FreeCAD & Parametric CAD

The concept the whole path rests on: a model is not a shape, it is a recorded sequence of operations that replays when an input changes. That single idea explains the feature tree, why edit order matters, why a model can break on a change that looks harmless, and why modelling with design intent is a real skill rather than a style preference. Everyone who struggles with FreeCAD later is usually someone who treated it as a shape editor here.

FreeCAD is unusual in that its structure is visible, and learners meet the structure before anyone explains it. Workbenches are separate toolsets over one document, objects have dependencies on other objects, and the recompute traverses that graph. Understanding it up front converts several confusing behaviours — a stale model, a workbench that cannot see another's object, a recompute that touches things you did not edit — into predictable consequences.

🔴 The skill FreeCAD lives or dies on, and the one most self-taught users never actually acquire. A sketch is a system of geometry plus constraints, and until it is fully constrained it has degrees of freedom that will move on you later. The habit being built is specific: drive dimensions with constraints that express what the part means, not just enough numbers to lock it. A Skill element, because the outcome is fully-constrained sketches produced consistently, and only practice produces that.

The mainstream modelling workflow: a single solid body grown and cut by a sequence of features driven by sketches. Covers the body-versus-part distinction that trips everyone, additive and subtractive features, and the discipline of one body per manufactured part. This is where the sketcher skill from the previous unit gets spent, so it must follow rather than precede it.

The escape from the trap of sketching on model faces. Attaching a sketch to a face is convenient and it welds that sketch to a piece of geometry that later edits can destroy; attaching it to a datum plane that is itself defined by design intent survives the change. This unit is the practical mitigation for the problem the next unit names, and teaching it before that unit means the fix arrives with the diagnosis.

Fillet, chamfer, draft and thickness are the features most likely to break a rebuild, because each depends on identifying specific edges and faces that a parameter change can move or eliminate. Covers where in the tree they belong, why a fillet applied early and a fillet applied late behave differently, and the practical rule of keeping them last so an upstream edit does not have to survive them.

🔴 FreeCAD's defining historical weakness, and it must be named explicitly because learners hit it, assume they made a mistake, and lose confidence in the tool. When the model regenerates, the internal names of edges and faces can change, so a feature that referenced 'that edge' now references a different one. Version 1.0 substantially mitigates it; defensive modelling habits still matter, and this unit teaches both what changed and what to keep doing anyway.

The unit that converts a model into a design. A named parameter sheet, expressions binding dimensions to those parameters, and derived values computed rather than typed. The assessable outcome is a model that regenerates correctly across a stated range of inputs — the same bar the code-CAD path sets, reached through a different interface, and the direct comparison between the two is worth making explicitly.

Multiple parts positioned by defined relationships. FreeCAD's history here is genuinely confusing to newcomers — a built-in assembly workbench alongside several long-standing community alternatives with incompatible approaches — so this unit has to cover the landscape honestly and then teach the current first-party path. Includes interference checking and exploded views, which are the practical reasons to assemble.

The deliverable that turns a model into something a workshop can make, and the step hobbyist workflows skip entirely. Projected views, sections, dimensioning that reflects how the part is measured rather than how it was drawn, tolerances, and a title block. Worth real time: for anything fabricated by someone else, the drawing is the contract and the model is a reference.

Which formats preserve the solid, which preserve the feature history (almost none), and which hand you a mesh you cannot meaningfully edit. Covers the practical reality of receiving a STEP file from someone else and having to work with dumb geometry, and the repair tools for the imported solids that arrive with defects.

A deliberate tour rather than depth: direct modelling for cases where a feature tree is overkill, sheet metal with bend allowances and flat patterns, finite element analysis for a first-order stress check, and toolpath generation for machining. The point is knowing what exists and roughly what it is for, so that later work does not get done the hard way in the wrong workbench.

How a project stays workable past a few dozen parts: naming, tree structure, external references between files, shared libraries of standard parts, and the resulting file dependencies. Unglamorous and load-bearing — a well-modelled part in a badly organised project is still unusable by anyone else, and that includes you in six months.

The capstone Skill element: a multi-part assembly driven from a parameter sheet, regenerating correctly across its range, with a complete drawing set someone could actually manufacture from. It is an artifact, it is assessed as an artifact, and no quiz closes it.