A pilot inspection cell combining a UR10e collaborative robot and a Zeiss ScanBox structured-light scanner cut orthopedic implant inspection cycle time by 45–50% and delivered sub-50 µm mesh resolution on features held to a ±50 µm tolerance, the authors report.

What did the pilot actually validate?

The validated portion covers two subsystems: an in-house spray module that deposits AESUB Yellow developer across polished stainless-steel surfaces, and the Zeiss structured-light scanner that captures a 3D mesh for CAD-overlay analysis. Industrial IoT, contextualized data streaming, and electronic device history record (eDHR) connectivity remain design-stage only and were not part of the pilot.

The spray rig uses stainless-steel atomizing components from Spraying Systems Co., mounted on the UR10e's end effector. Spray paths come from part-specific digital recipes stored on the robot controller. Pressure gauges and regulators hold flow consistent across cycles.

After coating, the robot repositions the part under the scanner. Fringe projection and stereo triangulation generate the mesh. A constrained global least-squares best-fit aligns measured points to the CAD nominal, with selected degrees of freedom locked to functionally relevant datum references—a method consistent with ISO 5459 best practice.

How repeatable is the measurement system?

The team ran a comparative Gauge R&R across 30 parts, three operators, and two surface-preparation methods (manual versus robotic spray), following NIST Gauge R&R guidance. Measured features included thickness, profile deviation along CAD cross-sections, flatness, and hole diameter.

A separate run on a standard optical-metrology artifact produced a %GRR between 5% and 20%—a result the authors describe as "consistent with acceptable repeatability for pilot-stage optical metrology studies and well below the tolerance range of the evaluated features." That figure sits comfortably inside the ±50 µm tolerance band.

Robotic spray also cut developer-coverage variability by more than half versus manual application. That uniformity translated directly into fewer false deviation flags during scanning.

Why structured-light rather than CMM or laser tracking?

Structured-light scanners generate dense point clouds from a single fringe projection, producing full-surface deviation maps in seconds rather than the point-by-point probe path that a coordinate measuring machine requires. Laser trackers and CT scanning still hold accuracy advantages on tight-tolerance datum features, but they trade throughput and integration flexibility for that resolution.

The reflective stainless-steel surfaces forced the developer-spray prerequisite. Less reflective substrates could skip the coating step entirely.

Which standards does the architecture target?

The pilot scopes against ISO 13485:2016, FDA 21 CFR Part 11, FDA 21 CFR Part 820, EU MDR 2017/745, and ISO 14971. Validation work included Gauge R&R plus the manufacturer's ScanBox optical-calibration routine executed at every shift change against a certified artifact.

What remains on the drawing board?

The authors describe, but did not pilot, an Industrial IoT layer sourcing metadata from barcode readers, MES interfaces, and environmental sensors, then binding it to inspection results before forwarding to an eDHR. AI-based mesh defect detection, statistical-process-control trend monitoring, digital-twin simulation, and biometric operator authentication remain future work requiring their own validation pipelines under change-control.

Does the architecture scale?

Modular cells and standardized digital recipes enable replication across sites. The authors outline cloud-based dashboards and centralized data lakes for enterprise benchmarking. Until the IoT layer and AI modules clear validation, scale-up stays constrained by the validated boundary: UR10e robot, in-house spray rig, and Zeiss ScanBox, all coordinated through deterministic digital I/O.

The structural choice—validated physical cells decoupled from the conceptual digital stack—mirrors the validation posture regulators expect. Whether the unbuilt IoT, AI, and eDHR modules clear that same bar in production remains the open question for adoption under ISO 13485 and FDA 21 CFR Part 11.