Dr. Vincent C. Paquit
Research Scientist in Electrical & Electronics System Research Division of ORNL
Friday, November 13, 2015 2:30-3:30
410 John D. Tickle Building
BIOSKETCH: Dr. Vincent C. Paquit is a research scientist in the Electrical and Electronics Systems Research (EESR) Division at the Oak Ridge National Laboratory (ORNL) pursuing research and development efforts in the Computer Vision and Image Processing area, with a predilection for multispectral and multidimensional data understanding. Before joining ORNL, he worked at the University of Burgundy (France) as an engineer in technology transfer for the Laboratoire Electronique Informatique Image (Le2i) for all commercial and technical applications in the fields of Electronic, Computer Science and Signal Processing. Since then, Dr. Paquit has been an active member of the Imaging, Signals, and Machine Learning (ISML) group, working on multiple projects and programs supporting two core missions of the Department of Energy: Energy sustainability and National Security. He is contributing to ORNL’s scientific endeavor by conceiving, designing and implementing complex computer vision and multidimensional imaging systems – combining both hardware and software development – to perform quantitative analysis of complex datasets and/or to make quantitative measurement of various objects. Currently, Dr. Paquit is the lead scientist for the Data-analytics Framework for Additive Manufacturing project at the Manufacturing Demonstration Facility (MDF). This project aims at better understanding additive manufacturing process for the purpose of process certification and control. His research interests include applied signal and image processing, algorithm development on GPU platform, 2D and 3D image segmentation, multispectral and hyperspectral imaging, biomedical imaging, pattern recognition, remote sensing data understanding, and machine learning. He has published numerous peer-reviewed articles, one book chapter, submitted multiple invention disclosures and patent applications, and served on program committees of several international conferences.
ABSTRACT: The Manufacturing Demonstration Facility (MDF) at Oak Ridge National Laboratory (ORNL) is DOE’s leading research center established to provide industry with affordable and convenient access to facilities, tools and expertise to facilitate rapid deployment of additive manufacturing (AM) technologies for metals and polymers printers. This presentation will focus on 3D metal printing.
In the last decade, powder bed based electron beam melting has emerged as a potential technique for fabricating larger volume parts at relatively higher deposition rates compared to laser based AM techniques. However, the printing process is a very complex and dynamic mechanism that relies on a stable and synergetic combination of parameters linked to material properties, printer characteristics, and object geometry design. Currently, the relation between process parameters and formation of defects such as porosities are not clearly understood. To close this knowledge gap, the MDF is currently pursuing a significant effort on developing data analytics techniques, processing platforms, and cross disciplinary discovery tools for additive manufacturing process understanding. In particular, we are developing in-situ process monitoring techniques to combine image data for both near infrared (IR) imaging and high speed IR imaging that allows capturing information from each layer as it is deposited; and from sensor data collected from hundreds of sources integrated in the printing system to monitor the process behavior. The in-situ process data is being implemented into Dream.3D, a three-dimensional visualization tool, in conjunction with the Air Force Research Laboratory to enable visual representation of the build data from electron beam AM technologies. Additional information related to process intent data and post inspection data such as X-Ray computed tomography and microstructural information can also be implemented into the three-dimensional framework. Our objective is to combine and analyze these multimodal datasets to assess the printing processes with an unprecedented granularity, to quantify the nature and volume fraction of the defects, and to map geometric inaccuracies and surface roughness in the part and their variations as compared to the input CAD models. The presentation will discuss the advances made by MDF in data analytics for AM techniques till date.