The engineering decision
Release production against a controlled design, demonstrated process window and traceable acceptance plan.
Have these inputs ready
- Released drawings and bill of materials
- Critical characteristics and process risks
- Pilot data across lots and operators
- End-of-line method and configuration records
Evidence changes at each release gate
Scroll horizontally to view the full diagram →
Separate design verification from process verification
A carefully assembled prototype may use selected cells, one operator and repeated manual corrections. The production process introduces lot variation, fixtures, tooling wear, weld setup, torque variation and software loading. The approved design needs a reproducible way to control those variables.
Design validation asks whether the construction meets the application requirement. Process validation asks whether the manufacturing system repeatedly builds that construction. End-of-line testing screens released units, but it cannot economically recreate every lifetime, environmental or destructive test.
| Gate | Question | Release evidence |
|---|---|---|
| Engineering build | Does the architecture work? | Schematic, load tests and resolved design risks |
| Design validation | Does the design meet its service envelope? | Controlled samples and verification matrix |
| Production pilot | Can the intended process repeat it? | Process settings, fixture checks and pilot distributions |
| Repeat supply | Does each batch conform? | Release records, trend review and change control |
A passing sample does not establish process capability
Consider a hypothetical characteristic with specification limits of 8 and 12 units, a process mean of 10.5 and standard deviation of 0.5. Cp is 1.33, but Cpk is 1.00 because the process is off-center. The smaller one-sided distance determines Cpk.
The conventional interpretation of these indices assumes a stable process, approximately normal data and independent observations. A short pilot or autocorrelated measurements can make the estimates misleading. Review chronological data, measurement uncertainty, sample size and distribution before attaching a capability label.
| Input / result | Value | Interpretation |
|---|---|---|
| Specification | 8 to 12 | Engineering limits |
| Mean / standard deviation | 10.5 / 0.5 | Illustrative stable-process estimates |
| Cp | 1.33 | Potential spread relative to width |
| Cpk | 1.00 | Upper side is limiting |
Turn failure modes into manufacturing controls
Use design and process risk analysis to identify characteristics that need prevention, in-process monitoring or final verification. For a weld, that may include a qualified process window, equipment checks, setup verification and sampled destructive evidence. A voltage reading at the pack output does not validate weld strength.
For BMS programming, retain hardware, firmware and parameter revisions separately. Verify the loaded configuration by readback or a checksum tied to the approved file. A unit can communicate normally while carrying incorrect protection settings.
Define containment when a process drifts: stop release, identify affected serials, inspect or rework under approval, and document the disposition. A reworked unit needs the relevant verification repeated; a final green indicator should not erase its history.
Avoid release gaps
| Risk | Control | Record |
|---|---|---|
| Cell lot variation | Incoming distributions and matching | Cell/lot-to-pack mapping |
| Joining variation | Qualified settings and verification | Machine/fixture revision and results |
| Incorrect configuration | Controlled programming and readback | Firmware and parameter checksum |
| Missed insulation defect | Defined inspection and electrical tests | Method, limits and measured results |
| Unapproved substitution | Engineering change review | Approved deviation and affected serials |
Measurement & validation procedure
| Test / review | Record | Acceptance condition |
|---|---|---|
| Review pilot variation | Time-ordered data across intended process conditions | Stable process and justified acceptance method |
| Challenge detection methods | Known reference conditions and fixture checks | Tests detect the failures they are assigned to screen |
| Trace a released unit | BOM, software, process and final results | The actual build can be reconstructed |
Continue this engineering task
Related engineering guidance
Prepare the evidence for review
- Prototype report and open issues
- Pilot process data
- Critical characteristics and control plan
- BOM, firmware and EOL revisions
Technical sources & claim support
References support the stated engineering principles. Worked examples use their explicitly stated assumptions.
- NIST/SEMATECH: Assessing Process Stability ↗
Stability of process mean and variance; chronological control charts
Statistical basis for checking stability before interpreting capability.
- NIST/SEMATECH: What is Process Capability? ↗
Cp, Cpk and assumptions for capability indices
Indices require an appropriate, stable process model and adequate data.
- NIST: Metrology and process control, dealing with measurement uncertainty ↗
Measurement uncertainty and manufacturing risk
Basis for considering uncertainty in acceptance decisions.
- ISO 9001:2026 — Quality management systems: requirements and official publication status ↗
Quality management systems; consistency of products and services
Quality-system context for documented information, process control and traceable production records.
Key claims and evidence
| Claim | Basis | Reference |
|---|---|---|
| Stability must be assessed before capability interpretation | documentation | NIST/SEMATECH: Assessing Process Stability |
| Cp and Cpk compare process spread/centering to limits | documentation | NIST/SEMATECH: What is Process Capability? |
| Proposed battery gate and control records are engineering recommendations | workflow | engineering workflow |