish this simply by placing the hole feature between the block feature and the
shell feature and it is done. In the non-history based modeler, a boss would have to be
manually placed, and then a hole placed through it and the bottom surface, or the shell
deleted, the hole placed, and the shell redefined, thus no dependency between features.

Traditional history based models where difficult to change if extensive changes were needed
(because of the dependency), and imported geometry was typically unmodifiable (because it
had no history). That is no longer the case. Systems like Solid Edge give you the ability
to use the history based modeling along with tools (direct edit) that allow modification
without disturbing the history or if the history does not exist.

Although it is hard to explain the differences, industry has leaned heavily to the history
based modeling approach based on the leading CAD systems in use today.

[/cite]

dy

 

[zissou]

Good reference, Mr Young.

With SE being a history-based modeler, if you use this "bullet-proof" technique - it seems as if, with a bit of work, you're reaping the benefits of both historical and non-historical modeling. When you start using the variable table to control the model - it seems as if you could really see some benefits.

 

[psm38]

That sounds a lot like David Breitstein's philosophy and one I follow religiously. The idea is to have only as many parent/child relationships that are truly needed when creating a model or assembly. This makes your models and assemblies more robust and predictable. Using variable tables to link features is by far the best way to go.

 

[beachcomber]

I'm not totally convinced about this, although I can see the benefits.
At the moment I'm trying to re-order a complex model to make it easier to modify in the future and to give a correct one or two features. It's a real nightmare - but that's because of the way it's modelled rather than the parent-child relationships.
Most users I've come across barely know of the existence of the variable table, let alone how to use it!!!

bc