In-situ material classification in sheet-metal blanking using deep convolutional neural networks
- 106 Downloads
Sheet-metal blanking is a class of metal fabricating processes that separate a metal workpiece from a primary metal sheet through a shearing process. Industry experts observe fluctuations in tool life and product quality, which is associated with fluctuations in microstructural parameters between and along material coils. With a methodology, that provides reliable information about the current material state in situ through a non-destructive testing method, process parameters, such as forces, speed, and lubrication could be adapted accordingly. This would enhance the productivity of sheet-metal blanking processes or other sheet-metal forming processes with a high process adaptability. The findings of this paper suggest that fluctuations of microstructural parameters along a material coil follow a pattern that is detectable for ferromagnetic materials using magnetic barkhausen noise emission. The preprocessed signal of 18 specimens stemming from six different coils have been encoded to recurrence plots and classified with a deep convolutional neural network with regard to the position on the coil that the respective specimen stem from. The neural network was able to reliably distinct between the recurrence plot of a specimen stemming from the beginning, the middle, or the end of a material coil. Through an advanced labelling approach and classifying for specific microstructural parameters, the same methodology could possibly be used to detect certain material states and quality deficits and to adjust the sheet-metal blanking process accordingly to enhance tool life and product quality.
KeywordsSheet-metal blanking Magnetic barkhausen noise Convolutional neural networks Industrial internet of things
The authors would like to thank the German Research Foundation DFG for the kind support within the Cluster of Excellence “Internet of Production” (Project ID: 390621612). Furthermore the authors would like to thank the Federal Ministry for Economic Affairs and Energy for the kind support within the research project “Mittelstand 4.0 – Digitale Produktions- und Arbeitsprozesse”.
- 1.Klocke F (2014) Manufacturing processes 4: forming. Springer Science & Business MediaGoogle Scholar
- 2.Harsch D, Fischer P, Berisha B, Heingärtner J, Renkci Y, Hora P (2018) Considering fluctuations of material properties, stainless steel 1.4301, on manufacturability of kitchen sinks. In: IOP Conference Series: Materials Science and Engineering, IOP Publishing ,vol. 418, no. 1, p. 012113Google Scholar
- 3.Siegert K (ed) (2015) Blechumformung: Verfahren. Springer, Werkzeuge und MaschinenGoogle Scholar
- 6.Hull JB (2015). Non-destructive testing. Macmillan International Higher EducationGoogle Scholar
- 10.Wang Z, Oates T (2015) Encoding time series as images for visual inspection and classification using tiled convolutional neural networks. In: Workshops at the Twenty-Ninth AAAI Conference on Artificial IntelligenceGoogle Scholar
- 12.Eckmann JP, Kamphorst SO, Ruelle D (1995) Recurrence plots of dynamical systems. World Sci Ser Nonlinear Sci Ser A 16:441–446Google Scholar
- 16.He K, Zhang X, Ren S, Sun J (2016) Deep residual learning for image recognition. In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 770–778Google Scholar
- 17.Loshchilov I, Hutter F (2017) Fixing weight decay regularization in adam. arXiv preprint: 1711.05101Google Scholar
- 18.Smith LN (2017) Cyclical learning rates for training neural networks. In: 2017 IEEE Winter Conference on Applications of Computer Vision (WACV). IEEE, pp. 464–472Google Scholar
- 19.Glebke R, Henze M, Wehrle K, Niemietz P, Trauth D, Mattfeld MBA, Bergs T (2019) A case for integrated data processing in large-scale cyber-physical systems. In: Proceedings of the 52nd Hawaii International Conference on System SciencesGoogle Scholar