Workshop practice
Why your roughing strategy cuts through islands
The model is correct, the geometry is correct, yet the toolpath crosses something it should not touch. It is not a software fault: that strategy is not looking at what you think it is.
You program a 2D roughing operation. Select the outside contour, define the depth, simulate… and the tool calmly crosses a boss that is there, modelled, visible on screen and part of the same solid.
The natural reaction is to suspect the model. You check it; it is perfect. Change a parameter; same result. Try another 2D strategy; still the same.
There is nothing wrong with the model. The strategy is simply not looking at it.
What each type of strategy sees
Here is the distinction that explains everything and is almost never stated clearly:
2D strategies work with geometry chains you select. Not with the solid. To them, the model is scenery: what exists is the list of edges or contours you supplied. If a boss is not on that list, it is not there as far as the strategy is concerned, and the cutter crosses it without remorse.
3D strategies work with the body. They evaluate the solid and calculate around what they find. A modelled boss is respected even if you have not explicitly selected it.
That is why the same job, with the same model, works or fails depending on the strategy family you choose.
Two routes to the solution
Route 1: change strategy. If you need it to respect the solid, use 3D adaptive roughing. That is what it does by definition.
Route 2: select everything it needs. If you want to stay with 2D for any reason, manually add every chain it must respect. It works, but is fragile: when you modify the model and add a new boss, the strategy will not notice and you have to remember.
For a part that will evolve, route 1 is safer.
The same misunderstanding, three other forms
Once you understand that the problem is what the strategy can see, several apparently arbitrary software behaviours turn out to be the same thing:
The strategy stops producing passes where the surface becomes horizontal. Strategies working by vertical projection need slope to calculate increments. As the surface flattens, pass spacing grows dramatically and it looks as though passes were “removed”. They were not removed: they were calculated as requested.
The strategy skips passes at the wall-to-bevel transition. If you selected the bevel but not the adjacent walls, the strategy has no continuity there. Add the neighbouring faces and the transition resolves itself.
A parallel-pass strategy loses quality when the slope changes, settings aside, because its spacing is constant in plan view rather than along the actual surface.
In all three cases, the reading is the same: the toolpath follows directly from what you selected. The software does not decide on its own.
For bevels around a closed perimeter
One specific case worth knowing to save time: when a bevel runs around an entire part, projection strategies will struggle somewhere along the path, because a closed perimeter inevitably contains horizontal stretches.
What works there is a ramp strategy following the perimeter. It does not project: it follows. The slope of each segment therefore does not matter.
The general rule
Before touching a parameter, ask: what is this strategy calculating from — the geometry I supplied, or the body?
If it is geometry, the result is only as good as your selection. If it is the body, you can trust the model.
Almost every roughing path that cuts through something, and almost every inexplicably disappearing pass, comes from answering that question incorrectly.
At 7Axis Industrial we manufacture parts, tooling and short runs. Programming the strategy correctly is part of the part's price, even if it does not appear as a line on the quotation.