Benchmarking machine learning approaches in Industrial Punching: A multi-modal Dataset

A dataset derived from multiple sensors recorded during an industrial punching process is introduced here. The primary objective of this dataset is to facilitate insights into tool wear dynamics and the correlation of multiple parameters within the punching operation. The data consists of acoustic emission and cutting force signals capturing the tool wear throughout the punch’s life. Additionally, data related to the thickness of the strip material has been recorded. Contrary to previous data records, each manufactured cutting surface is visually documented through images within the process. The Dataset therefore renders multi-modal. In recent studies, the collected data is often not publicly available, which results in a lack of universal benchmark.

This dataset aims to establish a basis for developing machine learning approaches in the domains of stamping, punching and blanking. Consequently, it aims to contribute to the refinement of tool wear prediction models and the optimization of strategies within industrial punching operations. This paper shows the data accusation methods used for the dataset and the structure of the data. Furthermore, use cases of the data are shown.

For Information about the data collection, go to: Benchmarking machine learning approaches in Industrial Punching: A multi-modal DatasetFor the Structure of the data, go to: Datasheet for DatasetSupporting Repository for data handling: https://github.com/LepusMaximus/MultiModPunchFunding: The Authors gratefully acknowledge financial support from the interdisciplinary technology network Efficient Production Technology (EffPro), co-funded by the European Union and the Free State of Bavaria from the European Fund for regional development (ERDF) and supported by the University of Applied Sciences Kempten and the Bavarian Ministry of Science and Art.