Johannes-Keppler-Universität Linz is launching a research series on machine learning and signal processing in steelmaking, together with voestalpine Stahl, Kallanish notes.
The project at the University’s Christian Doppler Laboratory for Signal Processing and Machine Learning in the Steel Industry (CD) institute will run until 2032. It has been endowed with €2.7 million ($3.2m) by Austria’s Federal Ministry of Economy, Energy and Tourism.
The laboratory will work on the development of theoretical principles and algorithms to improve signal processing for monitoring steel manufacturing processes for the next few years.
“Production processes – such as those at voestalpine Stahl GmbH – are monitored by sensors whose signals are processed by specialised algorithms,” says project head Oliver Lang. He explains there are various kinds of signals. “One type of signal that occurs frequently are so-called approximate periodic signals,” he identifies as one example. “There has been very little research on these signals in the past and we only have very few algorithms that are able to process them.”
The reason these occur rather often at strip mills is because of the many continuous processes involved, Kallanish understands. “The rotations and oscillations create these almost periodic signals, but they also cause a lot of critical interfering signals,” Lang concludes.
