Christoph Kugler is the group leader of digitalization at the SKZ Plastics Center in Würzburg, Germany. Among other things, the materials scientist explores the use of machine-based learning models for applications in plastics processing. Christoph, how can artificial intelligence in production contribute to greater sustainability and recyclability of plastics? There are several approaches. Let’s take the use of recyclables in industrial mass production as a first example. The quality of recyclable waste often varies more than that of virgin materials. This sometimes leads to irregularities in the production process, which makes it difficult for many companies to use recyclates. That’s why we’re trying to digitally monitor the process in the machine so that the production process of a component will automatically adjust if certain quality parameters are not met. This is where artificial intelligence can come in handy. Would this ideally help increase the rate of recyclables used in complex plastic products? This is where the potential lies. If AI is able to permanently optimize a production process, this will also increase the percentage of recycled materials used. Sounds good. Where’s the catch? Artificial intelligence needs to learn. It has to be trained over an extended period of time with production data. And no company likes to run experiments on its production facilities, at least not over a long period of time. After all, the machines are there to earn money. One idea for solving this problem is to use data originating from production facilities of many different plastics processors to train an AI program. To prevent producer A from gaining insight into the data of producer B, special methods such as federated learning are used. This allows an AI model to be trained on several systems without companies having to share their data openly. The Hamburg-based company Katulu specializes in this form of machine learning. We are currently working with them on several pilot projects. We’ve talked about production. But does AI also offer opportunities in the design and development of more sustainable plastic products? Definitely. I recently met with Digimind, a Berlin company that uses AI to reduce weight in packaging design in order to save material. They take the product’s CAD data and optimize it through AI. But even that has to be fed into the machine first … Exactly. The more data of packaging products we have available, the better it works in training the overall system. And that brings us back to shared learning. This is where we often fail in practice. There is simply not Christoph Kugler talks about plastics and artificial intelligence INTERVIEW 10
RkJQdWJsaXNoZXIy ODI5MzU=