How Should Process Control Limits Be Set for Repeat CNC Lots?
How Should Process Control Limits Be Set for Repeat CNC Lots?
Process control limits for repeat CNC lots should be set from the characteristic, measurement system, subgroup, process state, and reaction rule rather than copied from a generic template. Monitoring is useful when it detects meaningful drift before the lot escapes, but a stable chart does not automatically guarantee a drawing tolerance. The buyer should identify the CTQs, define sampling and stop conditions, and keep the data tied to revision, unit, method, and state.
Choose Characteristics With Consequence
Rank features by their effect on fit, motion, sealing, load transfer, safety, interchangeability, and appearance. For each CTQ, state nominal or limit, datum frame, units, method, instrument, support condition, and whether the part is restrained, free, assembled, coated, or cleaned. A location result in one alignment does not silently prove another relationship; a visual check does not establish size or form.
The paired plate images show a rectangular surface, two narrow slots, and an edge feature from two angles. They cannot establish flatness, thickness, material, slot function, or tolerance. Use the geometry to ask which surface establishes the datum, whether slots locate a mating part, and how the edge is supported during measurement.
Set Subgroup and Frequency
Define the sample size, interval, subgroup boundary, and event that expands inspection. Parts after a tool replacement, new material heat, fixture change, operator handoff, or long pause may not belong to the same evidence population. Mark that boundary in the lot index. A frequency that is reasonable for a noncritical surface may be inadequate for a locating slot or sealing feature.
Use 100% checks where the consequence and method justify them, and use sampling where its basis is documented. The plan should state who owns the check, where the result is stored, and how a missing or invalid result is handled. Do not combine data from different states or revisions simply because the values look similar.
Define Limits and Reaction
Distinguish specification limits, control limits, warning signals, and stop criteria. A trend toward a limit may require process review or more frequent checks; an exceeded limit requires containment and disposition. State how much time or how many units are affected when a signal appears. Keep the original result and document any correction or offset change.
A measurement anomaly should trigger review of instrument, fixture, alignment, temperature, units, and operator state before a replacement result is accepted. If the method changes, perform the approved equivalence or requalification review. A CMM, gauge, probe, or vision system reports within its method boundary and cannot silently prove a different characteristic.
Write the Control Plan Clearly
| Plan field | Required decision | Common boundary |
|---|---|---|
| Characteristic | What failure matters? | Do not rank by ease of measurement |
| Method and state | How and when is it valid? | Do not mix restrained and free state |
| Frequency | What sample represents the lot? | Reopen after an event boundary |
| Reaction | Who stops and who approves? | Do not overwrite failed data |
The table is a control-plan aid, not a universal SPC prescription. A quality inspection planning reference can help organize ownership, but the drawing and risk review set acceptance.
Ask for Control Evidence in the RFQ
Provide revision, CTQs, risk ranking, measurement state, sample identity, records, and reaction authority. Ask suppliers to state proposed frequencies, chart or trend fields, tool and fixture triggers, and the event that expands inspection. Require notification before changes to program, tool, fixture, material, route, processor, inspection method, packaging, or ownership.
Do not accept a generic “SPC available” statement. Ask for an anonymized control-plan or record format showing characteristic, state, frequency, owner, reaction, and revision. This makes process control understandable when volume, shift, or line changes.
Buyer Action
Approve CTQ ranking, method, state, frequency, limits, and reaction before production. At receipt, verify that records identify the revision, units or samples, method, and lot boundary. Treat process control as bounded evidence and reopen the plan when geometry, material, fixture, tool, or demand changes.
Review the control plan when the subgroup definition changes. Mixing day and night shifts, different material heats, or separate machines can hide a shift that matters to the buyer. Keep subgroup labels in the record, and document why data may be combined or must remain separate. If the plan uses a warning signal, state who reviews it and how quickly the review occurs. A warning without an owner is not a reaction plan.
Calibration status belongs beside the result, not in a separate folder that cannot be reconciled. Record instrument identity and validity for each critical check, and hold data collected after an expired or questionable calibration until quality review determines whether a repeat measurement is needed.
