Cobots in the Foundry
What are the opportunities, limits, and prospects of intelligent automation?
For decades, foundry automation has meant only one thing: traditional industrial robots—large, fast, powerful, and rigidly segregated from the operator behind fences and physical barriers. Machines designed to do the same thing over and over, in the shortest time possible, without exception. This model has worked well in high-volume, low-variability production contexts, but it struggles to meet the needs of a modern foundry, where batches are getting shorter, format changes are multiplying, and part variability is a constant to be dealt with every day. These preliminary considerations should be kept in mind when answering the question: “Why cobots in the foundry?”.
Before discussing how cobots are applied in the foundry.
To these considerations, it is necessary to add what is, in fact, a mere observation—namely, the structural pressure that those working in this sector know well. We are well aware that finding operators willing to perform physically demanding, repetitive tasks in contact with dust, heat, and vibrations is increasingly difficult. This is not a temporary problem, but a demographic and cultural shift destined to intensify.
In this scenario, there is growing interest in collaborative robots, so-called cobots: machines designed not to replace the operator behind a separate cage, but to work alongside them, in the same space, on the same bench. Cobots in the foundry are the answer to a specific need.
But beware: not everything that can be robotized can also be made collaborative. This distinction is perhaps the most important thing to understand before approaching any automation project involving the introduction of cobots in the foundry, and we will return to this point several times. Also, be careful not to be fooled by those who make things too simple: listen to our experience, it’s worth it!
1. What are cobots and why cobots in the foundry
The term “cobot” comes from the contraction of collaborative robot and identifies a category of industrial robots specifically designed to operate safely in proximity to humans, without the need for physical separation barriers.
The concept is not new: the term was coined in 1996 by professors at Northwestern University who were developing robotic assistants for General Motors to help workers handle heavy components. Further information at this link. Since then, technology has made enormous strides in terms of accessibility, ease of programming, and integration with digital systems.
From a technical point of view, what distinguishes a cobot from a traditional industrial robot is the presence of force and torque sensors distributed across the joints, which allow the machine to perceive unexpected contact and stop in fractions of a second. The international reference regulations are represented by the ISO 10218-1 and 2 standards and the ISO/TS 15066 technical specification, which define the methods for safe collaboration, the force and pressure limits acceptable for direct contact, and the procedures for risk assessment.
In terms of physical performance, the market has evolved rapidly. Years ago, cobots were relegated to small sizes, with payloads up to 5 kg and operating reaches up to 900 mm. Today, all major manufacturers are significantly expanding their range. A recent example is ABB, which introduced the new POWA line: a wide range of reaches and payloads, with a working speed never before seen on a collaborative robot: cobots are coming to the foundry!
An often underestimated advantage concerns design: cobots are designed to be slim and with rounded profiles, which makes them physically less bulky than an industrial robot of equal payload. This often translates into more compact and efficient cell layouts, with the possibility of operating in close contact with operators without requiring—subject to risk analysis on the application—safety barriers and fences.
Key market players include Fanuc, ABB with the YuMi, GoFa, and POWA families, and KUKA with the LBR iiwa series. This is a rapidly expanding market: according to estimates by Interact Analysis, global cobot shipments will grow at a CAGR of 20% between 2025 and 2029, reaching approximately 125,000 units per year.
2. Cobots in the foundry: yes, but where?
It is possible to precisely identify the application areas where cobots in the foundry bring concrete and measurable value.
Cobots in the foundry: machine loading and unloading
This is the most widespread application and, in many cases, the one with the best return on investment. Machining centers, presses, automatic deburring machines, washing systems: all these machines require a cyclic pick-and-place operation of the part that is physically monotonous, potentially dangerous due to crushing or burn risks, and does not require high forces.
A technical curiosity: did you know that in these applications, the cobot can operate in reduced mode when an operator enters the cell, automatically limiting force to avoid injury, and resume full speed as soon as the space is clear? Now you know!
Cobots in the foundry: assembly operations and insert mounting
The mounting of threaded inserts, bushings, and reference pins in castings is a high-variability, low-force operation: an ideal context for a cobot equipped with force control. The ability to “feel” resistance during insertion allows for real-time detection of dimensional anomalies or incorrect positioning, effectively making the cobot function as a process control tool.
Cobots in the foundry: quality control with artificial vision
Testing and inspection applications have seen growth exceeding 20% in the cobot sector, fueled by integration with vision systems for precision tasks. In the foundry, this translates into checking for surface defects, automatic dimensional verification with calibrated cameras, and inspection of critical areas that the operator is unable to reach with constant continuity and repeatability. More information on computer vision in the foundry in this article.
