Tuesday, May 12, 2020 2:35 pm
2:35 pm: Deep Predictive Analytics for Additive Manufacturing Process Modelling for EOS
Speakers: Dr. Jonathan Masci, Director of Deep Learning, NNAISENSE Harald Krauss, Technical Project Manager, EOS
3:05 pm: How Bühler and the Swiss Data ...
Speakers: Clément Lefebvre, Data Scientist, Swiss Data Science Center Matthias Graeber, Head of Data Science, Bühler
Selective laser melting is a cutting-edge additive manufacturing technology that builds metal parts of almost limitless geometric complexity using a powerful laser to selectively melt layers of metallic powder onto previous layers, one 2D slice of the part at a time. In this process, the distribution of laser energy across the layer is a key factor determining the material properties of the part including whether it contains costly defects that can render it functionally useless. In an industry first, NNAISENSE has developed a deep network model that accurately predicts the heat distribution of the next layer, in real-time, based on job specific parameters. The thermal map generated by the model can be used to (1) detect deviations from predicted behavior, signaling the onset of a potential defect, (2) control the laser intensity based on the expected heat at each location to avoid defects and optimize material properties. In this talk, Jonathan and Harald will begin by providing general background related to the challenges of applying machine learning-based analytics to real-world industrial problems in general, and in relation to the particular use case in question. They will then discuss the approach they used to implement and train the model, and present the latest results. Finally, they will wrap up with an outlook on future directions for the application of AI to Industry 4.0.