Introduction
Tolerance stack-up analysis is one of the most practically important skills in mechanical design, and one of the most inconsistently taught. The concept is straightforward — individual part tolerances accumulate when parts are assembled, and the combined variation can exceed the available clearance or produce interference where none is intended. But applying the analysis rigorously, choosing the right method (arithmetic vs. statistical), and using the results to drive design decisions rather than just to check compliance is a skill that develops through practice.
This article walks through both the arithmetic and statistical approaches to tolerance stack-up, illustrated with assembly examples, and covers how to use the results to make better design decisions.
Why Stack-Up Analysis Matters
Consider a simple assembly: a shaft inserted into a bore, with a snap ring retaining it axially. The axial position of the snap ring groove relative to the bearing shoulder face depends on: the shaft length tolerance, the groove position tolerance, and the assembly datum consistency. Each individual tolerance may look reasonable. The combined worst-case variation may be larger than the available axial clearance — meaning some fraction of assembled units will have the snap ring interfering with the bearing face.
Stack-up analysis answers the question: given the individual tolerances specified, what is the worst-case (or statistically expected) variation in the critical assembly dimension? And if that variation is too large, which tolerances need to be tightened?
Step 1: Define the Critical Assembly Dimension
Every stack-up analysis starts with identifying the critical dimension — the gap, clearance, or interference that the analysis is trying to characterize. This is called the “resultant characteristic” or “assembly gap.” Be specific: not “the clearance at the bearing” but “the axial gap between the snap ring face and the bearing inner race face in the worst-case assembled condition.”
Sketch the assembly and draw a loop — a closed loop through all the parts and features that contribute to the critical dimension. The loop must form a closed chain from one side of the critical gap, through all contributing features, back to the other side. Every dimension in that loop is a contributor to the stack-up.
Step 2: Establish the Dimension Chain
List every dimension that contributes to the loop, with its nominal value and bilateral tolerance. Designate each dimension as either a positive contributor (dimensions that increase the gap as they increase) or a negative contributor (dimensions that decrease the gap as they increase). The sign assignment depends on the direction each dimension acts in the loop.
Example dimension chain for an axial assembly:
- Shaft length: 80.00mm ±0.1mm (negative contributor — longer shaft closes the gap)
- Housing bore depth: 82.00mm ±0.15mm (positive contributor — deeper bore opens the gap)
- Retaining clip thickness: 1.50mm ±0.05mm (negative contributor — thicker clip closes the gap)
Nominal gap = 82.00 − 80.00 − 1.50 = 0.50mm
Step 3: Arithmetic (Worst-Case) Method
The arithmetic method sums all tolerances directly to find the absolute worst-case variation:
Total tolerance = Σ|tolerance_i|
For the example above: Total tolerance = 0.10 + 0.15 + 0.05 = 0.30mm
Worst-case gap: 0.50 − 0.30 = 0.20mm minimum, 0.50 + 0.30 = 0.80mm maximum
The arithmetic method guarantees that every assembled unit will have a gap between 0.20mm and 0.80mm. This is the appropriate method when:
- 100% of assembled units must meet the requirement (safety-critical assemblies, sealed systems)
- The number of contributors is small (typically fewer than four or five)
- Part tolerances are correlated (e.g., all parts from the same batch)
The limitation: with many contributors, the arithmetic method is overly conservative — the probability of all contributors simultaneously at their worst-case values decreases rapidly with the number of terms.
Step 4: Statistical (RSS) Method
The root-sum-of-squares (RSS) statistical method assumes tolerances are independent and normally distributed, and combines them statistically:
Total tolerance (RSS) = √Σ(tolerance_i²)
For the example: Total tolerance (RSS) = √(0.10² + 0.15² + 0.05²) = √(0.01 + 0.0225 + 0.0025) = √0.035 = 0.187mm
Statistical gap range: 0.50 ± 0.187mm = 0.313mm to 0.687mm (at ±3σ, covering ~99.73% of assemblies)
The RSS method is appropriate when:
- A small fraction of out-of-tolerance assemblies is acceptable (high-volume commercial products)
- Part dimensions are truly independent and approximately normally distributed
- There are enough contributors (typically five or more) for the central limit theorem to apply
Comparison of Methods
| Characteristic | Arithmetic (Worst-Case) | Statistical (RSS) |
|---|---|---|
| Guarantees all assemblies pass? | Yes (100% compliance) | No (~99.73% at ±3σ) |
| Conservative? | Very (often over-tightens tolerances) | Less conservative, more realistic for volume production |
| Requires statistical assumptions? | No | Yes (normal distribution, independence) |
| Best used for | Safety-critical, small chains, low volume | High-volume commercial, large chains, non-critical |
| Sensitivity to non-normal distributions | None | Significant — verify distribution assumption |
Using Stack-Up Results to Drive Design Decisions
The analysis is only useful if it changes something. After completing a stack-up analysis, apply the results as follows:
- If worst-case gap is negative (interference) when zero or positive is required: The tolerance chain must be tightened. Use sensitivity analysis to identify which contributor has the highest impact: the contributor with the largest tolerance is usually (but not always) the best target for tightening. Calculate how much each contributor needs to tighten to bring the worst-case gap positive, and select the combination with the lowest manufacturing cost impact.
- If worst-case gap is acceptable but statistical gap is negative at a significant probability: Determine the acceptable defect rate and design to the corresponding sigma level. Or, restructure the design to eliminate contributors from the chain (shorter loop = smaller stack-up for the same individual tolerances).
- If nominal gap is very large relative to the tolerance: The tolerances may be tighter than necessary. Loosen tolerances while keeping the minimum gap above the required minimum — this reduces cost without compromising function.
FAQ
Q: How do I handle angular tolerances in a stack-up analysis?
A: Angular variations introduce linear displacement at a distance from the pivot point. To incorporate angular tolerances: multiply the angular tolerance (in radians) by the moment arm (distance from the pivot to the critical gap location) to convert the angular variation to a linear contribution. For small angles, this linear approximation is valid. For large angular variations (more than a few degrees), the trigonometric relationship must be used exactly.
Q: How many significant figures should I carry in a tolerance stack-up calculation?
A: Carry at least one more significant figure through intermediate calculations than appears in the final result. For most mechanical stack-ups working in millimeters, three decimal places in intermediate results and two in the final gap value is appropriate. Be aware that the precision of the result is only as good as the precision of the input tolerances — a result of ±0.003mm from a chain including a ±0.1mm contributor is false precision.
Q: My analysis shows the stack-up is marginal — barely passing worst-case. Should I be concerned?
A: Yes. A marginal worst-case result means that any shift in the process — a supplier changing their process mean, a tool wearing differently, a fixturing change — can push assemblies out of tolerance. Robust designs have comfortable margins, not marginal ones. Design to have worst-case margins that are clearly positive (not near-zero), and if that requires tightening tolerances, the cost is justified by the quality and reliability benefit.
コメント