Steelmaking has always presented a difficult engineering challenge: some of the operations that are most critical to production are also among the most hazardous for people.
Molten metal, extreme temperatures, heavy coils, moving equipment, confined spaces, toxic gases, repetitive manual tasks and high-energy machinery are part of everyday steel production.
The objective of industrial robotics in this environment should therefore not be simply to replace manual labor.
The more strategic question is:
Which tasks should no longer require direct human exposure?
This distinction is important.
A robot that merely automates an inexpensive manual operation may deliver limited economic value. A robotic system that removes an operator from a high-temperature casting platform, a hazardous inspection area or a heavy-material handling operation can simultaneously improve safety, process consistency, productivity and operational reliability.
This is why robotics is becoming increasingly relevant to the steel industry.
The World Steel Association has identified a rising trend in the use of artificial intelligence, robotics and other technologies as tools for reducing occupational risk. Its 2026 safety data also shows why the issue remains important: steel manufacturing still involves intrinsic hazards associated with hot metal, moving machinery, mobile equipment, product handling, gases and other industrial processes.
Modern steel robotics therefore represents more than automation.
It represents the transition from:
human exposure + manual execution
to:
remote supervision + automated execution + data-driven control.
Why Robotics Has Particular Value in Steel Manufacturing
Robotics is used throughout manufacturing, but steel plants create an unusually demanding environment for automation.
A conventional industrial robot may operate successfully in a controlled automotive assembly plant. A steelmaking application can require that the same basic technology tolerate radiant heat, scale, dust, vibration, water, fumes, heavy payloads and proximity to molten metal.
That changes the engineering problem.
Robotic applications in steel plants generally create the greatest value when at least one of four conditions exists:
| Condition | Why Robotics Can Add Value |
|---|---|
| High occupational risk | Removes personnel from hazardous areas |
| Repetitive operation | Improves consistency and reduces ergonomic exposure |
| High precision requirement | Reduces variability and operator-dependent execution |
| Difficult access | Enables inspection or intervention where human access is dangerous or costly |
The strongest projects often combine several of these conditions.
Sampling molten steel, for example, is repetitive, requires consistent positioning and exposes personnel to extreme heat. Automating it can therefore produce several benefits simultaneously.
This leads to a useful principle:
The best robotics projects in steel plants often begin with exposure reduction, not labor reduction.
Where Robots Are Being Used in Steel Plants
Robotics can be applied across almost the entire steel production chain.
Typical applications include:
- furnace and molten-metal operations;
- continuous casting;
- sampling and temperature measurement;
- slag and dross handling;
- rolling and finishing;
- coil and product handling;
- trimming and sample preparation;
- inspection;
- maintenance;
- logistics;
- packaging;
- hazardous-area monitoring.
The robot architecture varies according to the application.
Industrial articulated robots remain important, but steel plants increasingly use autonomous mobile systems, quadruped inspection robots, machine-vision systems and, more recently, experimental humanoid robots.
The term robotics is therefore becoming broader than the traditional image of a six-axis robotic arm inside a fenced cell.
Robotics Around Molten Metal
Few areas demonstrate the safety value of robotics better than operations near liquid steel.
Workers in these environments can be exposed to radiant heat, splashing, fumes, moving equipment and unexpected process events.
Robotic systems can perform operations such as:
- temperature measurement;
- sample taking;
- probe insertion;
- shroud manipulation;
- casting-powder handling;
- furnace-related manipulation;
- oxygen-lancing support;
- repetitive metallurgical measurements.
Primetals Technologies developed its LiquiRob platform specifically for harsh ironmaking and steelmaking environments and for tasks performed with or close to liquid metal. Applications include electric furnaces, converters, secondary metallurgy and continuous casting.
The important point is not simply that a robot can reproduce an operator’s movement.
Automation also makes the procedure repeatable.
Probe depth, positioning, timing and sequence can become controlled process variables rather than operator-dependent actions.
Therefore:
Robotics can improve both occupational safety and measurement consistency.
Continuous Casting: Removing People From High-Risk Areas
Continuous casting is one of the clearest examples of this principle.
Activities around the ladle, tundish and casting platform may involve high temperatures, liquid steel and moving equipment.
Robots can perform operations that previously required personnel to work close to these hazards.
A documented example comes from Ternium Brazil.
According to worldsteel’s Safety and Health Excellence Recognition programme, Ternium has used robotic operation in continuous casting since 2018 to reduce human exposure during steel transfer from the ladle to the tundish.
