Tolerance stack-up analysis predicts how part tolerances combine in an assembly and whether the final fit, clearance, or alignment will meet requirements. Engineering students need this reference because real manufactured parts are never exactly nominal. A clear stack-up method helps connect drawings, GD&T limits, process capability, and design risk.
This cheat sheet summarizes the formulas and decision rules used in mechanical design and manufacturing review.
The core idea is to define an assembly response, write it as a function of part dimensions, then combine the effects of each tolerance. Worst-case analysis assumes every part reaches its most unfavorable limit at the same time. RSS and statistical analysis assume independent variation and combine standard deviations or equivalent tolerance bands.
Sensitivity coefficients show how strongly each dimension affects the final assembly result.
Key Facts
- Assembly response can be written as Y = f(x1, x2, ..., xn), where Y is the gap, flushness, position, or other critical output.
- For a linear stack, Y = c1x1 + c2x2 + ... + cnxn, where each ci is a sensitivity coefficient that gives direction and scale.
- Worst-case tolerance is T_wc = |c1|T1 + |c2|T2 + ... + |cn|Tn, which protects against all parts being at their worst limits.
- RSS tolerance is T_rss = sqrt((c1T1)^2 + (c2T2)^2 + ... + (cnTn)^2) when tolerances are independent and centered.
- If a bilateral tolerance ±T represents a normal 3 sigma spread, then sigma = T / 3 for that dimension.
- For independent normal variables, sigma_Y = sqrt((c1sigma1)^2 + (c2sigma2)^2 + ... + (cnsigman)^2).
- Clearance at worst case is C_min = C_nominal - T_wc, and interference risk exists if C_min is less than 0.
- Percent contribution for a statistical stack is Contribution_i = (cisigmai)^2 / sigma_Y^2 times 100 percent.
Vocabulary
- Tolerance Stack-Up
- A calculation that predicts how individual part tolerances combine to affect an assembly dimension or functional requirement.
- Worst-Case Analysis
- A stack-up method that assumes every part dimension is at the limit that produces the most unfavorable assembly condition.
- RSS Analysis
- A root-sum-square method that combines independent tolerance effects statistically rather than adding all limits directly.
- Sensitivity Coefficient
- A multiplier that shows how much the assembly result changes when one input dimension changes.
- Clearance
- The positive space or gap between assembled parts, usually checked to make sure interference cannot occur.
- Process Capability
- A measure of how well a manufacturing process can hold dimensions within tolerance compared with its natural variation.
Common Mistakes to Avoid
- Adding nominal dimensions without sign direction, then adding all tolerances afterward is wrong because each dimension may increase or decrease the final gap differently.
- Using RSS analysis for dependent dimensions is wrong because RSS assumes independent variation, and correlated features can vary together in the same direction.
- Treating every drawing tolerance as 3 sigma without confirmation is wrong because tolerance meaning depends on company standards, supplier capability, and inspection assumptions.
- Ignoring sensitivity coefficients is wrong because a small part tolerance can dominate the stack if its geometry gives it a large effect on the assembly response.
- Checking only the nominal clearance is wrong because manufactured parts vary, and the minimum or statistical clearance may fail even when the nominal design looks acceptable.
Practice Questions
- 1 A clearance has nominal value 2.50 mm and three independent contributors with worst-case tolerances ±0.20 mm, ±0.15 mm, and ±0.10 mm, all with sensitivity 1. Find the worst-case minimum clearance.
- 2 For the same three contributors, calculate the RSS tolerance using T_rss = sqrt(T1^2 + T2^2 + T3^2).
- 3 An assembly output is Y = A - B + 2C. If A = 40.00 ± 0.10 mm, B = 12.00 ± 0.05 mm, and C = 3.00 ± 0.02 mm, find the nominal value of Y and its worst-case tolerance.
- 4 Explain why a designer might choose worst-case analysis for a safety-critical latch but RSS analysis for a consumer product snap-fit.
Understanding Tolerance Stack-Up Analysis Reference
A stack begins with the actual physical path through an assembly. Trace from one functional surface to the other. Every dimension on that path must have a clear role.
A length may add material into a gap. A recess may remove material from it. The sign comes from this physical effect, not from whether a number looks positive on a drawing.
This is where many errors begin. Draw a simple one dimensional sketch and mark the two end surfaces. Then imagine one part growing while all others stay fixed.
If the output grows, that input has a positive effect. If the output shrinks, it has a negative effect. This check is more reliable than memorising sign rules.
A drawing dimension is not always the variation that reaches the assembly. Datum choices matter. A hole located from one datum may be measured from another surface in the stack.
Flatness, perpendicularity, position, and runout can change the functional result even when every size dimension is acceptable. For example, a pin may fit through a hole by size, yet fail to enter because the hole axis is displaced. Convert each relevant geometric control into its effect on the requirement.
This often needs a worst orientation or a small geometric model. Do not put every drawing tolerance into one calculation without checking whether it moves the functional surfaces.
Statistical results depend on evidence about the process. Independent variation is a useful model only when one part feature does not move with another. Features machined in the same setup can be correlated.
Temperature can shift many dimensions in the same direction. Tool wear can create a gradual offset rather than a centered random spread. In these cases, a simple RSS result can be too optimistic.
Check inspection records, process capability studies, and the way parts are paired during assembly. A tolerance stated on a drawing is a permitted limit.
It does not automatically describe the actual distribution of produced parts. The assumed spread and centering should be stated clearly in any analysis.
Contribution results guide design effort. A feature with the largest contribution is often the best place to improve the assembly, but improvement may mean several things. It might need a tighter tolerance, a more capable process, a different datum, or a design change that reduces its leverage.
Tightening a low contribution dimension usually adds cost without solving the main risk. Leave room for gauges, wear, coating thickness, thermal expansion, and service loads when they affect the fit. A clearance that is barely positive on paper may still be unreliable in production.
Record the requirement, assumptions, units, tolerance source, and acceptance limit beside the calculation. Another engineer should be able to follow the chain from the drawing to the final decision.