[ad_1]
“We are working on laser cladding with powdered metal,” Palalić’ explains. In this process, the heat source is a high-power laser, which “melts specific areas of the workpiece while simultaneously adding an inert gas mixed with fine metal powder, which also melts and bonds with the component’s surface, a process that is applied layer by layer.”
This process can combine different materials in a single component or even repair certain areas or parts. It also makes it possible to create geometries that it would be impossible to produce using traditional metalworking methods, whereby barely any waste is produced.
Yet, some downsides exist: “It is almost always necessary to rework any component that has been manufactured using additive manufacturing by removing material, specifically through machining processes such as milling, drilling, and grinding,” Palalić explains.
This hybrid production method places high demands on the process. To ensure that the components manufactured for such applications as aerospace or medical technology do what is expected of them, the process chain’s individual steps must be precisely coordinated.
However, as the researcher explains, there is still a great deal of uncertainty about the mechanical properties of these components because the process of applying and removing material is extremely complex.
“A huge number of correlating parameters and effects still need to be studied and whose interactions cannot be described in analytical terms. This is why it is appropriate to use machine learning in this case, and not just to obtain information about what happened during a given process, but also to enable us to make predictions about the quality of the component that were previously impossible.”
Palalić’s ultimate goal is to develop a virtual model, or a kind of digital twin, of both the component and the manufacturing process, to facilitate the optimization and monitoring of the process, e.g. by identifying errors. Her first step is to install sensors in the machine after which she uses SE to create an efficient software program.
As Palalić’ explains: “the machine learning algorithms need to be fed with processed sensor data.” The third thing she has to do is correctly select, parameterize, and then apply the algorithms for the particular machine learning AI method and generate a digital model.
Table of Contents
AI literacy as a basic skill
Wirzberger emphasizes the need to qualify students from non-informatics disciplines in AI and SE and explains, “If our goal is to equip the skilled workforce of the future properly, we need to incorporate AI into study programs across the board.”
Another special feature of the academy is that doctoral students such as Palalić co-supervise the AISA seminars and advise students on AI or SE issues. “For example,” says Wirzberger, “if a student would like to use an AI-powered algorithm in the context of a master’s thesis and has questions about it, we’re the ones who can provide assistance and feedback.”
Increasingly, AI literacy is becoming a basic skill everyone must master. I need to understand what drives these systems; how much trust I can place in them, and at what point would I need to apply particularly critical thought?”
In her doctoral thesis, which Wirzberger and Becker jointly supervise, Nadine Koch wants to find out how this competence could be taught more effectively. “It’s about developing AI didactics for people with no expertise in AI,” Wirzberger explains.
“How can I systematically and clearly communicate to non-AI professionals how to select a particular algorithm, recognize its advantages and disadvantages, and how it can be used and evaluated? Can I use well-established forms from teaching/learning studies or computer science didactics? How should I adapt them?” Because once we have answers to these questions, even more skilled workers will soon be able to start their professional lives being AI-competent.
Source: University of Stuttgart
[ad_2]
Source link