The robotic system performs tasks including:
- ladle-shroud manipulation;
- application of tundish insulation powder;
- probe operations for temperature, hydrogen and sampling;
- ladle-shroud handling;
- oxygen lancing.
Worldsteel reported risk reductions of 91% for hot ladle-shroud manipulator handling and 88% for steel leakage from the ladle in the evaluated activities.
This is precisely the type of application where the economic justification should not be based solely on headcount.
The project affects:
safety + process consistency + labor exposure + operational continuity.
A Practical Example: Robotic Shroud Manipulation
Another example comes from Siam Yamato Steel in Thailand.
The company implemented a customized LiquiRob system at its Map Ta Phut plant for shroud manipulation on a bloom caster.
The solution incorporates a 3D camera system for accurate positioning, allowing the robotic system to perform the operation while improving occupational conditions for operators. Primetals reported a significant improvement in safety and occupational health following implementation.
This illustrates another important trend.
Industrial robots are increasingly being combined with:
robotics + machine vision + sensors + process automation.
The robot provides movement.
Sensors determine the environment.
Vision identifies position.
Automation logic determines the sequence.
Process data confirms whether the operation was completed correctly.
That combination is much more powerful than the robot alone.
Robots in Rolling and Finishing Operations
Downstream steel processing offers another large field for robotics.
Typical opportunities include:
- sample cutting;
- trimming;
- marking;
- grinding;
- deburring;
- strapping;
- packaging;
- dimensional inspection;
- material positioning;
- palletizing.
These processes frequently combine repetitive motion with hot material, sharp edges or heavy products.
A particularly relevant 2026 example is Hyundai Steel’s Dangjin wire rod mill.
Primetals Technologies completed commissioning of TrimRob, a foundry-rated robotic coil trimming and sampling system.
The system automates trimming and sampling and was designed to remove operators from repetitive, strenuous and unsafe working conditions.
This demonstrates how robotics is moving beyond conventional assembly-type applications.
A robotic cell can become part of the metallurgical production flow itself.
Heavy Material Handling: Where Automation Becomes More Difficult
Steel logistics is fundamentally different from handling consumer goods.
A steel coil can weigh tens of tonnes.
Slabs and billets have substantial thermal and mechanical hazards.
Consequently, automated material handling requires more than navigation software.
The system must consider:
- payload;
- center of gravity;
- braking distance;
- floor condition;
- traffic segregation;
- crane interfaces;
- collision avoidance;
- product identification;
- warehouse logic;
- process sequencing.
Automated guided vehicles (AGVs), autonomous mobile robots (AMRs), automated cranes and specialized transport platforms can all participate in this environment.
The value proposition is not simply transportation without a driver.
The greater opportunity is the creation of an integrated material-flow system in which production orders, product identification, storage position and movement are digitally connected.
Robotic Inspection: Sending the Machine Before the Person
Not every robot in a steel plant needs to manipulate material.
Inspection is becoming one of the most important robotic applications.
Robotic inspection platforms can include:
- quadruped robots;
- drones;
- crawlers;
- remotely operated vehicles;
- camera-equipped mobile platforms;
- thermal inspection systems;
- gas-detection robots.
These technologies are particularly useful where the inspection itself exposes personnel to risk.
Examples include:
- confined spaces;
- high-temperature areas;
- gas facilities;
- difficult elevations;
- equipment requiring scaffolding;
- zones near operating machinery.
Hyundai Steel, for example, has used the quadruped robot Spot to reduce worker exposure to explosion-related risks. Worldsteel reported a target application covering 156 gas facilities and 2,927 locations in confined areas at the Dangjin Steelworks.
The underlying principle is simple:
If information can be collected without placing a person in the hazardous area, human entry should be challenged as the default inspection method.
From Robotic Inspection to Predictive Maintenance
Inspection robots become even more valuable when their data is integrated with condition monitoring.
A robot carrying thermal cameras, acoustic sensors, gas detectors or conventional cameras can collect repeated observations along a defined route.
Those observations can then be compared over time.
Instead of asking:
“Is there a problem today?”
the plant can begin asking:
“How is this condition changing?”
This enables trending of:
- temperature;
- vibration;
- leakage;
- corrosion;
- refractory condition;
- mechanical wear;
- abnormal noise;
- gas concentration;
- visual degradation.
At this point, robotics connects directly with predictive maintenance.
The robot is no longer merely replacing the inspector’s movement.
It becomes a mobile sensor platform.
Robots, Machine Vision and Artificial Intelligence
Traditional industrial robots depend heavily on predetermined coordinates.
Modern robotic systems increasingly use machine vision and AI to interpret less structured environments.
