Overview: Why Collision-Only Assessments Fail
The most common defect in a UR10 (or any collaborative robot) risk assessment is scope collapse: the team fixates on human-robot contact forces and pressures, closes the file, and ships the cell. Vendor-supplied assessment templates encourage this because their worked examples are almost entirely collision scenarios. A collaborative arm removes fencing; it does not remove trip hazards from unsecured cabling, an under-anchored pedestal, sharp tooling edges at torso height, missing signage, or an unsafe maintenance access position.
Treat the collaborative arm as one hazard source inside a machine. The assessment object is the application — robot, end effector, workpiece, fixture, conveyor, guarding, pendant, electrical enclosure, and the people who touch all of it across its life.
Assessment Structure
Build the assessment as a matrix: life-cycle phase × task × hazard category. Every populated cell gets a hazard entry, a risk estimate before mitigation, a control measure, and a residual risk estimate after mitigation.
Life-Cycle Phases (Rows)
- Transport, installation and commissioning — lifting the arm and controller, pedestal anchoring, first power-up, payload/CoG entry, TCP definition, safety configuration entry and password protection.
- Teaching / (re)programming — freedrive handling, jogging near fixtures, program test at reduced speed, operator standing inside the reach envelope.
- Normal production — part load/unload, cycle start, cycle stop, cell entry frequency, throughput-driven shortcuts.
- Fault finding and recovery — clearing jams, retrieving dropped parts, protective-stop reset, restart from an unknown pose.
- Cleaning and maintenance — joint or tool service, lubricating adjacent machinery, energy isolation (electrical, pneumatic, gravity, stored spring), reaching over the robot.
- Modification and decommissioning — tool changeover, program version change, removal and disposal.
Fault finding, cleaning and changeover generate the highest-severity findings in most cells, because these tasks put a person deep inside the envelope with automatic mode potentially available.
Hazard Categories (Columns)
| Category | Typical UR10 application entries |
|---|---|
| Mechanical | Clamping/crushing against fixtures and walls, shear at joints, sharp end-effector and workpiece edges, dropped payload on gripper power loss, tool ejection at speed, unstable pedestal or cart, pinch between arm and conveyor guard |
| Electrical | Enclosure access while energized, unbonded metal tooling, improper grounding of the cell frame, cable insulation abrasion at drag points, arc flash at upstream disconnect, incorrect fusing of tool 24 V supply |
| Ergonomic | Awkward reach into fixtures, teaching posture with pendant, repetitive manual load/unload, poor lighting at the work point, mental overload from unclear HMI state indication |
| Environmental | Trip hazards from robot, tool and pendant cabling; spills and slip hazards; noise from adjacent equipment; fumes/coolant mist; heat; dust ingress; workspace intrusion by pedestrian traffic |
| Human factors / information | Missing or wrong signage, no collaborative-operation marking, undocumented residual risk, untrained temporary staff, unlabelled emergency stop locations |
Running the Session
Team size is a real control variable, not administrative trivia. Field practice converges on two to three participants: typically the controls/robot integrator, the process owner or line operator, and a safety or maintenance representative. Larger groups drift into speculative scenarios, dilute ownership of actions, and stall on documentation. Where broader input is legitimately needed (facilities, quality, ergonomics), take it as targeted written input rather than seating twelve people in one review.
- Define limits first. Space (reach envelope plus tool plus workpiece plus human approach zones), time (duty cycle, expected service life, maintenance intervals), and use (intended use plus reasonably foreseeable misuse — leaning on the arm, using it as a hand rest, entering to speed up a cycle).
- Walk the layout physically. Mark the envelope on the floor with tape, including tool and part extension, before estimating any risk.
- Identify hazards task by task. Work the life-cycle rows in order; for each task, sweep every hazard category. Record low-frequency events explicitly — "once a year" events still need an entry and a documented decision.
- Estimate risk using a consistent scheme (severity, exposure frequency/duration, probability of occurrence, possibility of avoidance). Use one scheme for the whole file; do not mix vendor scoring with an internal matrix.
