Special Session on ML in Industrial Applications at INDIN2023
We are pleased to share an update from the special session we organized at the 21st IEEE International Conference on Industrial Informatics (INDIN2023), which took place in Lemgo, Germany, from July 17th to 20th, 2023. The session spanned two days and focused on Machine Learning (ML) applications in the industrial sector. The central theme was the incorporation of domain knowledge and building trust in ML models, particularly in industrial contexts.
It’s aim was to discuss the application of explainable artificial intelligence (XAI) and interactive machine learning to empower process and domain experts in the ML development process, especially when dealing with high-dimensional, noisy, and severely unbalanced data typical in industrial settings.
The successful organization of the session was overseen by a team of accomplished individuals. Leading the efforts was Gianluca Manca, associated with ABB Corporate Research Center Germany and Helmut-Schmidt-University Hamburg, in the role of Principal Organizer. Assisting him were Marcel Dix from ABB Corporate Research Center Germany, Alexander Fay from Helmut-Schmidt-University Hamburg, Willem van Driel representing Signify and Delft University of Technology, and Carl Westin from Linköping University, all serving as dedicated organizers.
Presentations of 7 research papers were featured on topics such as semi-supervised variational autoencoders for regression in soft sensors, using prior knowledge to improve adaptive real-time exploration and optimization, and explaining deep neural networks for bearing fault detection with vibration concepts.Presenters came from respected institutes and companies such as ABB Corporate Research Center Germany, Siemens AG, Imperial College London, Helmut-Schmidt-University Hamburg, University of Sheffield, Technische Hochschule Ostwestfalen-Lippe, B. Tubbs & Associates Consulting, and Ludwig Maximilians Universität. We would like to congratulate the authors of the paper “Measuring the Robustness of ML Models Against Data Quality Issues in Industrial Time Series Data,” which won the best paper award at INDIN2023.
To end this article, we would like to extend our gratitude to the participants, presenters, and organizers for their contributions. The discussions and insights from the session are valuable to the ML community, particularly those working in industrial applications.
Please feel free to contact us for more information on the papers presented and the outcomes of the session.