A modern system for laser material processing supplies considerable amounts of data: On the one hand, the optical elements in the laser and in the processing head can be monitored, and on the other, process monitoring provides data from the interaction zone. Today, this data already makes it possible to monitor and document the quality of the individual processing steps, which helps users, for example, evaluate changes in the machining process over time or over many machines.
The same applies to the status of a machine or an entire series of laser systems - here, too, changes can be tracked over time or across a number of machines. This generates considerable amounts of data, which are processed locally or centrally. Furthermore, images consist of huge amounts of data, and artificial intelligence has already become established in evaluating them. The systems are trained to detect and clearly classify deviations from the norm or even errors or malfunctions.
Design laser systems for AI application
"These AI capabilities are well known, now it's a matter of building new machines that are designed from the outset to make the most of AI's potential," is how Professor Carlo Holly, one of the organizers of this year's AI for Laser Technology Conference, describes an important trend. "The sensor technology and the algorithms must fit together and ideally be planned in a coordinated manner from the outset," he adds.
The next goal is self-learning machines. They are to work in four steps: First, the sensors generate the data from the process. Then, the data is analyzed and made understandable, i.e. interpreted on the basis of existing data. In a third step, the system simulates how the results of the process will develop. For this purpose, the previous trend can be extrapolated or the influence of certain parameters can be simulated. This enables the fourth step: system control. So far, AI has been used mainly for quality monitoring and predictive maintenance of the machine. A closed control loop is the "next big thing."
Simplify process planning with AI
But as digital process data becomes increasing available, AI can do even more. The data from the past makes it possible to calculate a parameter range for a new machining task and a specific material in which the process should run well. When a new machining process is planned and set up, AI can save time and resources, and therefore money.
"AI can find an optimal solution within multi-parameter systems, for which a human would need much more time," Holly explains. With AI, new processes can be tested in the computer, as is known from digital twins in the field of Industry 4.0. The complexity here has another dimension: "With these high-tech systems, we are increasingly facing the challenge of finding junior staff with suitable multiple qualifications," says Holly. After all, anyone who wants to simulate and plan a system like this should have knowledge of mechanical engineering, computer science and physics. The presentations at the AI Conference highlight the need for an interdisciplinary approach and contribute to training the participants.
The fields of application are broad
"We're already seeing AI in use, but that's just the beginning" is how Holly describes the status quo. AI is used in the detection of defects in welds or in the sorting of metallic waste. At the third "AI for Laser Technology Conference" on November 23 and 24, 2023 in Aachen, several presentations will deal with applications, joining and additive manufacturing, among others. Naturally, software and infrastructure is also a major topic, which is why Microsoft's plenary talk is eagerly awaited. "In the end, it's all about integrating how systems, processes and algorithms are developed," Professor Holly sums up the direction of this year's conference.
For more information on the AI for Laser Technology Conference 2023, please see: https://s.fhg.de/G5Vc