Machine vision can help identify:
- product position;
- geometry;
- orientation;
- surface condition;
- obstacles;
- tool alignment;
- handling points.
Artificial intelligence can extend this capability by recognizing patterns and anomalies that are difficult to encode with fixed rules.
This is particularly important in steel plants because real production environments are not always perfectly repeatable.
Scale accumulates.
Material positions vary.
Lighting changes.
Surfaces reflect radiation differently.
Equipment wears.
The combination of robotics and perception therefore allows automation to move from rigid repetition toward greater adaptability.
Physical AI: The Next Stage of Industrial Robotics
A new term is increasingly appearing in industrial automation:
Physical AI.
Traditional AI primarily interprets information.
Physical AI connects intelligence with machines capable of acting in the physical environment.
In a steel plant, this could mean combining:
- artificial intelligence;
- machine vision;
- robotic manipulators;
- autonomous vehicles;
- process models;
- sensors;
- digital twins;
- industrial automation.
The objective is not merely for the system to recognize what is happening.
It is to recognize the condition and determine or execute an appropriate physical response.
This concept is already moving into steel-industry experimentation.
Humanoid Robots Enter the Steel Industry
In February 2026, POSCO Group announced a project with POSCO DX, POSCO Investment and Persona AI to test industrial humanoid robots in steelworks logistics.
The initial application concerns logistics management for steel coils.
Coils weighing approximately 20 to 40 tonnes require crane operations. The planned humanoid application involves collaboration with workers during activities such as attaching crane belts to coils.
POSCO explicitly identifies accident risk and repetitive-task musculoskeletal exposure as reasons for investigating the technology.
This does not mean humanoid robots are about to replace conventional industrial robots throughout steel plants.
Dedicated machines will often remain faster, simpler and more economical for repetitive tasks.
Humanoids become interesting where an environment was originally designed for humans and where rebuilding the entire infrastructure for conventional automation would be difficult.
That distinction matters.
Conventional Robots vs. Humanoids
| Characteristic | Conventional Industrial Robot | Humanoid Robot |
|---|---|---|
| Repetitive fixed task | Excellent | Usually unnecessary |
| High-speed production | Excellent | Currently limited |
| Existing human-designed environment | May require modification | Potential advantage |
| Heavy industrial maturity | High | Emerging |
| Flexibility across tasks | Moderate | Potentially high |
| Cost predictability | Relatively mature | Still uncertain |
| Steel-industry deployment | Established | Early-stage / pilot |
Therefore, humanoids should currently be viewed as an emerging industrial technology, not as a universal replacement for established robotic automation.
The Robot Is Only One Part of the Automation System
One of the most common misconceptions in robotics projects is focusing on the robot itself.
In practice, the robotic arm may be only one component of the investment.
A complete system can require:
- robot;
- end effector;
- tooling;
- machine vision;
- sensors;
- guarding;
- safety scanners;
- PLC;
- drives;
- communication network;
- SCADA integration;
- MES interface;
- product tracking;
- fixtures;
- cooling;
- protective enclosures;
- electrical infrastructure;
- software;
- commissioning.
For harsh steelmaking environments, additional protection may be required against:
- heat;
- dust;
- scale;
- water;
- electromagnetic interference;
- impact;
- fumes.
Therefore:
Never evaluate a steel robotics project from the robot purchase price alone.
The correct metric is the cost and performance of the complete robotic application.
Safety Does Not Become Automatic Because a Robot Is Installed
There is another critical misconception.
Removing a worker from one hazard does not mean the robotic system itself is risk-free.
Industrial robots introduce hazards associated with:
- unexpected movement;
- stored energy;
- crushing;
- trapping;
- end-effectors;
- maintenance intervention;
- restart conditions;
- programming;
- loss of communication;
- sensor failure.
OSHA specifically warns that lack of awareness of robotics hazards can have fatal consequences and highlights maintenance and entry into robot working ranges as important risk situations.
The safety engineering of the robotic cell must therefore be treated as part of the project from the beginning.
ISO 10218 and Robotic Safety
The international safety framework for industrial robots was substantially updated in 2025.
ISO 10218-1:2025 addresses safety requirements for industrial robots themselves.
ISO 10218-2:2025 addresses industrial robot applications and robot cells, including integration, commissioning, operation, maintenance and decommissioning.
This distinction is important for steel plants.
Buying a robot that complies with applicable robot requirements does not automatically make the completed application safe.
The integrated robotic system must be evaluated.