Cobots in the foundry: marking and traceability
Laser marking, reading, and application of Data Matrix codes: these are near-zero force operations with very high repeatability requirements, often integrated into stations where the operator is already busy with other activities. The cobot integrates perfectly into this type of cell.
Cobots in the foundry: light finishing: Polishing, brushing, deburring
Cobots are particularly suitable for operations requiring dexterity and sensitivity. Integrated force control allows for maintaining constant pressure on the tool regardless of the part geometry, compensating for dimensional variations in the casting. The forces involved—typically in the order of a few Newtons—are fully compatible with the collaborative philosophy and the payload limits of most cobots available on the market.
At TREBI, we have used cobots to automate machine tool tending and for sanding applications on irregularly shaped parts. Cobots in the foundry: let’s exploit the advantages offered productively.
3. Where the cobot in the foundry is NOT the best choice
This is the most important section of the article, and the one that, unfortunately, is often omitted from commercial literature. And as you know, we like to be blunt.
You won’t hear us shouting from the rooftops “hurray for cobots in the foundry,” or “cobots in the foundry are the solution to all problems!”
The fundamental principle is as follows: not all robotizable applications are also collaborative applications. We have already stated this, but it is time to go deeper.
Cutting and heavy deburring
Cutting castings or deburring parts with massive gates, thick burrs, or complex geometries requires machining forces that can reach and exceed 500 N impulsively. No standard range cobot is sized to absorb these loads without compromising path quality or triggering continuous safety stops. In these cases, structural rigidity is needed, and structural rigidity is incompatible with the philosophy of compliance on which collaborative safety is based.
Aggressive grinding
A 350 mm diamond disc on cast iron requires power in the order of 15 kW and contact forces that vary unpredictably depending on the tool, material, and angle of attack. A traditional industrial robot, with its mass and rigidity, handles these variations. A cobot is not the right tool, and not just because of the forces: the slim and light mechanics that make it collaborative are exactly what make it unsuitable for absorbing the vibrations generated by the tool.
Significant material removal
Any operation requiring significant volumetric removal falls within the domain of traditional industrial robots. It is not a matter of brand or configuration: it is a matter of physics. The cobot was born for applications where force is a sensing tool, not a power tool. When working forces rise, the cobot has neither the mechanics nor the software suitable.
The message is not that the cobot in the foundry represents an inferior technology: it is a different technology, with a precise application domain. Those who use it in the wrong domain get poor results and mistakenly convince themselves that the technology does not work. Those who use it in the correct domain get excellent results.
At TREBI, we have never used a collaborative robot to debur cast iron or to cut a casting. The application is extremely unreliable and the result often uncertain. We integrate cobots in the foundry only when the right conditions exist.
4. Cobots in the foundry: here are the advantages (without rhetoric)
I want to be direct and unconventional: the cobot in the foundry does not have a universal and extraordinary advantage over the industrial robot. One must always evaluate the application, not the robot itself.
If I am handling a sharp or dangerous object, I will still need to install physical protections to protect the operator. In that case, the collaborative advantage almost entirely vanishes. Similarly, some say that cobots offer greater production flexibility because they are simpler to program, but this is not entirely true. Cobots (and therefore also cobots in the foundry) are simpler to program when mediocre results are accepted. If high performance in terms of cycle time and production quality is desired, the programming complexity is substantially the same as an industrial robot. And the same graphical programming paradigms developed for cobots can also be applied to traditional robots.
Then there is the issue of operational availability: cobots are not designed to work 24/7 at sustained industrial rates over the long term. The slimmer mechanics, designed for compliance and lightness, are not oversized like those of an industrial robot. They are excellent tools for specific applications, not a universal solution.
The true advantage of the cobot lies in applications where controlled force is a driver of the operation itself. A cobot can mount a component accurately by detecting insertion resistance in real time, avoiding excessive forces that would damage the part or the mold. This is the kind of value that a traditional industrial robot cannot offer with the same integration.
5. Integration with Industry 4.0 and artificial intelligence
The cobot in the foundry, on its own, is already a useful tool. Integrated with digital systems, it becomes something more.
Adaptive artificial vision
New generation vision systems do not just identify the part: they compensate in real time for variations in position and orientation, allowing the cobot to operate on castings positioned in a way that is not perfectly repeatable. This solves one of the historical problems of foundry automation and opens the door to flexible cells without dedicated fixtures. Real production flexibility starts exactly from this step.
Operator assistance
Latest generation cobots can be paired with contextual digital instruction systems, screens, or projectors that guide the operator through the remaining manual phases, reducing distraction errors and significantly accelerating the training of new personnel.
