TRUMPF (world headquarters, Ditzingen, Germany, U.S. headquarters, Farmington, CT) is taking AI quite seriously. The subject is so important that the company plans to leverage AI in their products, services, and in their own internal processes.
Customer-oriented AI engagements will affect machine design and user services.
“We will have new solutions that work with AI,” says Stephan Mayer, CEO of Machine Tools at TRUMPF. “You can call it Industrial AI,” he says, meaning the acceleration of existing processes as well as developing new functionality to cutting, bending, and welding. “Things like pattern recognition are important. You want the AI to be something that replaces manual steps, something that speeds up processes.”
Some examples provided by Mayer include:
- “Applications for cutting that can look at the temperature of the laser, and do reflectivity monitoring.” (This already exists; an available camera system on a laser cutting machine looks at the temperature of the cutting spot and changes speed to maintain the most optimal cutting temperature.)
- “In quality control, an application can monitor the system and do a visual inspection; an example is a system that can watch a weld seam within the process [of doing it].”
- In a different mode, “ChatGPT can apply in service. We have hundreds of thousands of reports that we have put into a database.” If a tech encounters a problem, “ChatGPT goes in and researches the problem and provides the answers.” TRUMPF already has Service App, which runs on a phone and specifies problems by machine and type of error. Adding a ChatGPT app will allow that software to probe the “Big Data” databases collected by TRUMPF to give an answer and recommended action.
Some of the AI boon will show itself in TRUMPF internal operations. For instance in product management, three areas will show the most change. “We will see changes in development speed,” according to Mayer. “Typical software life cycles are five to ten years. We can design a new machine, take it to customers, and offer new software releases in three months. R&D changes to a three-month cycle, it’s faster development.”
“Second,” he continues, things are much more data-based. There is a need to collect data,” to seed the upcoming AI systems. “If a machine collides, we find out a lot of context, like user, material, etc. This is processing Big Data,” states Mayer, “and we need to build the infrastructure to handle it.” Meanwhile, the company is launching a program to test AI models through a collaboration with the University of Stuttgart near its worldwide headquarters in Ditzingen. TRUMPF will have access to the university’s “Hawk” supercomputer and will use that system to simulate machine functions and use large amounts of data to train AI models.
Mayer shares, “The third change we see involves our experts. We have built our teams with great expertise from individuals. We already had our [fabricating] engineers. We have hired AI experts. We now have teams that include AI experts and sheet metal experts, working in a world of connected machines. In R&D, things will be shifting to software and AI development.”
He notes that currently, there are about 100 TRUMPF employees who are connected to the AI effort in some way. The company expects that number to be equal to the number of TRUMPF employees sometime in 2029.
“Our corporate strategy is to be the technology leader,” states Mayer. “We want to have the highest uptime, and the maximum OEE (Overall Equipment Effectiveness). We want to provide speed; we want to provide the customer with optimal operation. Digitalization and artificial intelligence go hand-in-hand. For our cutomers, this means productivity and efficiency gains along the entire sheet metal process chain, regardless of the company’s size.”
When asked about using AI versus extensive algorithms, Mayer offers his thoughts about the key difference: “Your problem can only be solved when everything possible is included in the algorithm. With AI, you don’t need a 100% inclusion of possible problems. AI increases the robustness of automation.”
He expands on the last point: “AI can help you determine when a part should run. Using the collected database info, a shape of a part could dictate cutting the sheet during the day when people are around, because it could cause tip-ups.” Therefore, those jobs are first-shift jobs, rather than unattended, lights-out jobs.
While it is fulfilling—and tempting—to brainstorm new solutions, Mayer is very much a pragmatist when it comes to advancing automation. “When you ask yourself what to develop, start at the customer. Don’t digitize to digitize. Don’t apply AI just to apply AI. The biggest gap is the human interaction. Start with the customer’s issue.”
The customer focus will take some time. “We’re at the beginning of a long journey,” he says.