Automated Inspection Quality Decision: Positioning First
For automated inspection quality decision, start with the reference and production parts, feature and presentation range, interfaces, measurement decision criteria, and exception routes. A completed positioning move is not a release record. Validate datum, calibration, and signal prerequisites, retain part and feature identity, challenge invalid measurements, and use reference-sample trials plus inspection. This guide is an application-review method, not a project-specific release.
Keep position, signal, and verdict separate
- Define the physical input window before selecting or programming the robot.
- Treat each positioning or acquisition command as a request and each validated datum or sensor condition as permission.
- Keep part, joint, package, or result identity through every acquisition and decision transition.
- Force abnormal conditions during trials; a nominal cycle is not sufficient evidence.
- Release only against project-specific measurements, inspection, interfaces, and safety validation.
Automated Inspection Quality Decision: use the Position-Measure-Decide Ledger
The intended reader is quality and automation teams integrating robots with measurement or inspection equipment. The decision is to separate part location, tool position, measurement signal, acceptance rule, sample reference, result identity, and exception disposition. The common shortcut is using robot arrival or a completed scan motion as evidence that a valid measurement and quality decision occurred. That shortcut fails because A robot can reach the intended point while the sensor is unready, outside calibration, saturated, or measuring the wrong feature.
EVST’s Position-Measure-Decide Ledger keeps feature and part identity connected to datum, calibration, acquisition permission, signal evidence, quality decision, and result routing. It is intentionally stricter than a video review. Video can confirm that equipment and operation are present; it cannot establish a universal cycle, quality result, accuracy, stability, or throughput.
ISO 10218-2:2025 spans the lifecycle of robot applications and cells from integration through decommissioning. EVST includes inspection setup, calibration, automatic measurement, sample intervention, recovery, maintenance, and eventual changes. (ISO 10218-2:2025 — Industrial robots and robot applications — Part 2 ISO 12100:2010 — Safety of machinery — Risk assessment and risk reduction OSHA Robotics NIST Performance Assessment Framework for Robotic Systems)
Define the feature and measurement range first
For automated inspection quality decision, the input state includes part and feature definition, datum, presentation range, robot path, sensor and measurement principle, calibration plan, uncertainty needs, acceptance rule, reference samples, result routing, and target cycle. Inspection inputs should identify the measurable range, the authoritative specification, ownership of calibration and data, and the outcome of an invalid signal.
A reference part can give a clean scan while calibration, feature access, presentation, and signal range remain undefined. Those measurement limits belong in the requirement before programming. EVST expresses each measurement input through part and feature identity, valid signal range, calibration proof, timeout, unknown-result route, and reacquisition rule.
ISO 12100 places hazard identification and risk evaluation across relevant operating phases. EVST includes calibration, sample change, setup, inspection, cleaning, sensor service, and recovery in this station boundary.
Do not let robot arrival stand in for sensor readiness
The working process is to identify and locate the part, position the tool, confirm sensor readiness, acquire the measurement, validate the signal, apply the declared decision rule, bind the result to the part, and route every unknown state. The equipment set includes robot, end-of-arm measurement or inspection equipment, fixtures, sensors, calibration artifacts, reference samples, controller, data interface, guarding, safety controls, and result-routing equipment. These must be connected through explicit interfaces: part identity, fixture and datum valid, robot position achieved, sensor ready, calibration valid, acquisition complete, signal quality, decision result, traceability complete, and disposition available.
Robot in-position is not sensor ready, valid calibration, or a passing quality verdict. Gate acquisition and disposition with measurement evidence; time may support stabilization but cannot create a valid signal.
| Decision point | Required evidence | Reject the shortcut when |
|---|---|---|
| Input accepted | Identity and declared range are valid | The real part or state is unknown |
| Equipment permitted | datum, calibration, and signal prerequisites and safety conditions agree | Permission relies only on elapsed time |
| Process complete | The physical operation and data record are complete | Robot motion finished but result is missing |
| Result released | Acceptance rule and identity are linked | A generic OK cannot be traced to the active item |
| Restart allowed | A conservative state and failed prerequisites are revalidated | Recovery resumes from assumed history |
Challenge calibration and false-result behavior
Verification should cover positioning repeatability, feature coverage, sensor readiness, calibration status, signal validity, measurement-system analysis where applicable, acceptance logic, false result challenges, traceability, and recovery. The inspection trial names the part and feature set, reference and calibration condition, measurement method, verdict limits, retained fields, and unknown-result disposition. NIST organizes robotic assessment around observable requirements, metrics, and repeatable tests. EVST uses those principles to separate positioning, signal validity, and the resulting quality decision.
