How should you calculate tolerance stack-up in deep drawn assemblies?
Tolerance stack-up in a deep drawn assembly is calculated by defining the functional gap or interference, assigning every contributing dimension a nominal value and tolerance, and adding those contributions in the direction that matters. A first-pass worst-case analysis sums every bilateral tolerance: four contributors of ±0.05, ±0.08, ±0.10 and ±0.12 mm produce a stack of ±0.35 mm before any tightening. Statistical root-sum-square (RSS) analysis is √(0.05² + 0.08² + 0.10² + 0.12²), approximately ±0.166 mm, but it assumes independent, centered processes and is not a substitute for capability evidence. For drawn parts, include the formed blank, draw sequence, wall and depth variation, trimming or piercing, later machining, plating thickness and assembly reference surfaces. Record every contributor in the same unit and apply the correct sign according to whether it increases or reduces the functional gap. The correct method is selected from function and production evidence, not from the numeric result alone.
Why deep drawing creates a different stack-up problem
A flat stamped part usually contributes one thickness and one set of edge locations to an assembly. A drawn housing contributes a changing wall, a formed bottom, a trimmed rim, a depth and several features created after forming. Material thinning is not necessarily uniform, springback can change a diameter or flange angle, and a feature pierced before a later draw can move relative to the final datum. Plating or coating then adds another controlled or uncontrolled dimensional contribution.
That is why a stack-up cannot be built only from the final drawing's nominal dimensions. The manufacturing route matters. A wall produced by ironing has a different variation pattern from a wall machined after drawing, and a hole created in the progressive die is not measured from the same stable surface as a hole added after annealing. The stack should identify both the datum chain and the process step that creates each contributor.
Worst-case vs RSS: which method should you use?
| Method | Appropriate use | Evidence to request |
|---|---|---|
| Worst-case arithmetic | Safety-critical functions, low-volume builds, unknown process distributions or interfaces that must work with every permitted combination. | Every tolerance and datum; confirmation that the result stays inside the functional limit without relying on statistical averaging. |
| Root-sum-square (RSS) | High-volume production where contributors are independent, centered and supported by measured process capability. | Capability studies, control limits, sample size and evidence that no contributor is strongly biased or truncated. |
| Monte Carlo | Non-linear geometry, asymmetric tolerances, multiple mating parts or a stack where RSS assumptions are not credible. | Input distributions based on real measurements, not guessed standard deviations, plus a clear functional acceptance limit. |
RSS can make a marginal design look acceptable because it assumes that extreme values do not occur together. That assumption may be harmless in a stable, capable process and dangerous in a new programme with unknown distributions. Start with the worst-case result to establish the contract, then use statistical analysis to understand expected production behavior, not to hide a design that fails when all permitted parts are at their limits.
Which datums and callouts should control the stack?
- Functional seating face: locate from the surface that positions the housing in the assembly, not automatically from the trimmed rim.
- Functional diameter: make the inside or outside diameter the axis datum according to the mating condition, and state it on the drawing.
- Clock position: use a real hole, flat or notch as the tertiary datum so asymmetric features are measured repeatably.
- Measurement position: state where wall thickness, roundness and concentricity are evaluated; a drawn wall can vary by height and around the circumference.
- Post-finish condition: say whether a dimension applies before or after plating, coating, annealing or machining.
What should a buyer send for a stack-up review?
- The controlled assembly drawing, the individual component drawings and the functional gap or interference to be controlled.
- Datum scheme, critical dimensions, general tolerance block and any profile or position callouts.
- Material, thickness, temper, drawing route, heat treatment and finishing requirements.
- Known process capabilities or measurement data from the drawing, prototype or current supplier.
- The measurement method and the production stage at which each critical characteristic will be verified.
Key point
Build the stack from the functional interface, then follow the actual process route. Worst-case proves that every permitted combination can work; statistical analysis only describes the expected distribution after the process has demonstrated that its assumptions are true.
How does Balford review a deep drawn stack-up?
Zhejiang Balford Mechatronics Co., Ltd. reviews deep drawn and stamped parts against the drawing, process route and assembly function during DFM. The plant runs presses from 25 t to 350 t and deep draws up to a 250 mm maximum diameter, with tool design and build handled in house. A tolerance such as 0.01 mm is evaluated only for selected critical machined features where the drawing, material, geometry, process and inspection method support it; it is not a general forming tolerance. Bulk heat treatment, plating, passivation, phosphating and other listed finishing operations are performed by qualified external partners when required.
For the wider supplier-selection process, see how to evaluate a metal stamping supplier in China. For the forming route, use the deep draw metal stamping capability page.
Frequently asked questions
Should plating thickness be included in a tolerance stack-up?
Yes, when the coating changes a functional diameter, thread, press fit, flatness or sealing face. Include the specified coating thickness range and state whether the drawing dimension applies before or after finishing. A nominal coating value is not enough when the supplier must accept the full permitted range.
Can RSS analysis be used for a new deep drawn part?
RSS should be used only when its assumptions are supported by process data. A new part usually has no production distribution yet, so worst-case analysis is the safer contractual starting point. RSS can then be added after capability studies show that the contributors are independent and centered.
What is the most common missing input in a drawn-part stack-up?
The measurement location. A wall thickness, roundness or concentricity value has little meaning unless the drawing states the height, axial section and circumference where the value applies. The stack should follow those same measuring positions.
Does a 0.01 mm tolerance apply to every deep drawn dimension?
No. It may be achievable on selected critical machined features when the drawing, material, process and inspection method support it. A deep drawn wall, depth or formed radius has its own process capability and should not be assigned a general 0.01 mm tolerance without a drawing-specific review.