Tolerance Stacks and Functional Datums

An assembly clearance that looks comfortable at nominal dimensions can become ambiguous once every contributing tolerance, thermal effect, and coating thickness is added into the stack, because a nominal-dimension check answers a different question than a tolerance stack does, and the two can give opposite answers about whether an assembly will fit.

Worst-case stacking sums the full tolerance range of every contributor and guarantees fit as long as every part is within specification, at the cost of tolerances that are often tighter, and therefore more expensive to produce, than the application actually needs. RSS, or statistical, stacking assumes contributors vary independently and combines them by root-sum-square, which loosens individual tolerances but accepts a small statistical probability of an out-of-tolerance assembly rather than a hard guarantee.

The choice between worst-case and RSS is not a formula preference, it is a risk decision, and it depends on production volume, inspection capability, and the consequence of an occasional out-of-tolerance assembly. A low-volume or safety-relevant assembly generally justifies worst-case’s certainty despite tighter individual tolerances, while a high-volume assembly with proven process capability can often use RSS’s looser tolerances without meaningfully increasing failed-assembly rates.

A datum strategy decides which features locate a part relative to the assembly and, by extension, which dimensions actually matter for the function being protected. Choosing datums based on ease of measurement rather than based on the functional requirement can produce a technically compliant part that still does not assemble or function correctly, because the inspected dimensions were not the ones the function actually depends on.

Fastener access and clearance belong inside a tolerance stack, not outside it as an afterthought, because a hole pattern that is dimensionally correct at nominal can still be unreachable by a tool or unable to accept a fastener once the stack of hole position, fastener diameter, and mating clearance tolerances is worked through in the least favorable direction. Assembly sequence changes which stack direction actually matters, since a part installed last in a chain of assembled features inherits the accumulated position error of every part installed before it, not just its own individual tolerance.

Thermal expansion and coating thickness are contributors that a purely dimensional stack, done at room temperature on bare material, will miss. A stack that fits at assembly can bind or loosen across a real operating temperature range if the contributing parts expand at different rates, and a coating or plating thickness applied after the base dimension was toleranced adds a contributor that needs its own tolerance rather than being ignored as negligible.

Surface finish rarely appears in a purely dimensional tolerance stack, yet it changes the effective mating condition wherever two surfaces slide, seal, or transfer load against each other. A rougher-than-intended surface finish can consume part of a clearance through asperity contact under load, or can increase friction and wear at a sliding interface, in a way that a stack built only from nominal dimensions and their tolerances does not represent.

Tighter tolerance control reduces assembly and function risk but increases process capability demands, measurement burden, cost, and yield risk at the supplier, while looser tolerance eases production at the cost of a higher probability of a marginal or failed fit. Confirm the drawing’s functional intent and datum scheme, the actual GD&T controlling the relevant features, process capability evidence for the chosen tolerance class, thermal and coating effects across real operating conditions, and the assembly sequence, with qualified drawing and quality review before releasing a tolerance stack as final.