Key Characteristic Management

Key Characteristic Management (manufacturing-quality/as9103/key-characteristic-management)

Use when you must decide which characteristics on a drawing are key characteristics (KCs) and what variation management applies to them in aerospace production. A key characteristic is a feature whose variation affects fit, function, performance, or safety, so it gets a tighter control strategy than an ordinary characteristic. This leaf opens the as9103 pack directory under manufacturing-quality. KCs flow from design (GD&T callouts, toleranced in millimeters) through first article inspection (AS9102) into production variation management (AS9103 practice) and SPC. It pairs with the delta-fai leaf for change scoping, the first-article-inspection leaf for form 3 characteristic accountability, the statistical-process-control leaf for the chart and capability math, and the nonconformance-control leaf for escapes.

Domain quick reference

• AS9103 (IAQG/SAE) frames key characteristic identification and variation management of aerospace production processes; name plus paraphrase only, the standard text is proprietary and is not reproduced here.
• Decision rule table: a characteristic is a key characteristic when any single rule fires: (1) safety-critical flag, (2) customer designation, (3) fit/function impact with a mate, seal, or performance downstream impact, (4) a position or profile feature at or below the 0.1 mm tight tolerance threshold, or (5) two or more historical failures in process history. The reasons list the fired rules in table order.
• Risk score weights (0-100, full set sums to 100): safety 30, fit/function 25, tight tolerance 20, historical 15, downstream 10. The score is the sum of the weights whose signal fires; an override weight set may replace entries as long as it still sums to 100. Customer designation gates KC status by rule but carries no weight, so a customer-only KC ranks at 0.
• Cpk targets mirror common aerospace practice: 1.33 default, 1.67 for a safety-critical characteristic. This leaf assigns the target only; the X-bar/R chart and Cpk index math live in the statistical-process-control leaf.
• Control method rule table, first match wins: electrical features get 100 percent inspection (every-unit functional test), assembly-mate features get attribute control (go/no-go fit gage), two or more historical failures escalate to 100 percent inspection, a characteristic whose process capability is not yet demonstrated gets a gage study first, and everything else runs SPC variable charts (X-bar/R).
• Revalidation trigger table: tooling, process, and design changes revalidate a KC (delta FAI scoped to the affected characteristic plus a capability re-study); supplier and personnel changes do not, because they route to external-provider qualification and training control unless the controlled process itself changes.
• Units: tolerance_mm is millimeters throughout, one unit stated.

Workflow

1. Collect the characteristic records from the drawing callouts (GD&T tolerances, mm) and the process history (prior failures, known capability).
2. Validate the batch: classify_characteristic and the batch functions raise ValueError on a negative tolerance, an unknown feature type, a non-physical field, or an empty list.
3. Rate each record with classify_characteristic: verdict KC or non-KC with the fired-rule reasons.
4. Score the KCs with kc_risk_score (default weights, or pass an override weight set) and rank them with rank_key_characteristics.
5. Assign the variation plan with variation_management_plan: per-KC control method, Cpk target, sampling frequency, and verification gate.
6. For any production change, call change_trigger to decide whether the KC must be revalidated and what evidence is needed.
7. Produce the report lines with produce_kc_report.

Worked example

Six characteristic records from an actuator housing drawing and process history:

• BRG-BORE-01, bearing bore, hole, tolerance 0.025 mm, safety-critical, fit/function with performance downstream, capability known: KC (safety flag, fit/function), risk 65/100, plan SPC variable chart Xbar-R with Cpk target 1.67.
• HSNG-POS-02, mount hole position, position, tolerance 0.05 mm, fit with mate downstream, capability not yet demonstrated: KC (tight tolerance at or below 0.1 mm, fit/function), risk 55/100, plan gage study, Cpk target 1.33.
• CASE-EDGE-03, cosmetic edge break, other, tolerance 0.8 mm, no flags: non-KC, no rule fires.
• SEAL-FACE-04, seal face flatness, flatness, tolerance 0.05 mm, fit with seal downstream, capability known: KC (fit/function), risk 35/100, plan SPC variable chart Xbar-R, Cpk target 1.33.
• GROOVE-W-05, o-ring groove width, thickness, customer-designated, no other flags: KC (customer designation only), risk 0/100, plan SPC variable chart Xbar-R, Cpk target 1.33.
• PIVOT-PIN-06, pivot pin diameter, assembly-mate, tolerance 0.02 mm, fit with mate downstream, three historical failures: KC (fit/function, historical failures), risk 50/100, plan attribute go/no-go fit gage, Cpk target 1.33.

