Cobot Welding Cell Risk Assessment

Buying a welding cobot is one decision. Deploying it legally and safely is a different one, and it starts with a formal risk assessment. Most small and mid-size fabrication shops encounter the requirement for the first time when an integrator quotes a line item for “safety engineering” alongside the robot arm and welding package. Understanding what the assessment covers, why it cannot be skipped, and what a thorough version looks like puts you in a much stronger position to evaluate that quote and manage the commissioning process well.

What a Cobot Welding Cell Risk Assessment Actually Covers

A welding cobot cell is not a simple machine. It combines a powered robotic arm, a live welding process (MIG, TIG, or flux-core), high-voltage power equipment, ultraviolet arc radiation, and toxic fume generation, all inside a shared workspace that operators and maintenance technicians will enter regularly. The risk assessment treats all of these as a single system rather than examining the robot in isolation from the welding hazards.

In practice the assessment produces a structured analysis of every hazard the cell creates. Each hazard is rated by severity and likelihood. The assessment then identifies whether existing or planned controls bring each risk to an acceptable level and documents what residual risk remains after those controls are applied.

That document becomes the legal and contractual record that both integrator and end user can point to if an incident occurs. It also underpins the Declaration of Conformity required in most jurisdictions before a cell can be put into production. The safety resources on this site cover the regulatory context in more depth; this guide focuses on the practical process from hazard identification through to ongoing review.

Does ISO 10218 Apply to Welding Cobots?

Yes, and it applies regardless of how the robot manufacturer markets the product as “collaborative.” ISO 10218-1 sets requirements for the robot manufacturer; ISO 10218-2 sets requirements for the integrator building the cell. Both standards apply to every welding cobot installation.

ISO/TS 15066 then provides supplementary data for collaborative operation specifically: body-region force and pressure thresholds, contact modes (transient versus quasi-static), and guidance on calculating whether an application is genuinely safe for human-robot contact. The Technical Specification does not replace ISO 10218-2; it extends it with the biomechanical evidence base ISO 10218-2 references.

In North America, ANSI/RIA R15.06 is the domestic equivalent and aligns closely with the ISO 10218 series. Many North American integrators work to both simultaneously. For CE marking in the European Union and UK, compliance with ISO 10218 under the Machinery Directive (transitioning to Machinery Regulation 2023/1230) is the standard route.

One distinction that trips up buyers who have read cobot marketing material: a robot rated for power-and-force-limiting (PFL) collaborative mode does not necessarily operate in that mode during welding. Most welding programs run at speeds and with end-effector masses that exceed ISO/TS 15066 thresholds. For the duration of the weld cycle, the cell must be treated as a conventional industrial robot installation for guarding purposes, even if the same arm operates collaboratively during fixture loading.

Hazard Identification in a Welding Cobot Cell

Hazard identification is the first substantive phase of the assessment. An experienced integrator or safety engineer walks through the cell systematically, using a structured method such as the risk graph or risk matrix defined in ISO 12100, to build a comprehensive hazard list. For a welding cobot cell, that list typically covers:

• Mechanical hazards from robot arm motion: crushing and striking from the arm itself; entanglement or laceration at the end-effector and wire feed mechanism; pinch points at positioners and fixturing clamps
• Electrical hazards: welding power source voltage; cabinet wiring and terminals; earth return current paths through the workpiece and table
• Thermal hazards: spatter reaching operators or cable runs; hot workpiece temperature after the weld cycle ends; torch nozzle heat between welds
• Radiation: ultraviolet and infrared arc radiation, noting that the optical hazard zone extends well beyond the immediate weld area and affects anyone with line-of-sight to the arc
• Fume and gas: welding fume composition varies significantly by base metal and consumable; manganese and hexavalent chromium are specific concerns on stainless steels and certain alloy grades, with different exposure limits from mild steel fume
• Ergonomic and access hazards: operator posture during fixture loading and unloading; reach distances for maintenance access inside the cell; manual handling of fixtures or workpieces

The fume hazard deserves particular attention in the assessment. Fume extraction system design for welding workcells is its own engineering task, but its outputs must be explicitly referenced in the risk assessment. Extraction system capacity, capture velocity at the weld point, and filter specification are all control measures that reduce fume exposure risk. If they are absent or undersized, the residual fume risk remains unacceptable regardless of how well the mechanical and electrical hazards are controlled.

The Risk Assessment Process Step by Step

Once hazards are identified, the formal assessment follows an iterative loop defined by ISO 12100. The seven-step sequence is:

1. Determine the limits of the machinery. This covers the full task range the cell will perform, expected operator access patterns during normal operation and maintenance, intended use, foreseeable misuse, and the maintenance intervals that will require personnel to enter the cell.
2. Identify hazards. The systematic walkthrough described above, covering mechanical, electrical, thermal, radiation, fume, and ergonomic sources of harm.
3. Estimate risk. For each hazard, rate severity of potential harm (from minor reversible injury through to fatal) and the likelihood of harm occurring (probability of exposure multiplied by probability of the hazardous event).
4. Evaluate risk. Compare the estimated risk against an acceptable level. ISO 12100 does not set a numerical threshold; the evaluator applies professional judgement supported by reference standards and established industry practice.
5. Reduce risk. Apply measures in strict hierarchy order. Design out the hazard first. Where that is not practicable, add safeguards: interlocked guards, presence-sensing devices, safety-rated control functions. Only after engineering controls are exhausted should information for use be relied upon (warning signs, PPE mandates, operator training requirements).
6. Validate and verify. Confirm that each control measure performs as specified. This includes functional testing of safety-rated stops, measuring contact forces if PFL mode is used, checking interlock performance under simulated fault conditions, and verifying that restart procedures after an emergency stop are properly defined and tested.
7. Document. Produce the risk assessment report: hazard list, risk ratings before and after mitigation, a residual risk statement, and cited references to the applicable standards.

