Using the same ten parts with (\pm 0.1) mm, the RSS prediction is:
In the world of mechanical design and manufacturing, tolerance analysis is the quiet gatekeeper of quality. It asks a simple, expensive question: When we assemble these parts, will they fit? For decades, engineers used the simplest method—Worst-Case (linear) analysis—to answer this. But as products grew more complex and precision more costly, a superior statistical method emerged: Root Sum Square (RSS) analysis . Understanding RSS is not just about better math; it is about achieving the balance between risk, cost, and performance. The Problem with "Worst-Case" Imagine assembling a shaft and a bearing. The shaft is (10.0 \pm 0.1) mm, and the housing is (10.3 \pm 0.1) mm. A worst-case engineer asks: What is the smallest gap possible? They take the largest shaft (10.1 mm) and the smallest housing (10.2 mm) and find a gap of 0.1 mm. It fits. But what if we have ten parts stacked together, each with a (\pm 0.1) mm tolerance?
Where (T_i) are the individual tolerances (expressed as standard deviations or half-bands). This formula emerges from a beautiful fact: . rss tolerance analysis
| Metric | Worst-Case | RSS | | :--- | :--- | :--- | | Predicted assembly variation | (\pm 1.0) mm | (\pm 0.316) mm | | Required individual tolerance for same assembly variation ((\pm 0.316) mm) | (\pm 0.0316) mm | (\pm 0.1) mm | | Relative manufacturing cost (approx.) | High (tight tolerances) | Low (loose tolerances) | | Theoretical assembly failure rate | 0% (if all parts at extremes) | ~0.27% (beyond (\pm 3\sigma)) |
Worst-case analysis adds the absolute values of all tolerances. For ten parts, the total possible variation is (10 \times 0.1 = \pm 1.0) mm. To ensure assembly under the worst possible scenario, every part must be manufactured at the extreme edge of its tolerance simultaneously. This is profoundly unlikely. Worst-case forces designers to tighten individual tolerances dramatically—driving up machining, inspection, and rejection costs—to avoid a near-impossible event. It is the engineering equivalent of buying flood insurance for a house on the moon. RSS tolerance analysis replaces the linear sum with a geometric one: Using the same ten parts with (\pm 0
[ T_assembly = \sqrt10 \times (0.1)^2 = \sqrt0.1 \approx \pm 0.316 \text mm ]
[ T_assembly = \sqrtT_1^2 + T_2^2 + T_3^2 + \dots + T_n^2 ] But as products grew more complex and precision