The ranked KC order is BRG-BORE-01 (65), HSNG-POS-02 (55), PIVOT-PIN-06 (50), SEAL-FACE-04 (35), GROOVE-W-05 (0); the report shows 5 key characteristics and 1 non-key. A tooling change on BRG-BORE-01 revalidates it: delta FAI scoped to the bore plus a capability re-study, so the contract test asserts the 1.67 target lands on the safety-critical bore.

Verification

• Confirm classify_characteristic(bore_record) returns KC with the safety reason, and classify_characteristic(edge_record) returns non-KC.
• Confirm the worked-example risk scores: 65, 55, 50, 35, 0, in that rank order.
• Confirm variation_management_plan assigns Cpk 1.67 to the safety-critical bore and 1.33 elsewhere, and that the gage study lands on the capability-unknown position feature.
• Confirm change_trigger(bore_record, "tooling") returns verdict True with delta FAI and capability re-study evidence, while "supplier" and "personnel" return verdict False.
• Confirm a negative tolerance, an unknown feature type, an empty batch, and an unknown change type raise ValueError.
• Run the contract test offline: python3 scripts/test_key_characteristic_management.py (35 tests, deterministic).

Related leaves

• manufacturing-quality/as9102/delta-fai: change classification and delta FAI scope that a KC revalidation calls on.
• manufacturing-quality/as9102/first-article-inspection: form 3 characteristic accountability where every KC is verified at first article.
• manufacturing-quality/as9100/statistical-process-control: owns the X-bar/R chart limits and the Cpk math; this leaf sets the target only.
• manufacturing-quality/as9100/nonconformance-control: disposition when a KC variation escapes control.
• cross-cutting/tolerancing/gdandt-basics: GD&T callouts and tolerance interpretation feeding the KC record (cross-cutting).

Pitfalls

• Ranking a customer-designated KC by its risk score: customer designation gates KC status by rule but carries no weight, so a customer-only KC scores 0/100 - the risk rank is not the importance rank for designation-driven characteristics.
• Reading KC status from the risk score: any single rule firing (safety flag, customer designation, fit/function with downstream impact, a position or profile feature at or below 0.1 mm, or two or more historical failures) makes the characteristic key, and the reasons list the fired rules in table order.
• Escalating every tight-tolerance feature: the tight-tolerance rule applies to position or profile features at or below the 0.1 mm threshold, not to every close dimension on the drawing.
• Applying the control method rules out of order: the table is first match wins, so electrical features get 100 percent inspection, two-or-more historical failures escalate to 100 percent inspection, and an undemonstrated capability gets a gage study before any SPC variable chart is assigned.
• Computing Cpk here: this leaf assigns the target only (1.67 for safety-critical, 1.33 default); the X-bar/R chart limits and Cpk index math live in the statistical-process-control leaf.
• Treating supplier and personnel changes as revalidation triggers: tooling, process, and design changes revalidate a KC (delta FAI plus capability re-study), while supplier and personnel changes route to external-provider qualification and training control instead.
• Passing non-physical batches: negative tolerances, unknown feature types or change types, empty batches, duplicate ids and override weight sets that do not sum to 100 raise ValueError.

Behavior contract (gate 3)

Run the deterministic contract test (stdlib unittest, offline):

python3 scripts/test_key_characteristic_management.py

The test covers the five-rule decision table (safety flag, customer designation, fit/function with mate/seal/performance downstream, tight position/profile tolerance boundary at 0.1 mm, historical failure threshold), the worked-example verdicts and risk scores, the 0-100 score with override weight sets, ranking with ties broken by id, the per-KC plan with Cpk 1.67 for the safety-critical bore and the method rule table (Xbar-R, gage study, attribute, 100 percent inspection), the change revalidation rule table with evidence, the report content contract, and ValueError rejection of negative tolerance, unknown feature type, empty batches, duplicate ids, non-physical fields, and invalid weights.

Compliance

• Standards referenced, not reproduced: AS9103 is named for context but is not yet in standards-map.yaml, so the frontmatter standards id is the parent QMS standard AS9100 (production process control frame). AS9103 wording and tables are not reproduced.
• compliance: STANDARDS-REF, gated: false.