Gallery/Live trajectory/CATIA V5
Chamfering a turned boss
A dress-up feature applied to a revolved part: orbit to inspect, select the edge, invoke Chamfer, set the length, preview, commit.
Click this panel, then ↑↓ to step. Shift jumps 5, Home/End to the ends.
Every mark is a recorded action. Narration was captured on a separate stream and re-aligned at import, delayed by the 1589ms between the first video frame and the first audio sample.
Dress-up feature quality
A competent dress-up operation, executed without waste. Three orbits to expose the edge, the command hit directly, the value bracketed by two previews. The technique is sound. The parametric hygiene is not — every selection here is a geometric pick that will break the first time something upstream moves. And one criterion could not be judged at all, because the capture machine had no git and therefore produced no snapshot of the part. Good hands. Unproven intent.
Edge selection strategy
3 / 5 25% weightWas the chamfer applied to the correct topology — the edge itself, rather than a face or a whole feature — and picked in a way that survives an upstream edit?
The edge is picked directly in the 3D window, twice — once to stage the selection at 22.2s, once again inside the dialog at 31.6s. It works, and the feature commits cleanly. It also breaks the moment anything upstream moves, because a click in the viewport resolves to whatever happens to sit under the cursor at that instant. A propagation, or a reference to the parent feature, would survive an edit. This will not.
Chamfer parameterisation
3 / 5 25% weightIs the length/angle mode appropriate, and is the value defensible against the local wall thickness and the manufacturing process?
Length1 is nudged with five clicks in 973 milliseconds. No value is typed. Nothing in the session shows the wall thickness being measured or the mode being reconsidered — the dialog is used as it opened. The result looks plausible on the part. There is no evidence it was reasoned about, and a chamfer that is plausible is not the same as a chamfer that is right.
Preview discipline
4 / 5 15% weightDid the demonstrator preview the result and inspect it before committing the feature?
Preview is clicked twice: once at 35.4s before the value is touched, once at 39.9s after. That brackets the change, which is the right instinct — you see what you had and what you are about to get. The inspection orbit lands at 47.5s, after the commit rather than before it. Close to full marks, one step out of order.
Feature tree order
not assessable 20% weightDoes the chamfer sit in the right place in the tree — after the features it depends on, before anything that should inherit it?
There is nothing here to grade. Git was not installed on the capture machine, so no snapshot of the CATPart was taken at any step, and the specification tree is only legible in the corner of the video. Where the chamfer landed, and whether that placement was deliberate or incidental, cannot be recovered from this session. The 20% is excluded from the score rather than guessed at — a rubric that invents an answer where the data is silent is worse than one that admits the gap.
Navigation economy
4 / 5 15% weightDid the demonstrator reach the command and the geometry efficiently, or hunt through menus and orbit aimlessly?
Three orbit drags between 16.8s and 18.7s put the edge where it can be picked, and Chamfer is hit straight off the floating Dress-Up Features toolbar. No menu hunting, no wrong toolbar, no misfires. The edges of the session are looser than the middle: two taskbar clicks while switching into CATIA at the start, and a stray click into a Chrome window at 50.8s. The work itself is efficient. The session around it is not.
Usable as a demonstration of the operation. Not usable as an example of robust parametric practice.
Pairs come from the action space: every control that was available and not chosen forms one (preferred, rejected) pair against the control the expert took.