- Apply the reduction hierarchy: inherently safe design first (remove pinch points, deburr tooling, reduce payload, reposition fixtures), then safeguarding and complementary protective measures (speed and separation limits, safety-rated stops, area scanners, light curtains, mats, fixed guards on the process side), then information for use (signage, markings, manuals, training).
- Re-estimate residual risk and repeat until acceptable. Document the reasoning, not just the score.
- Verify and validate each measure on the actual cell, then sign, date, and version the document.
Documentation Template Sources
Do not start from a blank page. Practical starting points used by integrators include vendor-published collaborative-application assessment templates and workbooks from robotics industry associations (in North America, the Robotic Industries Association / A3 form set is a widely used baseline). Whichever form you adopt, extend the hazard column set beyond collisions before the first session, or the template's example bias will shape your findings.
Cell-Specific Checks Engineers Routinely Miss
| Check | Why it matters |
|---|---|
| Safety configuration lock | Safety limits (speed, force, planes, tool position) must be password-protected and the configuration checksum recorded in the file; an unrecorded change invalidates the assessment. |
| End-effector geometry | Robot-side force limiting does not compensate for a 3 mm-radius gripper finger or a sheet-metal part edge. Deburr, add radii, add compliant covers. |
| Payload release on stop | Define gripper behaviour on protective stop, emergency stop, and power loss. Spring-closed or mechanically latched grippers prevent part drop. |
| Pedestal / cart stability | Verify anchoring or ballast against worst-case dynamic loading and against a person leaning on the arm. Mobile carts need locked casters and a tip calculation. |
| Cable routing | Route controller, tool and pendant cables in trays or under covers. Cable-based trips are among the highest-frequency incidents in fenceless cells. |
| Approach zones and traffic | Assess pedestrian and forklift routes adjacent to the envelope; add floor marking, bollards or a low rail where traffic passes within reach. |
| Signage and marking | Mark the collaborative workspace, state residual risks, and identify emergency stop locations. Missing signage is a documentation nonconformity even when the mechanics are safe. |
| Energy isolation | Provide a lockable disconnect and document isolation of pneumatics, vacuum and gravity-loaded axes for maintenance tasks. |
| Adjacent machinery | The press, conveyor or CNC the robot tends usually carries higher-severity hazards than the arm itself. Assess the integrated machine, not the robot alone. |
Verification and Living Document Control
- Confirm every protective measure functions on the built cell: trigger each emergency stop and safety input, verify safety planes and reduced-mode transitions, and confirm stop behaviour from worst-case pose and speed.
- Where force/pressure limits are claimed, verify them by measurement on the actual tooling and worst-case contact geometry, and retain the measurement records with the assessment.
- Record training and hand-over: operators, maintainers and temporary staff must be briefed on the residual risks listed in the file.
- Re-open the assessment on any change to tooling, payload, program path, safety configuration, cell layout, adjacent equipment, or operator task allocation. Version-control the document and keep superseded revisions.
- Schedule a periodic review (typically annual, or per your internal management-of-change procedure) even without changes, to capture drift such as pallets stored in the envelope or removed floor markings.
FAQ
Which standard defines the risk assessment procedure for a UR10 cell?
The general machinery methodology is ISO 12100. Confirm the exact current edition and the applicable robot-specific standards for your region and contract, since older three-part numbering (ISO 12100-1/-2 with ISO 14121-1) still appears in circulated templates and causes citation errors.
Do I still need a risk assessment if the robot is collaborative?
Yes. Collaborative capability applies to the robot arm, not to the application. Grippers, workpiece edges, adjacent machines, cabling, stability and electrical hazards are unaffected by force limiting and must be assessed for the integrated cell.
How many people should be in the risk assessment team?
Two to three: the controls/integration engineer, the process or line owner, and a safety or maintenance representative. Larger groups generate speculative scenarios and stall the documentation; gather additional specialist input in writing instead.
What hazards get missed most often in collaborative cells?
Trip hazards from robot, tool and pendant cabling; unstable pedestals or carts; sharp end-effector and workpiece edges; payload drop on power loss; and missing signage or documented residual risk.
When must the assessment be redone?
On any change to tooling, payload, program path, safety configuration, layout, adjacent equipment or operator tasks — plus a scheduled periodic review to catch workspace drift such as stored material inside the envelope.