The loop is iterative. If a proposed control measure introduces a new hazard (a fixed guard that creates an additional pinch point, for example), that new hazard re-enters the process from step two. The assessment is not complete until every hazard has been reduced to an acceptable residual level.

What Force and Speed Limits Mean in Practice

For anyone who has read cobot marketing material, ISO/TS 15066’s force and pressure limits can sound reassuring: the robot stops before it can injure someone. In a welding cell the picture is more nuanced, and automation engineers we interviewed consistently flag that default settings are not a substitute for measurement.

The 15066 tables specify contact thresholds by body region. Thresholds for the head and neck are far lower than for the shoulder or upper arm. An engineer must calculate the force the robot exerts at the point of contact for the specific task geometry, compare it to the relevant body-region limit, and verify with a calibrated force-measurement device that the robot’s PFL settings actually stop it within that threshold under the actual load conditions of the application.

Beyond the measurement question, the practical implication for most welding cells is that PFL mode cannot apply during the weld cycle itself. The combination of welding wire, torch, and fixture geometry typically produces an end-effector mass and operating speed that exceed what ISO/TS 15066 permits for contact-safe operation. The cell therefore operates in two distinct modes: collaborative (PFL, low speed) during fixture loading and unloading, then industrial (safety-rated speed limit with area scanner or fence protection) during the weld program. Both modes require their own risk assessment treatment.

Is Arc Flash Part of the Welding Cobot Risk Assessment?

Yes, and it must appear explicitly in the assessment document, not as a sub-item under a general “welding hazard” category. Arc flash hazard management in welding cobot cells has its own engineering considerations, partly because maintenance access patterns in a cobot cell often differ from a conventional welding booth.

The risk assessment must address several arc flash-specific points. It should define the arc flash boundary for the cell based on arc energy and electrode type. It must specify the optical density of any viewing windows in the cell enclosure panels. It should confirm that interlocked access doors force the arc to stop before any door can be physically opened, and document the interlock failure mode behaviour. It must also mandate appropriate PPE for any task where a maintenance technician enters the arc flash zone while arc-capable equipment remains energised.

NFPA 70E covers the electrical arc flash side; the welding-specific optical hazard falls under ANSI Z87.1 and the applicable welding process standard. Both must be cited in the risk assessment documentation, not assumed to be covered by a general reference to welding standards.

Who Should Carry Out a Cobot Welding Risk Assessment?

ISO 10218-2 requires the risk assessment to be carried out by a competent person with adequate knowledge of the relevant standards and the specific application. It does not mandate an independent third-party assessor. In practice, the integrator typically leads the assessment because they hold the design information needed to evaluate control measures before and during the build.

For end users, the question is whether to accept the integrator’s assessment as-is or commission an independent review. On high-throughput cells, complex multi-robot layouts, or applications where a regulator is likely to inspect (food processing, pharmaceutical, automotive supply chain), a second engineering opinion before commissioning is generally worth the additional cost. On straightforward single-station cells in lower-risk environments, integrators report that a well-documented first-party assessment from a reputable integrator is accepted without issue.

When comparing integrators, asking each to describe their risk assessment methodology and the structure of their documentation is a useful differentiator. An integrator who cannot articulate the ISO 12100 loop or who offers a one-page “safety checklist” in place of a formal assessment is a signal worth taking seriously before signing a contract. The welding cobot price guide explains why safety engineering consistently adds a significant share to the total deployed-cell cost, and why it is a poor place to accept a lower quote.

Documentation, Validation and Ongoing Review

The risk assessment report is a live document, not a commissioning formality to be filed and forgotten. It should be stored with the cell’s technical file, made accessible to maintenance personnel, and reviewed whenever the cell changes in a meaningful way.

Changes that trigger a new or updated assessment include: programming a new part type that alters reach, speed, or force profile; changing the welding wire or shielding gas mix (both affect fume composition and potentially the thermal hazard profile); modifying fixturing in a way that changes operator access patterns or reach distances; adding or removing a positioner; and any change to the safety-rated control functions, including firmware updates on safety PLCs or safety-rated I/O modules.

Validation testing deserves more attention than it typically receives on site. Automation engineers we interviewed consistently flag that as-built installations deviate from design drawings in ways that affect safety function performance. Cable routing, sensor positioning, and emergency-stop circuit wiring are the most common divergence points. Documenting as-built deviations, retesting affected safety functions, and recording the results closes the gap between the design-stage assessment and what is actually operating on the shop floor.

Periodic review, even without changes, is sensible practice. Standards are updated on a rolling basis. Cells that were fully compliant at commissioning may need adjustment as revised versions of ISO 10218-2 or ISO/TS 15066 come into effect, or as the risk profile of the surrounding facility changes.