Tolerance stack-up analysis predicts how individual part variations combine to affect an assembly dimension such as a gap, clearance, preload, or interference fit. It matters because every manufactured part has allowable variation, and small deviations can add together to create a failed fit even when each part is within its own tolerance. Engineers use stack-up analysis to decide whether a design can be built reliably, inspected efficiently, and produced at a reasonable cost.
A stack-up begins by defining a dimension chain from one functional surface to another, then assigning each contributing dimension a sign based on whether it increases or decreases the final result. Worst-case analysis assumes all dimensions hit their most unfavorable limits at the same time, while statistical RSS analysis assumes independent random variations combine by root-sum-square. Tolerance allocation works in the reverse direction by dividing an allowable assembly variation among parts based on function, process capability, cost, and risk.
Understanding Engineering: Tolerance Stack-Up Analysis
A useful stack-up starts with function, not with a drawing dimension. Engineers first identify the two features whose relationship matters. They may be a wheel and a brake bracket, a circuit board and an enclosure wall, or a bolt and the far side of its mounting slot.
Every surface in the path between those features must be traced. This is called a dimension loop.
The loop must use real locating surfaces, often called datums. A dimension measured from an unrelated edge may look correct on paper yet contribute nothing useful to how the parts actually sit together.
Parts do not vary only in size. Their shape and position vary too. A hole can be in the correct diameter range but shifted away from its intended location.
A shaft can have the right diameter but bend slightly. A flat face can tilt. These effects are controlled with geometric tolerances such as position, flatness, perpendicularity, and runout.
In an assembly, location error can reduce clearance just as much as a size error. Engineers therefore convert relevant geometric error into its effect along the functional direction.
This is why datum selection and the order of assembly matter. A part held against one face behaves differently from the same part held against another.
The two common calculation methods represent different promises. A limit based method protects every possible combination allowed by the drawings. It is important for safety parts, one time assemblies, tightly controlled fits, and products where a failed unit cannot be accepted.
A statistical method predicts the spread of normal production when sources of variation are stable and largely independent. It can permit wider individual tolerances, which often lowers manufacturing cost. It is not permission to ignore risk.
Dimensions made in the same machining setup can drift together. Tool wear, temperature changes, shared fixtures, and supplier batches create correlation. When variation is linked, a simple statistical result can be too optimistic.
Tolerance allocation is a design decision with practical consequences. A tight limit should be placed where a process can hold it reliably and where inspection can verify it. Precision grinding may hold a small size range well, while a bent sheet metal flange may vary more because of springback.
Engineers often reserve the tightest control for a critical interface, then allow more variation in noncritical features. In school projects, the same idea appears when laser cut pieces must slot together, 3D printed parts need clearance, or a model car axle must turn freely. Check the intended fit at the extreme material conditions, consider burrs, coatings, paint, assembly force, and temperature.
A small clearance on a cold clean drawing can disappear after coating or heating. Good stack-up work ends with a clear inspection plan that measures the features that truly control function.
Key Facts
- Assembly result: R = d1 + d2 - d3 + d4, using signs from the dimension chain.
- Worst-case tolerance: T_wc = T1 + T2 + T3 + ... + Tn.
- Worst-case limits: R_min = R_nominal - T_wc and R_max = R_nominal + T_wc when T values are bilateral half-widths.
- Statistical RSS tolerance: T_rss = sqrt(T1^2 + T2^2 + T3^2 + ... + Tn^2).
- For independent normal dimensions, standard deviations combine as sigma_R = sqrt(sigma_1^2 + sigma_2^2 + ... + sigma_n^2).
- If a dimension has tolerance ±T and represents a 3 sigma process, then sigma = T / 3.
Vocabulary
- Tolerance stack-up
- A calculation that determines how part dimension variations combine to affect a final assembly dimension.
- Dimension chain
- The ordered path of dimensions that connects the two surfaces defining the functional assembly requirement.
- Worst-case analysis
- A method that assumes every part dimension is at its extreme limit in the direction that creates the largest or smallest assembly result.
- RSS analysis
- A statistical method that combines independent tolerances using the square root of the sum of their squares.
- Tolerance allocation
- The process of assigning allowable variation to individual part dimensions so the assembly requirement is met.
Common Mistakes to Avoid
- Adding all nominal dimensions without signs, because some dimensions increase the gap while others reduce it.
- Using RSS for dependent dimensions, because shared tooling, common datums, or process shifts can make variations correlated rather than independent.
- Treating total tolerance width as the plus-minus value, because a dimension listed as 20.00 ± 0.10 has a half-width of 0.10 and a full width of 0.20.
- Allocating equal tolerances to every part automatically, because critical surfaces, manufacturing processes, and cost sensitivity often require unequal tolerances.
Practice Questions
- 1 A gap is defined by G = A - B - C. If A = 50.00 ± 0.20 mm, B = 18.00 ± 0.10 mm, and C = 31.50 ± 0.15 mm, find the nominal gap and the worst-case minimum and maximum gap.
- 2 Four independent dimensions contribute to a critical length with bilateral tolerances ±0.05 mm, ±0.08 mm, ±0.04 mm, and ±0.06 mm. Calculate the worst-case tolerance and the RSS tolerance.
- 3 An assembly has a tight functional gap requirement, but one spacer is produced by a low-cost process with poor repeatability. Explain how tolerance allocation could be changed to improve yield without making every part more expensive.