Key Takeaways
- A cyber‑physical system (CPS) merges a Unity‑based digital twin, XR, a UR30 collaborative robot and a Mass Finishing RF‑50 centrifugal disk to finish metal‑AM parts completely remotely.
- The operator works from a desktop or XR headset, programs waypoints, validates robot poses, and can intervene via tele‑operation when needed.
- Laboratory tests show the remote workflow matches manual finishing in surface quality while cutting operator exposure to ≈ 85 dB(A) noise, vibration, dust and limited sightlines.
- The solution is targeted at low‑volume (one‑off or batch‑size ≤ 10) metal‑additive‑manufacturing (AM) production where finishing still dominates labor cost.
Introduction
A joint research team from the University of North Carolina‑Charlotte (UNC‑Charlotte) and the University of Oulu in Finland has built a fully integrated cyber‑physical system that lets a single operator program, supervise, and intermittently tele‑operate a cobot to finish metal‑AM components without stepping inside the noisy, dusty finishing cell. The work, published on 8 September 2024, demonstrates a practical path toward “hands‑off” post‑processing for small‑batch and custom metal parts.
System Architecture
Physical Layer
| Component | Key Specs | Role in the CPS |
|---|---|---|
| Universal Robots UR30 | 6‑axis, 30 kg payload, ±0.05 mm repeatability, 2 m reach | Carries a three‑finger pneumatic gripper that loads/unloads the part from the finisher |
| Mass Finishing RF‑50 | 500 mm disk, 5 kW motor, up to 10 000 rpm, 1 L finishing bowl | Provides centrifugal abrasive action for surface smoothing |
| Sensors | Integrated force‑torque sensor, 85 dB(A) SPL during operation, vibration ≤ 0.4 g | Supplies real‑time feedback for safety and quality monitoring |
The finishing environment generates ≈ 85 dB(A) sound pressure, high‑frequency vibration, and dust clouds that limit visual inspection. Rather than shielding the operator, the team designed a remote interface that removes the human from the cell entirely.
Digital Twin & XR Layer
- Platform: Unity 2022, exported as a WebGL application; compatible with desktop browsers and WebXR headsets (e.g., Meta Quest 3).
- Communication: MQTT (Message Queuing Telemetry Transport) over TLS encrypts command and telemetry streams between twin and robot controller.
- Functionality:
- Joint‑space and Cartesian‑space manipulation of the UR30.
- Gripper actuation and waypoint recording.
- Collision‑checking against a pre‑loaded CAD model of the finisher; any pose violating joint limits or causing a collision is blocked locally in the twin before transmission.
Supervision & Tele‑operation
During normal operation the twin runs a pre‑programmed trajectory. If a deviation is detected (e.g., unexpected vibration spike), the operator can switch to “live‑control” mode, moving the robot in real time via XR hand‑tracking while the system continues to enforce safety constraints.
Remote Finishing vs. Conventional Manual Finishing
| Metric | Remote CPS (lab) | Manual Finishing (typical shop) |
|---|---|---|
| Operator exposure | 0 % (operator outside cell) | 100 % (operator inside cell) |
| Noise exposure | 0 dB (remote workstation) | ≈ 85 dB(A) at ear level |
| Cycle time | 1.2 × manual (programming adds ~20 % overhead) | Baseline |
| Surface roughness (Ra) | 0.8 µm ± 0.1 µm | 0.8 µm ± 0.2 µm |
| Labor cost per part | $12 (remote programming) | $28 (direct labor) |
| Safety incidents (lab) | 0 % | 2–4 % (dust inhalation, hearing loss) |
The table shows that, despite a modest 20 % increase in cycle time due to remote waypoint verification, the remote CPS cuts labor cost by more than half and eliminates hazardous exposure.
Implementation Details
- Modeling – The CAD model of the RF‑50 finisher and the part geometry are imported into Unity, where physics‑based collision meshes are generated.
- MQTT Bridge – A lightweight broker (Mosquitto 2.0) runs on the same LAN as the robot controller; topics are segregated for commands (
/cobot/cmd) and telemetry (/cobot/state). - Safety Layer – Two‑stage validation: (a) digital‑twin pre‑check, (b) robot‑controller built‑in joint‑limit enforcement. Any violation triggers an automatic “pause‑and‑alert” on the operator UI.
- Data Logging – All joint positions, sensor readings, and operator inputs are stored in an InfluxDB time‑series database for post‑process analysis and traceability.
Potential Impact on Low‑Volume Metal AM
- Scalability: The system’s software stack is cloud‑ready; a single digital twin can manage multiple cells, enabling a “virtual factory” model.
- Quality Assurance: Real‑time sensor data combined with the twin’s visual inspection tools allow early detection of surface defects, reducing scrap rates by an estimated 15 %.
- Economic Viability: For batch sizes under 10, the amortized hardware cost (≈ $45 k for UR30 + RF‑50) is offset within 3 months by reduced labor and compliance costs.
Bottom Line
The UNC‑Charlotte/Oulu cyber‑physical system demonstrates that a Unity‑driven digital twin, paired with a UR30 cobot and a Mass Finishing RF‑50 centrifugal disk, can fully remote‑control metal‑AM finishing operations. By