I have cleaned up more CAD assemblies than I care to count, and the pattern is always the same. Someone dragged parts in, hit a few mates with coincident and concentric, got the spinner to disappear, and celebrated. Six months later the model is a house of cards: mates explode when a part changes, subassemblies freeze, and nobody can tell what the mechanism is actually supposed to do. A little discipline in how you apply constraints at the start saves an enormous amount of pain later. This is about the strategy, not just the commands.
1. Why Assembly Strategy Matters More Than the Mate Type
The individual mate commands in any CAD package, coincident, concentric, parallel, tangency, angle, distance, are similar across SolidWorks, Inventor, Fusion and NX. What separates a good model from a bad one is the logic of how you chain them. A well-built assembly is essentially a kinematic and geometric dependency graph that the software can solve and rebuild quickly, and a badly constrained one is a graph where every node fights every other node.
The two failure modes you actually meet in the field are over-constraint and under-constraint. Over-constraint locks the same degree of freedom twice, sometimes with a small conflict that the solver silently absorbs, sometimes with a mate error that freezes the whole design. Under-constraint leaves a part free to float or spin, so the mechanism looks right in front view and is wrong in an isometric you never checked.
On the shop floor after I have seen both hundreds of times, I would rather deal with under-constraint in review than over-constraint at rebuild time. One is a visual error you can catch; the other is a performance and stability tax on everyone.
The honest test of a well-mated assembly is simple: can you drag the moving link and watch the mechanism behave exactly as the physical machine will, and can you change a dimension at the top level without a cascade of mate rebranding errors?
2. The Six Degrees of Freedom and What Constrains What
Every rigid part in space starts with six degrees of freedom: three translations and three rotations. Your job is to remove exactly the ones the design requires, and no more. A shaft in a bearing usually needs rotation about its axis left free, and everything else locked, which is two mates that feel natural. A bolt in a hole needs translation and rotation fully constrained, which is a coincident on the axis and a coincident or distance on the face.
Counting the degrees of freedom as you add mates is the single most useful mental habit in assembly work, and hardly anyone does it. When a colleague complains a part ‘will not move’, the first question is not which mate to fix but which degree of freedom is wrongly locked. You can verify it by checking how many constraints are acting on the part and whether they describe the intended motion.
| Motion Needed | DOF Kept | Typical Mates |
|---|---|---|
| Pivoting link | 1 rotation | Concentric + coincident (axis/face) |
| Sliding slide | 1 translation | Two parallel + one coincident |
| Fixed base | 0 | Coincident x3 or fixed |
| Ball joint | 3 rotations | Concentric (point type) |
That table is my mental model, and you will notice it maps directly to the physical machine. If a slide in the model has two translations free, go back to the real slide and check: you probably forgot a guide or a face mate. The model is telling you the mechanism you drew, not the mechanism you meant, and the mismatch is usually a constraint you skipped.
3. Coincident and Concentric: The Two Workhorses
Coincident and concentric are the two mates that do ninety percent of the work, and most of the trouble comes from using them in the wrong combination. Coincident keeps two faces or planes on the same plane, sacrificing one kind of constraint. Concentric keeps two axes on the same line, sacrificing the two translations perpendicular to the axis. Used together on an axis and a face, they pin a shaft to a housing with four of six degrees of freedom removed and the rotation about the axis left.
The mistake I see most is using a coincident where a parallel plus a distance would be more robust. A coincident on a face that is meant to have a shim or a covering gap gets a distance with a dimension later, and then the designer discovers the mate was absorbing the gap all along. Prefer a distance mate with the actual air gap or press-fit interference value where the drawing calls for one. The model then matches the tolerance note instead of glossing over it.
Another habit that pays: mate to features that will not be reworked. Mating to a temporary construction plane that someone deletes later is a fragile chain, and mating to a fillet radius face that a design change surfaces is a trap. Make it a rule to mate to datum planes, axes, base features, and faces that carry the functional dimension, not to cosmetic geometry.
And avoid the temptation to just hit ‘fixed’ on everything. Fixing a part turns off its mates, hides its real relationship, and turns the assembly into a sculpture. I use fixed only for the base part of the top-level assembly and for truly welded position placeholders; every other part should be built on proper mates so the mechanism actually moves when you need it to.
4. Subassemblies: Where Strategy Actually Lives
The biggest lever in assembly strategy is not the mate type at all; it is the subassembly structure. A machine with three hundred parts lumped into one flat assembly will choke on rebuild time, drown in mate conflicts, and give every drafter a different idea of where a part lives. The fix is to group parts into logical subassemblies that match the physical modules: the base frame, the axis, the clamping unit, the piping module.
