Development of a semi-automated data acquisition and processing architecture for machine learning applications in grinding
Journal of Intelligent Manufacturing · 10.04.2026
DOI: 10.1007/s10845-026-02867-7 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Intelligent Manufacturing, online erschienen am 10.04.2026. Verfasst von Eike Reuter und 4 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 51 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Deutsche Forschungsgemeinschaft, RWTH Aachen University.
Abstract (Originalsprache)
Studies show that manufacturing companies that invest in automation and artificial intelligence can achieve productivity gains of up to 20% in production output and labor efficiency, driven by advances in data analytics and machine-learning technologies. In grinding processes, sensor-based process variables such as accelerations and acoustic emission signals are commonly used as inputs for machine-learning models to predict quality-related outcomes, including surface roughness and structural damage. However, industrial adoption is often limited by the lack of data acquisition and storage architectures tailored to data-driven applications, as well as by the high computational demands of many modeling approaches in production environments. To address these challenges, this work presents a modular architecture for grinding process monitoring combined with an efficient feature extraction and selection methodology based on Welch spectral analysis. The architecture enables structured data acquisition, centralized data storage, and consistent metadata management, while separating data processing, modeling and application layers. This modular design supports traceability and seamless integration of machine-learning models into industrial monitoring. Within this framework, physically interpretable frequency-domain features are extracted and systematically optimized. A hyperheuristic feature selection strategy is evaluated and compared with filter-based methods and a standalone Genetic Algorithm. The results show that filter-based approaches suffer from strong overfitting and limited generalization, while a standalone Genetic Algorithm improved robustness and predictive performance compared to filtering but remained dependent on problem complexity and population size. In contrast, the hyperheuristic consistently achieved superior robustness and predictive performance with reduced variance under blocked cross-validation.
Förderung
- Deutsche Forschungsgemeinschaft
- RWTH Aachen University
Zitieren
APA
Reuter, E., Killich, B., Breuer, P., Prinz, S., & Bergs, T. (2026). Development of a semi-automated data acquisition and processing architecture for machine learning applications in grinding. Journal of Intelligent Manufacturing. https://doi.org/10.1007/s10845-026-02867-7
Vancouver
Reuter E, Killich B, Breuer P, Prinz S, Bergs T. Development of a semi-automated data acquisition and processing architecture for machine learning applications in grinding. Journal of Intelligent Manufacturing. 2026. doi:10.1007/s10845-026-02867-7
BibTeX
@article{Reuter2026,
author = {Reuter, Eike and Killich, Björn and Breuer, Peter and Prinz, Sebastian and Bergs, Thomas},
title = {Development of a semi-automated data acquisition and processing architecture for machine learning applications in grinding},
journal = {Journal of Intelligent Manufacturing},
year = {2026},
doi = {10.1007/s10845-026-02867-7}
}
EUVIMED