TY - JOUR
T1 - In-process monitoring of porosity in additive manufacturing using optical emission spectroscopy
AU - Montazeri, Mohammad
AU - Nassar, Abdalla R.
AU - Dunbar, Alexander J.
AU - Rao, Prahalada
N1 - Funding Information:
at Pennsylvania State University: This work was partially supported by the Office of Naval Research, under Contract No. N00014-11-1-0668. This work was also partially supported by the Air Force Research Laboratory through America Makes under agreement number FA8650-12-2-7230. Any opinions, findings and conclusions or recommendations expressed in this publication are those of the authors and do not necessarily reflect the views of the Office of Naval Research, Air Force Research Laboratory, or America Makes. Funding at the University of Nebraska-Lincoln: One of the authors (PKR) thanks the NSF for funding his research through the following grants: CMMI-1719388, CMMI-1739696 and CMMI-1752069 (CAREER). Specifically, the concept of using spectral graph theory for modeling in metal additive manufacturing applications was funded through CMMI-1752069 towards a correct-as-you-build smart additive manufacturing paradigm. The authors thank the three anonymous reviewers whose constructive comments and suggestions have doubtlessly improved the rigor of this work. The authors also thank the Associtate Editor of the Transactions, Professor Zhenyu Kong for shepherding this work through the review process.
PY - 2020/5/3
Y1 - 2020/5/3
N2 - A key challenge in metal additive manufacturing is the prevalence of defects, such as discontinuities within the part (e.g., porosity). The objective of this work is to monitor porosity in Laser Powder Bed Fusion (L-PBF) additive manufacturing of nickel alloy 718 (popularly called Inconel 718) test parts using in-process optical emission spectroscopy. To realize this objective, cylinder-shaped test parts are built under different processing conditions on a commercial L-PBF machine instrumented with an in-situ multispectral photodetector sensor. Optical emission signatures are captured continuously during the build by the multispectral sensor. Following processing, the porosity-level within each layer of a test part is quantified using X-ray Computed Tomography (CT). The graph Fourier transform coefficients are derived layer-by-layer from signatures acquired from the multispectral photodetector sensor. These graph Fourier transform coefficients are subsequently invoked as input features within various machine learning models to predict the percentage porosity-level in each layer with CT data taken as ground truth. This approach is found to predict the porosity on a layer-by-layer basis with an accuracy of ∼90% (F-score) in a computation time less than 0.5 seconds. In comparison, statistical moments, such as mean, variation, etc., are less accurate (F-score ≈ 80%) and require a computation time exceeding 5 seconds.
AB - A key challenge in metal additive manufacturing is the prevalence of defects, such as discontinuities within the part (e.g., porosity). The objective of this work is to monitor porosity in Laser Powder Bed Fusion (L-PBF) additive manufacturing of nickel alloy 718 (popularly called Inconel 718) test parts using in-process optical emission spectroscopy. To realize this objective, cylinder-shaped test parts are built under different processing conditions on a commercial L-PBF machine instrumented with an in-situ multispectral photodetector sensor. Optical emission signatures are captured continuously during the build by the multispectral sensor. Following processing, the porosity-level within each layer of a test part is quantified using X-ray Computed Tomography (CT). The graph Fourier transform coefficients are derived layer-by-layer from signatures acquired from the multispectral photodetector sensor. These graph Fourier transform coefficients are subsequently invoked as input features within various machine learning models to predict the percentage porosity-level in each layer with CT data taken as ground truth. This approach is found to predict the porosity on a layer-by-layer basis with an accuracy of ∼90% (F-score) in a computation time less than 0.5 seconds. In comparison, statistical moments, such as mean, variation, etc., are less accurate (F-score ≈ 80%) and require a computation time exceeding 5 seconds.
UR - http://www.scopus.com/inward/record.url?scp=85074033501&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85074033501&partnerID=8YFLogxK
U2 - 10.1080/24725854.2019.1659525
DO - 10.1080/24725854.2019.1659525
M3 - Article
AN - SCOPUS:85074033501
VL - 52
SP - 500
EP - 515
JO - IISE Transactions
JF - IISE Transactions
SN - 2472-5854
IS - 5
ER -