"Everything Was in Spec but It Won't Fit" - Tolerance Stackup in Practice

Worst-case vs RSS method comparison, stackup calculation example, design margin strategies

"Every part was within tolerance, but the assembly doesn't fit." If you've heard this, tolerance stackup was ignored. When individual tolerances accumulate in the same direction, worst-case total error equals the sum of all individual errors.

1. Worst-Case Method

The most conservative approach — assumes all tolerances accumulate in the worst direction simultaneously.

Total stackup = Σ|individual tolerances|
Example: 5 parts in series, each ±0.05mm → total stackup = ±0.25mm
Pros: 100% assembly guarantee
Cons: Tolerances become very tight, driving up machining costs. Probability of all tolerances being worst-case simultaneously is extremely low.

2. RSS Method (Root Sum Square / Statistical)

Assumes each tolerance follows a normal distribution.

Total stackup = √(t₁² + t₂² + t₃² + … + tₙ²)
Example: Same 5 parts × ±0.05mm → √(5 × 0.05²) = ±0.112mm (≈45% of worst-case)
Pros: Realistic tolerances, lower machining cost
Cons: Statistically ~0.27% non-conformance (3σ). Use worst-case for safety/medical applications.

3. Design Strategies to Reduce Stackup

Reduce part count: Integrating 5 parts into 3 directly reduces stackup contributors.
Share datums: All parts referencing the same datum minimizes accumulated error.
Adjustable elements: Place shims or adjustment screws at final assembly to compensate accumulated error.
Tight tolerances only where functional: Don't specify ±0.05 everywhere — apply tight tolerances only to dimensions that directly affect function.

💡 Expert Advice
"The first thing I do when receiving an assembly drawing is calculate minimum gap. If minimum clearance goes negative, that assembly is physically impossible — no matter what the individual part tolerances say."

Tolerance Stackup — US Practice

For US engineering, ASME Y14.5 provides the dimensioning framework, but stackup analysis methodology is in supplemental texts: Bryan Fischer's Mechanical Tolerance Stackup and Analysis is the dominant reference. Common methods: worst-case (linear sum), RSS (root-sum-square, for ≥4 contributors), Monte Carlo (statistical, for high-precision).

Software Tools

Sigmetrix CETOL 6σ and Mitutoyo MeasureLink are common in US automotive PPAP processes. SolidWorks TolAnalyst (built-in) handles simpler 1D and 2D stacks. For aerospace work, Boeing and Lockheed mandate Monte Carlo analysis with at least 10,000 trials per dimension for cabin/structural assemblies. Document the analysis method on the drawing or in the design record per ASME Y14.41.