Skip to content

technical-reference.md

Latest commit

 

History

History
138 lines (110 loc) · 6.49 KB

technical-reference.md

File metadata and controls

138 lines (110 loc) · 6.49 KB
Raw

Technical reference

@casys/mcp-tolerance is a deterministic TypeScript calculation provider. It does not call an LLM, execute user code, contact an upstream service, or infer a design requirement.

Result families

Operation Input intent Returned evidence
tolerance_it Fundamental tolerance at a nominal diameter IT width and selected diameter range
tolerance_limits One supported hole or shaft class Upper/lower deviation, IT width, and fundamental deviation
tolerance_fit Base-hole H class with a supported shaft class Member limits, fit classification, and signed clearance/interference extremes
tolerance_fit_analyze Any supported uppercase hole class with any supported lowercase shaft class Member limits, fit classification, and signed clearance range
tolerance_stackup Linear chain of bounded contributors Nominal, worst-case and RSS intervals, plus contributor terms

Inputs and outputs use closed schemas. Results include provenance, not_checked, and an empty violations array. violations stays empty because the provider accepts no application acceptance threshold against which to form a verdict.

Supported ISO 286 positions

  • Hole positions: C, D, E, F, G, H, JS, K, M, N, P, R, S.
  • Shaft positions: c, d, e, f, g, h, js, k, m, n, p, r, s, u.
  • IT grades: 1–18.
  • Nominal diameters: greater than 0 mm and at most 500 mm.

tolerance_fit deliberately accepts only the base-hole position H. Use tolerance_fit_analyze for other supported hole positions, including shaft-basis examples such as G7/h6. Mathematical support does not mean that a pair is preferred or suitable for a load, material, process, or assembly method.

Units, signs, and ranges

  • Nominal diameters and stack-up nominal dimensions use millimetres.
  • ISO deviations and contributor offsets use micrometres (1 µm = 0.001 mm).
  • For holes, EI is lower and ES is upper.
  • For shafts, ei is lower and es is upper.
  • Absolute size is nominal_diameter_mm + deviation_um / 1000.
  • Signed clearance is hole size - shaft size: positive means clearance; negative means interference.
  • Diameter ranges are lower-exclusive and upper-inclusive. [18, 30] means 18 < diameter <= 30 mm.

The clearance and interference fields returned by tolerance_fit are signed views of the same extremes. A negative max_interference_um in a clearance fit does not mean that interference occurs. Prefer the literal fit classification or the signed minimum/maximum clearance.

1D stack-up semantics

Each contributor supplies a nominal dimension in mm, non-negative plus_um and minus_um offsets, and a direction. +1 adds to the assembly result; -1 subtracts from it.

{
  "contributors": [
    {
      "name": "housing depth",
      "nominal_mm": 10,
      "plus_um": 50,
      "minus_um": 50,
      "direction": 1
    },
    {
      "name": "component stack",
      "nominal_mm": 9.8,
      "plus_um": 30,
      "minus_um": 20,
      "direction": -1
    }
  ]
}

This chain has a nominal gap of 0.200 mm and a worst-case interval of 0.120–0.270 mm. The returned contributor_breakdown preserves input order and contains the signed nominal and worst-case/RSS terms used in the aggregates. It is reconstructable accounting, not a sensitivity ranking.

Worst-case covers every combination only when every supplied dimension stays inside its stated bounds and the chain includes every material contributor. RSS is a statistical estimate requiring independent, centred processes and a consistent sigma convention. The provider does not infer distributions, covariance, process capability, confidence, yield, or an acceptance threshold.

Formula and table provenance

The implementation references ISO 286-1:2010 as follows:

  • IT1–IT12: Table 1 values by main diameter range.
  • IT13–IT18: i = 0.45 * D^(1/3) + 0.001 * D, using the selected range's geometric-mean diameter and the documented tiered rounding.
  • Shaft c–g: Table 4 upper deviations, including the finer c sub-ranges.
  • Shaft h: es = 0; shaft js: symmetric around zero.
  • Shaft k, m, n, p, r, s, u: Table 5 lower deviations, including the finer r, s, and u sub-ranges.
  • Hole H: EI = 0; hole JS: symmetric around zero.
  • Hole C–G: derived from the corresponding shaft upper deviation.
  • Hole K, M, N, P, R, S: derived from the corresponding shaft lower deviation.

The committed fixture gate cross-checks expectations transcribed from Tables 1, 4, and 5 against the engine and fails closed on missing rows, cells, sub-ranges, or divergent values:

deno task check:iso286-fixtures

MCP App contracts

The exported view-app-manifest declares these whole-view resources:

  • ui://mcp-tolerance/limits-viewer
  • ui://mcp-tolerance/fit-viewer
  • ui://mcp-tolerance/stackup-viewer

Each surface owns one semantic business object. Identity, readings, scope notes, and provenance stay together. There is no artificial navigation or inferred verdict. Recorded projections are accepted only through their exact read-only viewer.session.apply schema and projection fingerprint.

Boundaries

The fit calculations do not model or verify actual measurements, inspection uncertainty, surface roughness, form or geometric tolerances, thermal expansion, material behaviour, lubrication, assembly force, press-fit stress, retained stress, wear, or fatigue.

The stack-up calculation is scalar and linear. It has no vector loop, datum scheme, assembly sequence, geometric model, automatic link from a fit result, or Monte Carlo simulation. Add every relevant effect explicitly or use a more complete analysis method.

The project implements the named 2010 edition. It is not an official ISO publication or certification service. For regulated or safety-critical work, verify the calculation independently against the licensed standard and the governing design requirements.