ISO/TS 15066 Force Limits for Welding Cobots: What the Standard Actually Requires

Updated: August 13, 2026

ISO TS 15066 cobot welding safety is not a matter of typing one force number into a robot controller. Defensible welding applications start by identifying every foreseeable contact, classifying the body region and contact type, measuring the applicable force and pressure, and preserving the tested configuration as evidence.

What ISO/TS 15066 Requires, and What It Does Not

One common application-risk mistake is treating an industrial standard as a product label. ISO’s catalogue entry for ISO/TS 15066:2016 still lists the document as published and current, while its lifecycle stage says it is to be revised. That status does not certify an individual welding cell.

Within this standards set, the technical specification adds collaborative robot safety guidance while ISO 10218 carries broader safety requirements for industrial robots and their applications. Collaborative operation may also use measures such as a safety-rated monitored stop or speed and separation monitoring; those methods are distinct from power and force limiting, and no cobot safety or safety standard label selects the right method automatically.

Normalize RFQ language before choosing a test

Specifications often mix robot safety standards, product descriptions, and control-language shorthand. Treat the following phrases as prompts to define scope, edition, evidence, and ownership—not as proof that a requirement applies or that safety compliance has already been established.

7-Node Force-Limit Decision Chain

Limits sit at the end of the decision chain, not the beginning. If a team starts with “What force number can we enter?”, it has skipped the questions that determine whether contact is acceptable and which criterion applies.

1. Define the human task, including production, loading, teaching, debugging, maintenance, repair, and recalibration.
2. Remove avoidable contact by changing the layout, access, fixtures, sequence, or safeguards before trying to justify a contact with a number.
3. Identify every moving interface: the arm, torch, wire, cable dress, workpiece, positioner, fixture, and carried part.
4. Classify the event by deciding whether the person can recoil from a transient impact or may be trapped in a quasi-static contact.
5. Select the body region and geometry because a broad padded surface and a narrow workpiece edge don’t create the same pressure distribution.
6. Freeze and reproduce the worst case, recording payload, tool, path, speed, reach, direction, effective mass, fixture, and counter-surface.
7. Compare every applicable quantity. Where the selected criteria specify force and pressure, both must pass; otherwise redesign the contact or add safeguarding.

Pass-Both Rule for Applicable PFL Contacts

Force describes the overall interaction, while pressure depends heavily on the area and shape over which that force acts. Broad compliant padding can reduce peak pressure without necessarily reducing total force; a narrow torch bracket or workpiece corner can concentrate pressure even when the force reading looks modest.

“As soon as either the force or the pressure limit is exceeded, the requirements are not fulfilled.”

DGUV FB HM-080, guidance for power-and-force-limited robot systems

That qualification matters. NIST’s biomechanical-limits workshop record notes that pain onset can be expressed through quantities such as force or normal stress and that impactor shape affects validity. Apply the pass-both rule where the selected PFL criteria set both limits; do not turn it into an unsupported universal statement.

Classify Every Contact With a Contact-Map Test Matrix

Free impacts and trapping events aren’t interchangeable. Use this matrix as a planning template, not a prefilled compliance record; replace every row with the application’s real tasks, surfaces, body regions, configurations, criteria, and measurements.

How to Build a Defensible Validation Test

The contact map defines what a credible test must reproduce: the assessed contact, relevant body response, and worst-case application energy. Controller screenshots can’t show local pressure, trapping geometry, instrument response, or whether the tested program and payload match production.

1. Freeze the bill of configuration. Record the robot, payload, torch, wire, cable dress, workpiece, fixture, positioner, program revision, safety parameters, and cell layout.
2. Derive test points from the risk assessment. Convenient points on the arm don’t replace the surfaces a person can actually encounter.
3. Select the measurement method. The ISO/PAS 5672:2023 catalogue describes the document’s force-and-pressure test-method scope and device-characteristic scope.
4. Match the simulated body response. Document the biofidelic element, contact area, mounting, instrument identity, calibration, and validity for the selected scenario.
5. Reproduce the worst case. Test reach, direction, speed, effective mass, stopping behavior, trapping clearance, and counter-surface under the frozen configuration.
6. Capture raw evidence. Preserve force and pressure traces, acquisition rate, filtering, repeats, uncertainty, setup photos, and the analysis method.
7. Close the loop. If a criterion fails, change the energy, geometry, surface, path, clearance, or safeguarding, then repeat the affected test and approve the final result.

Nine-record validation handover ledger

A validation result becomes reusable project evidence only when the test record connects the assessed scenario to the approved configuration. The following ledger separates the records needed to reproduce a result from the events that should reopen it.

A Passing Contact Test Does Not Make Welding Safe

Power and force limiting addresses assessed human-robot contact. Beyond that test plan, welding introduces a second hazard layer, so the risk assessment must independently control the arc, fumes, electrical energy, heat, spatter, sharp wire and workpiece edges, fixtures, positioners, cylinders, and stored energy.

OSHA’s Technical Manual on industrial robot systems separates system hazards, task states, and lifecycle responsibilities. NIOSH likewise treats welding fume and manganese exposure as a process-dependent occupational hazard, not a robot contact limit.

Acceptance Evidence and the Change-Control Tripwire

Handover should not reduce the result to “the cobot is compliant.” A useful Acceptance Dossier shows which application configuration and contact scenarios were assessed, how they were tested, what passed, what residual risks remain, and which changes reopen the assessment.

Who owns which evidence?

What activates the Change-Control Tripwire?

• Torch, electrode, end effector, or adapter
• Payload, workpiece, fixture, or positioner
• Cable dress or hose routing
• Path, speed, reach, direction, or safety parameter
• Cell layout, access, screen, guard, or monitored zone
• Robot, controller, firmware, or safety software
• Repair affecting stopping or contact behavior
• New production, teaching, cleaning, or maintenance task

Revalidation should be scoped to the affected scenarios, and the decision must be recorded. A change that alters speed, effective mass, geometry, trapping clearance, access, or task behavior can invalidate a result even when the robot model remains unchanged.

3-Gate Procurement Framework

Because every passing test is configuration-bound, buyers should purchase an integration scope and acceptance record, not only a cobot arm. These three gates keep a request for quotation focused on evidence that can survive commissioning and later changes.

1. Elimination gate: Can the contact, trapping point, exposed sharp feature, or hazardous access be removed by design?
2. Validation gate: Can every retained contact be reproduced, measured, documented, and passed under the worst credible configuration?
3. Process-hazard gate: Are welding and maintenance hazards controlled independently, and are those controls included in site acceptance?

When comparing a collaborative welding robot system, ask which application evidence the supplier will deliver with the cell. Product specifications are inputs to integration; they aren’t substitutes for the final risk assessment and acceptance tests.

Frequently Asked Questions

Specify the Evidence Before You Specify the Speed

ISO/TS 15066 does not ask a buyer to accept a “safe cobot” label at face value. Project teams must make application-level decisions: remove avoidable contacts, classify those that remain, measure the applicable quantities under worst-case conditions, control welding hazards separately, and preserve a configuration-bound record.