Poka-Yoke for Welded and Machined-Part Assembly

Poka-yoke, or error-proofing, uses a simple control to prevent a mistake or detect it before nonconforming work moves forward. Start with one documented defect mode, such as a bracket placed in the wrong orientation. The example below is fictional; it is a design and validation worksheet, not a real automotive supplier case or a substitute for engineering approval.

Describe the defect before choosing a device

Separate the mistake from the resulting defect. Record the part and drawing revision, operation, intended condition, observed failure mode, where it was made, where it was detected, and how much affected work may be in scope. Contain suspect parts through the quality system. Use failure mode and effects analysis (FMEA) when appropriate to review severity, occurrence, and detection with the responsible team. Follow the site’s scoring rules; do not treat a generic risk-priority score as automatic approval.

Illustrative wrong-orientation error-proofing concept

Panel	Work condition	Control idea	Question to resolve
1. Before	Two similar brackets can be loaded with the weld flange reversed	Map part presentation and the exact orientation error at the fixture	Can the error be prevented by part geometry or a safe keyed location?
2. Trial device	Fixture accepts the approved orientation and makes the wrong orientation visibly detectable	Consider a qualified physical key or a verified presence/orientation sensor	Can the device damage a part, create a pinch hazard, or pass a wrong variant?
3. Verified work	Correct variant loads, is clamped, and proceeds under approved welding controls	Document inspection, maintenance, bypass control, and reaction to a failed check	Who verifies the control after setup, repair, or revision change?

The diagram describes a concept only. A qualified engineer must review fixture geometry, clamping, loading ergonomics, guarding, and the applicable welding procedure. Do not alter a qualified process or defeat an interlock to achieve a production target.

Define the critical feature and the escape point

Start with the feature that makes the orientation matter. A bracket may have a critical weld location, hole pattern, mating face, or clearance requirement. Record the part number and revision, the intended orientation, the operation where an error can enter, the next operation that could receive it, and the existing inspection point. State what counts as an escape: for example, a reversed bracket that passes the current check and reaches the next assembly operation. The exact definition must match the drawing and the site’s quality procedure.

Map the work sequence from part presentation through locating, clamping, welding, cooling, and inspection. Mark each opportunity to rotate the part, confuse a similar variant, skip a locator, or bypass a check. Use observed work and nonconformance records. If the team has only one reported defect, label the count and period; do not turn it into a stable defect rate. Record good parts that the current inspection rejects as well as defects it misses.

Review the risk with the people who design, weld, inspect, and maintain the process. FMEA can help organize failure modes, effects, causes, current controls, and actions. Follow the site’s approved method and escalation rules. A risk score prioritizes discussion; it does not replace required customer approval, process qualification, safety assessment, or containment.

Baseline record before designing a poka-yoke

Record	Example for a mirrored bracket	Why it matters
CTQ feature	Weld flange faces the specified side of the assembly	Connects the control to the approved drawing requirement
Error opportunity	Two mirrored brackets can be presented to the fixture	Identifies when prevention or detection must act
Existing control	Operator visual check before tack welding	Shows the current inspection point and its limits
Escape point	Reversed part can reach downstream assembly if missed	Defines the latest safe point to detect or contain
Baseline evidence	Defect count, inspected quantity, time period, revisions, and shifts	Prevents unsupported rates and mixed populations

Compare prevention and detection options

Countermeasure selection for a wrong-orientation defect mode

Option	How it helps	Failure mode to test
Distinct part geometry or keyed fixture	Can make an incorrect orientation physically unable to seat	Wear, damage, similar variants, forceful loading, and safe ergonomic access
Presence or orientation sensor	Can inhibit the next operation when the defined condition is absent	False accept, false reject, dirty sensor, wiring fault, bypass, and reset behavior
Visual cue or locator label	Can help workers identify the correct orientation before loading	Visibility, revision changes, lighting, and whether the cue is routinely followed
Inspection check	Can detect a defect at a defined point before downstream use	Inspection coverage, measurement method, human error, and escape to later work

Prefer prevention when the team can make the mistake impossible without adding risk or unreasonable effort. A warning or inspection control can still be useful, but it does not prove the defect cannot escape. Define how affected material is identified and controlled when a device fails.

Write the stop and release logic

For a mechanical key, define the locating surfaces and tolerances that allow the correct part to seat while preventing the reversed part from reaching the approved clamp position. Confirm that fixture wear, weld spatter, distortion, or an unauthorized shim cannot create a false fit. Check that the operator can load and remove the part without reaching through a hazard zone or forcing the work against the locator.

For a sensor, define the exact input that proves the correct part and orientation. A part-present switch may confirm presence but not orientation. A vision check may depend on lighting, surface condition, camera position, and software configuration. The control should prevent the next process step when the required condition is absent or uncertain, then give a clear fault indication and authorized recovery path. Do not rely on a green light unless the team has verified what signal causes it.

Specify who can reset the device and what evidence is required. The operator may clear a known obstruction only after the work is safe and the device returns to its defined normal state. Maintenance may repair a sensor; engineering or quality may need to approve a changed location or tolerance. If the control is bypassed for maintenance, identify affected parts and apply the approved alternate inspection until the control is restored. The bypass log should include start time, reason, authorization, scope, and verified return to service.