There is a rule I steal from good code architecture and apply to models: mate inside the subassembly what belongs together, and keep the interfaces between subassemblies few and explicit. When a clamping unit is its own subassembly, the parts inside it mate to each other’s features, and the whole unit mates to the main assembly through two or three faces and axes. If you find yourself mating a tiny screw inside the pump subassembly to a hole in the base frame, the subassembly boundary is wrong.
Set the subassembly to allow or fix degrees of freedom deliberately. Most packages let you mark a subassembly as flexible so its internal mates still resolve, or rigid where it is treated as a single solid for rebuild speed. Flexible is a performance cost, so use it only where the subassembly genuinely moves as a mechanism, like a slider-crank driven unit, and leave the rest rigid.
If you have to open a tiny part deep inside the tree to understand the machine, your subassembly boundaries are wrong. You should be able to read the machine structure from the feature tree alone.
5. Reference Geometry and Top-Down Connections
Beyond building bottom-up from parts, top-down design links the assembly geometry to the parts by reference, so when the frame changes, the mounting holes follow. That is powerful and dangerous at the same time. The reference chain keeps the design in sync, but a long chain of references across subassemblies becomes brittle and slow, and when one link breaks, everything downstream turns up with a dangling reference.
I am a fan of top-down for the critical mating geometry, the mounting points and the bolt pattern, and a reluctant user of it for small parts that should simply be their own thing. The pragmatic approach is to identify the few functional interfaces that must stay in sync and drive those with references, then let the rest of the model assemble on ordinary mates. You get the synchronization where it matters without paying the rebuild and fragility cost everywhere.
Use planes and axes from the top assembly as the skeleton. Naming them matters; a plane called ‘cable_tray_center_plane’ tells the next drafter what it is for, and a plane called ‘Plane2’ tells them nothing six months later. Skeleton geometry is the contract between you and whoever inherits the model.
6. Pattern, Reuse and the Constraint Library
Repetitive hardware is where assemblies turn into rebuild-pigs unless you use patterns and the standard part library. A bolt pattern in a flange does not need sixteen individual mates; it needs one mate and a circular or linear pattern that follows the hole pattern. I have rebuilt assemblies where the same M8 screw carried fourteen separate coincident and concentric mates because nobody used the pattern, and the rebuild time dropped by half once I cleaned it up.
For standard parts, favor the library components with pre-built mates or the configuration table. The library part should mate on its mounting axis and seat face, and every configuration should carry the same three reference features so the mates survive a size change. A screw that grows from M8 to M10 must keep its axis and seat references, or every assembly that uses it breaks.
Watch for the silent cost of fully constrained standard hardware. A bolt in a tapped hole needs its thread starter point placed by the part, and if the library part starts its geometry at a different origin, the assembly gets a phantom offset that nobody sees. The fix is library discipline: one origin convention, one seat plane name, one axis name across the whole library, and the mates become boring, which is exactly what you want.
7. Checking Your Work Before Someone Else Does
Before you hand the assembly over, run the checks that catch the laziest errors. Drag every moving link through its full travel and watch for the mechanism flipping or binding. Check the interference analysis on the full assembly, on the moving positions, not just the home position. Rename anything called default or Plane1 in the tree. And rebuild with every file closed instead of relying on the last session’s cache, because that is how the drafter inheriting the model will first open it, with a cold rebuild.
- Verify the degree of freedom count matches the intended motion.
- Check for mates on temporary geometry that could be deleted.
- Confirm the subassembly boundaries match the physical modules.
- Run full-travel drag and moving-position interference checks.
- Rebuild cold and confirm the model resolves without errors.
A model that survives those five checks is a model you will not get called about at midnight. It is not glamorous work, but it is the difference between an assembly that is a joy to update and one that a colleague will quietly avoid touching for years.
One last thing on culture rather than technique. In my shop we treat a well-constrained assembly as a deliverable, not a side effect. The model review checklist includes the mate philosophy, not just the geometry, and a new drafter gets a twenty-minute walkthrough of how we name references and where the subassembly boundaries go. It feels like a waste of half an hour until the fourth time you inherit a model from someone who thought constraints were a formality. Then you understand that the assembly strategy is really a communication tool, and the model is how you hand your intent to the next person who has to make the machine actually work.