Steelmaking also introduces application-specific hazards that require additional risk assessment and engineering controls. ISO 10218-1:2025 does not itself cover hazards arising from handling loads such as molten metals, while ISO 10218-2:2025 addresses the safety of the integrated robot application and cell. Steelmakers must therefore evaluate these process-specific hazards as part of the complete robotic-system risk assessment.
A Better Method for Selecting Robotics Projects
Many automation projects begin with the wrong question:
“Where can we install a robot?”
A better sequence is:
“Which operational problem should we eliminate or reduce?”
A practical screening methodology is:
| Step | Key Question |
|---|---|
| 1. Identify exposure | Where are workers exposed to significant hazards? |
| 2. Quantify the problem | What injuries, downtime, variability or cost does it create? |
| 3. Define the task | What physical actions must be automated? |
| 4. Evaluate environment | Heat, dust, payload, reach, access, cycle time? |
| 5. Check technical feasibility | Can robotics perform the task reliably? |
| 6. Design safety architecture | How will people and equipment be protected? |
| 7. Integrate data | PLC, SCADA, MES, tracking, quality systems? |
| 8. Establish baseline | What is current performance before automation? |
| 9. Run pilot | Validate under actual plant conditions |
| 10. Measure results | Safety, quality, productivity, availability and cost |
This prevents technology from becoming the objective.
The objective remains industrial performance.
How to Calculate the ROI of a Robotics Project
Robotics ROI should include more than labor savings.
A more complete framework is:
Annual Robotics Benefit =
Labor and exposure reduction
- productivity gains
- quality gains
- avoided downtime
- reduced inspection or access cost
- lower rework and scrap
- safety-related economic benefits
− additional operating and maintenance costs.
Then:
ROI = (Annual Benefit − Annualized Project Cost) / Annualized Project Cost
But financial ROI alone may still underestimate the project.
Suppose a robot removes workers from an operation with a low probability but potentially catastrophic consequence.
Historical accident cost may be zero simply because the accident has not yet occurred.
That does not mean the risk has zero economic value.
Therefore, robotics projects involving serious hazards should include a formal risk-reduction assessment, not merely conventional payback.
A Strategic Robotics Prioritization Matrix
A practical portfolio can classify potential projects according to two dimensions:
Operational Value and Human Exposure.
| Human Exposure | Operational Value | Priority |
|---|---|---|
| High | High | Very High |
| High | Medium | High |
| Medium | High | High |
| Low | High | Evaluate on ROI |
| High | Low | Evaluate primarily on safety |
| Low | Low | Low |
This helps prevent a common mistake: automating easy tasks while dangerous operations remain manual.
Common Mistakes When Implementing Robotics in Steel Plants
Mistake 1 — Starting with the robot instead of the problem
Selecting equipment before defining the industrial problem frequently produces expensive automation with weak business value.
Mistake 2 — Using labor reduction as the only justification
Safety, quality, process stability, availability and reduced access requirements may be more valuable than headcount savings.
Mistake 3 — Underestimating the environment
Heat, dust, scale, vibration and water can dramatically affect robotic reliability.
Mistake 4 — Ignoring the end effector
The robot moves.
The end effector performs the work.
Tooling design can determine whether the application succeeds or fails.
Mistake 5 — Automating an unstable process
Robotics does not automatically correct poor process design.
Automating instability can simply reproduce the problem faster.
Mistake 6 — Ignoring maintainability
A robotic cell that only an external integrator can troubleshoot may create a new source of downtime.
Mistake 7 — Failing to integrate automation and operations teams
Successful projects require production, maintenance, automation, safety, IT/OT and engineering to work together.
Mistake 8 — Treating safety guarding as an afterthought
Safety architecture must be part of the initial system design.
Mistake 9 — Scaling before proving reliability
A pilot should demonstrate cycle time, availability, maintainability and safety before plant-wide replication.
Mistake 10 — Measuring installation instead of results
Installing a robot is not a KPI.
The relevant KPIs are what changed afterward.
What Should Be Measured After Implementation?
A robotics project should establish its baseline before commissioning.
Typical indicators include:
| Area | Possible KPI |
|---|---|
| Safety | Exposure hours eliminated |
| Safety | Risk level before vs. after |
| Productivity | Units or tonnes per hour |
| Reliability | Robot/cell availability |
| Quality | Defect or rework rate |
| Process | Cycle-time variation |
| Maintenance | MTBF / MTTR |
| Labor | Manual intervention hours |
| Cost | Cost per tonne or operation |
| Financial | Payback / ROI |
Without a baseline, claims of improvement become subjective.
Robotics and the Future Steel Workforce
The relationship between robotics and employment is more complex than “robots replace workers.”
Some manual activities will clearly disappear.
