Semi-supervised Monitoring of Laser powder bed fusion process based on acoustic emissions

ABSTRACT

Metal-based Laser Powder Bed Fusion (LPBF) suffers from a lack of repeatability and is challenging to model, making their quality monitoring essential and demanding. The reason lies in the high dynamics taking place during the interaction of the laser with metallic powders. To bring this technology to mass production, industries are only interested in the process regime where the built layer's quality meets their standards. All other process regimes leading to poor mechanical properties and/or defect formation such as balling, Lack of Fusion (LoF) pores, keyhole pores, delamination, and crack propagation irrespective of their different regimes are considered anomalies. Today, the common methodology for monitoring uses conventional/supervised Machine Learning (ML) algorithms for the classification task. However, it requires collecting a balanced dataset corresponding to each investigated regime from the sensors, which is very expensive and time-consuming. As an alternative, the article proposes a semi-supervised approach where the defect-free regime can be differentiated from the anomalies by familiarising the ML algorithms only with the distribution of acoustic signatures corresponding to the defect-free regime. This work presents two generative Convolutional Neural Network architectures based on Variational Auto-Encoder and General Adversarial Network. As a result, we could classify the anomaly regimes with 96 and 97% accuracy, respectively.

KEYWORDS:

• Laser powder bed fusion
• monitoring
• laser processing
• Convolutional Neural Network
• acoustic emission

Acknowledgement

The authors would like to acknowledge the financial support of the project MoCont from the programme of the Strategic Focus Area Advanced Manufacturing (SFA-AM), a strategic initiative of the ETH Board. RDD and RL gratefully acknowledge the generous sponsoring of PX Group to their laboratory.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The authors would like to acknowledge the financial support of the project MoCont from the programme of the Strategic Focus Area Advanced Manufacturing (SFA-AM), a strategic initiative of the ETH Board. RDD and RL gratefully acknowledge the generous sponsoring of PX Group to their laboratory.

Notes on contributors

Vigneashwara Pandiyan

Vigneashwara Pandiyan is currently a Postdoctoral Researcher at Laboratory for Advanced Materials Processing, ETH Empa Swiss Federal Laboratories for Materials Science and Technology. He completed his Ph.D. from Nanyang Technological University, Singapore under Rolls-Royce @ NTU corp lab. He has 5+ years of experience as a researcher in manufacturing on characterisation, investigation, optimisation, and process development. Prior to joining Empa, he was a research scientist in A*Star – Agency for Science, Technology and Research ARTC, Singapore. He now concentrates on developing, validating, and implementing machine learning models for in-process sensing of manufacturing processes for anomaly detection and process automation based on sensor signatures.

Rita Drissi-Daoudi

Rita Drissi-Daoudi received her master degree in Mechanical Engineering in solid and structure mechanics with a minor in Material Sciences in 2018 from École Polytechnique Fédérale de Lausanne (EPFL). She is currently working toward her Ph.D. thesis on in-situ acoustic emission monitoring of the laser powder bed fusion process. at École Polytechnique Fédérale de Lausanne (EPFL), at the Thermomechanical Metallurgy Laboratory, Her research includes signal processing and machine learning but is mainly emphasized on defects formation understanding and laser material processing

Sergey Shevchik

Sergey Shevchik received the master’s degree in control from Moscow Engineering Physics Institute, Moscow, Russia, in 2003, and the Ph.D. degree in biophotonics from General Physics Institute, Moscow, in 2005. He was a Postdoctoral Researcher with General Physics Institute until 2009. Between 2009 and 2012, he was with Kurchatov Institute, Russia, developing image processing for human-machine interfaces. From 2012 to 2014, he was with the University of Bern, Bern, Switzerland, focusing on computer vision systems and multi-view geometry. Since 2014, he has been a Scientist with the Swiss Federal Laboratories for Material Science and Technology, Thun, Switzerland, developing signal processing and machine learning for industrial automatisation. His research interests include signal processing and machine learning for industrial automatisation.

Giulio Masinelli

Giulio Masinelli received the B.Sc. degree in electrical engineering from the University of Bologna, Bologna, Italy, in 2017, and the master's degree in electrical engineering from École Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland, in 2019. He is currently working toward his Ph.D. thesis with EPFL and the Swiss Federal Laboratories for Material Science and Technology (Empa), Switzerland, mainly developing machine-learning algorithms for data analysis and industrial automation. His research interests include signal processing and machine learning with an emphasis on deep learning applied to embedded systems.

Tri Le-Quang

Tri Le-Quang received the B.S. degree in applied physics from Vietnam National University, Ho Chi Minh City, Vietnam, in 2007, the M.S. degree in optics from Friedrich-Schiller-Universität Jena, Germany, in 2013, and the Ph.D. degree in material engineering from the Instituto Superior Tecnico Lisboa, Portugal, in 2017. Since then, he has been holding a Postdoctoral position with the Laboratory of Advanced Materials Processing, Swiss Federal Laboratories for Materials Science and Technology (Empa). His research interests include laser material processing, laser technology, and in situ monitoring.

Roland Logé

Roland Logé is a full-time professor currently working at the Materials Institute, École Polytechnique Fédérale de Lausanne (EPFL), Lausanne. Research activities are focused on the control and design of microstructures in metals and alloys through a combination of thermal and mechanical treatments. Phenomena of interest include recrystallization, grain growth, twinning, texture evolutions, precipitation, phase transformations, variable temperature conditions, and the possibility of concurrent plastic deformation. Different types of metal forming operations are investigated together with the selective laser melting process and its combination with laser shock peening.

Kilian Wasmer

Kilian Wasmer received the B.S. degree in mechanical engineering from Applied University, Sion, Switzerland, and Applied University, Paderborn, Germany, in 1999. He received the PhD degree in mechanical engineering from Imperial College London, London, U.K., in 2003. Since 2004, he has been with the Swiss Federal Laboratories for Materials Science and Technology (Empa), Thun, Switzerland, where he started to work on control of crack propagation in semiconductors. He currently leads the group of dynamical processes in the Laboratory for Advanced Materials Processing at Empa, Thun, Switzerland. Previously, he had focused his work on process development, process monitoring, and quality control via in situ and real-time observation of complex processes using acoustic and optical sensors in various fields such as in tribology, fracture mechanics, and laser processing. His research interests include materials deformation and wear, crack propagation prediction and material-tool interaction in particular laser material processing (e.g. welding, cutting, drilling, marking, cladding, additive manufacturing, etc.). Dr Wasmer is in the Director committee for additive manufacturing of Swiss Engineering. He is also a member of Swiss tribology, European Working Group of Acoustic Emission, Swissphotonics, and Deutsche Gesellschaft für Zerstörungsfreie Prüfung.