Removing the Light Curtain from a Palletising Cobot Cell
A Hong Kong plastics moulder has installed a 20 kg payload collaborative robot to palletise 8 kg totes at the end of a 12 m conveyor, 900 totes per hour over two shifts. The cell was originally fenced with a Type 4 light curtain mounted 900 mm from the robot's reach envelope. Every interruption needs a manual reset, and that happens about 40 times per shift because operators must clear rejected totes from a bench placed inside the cell. The production manager wants the curtain disconnected, relying instead on the robot's built-in power-and-force limiting, which the integrator has enabled. Your measurements: closest-approach speed 1.2 m/s, robot stopping time 400 ms, safety input and valve response 30 ms, and a peak contact force of 220 N at 240 N/cm² on a 1.5 kg sprung probe applied to the chest. The integrator's risk assessment is two pages long and concludes 'cobot - inherently safe'. No other guarding exists. The safety circuit is controlled by a safety PLC rated PL d.
1. Evaluate whether the curtain can be removed on the basis of the integrator's risk assessment and the robot's power-and-force limiting.
Read the worked answer
'Cobot - inherently safe' is not a risk assessment. ISO 10218-1 addresses the robot; ISO 10218-2 addresses the application - the gripper, the 8 kg tote, the conveyor transfer, the pallet stack and the people who walk into the cell. A two-page document that does not analyse the task, the payload or the approach directions does not meet ISO 12100 / ISO 10218-2 requirements and cannot support removing a protective device.
Do the measured forces permit collaborative operation? Not at this operating point. ISO/TS 15066 sets body-region-specific quasi-static contact limits; for the chest these are of the order of 140 N and 50 N/cm². The measured 220 N and 240 N/cm² exceed both by a wide margin, so power-and-force limiting is not achieved.
A second, separate problem: the measurement method matters. The Annex A method prescribes probe mass and spring constant per body region, so a 1.5 kg sprung probe is only a valid model for the region whose effective mass and stiffness it matches. Document which region it represents and measure the others properly - a report that says 'chest' with the wrong probe will be rejected by any competent reviewer.
Hazards the curtain is currently masking
- An 8 kg tote plus gripper is a large moving mass with corners, and there are crush and shear points between the tote and the pallet. The robot's internal force limiting does not protect against a hazard created by the payload.
- 40 interruptions per shift is a design defect, not an operator habit. The reject bench is inside the restricted space, so the cell is being defeated by its own layout.
Conclusion: the curtain stays until a task-based risk assessment, validated by a competent person, shows either that the hazards are eliminated or that the power-and-force limits are met with measured evidence at every approach. Neither condition is currently satisfied.
2. Calculate the minimum safety distance for the light curtain and propose engineering alternatives that would meet the production need without defeating the device.
Read the worked answer
Minimum distance calculation (ISO 13855) S = K × T + C, with the approach speed K = 1600 mm/s (valid when S > 500 mm).
- Total response T = robot stop 0.400 s + safety input and valve response 0.030 s = 0.430 s (add the curtain's own response time if it is not already included in the safety PLC figure).
- Intrusion factor C for a vertical light curtain: C = 8(d − 14) for detection capability d ≤ 40 mm; C = 850 mm for d > 40 mm.
| Detection capability | C | Required S |
|---|---|---|
| 14 mm | 0 mm | 1600 × 0.43 = 688 mm |
| 30 mm | 128 mm | 816 mm |
| > 40 mm / perimeter detection | 850 mm | 1538 mm |
So the installed 900 mm is adequate only if the curtain's detection capability is around 40 mm or finer, the reset device is positioned outside the field with a full view of the cell, and there is no reach-over, reach-under or reach-around. If the curtain is a 50 mm or coarser type, or if the field can be entered from the side, 900 mm is short and the device is not providing the protection it appears to provide.
Engineering options, in order of preference
- Move the reject bench out of the cell. This removes most of the 40 interventions per shift and costs almost nothing. Add a second reset station outside the field, positioned so the operator can see the whole cell.
- Safety-rated speed and separation monitoring. Reduce speed in the collaborative zone - force and pressure fall steeply with speed - until measured contact at every body region is below the ISO/TS 15066 limit. Implement it as a safety-rated function on the PL d safety PLC. This is usually the cheapest route to genuine collaborative operation.
- Replace the vertical curtain with a horizontal-plane safety laser scanner and mute only the tote path, with the muting logic designed per IEC 62046 and arranged so that a person cannot enter behind the muted material.
- Fixed guards around the pallet stack and the conveyor transfer, leaving the collaborative function only where it is actually needed.
Evidence required: task-based risk assessment; validated safety functions with measured stopping time; force and pressure measurements using the prescribed Annex A method with calibrated instrumentation; updated operating instructions and training records; and a revised declaration for the cell.
3. The production manager repeats the instruction to disconnect the curtain, saying it is 'only a formality'. Set out your response.
Read the worked answer
Response: I would not disconnect it, and I would say so in writing the same day.
Legal As the engineer responsible for the plant, the duties under the Factories and Industrial Undertakings Ordinance (Cap. 59) and the Occupational Safety and Health Ordinance (Cap. 509) require plant and work systems to be safe so far as reasonably practicable. Disabling a protective device without a risk assessment and without equivalent protection converts a compliant cell into an unguarded machine. If someone is injured, the instruction transfers very little liability to the person who gave it.
Technical The evidence is against it: 220 N and 240 N/cm² at the chest, against limits in the order of 140 N and 50 N/cm². 'Only a formality' is contradicted by measurements taken in that cell.
Commercial Altering a CE/UKCA-marked assembly or its safety circuit without re-validation can void the integrator's declaration of conformity, the machinery insurer's cover and the customer's supplier approval. That is a bigger cost than 40 resets a shift.
How to handle it
- Reply in writing: the measured figures, the clause of ISO/TS 15066 they breach, the fact that you cannot authorise the change, and the alternative you propose. Copy the safety officer and the plant manager.
- Solve the production problem so the pressure disappears: relocate the reject bench, add a second reset station outside the field, or run a temporary reduced-speed collaborative mode with measured verification while the layout change is engineered. Give a date.
- Raise it as a management-of-change item. Any change to a safety function requires the same validation as the original - that is the correct process route, and it is also the one that gets the change actually made.
- If the instruction is repeated, escalate to the duty holder or director responsible for EHS. If it is still insisted on, put the objection in writing and consider reporting the unguarded machine to the Labour Department.
Refusing to sign a false declaration is the minimum. The professional obligation extends to not leaving an unsafe machine running while the argument continues.