Tolerance Stack Calculator

A tolerance stack-up traces a chain of dimensions across an assembly and asks how much the closing gap can vary once every part is somewhere inside its own tolerance band. The nominal result is the signed sum of the nominal dimensions; the variation depends on which combination rule you accept. Worst case simply adds the tolerances, giving the arithmetic extreme that will occur if every part happens to sit at the same end of its band simultaneously. Root sum square instead treats each contributor as an independent random variable and adds variances, so the total is sqrt(t1^2 + t2^2 + ... + tn^2). For four parts at plus or minus 0.05, 0.08, 0.10 and 0.06 mm, worst case gives 0.29 mm but RSS gives 0.15 mm - roughly 48 percent tighter for the same drawings. The gap widens as the chain lengthens: with n equal tolerances, worst case grows as n while RSS grows only as sqrt(n). Because RSS combines squares, contribution is proportional to the square of each tolerance, so the largest single tolerance dominates the result and is where tightening pays. The sigma multiplier scales the RSS result to a target conformance level, 3 sigma corresponding to about 99.73 percent and 6 sigma to about 3.4 parts per million. The assumptions behind RSS are strict and often violated in practice: it presumes each dimension is normally distributed, centred on nominal, and statistically independent of the others. Parts machined in one setup, moulded in one cavity, or sorted by a supplier break independence or centring, and processes that drift produce biased means. Worst case remains the right choice for safety-critical joints, for chains of fewer than about four contributors, and where no rejects are tolerable; Monte Carlo simulation with real process data is preferred when distributions are skewed or correlated.

Tolerance analysis grew out of the shift from fitted craft assembly to interchangeable manufacture. Honore Blanc demonstrated musket locks with interchangeable parts in France in the 1780s, and the idea was pursued at the United States armouries at Springfield and Harpers Ferry through the first half of the nineteenth century, where the practical enabler was not arithmetic but gauging - go and no-go gauges that defined a permissible band rather than a single dimension. Once limits were written on drawings, the question of how limits accumulate across an assembly became unavoidable, and early practice was purely worst case because it needs no statistics. The statistical alternative followed the spread of process control. Walter Shewhart at Bell Telephone Laboratories introduced the control chart in 1924 and established the idea that a stable manufacturing process produces a predictable distribution, which is the foundation the root sum square method rests on. Aerospace and automotive engineering adopted RSS stacking through the mid-twentieth century as assemblies grew too long for worst case to be economic. Drawing conventions were codified in parallel: American military standard MIL-STD-8 in 1949, then ANSI Y14.5 in 1966 and its successors ASME Y14.5M-1982, Y14.5M-1994, Y14.5-2009 and Y14.5-2018, which brought geometric dimensioning and tolerancing into the same framework. The sigma multipliers used here entered general engineering vocabulary through the Six Sigma quality programme developed at Motorola in the 1980s, credited to engineer Bill Smith.