Fillets Make Your Parts More Expensive [video]
youtube.com
youtube.com
This video should be required viewing for designers who send plans to machine shops. It really helped fill a ton of gaps in my knowledge, and I’ve been designing for over a decade, with a machinist best friend. Excellent presentation from both perspectives.
Here are a few of the real heavy hitting insights:
From day one, I would list off the expensive features. Each click adds very significantly to the part cost.
Loft, Sweep, Fillet, Blend, and friends...
Any curve of a degree greater then conic in 2D or 3D wireframe sketch or curve mode.
Just that short list saves a ton. And most design requirements are easily met without them.
Now, if it is needed, yes! By all means, use the feature. That is what it is there for. But if it can be avoided, the resulting part is much cheaper and can be often easily made with both more methods and the simpler, cost effective methods.
Other things I would include:
When recreating existing parts, parse design intent first, then gather dimensions, then model. You will need both the least amount of dimensions and or can work more easily with the ones you can acquire easily.
On new designs, build around a local origin, leveraging symmetry where possible. Bonus: locating features relative to that origin, or one transformed from it, means you can often introduce serious topology changes without having a ton of broken references and conflicting constraints.
Put simply: avoid the sketch on face feature to improve on model flexibility.
Should your CAD system support boolean cut, join, intersect, add, partition, or most of those, consider building core shapes relative to a common origin, then perform booleans to arrive at desired shape. You gain huge flexibility in topology changes and will suffer few to no dependencies on feature edits. It can take a bit longer to build this way. If time constraints are intense, reserve this technique for models more likely to see significant change.
In the assembly context:
Use coordinate systems transformed from root ones to establish key points of contact and articulation.
If necessary, design in place to sort where those are, then roll back history, place the coordinate systems, then build relative to those.
Each constrained item has 6 degrees of freedom to resolve. On large assemblies this can grow impractical to solve due to CPU sequential compute limits. Nobody builds really big things fully constrained. This is why.
Partially constrained assemblies are effective. Constrain key mechanisms and or critical items. Position all others and use sub assembly structure to further improve overall efficiency and productivity. Like the model technique above, reserve constraints for areas that need to adjust dynamically, position and check the rest.
There are a lot more, but those save a lot of dollars, both up front in design as well as revisions, derivative designs, and manufacturing.
If you can, choose a hybrid CAD system that supports all four of these, and or as many of these modeling modes as possible:
Non history, non parametric (direct modeling)
Non history, parametric (Siemens calls this variational)
History, non parametric
History, parametric.
Each of these has serious advantages depending on what you are doing and whether it requires you make use of data that originated outside of the system you are using.
As of my last check, NX and CATIA fully support all model paradigms. Many support two or three. A few only do the last one. (Solidworks*)
Often, when originating designs, history, parametric is all you need and powerful. The more you must work with others and their data to either manufacture or make derivative designs, the more important those other paradigms are.
That's it!
If nothing else, realize modern CAD tools have a ton of features. That does not mean they make sense to use.
*Nothing wrong or bad about Solidworks. It is a fine tool with a lot of great features. It does however present difficulty when working with outside data, and is a great example for the point I made here.
Where a part can be expressed with some combination of analytical geometry eg:
Cone, torus, sphere, trapezoid, cube, cylinder, etc... we have long established means and methods of production available and they are generally efficient
; however,
shapes derived from both curves of a higher degree than conic, and or paths that are themselves non analytical in nature, or employ multiple profiles blended, tend to become general b-spline geometry types which require higher order means and methods.
Secondly, inspection of these higher order parts often is more expensive as well. Simple analytical means will not suffice. Templates, CMM machines, point sampling, scanning are all likely solutions, all typically more expensive and computationally heavy requiring computer assistance today. (We used to make these sorts of things without computers, and doing so required serious skill not always available.)
A simple fillet as opposed to a chamfer, requires more machining given more care and requiring higher order path planning software and or operators able to employ that software to machine successfully.
These same dynamics apply to the geometries I mention above and the cost tends to multiply across multiple manufacturing centers or cells as well as downstream operations, such as inspect.