But robotic steel plants also require:
- automation engineers;
- robot programmers;
- maintenance specialists;
- machine-vision technicians;
- data engineers;
- reliability engineers;
- cybersecurity professionals;
- system integrators;
- operators capable of supervising automated systems.
The workforce therefore shifts from direct execution toward supervision, maintenance, analysis and optimization.
This transition requires training.
A technically advanced robot with insufficient internal maintenance competence can become a sophisticated source of downtime.
What Comes Next for Robotics in Steel?
The next stage will probably not be defined by a single type of robot.
Instead, several technologies will converge:
industrial robotics + autonomous mobility + machine vision + smart sensors + edge computing + AI + digital twins.
Quadruped robots can inspect.
AGVs and autonomous systems can transport.
Industrial arms can manipulate.
Vision systems can locate and inspect.
AI can interpret.
Digital twins can model.
And humans can increasingly supervise the system from safer environments.
POSCO’s recent work illustrates this direction. In addition to its humanoid initiative, the group has demonstrated quadruped robots operating in high-temperature steelworks environments for gas-leak detection and equipment inspection.
The result is not simply a “robotic plant.”
It is a progressively autonomous industrial system.
Frequently Asked Questions
What are the main uses of robots in steel plants?
Robots are used for molten-metal operations, sampling, temperature measurement, casting operations, trimming, inspection, maintenance, material handling, logistics, packaging and other hazardous or repetitive tasks.
Why is robotics particularly valuable in steelmaking?
Steelmaking combines hazardous environments with repetitive, precision-dependent processes. Robotics can therefore improve safety and process performance simultaneously.
Can robots operate near molten steel?
Yes, but the robotic system must be specifically engineered for the environment. Protective enclosures, cooling, specialized tooling, sensors and application-specific safety measures may be required.
Are AGVs and AMRs considered robotics?
Yes. Mobile robotic systems are increasingly important in industrial logistics, although their design and safety requirements differ from fixed industrial robot cells.
Can robots reduce steelmaking defects?
They can contribute to defect reduction where consistent positioning, measurement, handling or inspection affects quality. Robotics should be integrated with appropriate sensors and process-control systems.
Are collaborative robots suitable for steel plants?
They can be suitable for selected lower-risk applications, but the term “collaborative” does not automatically make an application safe. Risk assessment remains necessary.
Are humanoid robots already operating throughout steel plants?
No. Humanoid robotics in steelmaking remains an emerging field. Current projects such as POSCO’s 2026 initiative should be viewed as industrial trials and development programs rather than mature plant-wide deployment.
What is Physical AI?
Physical AI combines artificial intelligence with machines capable of sensing and acting in the physical world. In industrial environments it can connect AI, robotics, vision, sensors and process automation.
What is the best first robotics project for a steel plant?
Usually a well-defined operation with measurable risk or operational loss, technically feasible automation and a clear baseline. High-risk repetitive tasks are often strong candidates.
Should robotics ROI include safety?
Yes. Conventional ROI should be supplemented by risk-reduction analysis, particularly where robotics removes personnel from potentially severe hazards.
Conclusion: The Best Robot Is the One That Removes the Right Problem
Robotics can improve productivity in steel plants.
It can improve quality.
It can automate material flow.
It can enable inspection.
But its greatest strategic value may be simpler:
Robotics allows steelmakers to redesign operations so that people no longer need to perform certain tasks in dangerous environments.
That changes how robotics projects should be evaluated.
The objective should not be to maximize the number of robots installed.
It should be to identify where robotic technology can produce the greatest combination of:
safety + reliability + quality + productivity + economic value.
Steel plants will continue to contain heat, heavy material and high-energy processes.
Those physical realities will not disappear.
What can change is the amount of direct human exposure required to operate them.
And that is where robotics can create one of its most important competitive advantages.
Sources and Further Reading
- World Steel Association — Safety and Health in the Steel Industry: Data Report 2026
- World Steel Association — Safety and Health Excellence Recognition 2024
- Primetals Technologies — LiquiRob: Higher Efficiency, Quality and Safety
- Primetals Technologies — Robotics for Safer Casting at Siam Yamato Steel
- Primetals Technologies — TrimRob at Hyundai Steel Dangjin
- POSCO Group — Humanoid Robots for Steelworks Logistics
- World Steel Association — Hyundai Steel Quadruped Robot Safety Case
- ISO — ISO 10218-1:2025 Industrial Robot Safety Requirements
- ISO — ISO 10218-2:2025 Industrial Robot Applications and Robot Cells
- OSHA — Robotics Hazard Recognition