Artificial intelligence for process control
This is the most interesting front, and also the least mature. The cobot introduces a concrete possibility when paired with AI: machine learning models applied to force and position data recorded during processing can detect process anomalies—a tool wearing out, a part out of tolerance, a variation in material—before these anomalies produce visible scrap. The transition from reactive quality control to predictive quality control is the medium-term goal for foundries investing seriously in this direction. An overview of artificial intelligence applications in the foundry is available in this article, which we recommend reading.
6. Economic aspects: when the cobot in the foundry is worthwhile and when it is not
The cobot in the foundry is worthwhile when:
- Force control is important
- Programming the robot by dragging it reduces time
The traditional industrial robot is the right choice when:
- production rates are high and cycle time must be minimized
- process forces are significant: heavy deburring, grinding, material removal
- production is continuous on a single product with low variability
- the machine is sized to work 24/7 for more than ten years
7. The future of cobots in the foundry
The development trajectories are clear and convergent. New generation cobots are pushing payload limits upwards, with increasingly larger and more performing models. This progressively broadens the application domain towards operations that today still require an industrial robot.
Programming is becoming increasingly accessible: lead-through programming systems, where the operator physically guides the arm to teach it the path, are evolving towards natural language interfaces that allow defining an operation by verbally describing the goal. Setup time is further reduced, increasing the real flexibility of the cells.
Native integration with low-cost 3D vision systems and IIoT platforms will allow for building increasingly autonomous cells, capable of adapting to part variability without operator intervention. Human-machine collaboration will become more sophisticated: no longer simply “the cobot stops if I touch the arm,” but systems that anticipate the operator’s movement, coordinate shared action on a part, and dynamically distribute tasks based on who is present in the cell.
8. But… beware of dream sellers
There are cobot sellers on the market who propose a shortcut: “Instead of buying an industrial robotic cell that costs €150,000, take a cobot, put it on an aluminum profile bench, use manual tools, and for €30,000 you’ve done the application.” This is a massive mistake, and it’s worth explaining why.
Cobots in the foundry: the productivity problem.
An industrial machine integrates PLC, HMI, and hardware solutions designed to make the operation simple, reliable, and repeatable over time. When comparing costs, one must compare the same thing. A bare cobot on a bench is not a robotic cell: it is just a mechanical arm.
Safety.
When purchasing a robotic cell from a manufacturer, one also purchases the expertise of those who designed it. The manufacturer is responsible for compliance with the Machinery Regulation, has conducted the risk assessment, drafted the documentation, performed the testing, and applied the CE marking. This process is long, complex, and expensive—but it is what guarantees that the machine is made according to the state of the art and that, in case of a problem, there is someone who is professionally accountable for the work performed.
In the event of an accident, you are aware of how the machine was designed and built, and you are better protected from legal consequences.
A cobot is still a machine capable of causing harm if integrated incorrectly. Installing a cobot means building a machine, and a machine must be certified. The collaborative robot, on its own, does not make the application safe. In the event of an accident, the responsibility falls entirely on the company that installed and uses the system.
Cobots in the foundry: how Trebi approaches a project.
When a customer brings us a foundry application and asks if it makes sense to use a cobot, our starting point is not the technology: it is the application. We ask ourselves some concrete questions. What forces are involved? How variable is the part? What is the required rate? Must the operator stay in the cell during the cycle, or can they leave? Are there tools or parts that, regardless of the type of robot, still require physical protections?
Only after answering these questions do we decide if the cobot is the right tool, and what configuration makes sense. In some cases, the answer is a traditional industrial robot. In others, it is a cobot. In still others, it is a hybrid solution, where the cobot handles low-force operations and the industrial robot handles high-force ones, on the same part and in an integrated manner.
When the cobot is the right choice, we integrate it into a complete machine: mechanical structure designed for the foundry environment, electrical panel according to standards, internally developed software, vision system where necessary, CE documentation. The customer does not receive a robot arm on a counter: they receive a tested, marked machine ready to work.
Our personal conclusions
Cobots in the foundry will not replace traditional industrial robots in the most demanding foundry processes. This is not their role, and pretending it is is the quickest way to a failed project. Instead, they are a tool for doing things that are not done today, or that are done poorly: using force as an application driver, monitoring it in real time, understanding what is happening in the process.
The choice between a collaborative robot and an industrial robot must not be driven by technological fashion, by the suggestion of an application case seen at a trade fair, nor, above all, by sellers promising low-cost DIY solutions. The guide must be the precise analysis of the specific application.
And when that analysis leads to the cobot, the results can be excellent. But only if you know exactly why.