Stop and release rules for an orientation control

Condition	Required system response	Release evidence
Correct part and orientation confirmed	Allow the approved operation to continue	First-off check and required process record
Wrong orientation or missing part	Block the next process step and identify the cause	Correct loading confirmed by the defined check
Sensor or fixture fault	Stop affected work; contain parts since last verified good check	Repair, functional challenge, and authorized release
Unapproved bypass request	Do not bypass; use the site’s escalation and alternate-control process	Documented authorization and traceable inspection, if approved
Part revision or variant change	Hold setup until the approved control covers the new variant	Change review, challenge tests, and updated work instruction

Validate the control before release

1. Approve the design. Engineering, safety, quality, and production review the fixture or sensor against the controlled drawing and process requirements.
2. Define test conditions. List each approved part variant, correct orientation, representative wrong orientation, sensor-failure state, and any setup or revision condition.
3. Use controlled test pieces. Mark and segregate known-good and intentionally incorrect samples; never introduce an unmarked test part into production.
4. Record each response. For each trial, record the sample, expected response, actual response, device state, operator, date, and disposition.
5. Check false accepts and false rejects. A false accept is an incorrect condition allowed to proceed; a false reject is a correct condition blocked. Investigate both before release.
6. Set verification and maintenance. Define who checks the device at setup and after repair, what happens on failure, how bypass is controlled, and how the system restores normal operation.

Use an engineering-approved validation plan to determine the number of trials and acceptance criteria. A small demonstration sample is not proof of long-term reliability or a statistically valid escape rate. Continue monitoring actual defect and nuisance-stop records after release.

Work an illustrative 100-trial validation

Illustrative validation design — not observed production performance. Before installing a fixture key on a production cell, the team prepares known-good test pieces and safe, clearly identified challenge pieces. The challenge set includes the correct orientation, reversed orientation, and no-part condition. The test takes place outside production or under a controlled, approved validation plan so no deliberately incorrect piece can enter customer work.

For illustration, the team runs 100 challenges: 50 correctly oriented parts, 25 reversed parts, and 25 no-part conditions. The fixture allows 48 of the 50 correct parts to proceed and blocks 2 correct parts because a locator is misaligned. It blocks all 25 reversed parts and all 25 no-part conditions. In this designed sample, wrong-condition detection coverage is 50 ÷ 50 = 100%. The false-reject fraction for correct parts is 2 ÷ 50 = 4%. These values describe only this controlled test set. They do not establish a field escape rate or predict long-term reliability.

Illustrative challenge-test results for a keyed fixture

Challenge condition	Trials	Allowed to proceed	Blocked	Interpretation
Correct orientation	50	48	2	Two false rejects require investigation before release
Reversed orientation	25	0	25	All designed wrong-orientation challenges were blocked
No part / incomplete seat	25	0	25	All designed absence challenges were blocked
Total	100	48	52	Record each challenge and the expected and actual response

The two false rejects are not harmless. The team investigates whether the fixture key, part variation, burr, weld distortion, or loading method caused the failure. If it revises the key, it repeats the applicable challenge tests and documents the new configuration. Acceptance criteria must be approved before testing; do not redefine a failure as acceptable after seeing results. Any challenge piece stays identified and is accounted for after the test.

For a sensor-based control, add fault challenges such as disconnected input, dirty lens, lost communication, and power recovery where the design makes them relevant. Verify that uncertain or failed states stop the process safely, that the alarm identifies the condition, and that reset does not release an unverified part. An independent qualified reviewer should confirm the test plan for safety-related controls or critical characteristics.

Maintain the control after release

Put a functional check into setup or first-off standard work. The check should use an approved test piece or method, state the expected result, and be recorded. Define the interval based on risk, device behavior, maintenance needs, and process changes. Inspect the locating surfaces for wear or damage. For electronic controls, verify sensor alignment, wiring, configuration, and alarm response after a repair or program change.

Trend both defects and false stops. A low defect count can mean the control works, the process changed, or inspection stopped recording the condition. A high false-reject count can cause operators to ignore alarms or request bypasses. Review the actual records with production, quality, engineering, and maintenance. If a part revision, variant, material, fixture, weld procedure, or inspection method changes, route the update through formal change control and revalidate the affected conditions.

Illustrative rework-cost example

Assume a fictional shop finds 8 orientation errors in 400 inspected assemblies during a defined month. If each rework requires 12 minutes at a loaded labor rate of $42 per hour, the direct labor estimate is 8 × 12 ÷ 60 × $42 = $67.20. This excludes inspection, material, delay, customer impact, and device cost, so it is not the total cost of poor quality. Record actual local values before evaluating a countermeasure.

Check performance without hiding risk

• Count defects by verified mode, operation, product revision, and detection point.
• Track correct parts blocked by the control and incorrect conditions accepted by it.
• Record downtime, bypass requests, maintenance, and false alerts.
• Revalidate after a fixture change, sensor repair, software update, product revision, or process change.
• Keep required inspection and customer controls unless the authorized quality system approves a documented change.

For quality at the source and process acceptance, see the Lean and quality guide. To give operators a defined escalation signal when a fixture or process is abnormal, see the Andon response guide.

Poka-yoke implementation checklist

1. Confirm the part, revision, CTQ, defect definition, and escape point.
2. Contain suspect product and establish the current defect and inspection baseline.
3. Map each orientation opportunity and identify prevention and detection options.
4. Review fixture geometry, process qualification, ergonomics, guarding, and failure modes with authorized owners.
5. Define normal, wrong-part, missing-part, fault, bypass, and revision-change responses.
6. Run and record approved challenge tests with identified test pieces; investigate false accepts and false rejects.
7. Assign setup verification, maintenance, monitoring, and revalidation owners before release.

Sources

• Lean Enterprise Institute: Poka-yoke — describes prevention and warning/error detection approaches.
• Lean Enterprise Institute: Problem solving — emphasizes verifiable facts and iterative countermeasure testing.