Production plants are highly complex systems. Identifying the optimal sequence of machines and production steps not only saves a lot of money, but also contributes to energy and resource efficiency. Using Infineon Technologies Austria AG as an example, researchers at Lakeside Labs GmbH and the University of Klagenfurt are developing new algorithms to improve the efficiency of factories.
The semiconductor industry features some of the most complex production facilities in the world: Individual steps are performed at anywhere from 400 to 1,200 different machine stations. In many cases, these plants produce more than 1,500 products in around 300 different processing steps. But who decides which steps should be carried out at which machine and in which sequence in order to be able to produce as quickly and qualitatively as possible, while also saving energy and resources?
“Human intelligence cannot solve these kinds of optimisation problems, and even the linear optimisation methods available to date are now reaching their limit given the highly complex, large and dynamic search space. It simply takes far too long to compute,” Martin Gebser (University of Klagenfurt) explains. Martin Gebser (Department of Artificial Intelligence and Cybersecurity) is a research partner and is collaborating on the FFG-funded project SwarmIn, led by Lakeside Labs GmbH, along with Vienna University of Economics and Business. Other partners from industry are Messfeld GmbH, Novunex GmbH and, of course, Infineon Technologies Austria AG. Most of these optimisation problems are so-called NP-hard problems that cannot be solved with conventional deterministic algorithms. The goal of SwarmIn is therefore to develop new algorithms that lead to improved solutions and, above all, increase the energy and resource efficiency in factories of this kind.
The project team can draw on extensive previous experience with so-called bio-inspired algorithms. These mimic the behaviour of real-life swarms such as birds, fish or ants. The result is swarm intelligence in the form of computationally intelligent, interactive multi-agent systems. SwarmIn aims to create a radically new architecture in order to combine different AI approaches from combinatorial optimisation and swarm intelligence for the first time. In addition, humans will be involved as swarm participants. This will result in software libraries that can be used for the application domain Industry 4.0.
But why is the behaviour of ants, to take one example, helpful in developing production facilities that are configured as ideally as possible? Melanie Schranz (Lakeside Labs GmbH) explains: “The ant colony is organised so that it can adapt to new challenges in the best possible way, it can withstand numerous disruptions and it can grow or shrink in size relatively easily. Adaptivity, robustness and scalability are therefore decisive advantages of swarm intelligence, which we also need for the systems we are investigating.”
This project is funded by the FFG and the Federal Ministry for Climate Protection, Environment, Mobility, Innovation and Technology (BMK) as part of the FFG programme “Production of the Future”.