The inspection record binds part and feature identity to fixture, sensor, calibration, software version, time, signal validity, verdict, disposition, and reacquisition action. Missing evidence produces a hold.
Create bad signals without losing part identity
| Trial | Forced condition | Expected controlled response |
|---|---|---|
| 1 | robot positioning completes while sensor readiness or calibration is invalid | Put the active part or joint on hold; record the physical state; revalidate the failed prerequisite before retry. |
| 2 | the signal is acquired from the wrong feature or outside its valid range | Put the active part or joint on hold; record the physical state; revalidate the failed prerequisite before retry. |
| 3 | the acceptance rule cannot be linked to the active part or sample state | Put the active part or joint on hold; record the physical state; revalidate the failed prerequisite before retry. |
| 4 | positioning error and quality result are collapsed into one ambiguous alarm | Put the active part or joint on hold; record the physical state; revalidate the failed prerequisite before retry. |
An invalid acquisition exposes conflicts among robot, sensor, controller, and quality database. Assign ownership for feature identity, calibration, signal validity, verdict, timeout, and disposition.
Count calibration and sample handling in inspection time
The hazard scope includes robot motion, pinch points, probes or tools, lasers or other sensing energy where used, sharp parts, unexpected restart, and access during calibration, sample change, or recovery. The cycle model includes part identification, datum confirmation, positioning, sensor readiness, acquisition, validation, decision, result recording, routing, calibration checks, and recovery. Record part handling, datum confirmation, positioning, calibration checks, acquisition, validation, decision, routing, sample change, and recovery. One fast scan does not establish inspection capacity.
According to the cited safety and process standards, safeguards and operating procedures depend on the real application. EVST therefore treats the final robot model, tooling, layout, protective measures, controls, process qualification, inspection, and recovery procedure as project deliverables rather than video claims.
Measurement facts needed before cell review
- part, feature, datum, and presentation specifications
- sensor principle, calibration, uncertainty, and sample requirements
- robot path, fixture, interface, result, and traceability definitions
- acceptance, false-result challenges, exception routing, safety, and target-cycle requirements
With measurement inputs, EVST can examine robot access, sensor and fixture selection, calibration, data interfaces, safeguards, decision logic, sample trials, and result routing. Unverified fields stay explicit until reference tests or analysis close them.
Related EVST engineering resources
- Collaborative robot application planning
- Industrial robot architecture and application range
- Payload, reach, and robot-type selection
Questions quality teams ask about automated inspection
Is robot repeatability the same as measurement capability?
No. Robot positioning is one contributor. Sensor principle, calibration, feature access, fixture behavior, signal quality, uncertainty, samples, and the decision rule determine the measurement result.
What should the station record for every part?
Retain part and feature identity, datum status, calibration state, acquisition result, signal validity, decision rule, verdict, time source, and final disposition.
Why test an intentionally bad reference sample?
It shows whether the station can detect a known nonconforming condition and whether the result reaches the correct route without being confused with a positioning fault.
What should happen to an invalid signal?
Return an unknown measurement state and hold the part. Never convert saturation, missing data, or expired calibration into an automatic pass.
References
- ISO 10218-2:2025 — Industrial robots and robot applications — Part 2: integration, commissioning, operation, maintenance, and decommissioning of industrial robot applications and cells.
- ISO 12100:2010 — Safety of machinery — Risk assessment and risk reduction: hazard identification, risk evaluation, risk reduction, documentation, and verification across machinery life-cycle phases.
- OSHA Robotics: robot-system hazards, safeguarding context, and related OSHA resources.
- NIST Performance Assessment Framework for Robotic Systems: observable requirements, metrics, and repeatable test methods for robotic-system performance assessment.