How to Ensure Robot Safety: ISO 10218 and ISO/TS 15066

Step-by-Step Guide

1. 1

   Understand the Safety Standards Landscape

   Key standards: ISO 10218-1 (robot manufacturer requirements), ISO 10218-2 (system integrator requirements), ISO/TS 15066 (collaborative operation specifics), ISO 13849 (safety-related control systems), and regional regulations (OSHA in US, Machinery Directive in EU). These standards define what is legally required for safe robot deployment.

   text Copy

   Robot Safety Standards:
   ISO 10218-1: Robot manufacturer requirements
   ISO 10218-2: System integrator requirements
   ISO/TS 15066: Collaborative operation
   ISO 13849-1: Safety control system design (PL)
   ISO 12100: Risk assessment general principles
   EU Machinery Directive 2006/42/EC
   OSHA 1910 Subpart O (US)

2. 2

   Conduct a Risk Assessment

   Risk assessment is mandatory before deploying any robot. Process: 1) Identify hazards (crushing, impact, pinching, unexpected motion), 2) Estimate risk severity and probability, 3) Select safety measures (eliminate, guard, control, inform), 4) Validate effectiveness, 5) Document. Consider all operating modes (normal, maintenance, teaching, fault). Use ISO 12100 methodology.

   💡

   Tip: Involve operators in the risk assessment — they know the practical hazards that engineers miss. Document everything; in case of an accident, the risk assessment is your legal evidence of due diligence.
3. 3

   Implement Safety-Rated Monitored Stop

   The simplest collaborative mode: when a human enters the collaborative workspace, the robot stops (safety-rated stop, Category 0 or 1). The robot remains stopped while the human is present and resumes when the human leaves. Requires safety-rated sensors (light curtains, area scanners, safety mats) and a safety controller (SIL 2/PL d minimum). Used for hand-guiding and inspection tasks.

   ⚠️

   Warning: Safety-rated stops must be implemented with safety-certified components — standard PLCs and sensors are NOT sufficient. Use safety controllers (Pilz PNOZ, Allen-Bradley GuardLogix) and safety sensors (SICK microScan3, Keyence SZ-V).
4. 4

   Implement Speed and Separation Monitoring

   The robot adjusts speed based on human proximity. A safety-rated area scanner tracks the human position. Zones: green (far — full speed), yellow (approaching — reduced speed), red (close — stop). The robot maintains a minimum protective distance calculated from robot speed, human speed, and stopping time. This enables continuous operation while the human is present — more productive than safety-rated stop.

   text Copy

   Speed and Separation Monitoring:
   
   Protective distance: S = (v_robot + v_human) × T_stop + S_safety
   
   Zones:
     Green (>2m):   Full speed operation
     Yellow (1-2m):  Reduced speed (250mm/s max)
     Red (<1m):      Safety-rated stop
   
   Required: safety area scanner (SICK, Keyence)
             safety controller (PL d / SIL 2)
             robot speed monitoring

5. 5

   Implement Power and Force Limiting

   The most advanced collaborative mode: the robot is inherently safe by design. Force and speed are limited so that accidental contact causes no injury. ISO/TS 15066 defines biomechanical limits for 29 body regions (max force and pressure). Requirements: force/torque sensing (joint or end-effector), collision detection, and speed limitation. Cobots (UR, Franka, KUKA iiwa, ABB GoFa) are designed for this mode.

   💡

   Tip: ISO/TS 15066 biomechanical limits: max transient force on the hand = 140N, max pressure = 260 N/cm². These limits vary by body region — the face and neck have the lowest thresholds. Always test with a calibrated force measurement device.
6. 6

   Design Safety Control Systems

   Safety control systems must meet ISO 13849 Performance Level (PL) requirements. PL d (SIL 2) is the minimum for robot safety. PL e (SIL 3) is required for high-risk applications. Use safety controllers, safety-rated I/O, and safety communication (PROFIsafe, CIP Safety). Safety circuits must be separate from standard control — no single point of failure can compromise safety.

   ⚠️

   Warning: Never use standard PLCs or standard sensors for safety functions. Safety components have redundant channels, self-monitoring, and certified failure modes. A standard sensor failing could result in the robot not stopping when a human is present.
7. 7

   Select Safety Sensors

   Safety sensors detect human presence in the robot workspace. Types: 1) Light curtains (Type 4, ESPE) — create a detection plane, 2) Area scanners (SICK microScan3, Keyence SZ-V) — 2D LiDAR with safety certification, 3) Safety mats — pressure-activated floor mats, 4) Safety doors/gates — interlocked access points, 5) 3D safety cameras (SICK safeVisionary) — volumetric detection. Select based on workspace geometry and required detection capability.

   

8. 8

   Implement Collision Detection

   For power and force limiting, the robot must detect collisions and stop immediately. Methods: 1) Joint torque sensors (Franka, KUKA iiwa) — detect external torque deviations, 2) Motor current monitoring (UR) — estimate contact force from motor current, 3) External F/T sensors (ATI) — measure end-effector forces, 4) Tactile skins — distributed pressure sensors on the arm. Response time must be <80ms to prevent injury.

   python Copy

   # Collision detection via joint torque monitoring
   import numpy as np
   
   class CollisionDetector:
       def __init__(self, threshold=5.0):  # 5 Nm threshold
           self.threshold = threshold
           self.baseline = None
       
       def update(self, joint_torques, commanded_torques):
           external = np.array(joint_torques) - np.array(commanded_torques)
           if np.max(np.abs(external)) > self.threshold:
               return True  # Collision detected — stop immediately
           return False

9. 9

   Validate and Verify Safety

   Safety validation is mandatory before operation. Tests: 1) Emergency stop response time (must meet calculated stopping time), 2) Safety function activation (all sensors trigger correctly), 3) Force/pressure measurement (within ISO/TS 15066 limits for PFL mode), 4) Fault injection (verify safety system handles component failures), 5) Speed monitoring accuracy. Document all tests — this is your compliance evidence.

   💡

   Tip: Use a calibrated force/pressure measurement device (like the Robotouch from Pilz) to verify PFL biomechanical limits. Self-reported force values from the robot are not sufficient for compliance validation.
10. 10

    Train Operators and Maintenance Staff

    Safety is not just engineering — it is also operational. Training requirements: 1) Operators — recognize safety zones, understand stop conditions, know emergency procedures, 2) Maintenance — lockout/tagout procedures, safe teaching modes, 3) Supervisors — risk assessment updates when processes change. Document all training. Regular safety drills ensure readiness for emergencies.

    ⚠️

    Warning: Most robot accidents occur during maintenance and teaching, not normal operation. Maintenance mode often bypasses safety functions. Implement strict lockout/tagout procedures and never enter the robot workspace without disabling the system.
11. 11

    Plan for Incidents and Compliance

    Despite all measures, incidents may occur. Prepare: 1) Incident response plan (first aid, reporting, investigation), 2) Near-miss reporting culture (catch issues before they cause injury), 3) Regular safety audits (annual minimum), 4) Update risk assessment when processes change, 5) Maintain compliance documentation (CE mark, OSHA records, ISO certificates). Insurance may require evidence of safety compliance.