> ## Documentation Index
> Fetch the complete documentation index at: https://docs.cosmon.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Tolerance Stack-up

> Verify assembly-level requirements against individual part tolerances, directly from your CAD models and drawings.

## Overview

A tolerance stack-up checks whether the tolerances assigned to individual parts still let the assembly meet its functional requirement — a gap that must stay open, a shaft that must turn freely in a bore, a stack of shims that must land within a target range. Building the chain by hand means tracing every contributing dimension across multiple parts and drawings, then working the arithmetic worst-case or statistical.

Nexus builds the dimension chain directly from your CAD models and drawings, runs the stack-up, and tells you whether the requirement holds — and if it doesn't, which tolerances are driving the overrun.

## What Nexus can help with

* **Worst-case stack-up** — Sum contributing tolerances at their extremes to find the guaranteed bound on the resulting dimension.
* **Statistical (RSS) stack-up** — Combine contributing tolerances in root-sum-square to estimate the realistic variation, assuming independent, normally distributed contributors.
* **Chain identification** — Trace the dimension chain across multiple parts in an assembly, including GD\&T call-outs, and confirm nothing in the path was missed.
* **Sensitivity analysis** — Rank contributors by how much of the total variation each one accounts for, so you know which tolerance to tighten first.
* **Tolerance allocation** — Given a target assembly requirement, recommend a tolerance for each contributor that meets it without over-tightening the whole chain.

## Example prompts

<AccordionGroup>
  <Accordion icon="ruler-combined" title="Worst-case 1D stack-up">
    ```text theme={null}
    Run a worst-case tolerance stack-up on the gap between the bracket face and the housing wall in C:/Projects/Pump/pump_asm_v2.SLDASM. The chain runs through: bracket thickness (25 ±0.1mm), spacer height (10 ±0.05mm), and housing wall position (40 ±0.15mm). The gap must stay between 3mm and 7mm. Report the worst-case min and max gap and whether the requirement is met.
    ```
  </Accordion>

  <Accordion icon="chart-line" title="Statistical (RSS) stack-up">
    ```text theme={null}
    Run an RSS stack-up on the same gap chain, assuming each contributor is normally distributed with the drawing tolerance at ±3σ. Report the resulting mean, standard deviation, and the 3σ range. Compare against the worst-case result from the earlier run.
    ```
  </Accordion>

  <Accordion icon="magnifying-glass-chart" title="Sensitivity / contributor ranking">
    ```text theme={null}
    For the gap stack-up we just ran, rank the three contributors by their share of the total variation. Tell me which one tolerance I'd need to tighten to bring the 3σ range under 3.5mm, and to what value.
    ```
  </Accordion>

  <Accordion icon="sliders" title="Tolerance allocation from a target">
    ```text theme={null}
    The bore-to-shaft clearance must be 0.05–0.15mm (RSS, 3σ). The chain has three contributors: bore diameter, shaft diameter, and bushing wall thickness. Assume equal contribution from each and recommend a tolerance for each dimension that satisfies the requirement.
    ```
  </Accordion>

  <Accordion icon="layer-group" title="Stack-up from GD&T on a drawing">
    ```text theme={null}
    Attach C:/Projects/Bracket/bracket_asm.pdf. Build the dimension chain for the position tolerance stack between the two mounting holes and the datum edge, including the position tolerance (⌀0.2mm at MMC) and datum feature simulator. Run a worst-case stack-up and report the resulting positional bound.
    ```
  </Accordion>
</AccordionGroup>

## Tips for tolerance stack-up prompts

* **State the requirement and its direction** — say what the resulting dimension is (a gap, a clearance, an offset), its target range, and which direction is critical (minimum, maximum, or both).
* **List every contributor explicitly** — name each part, the dimension, and its tolerance. Nexus can trace a chain from the model, but stating it upfront removes ambiguity about which path through the assembly counts.
* **Say worst-case or statistical** — the two methods give very different answers. Worst-case guarantees the bound; RSS estimates the realistic spread assuming independent, normally distributed contributors. State which one your requirement calls for.
* **Include GD\&T, not just size tolerances** — position, concentricity, and MMC/LMC modifiers change the stack-up math. Reference the datum scheme so Nexus applies the right bonus tolerance.
* **Give process capability if you have it** — real distributions are rarely centred on nominal. If you have Cpk or historical measurement data for a contributor, share it — the RSS result will be more realistic than assuming ±3σ equals the drawing tolerance.
* **Ask for the driving contributor, not just the answer** — when a stack-up fails, "which tolerance should I tighten" is usually more useful than the raw number.