Brian Giera | 17-ERD-037
As with most advanced manufacturing (AM) systems, analysis of AM sensor data currently occurs post-build, rendering process monitoring and rectification impossible. Supervised machine learning offers a route to convert sensor data into real-time assessments; however, this requires a wealth of labeled sensor data that traditionally is too time-consuming and/or expensive to assemble.
This project solved this critical issue in a variety of advanced manufacturing systems. Machine learning algorithms were developed and implemented to enable automated quality assessment and, in some cases, rectification. Machine-learning-based algorithms capable of automated detection were deployed in a host of AM technologies, such as laser powder bed fusion, direct ink writing, and microfluidic platforms used for feedstock production. The common thread within these systems is that routinely collected sensor data (e.g., high-speed video and pyrometer readings) contains pertinent information about the state of the system that can be converted into actionable information in real time via machine learning. Successful implementation of these machine learning algorithms will reduce time and costs by automating quality assessments, and will lead to process control at Lawrence Livermore National Laboratory and within the greater DOE complex.
This project supports the Laboratory's high-performance computing, simulation, and data science core competencies by developing new machine learning capabilities associated with process control of data-intensive additive manufacturing. This project also supports Livermore's advanced materials and manufacturing core competency by automating detection and rectification of AM defects, which enables rapid build qualification and informs optimal selection of process parameters. The AM systems impacted by this project are programmatically relevant as stockpile stewardship requires rapid build qualification for large-scale production of high-quality parts.
Publications, Presentations, and Patents
Chu, A., et al. 2019a. "Automated Detection and Sorting of Microencapsulation via Machine Learning." Lab on a Chip 19 (10): 1808-1817. doi:10.1039/C8LC01394B. LLNL-JRNL-733470
——— 2019b. "Image Classification and Control of Microfluidic Systems." Proceedings of SPIE 11139(06), Applications of Machine Learning, San Diego, CA, September 2019. doi:10.1117/12.2530416. LLNL-JRNL-748383
——— 2019c. "Image Classification of Clogs in Direct Ink Write Additive Manufacturing." 2019 18th IEEE International Conference On Machine Learning and Applications, Boca Raton, FL, 2019. doi:10.1109/ICMLA.2019.00218. LLNL-CONF-789249
Giera, B. 2015. Rapid Closed-Loop Control Based on Machine Learning. US Patent Application 2017/0144378 A1.
Giera, B., et al. 2019. Automated Control of Microfluidic Devices Based on Machine Learning. US Patent 10,408,852 B2.
Yuan, B., et al. 2018. "Machine-Learning-Based Monitoring of Laser Powder Bed Fusion." Advanced Materials Technologies 3 (12): 1800136. doi:10.1002/admt.201800136. LLNL‐JRNL‐748383
——— 2019. "Semi-Supervised Convolutional Neural Networks for In-Situ Video Monitoring of Selective Laser Melting." 2019 IEEE Winter Conference on Applications of Computer Vision, 744-753. Waikoloa Village, HI, January 2019: IEEE. doi: 10.1109/WACV.2019.00084. LLNL-PROC-754101