While laser powder bed fusion technology has gained widespread adoption across various industries, its susceptibility to low repeatability and manufacturing challenges often results in defect formation. Consequently, there is a growing interest in the development of in-process monitoring systems to detect defects as fabrication progresses. This necessitates robust correlations between process events and actual defects, based on an accurate comparison between datasets acquired during in-process monitoring and post-process measurements. This work explores the correlation between off-axis long-exposure process monitoring and aligned reference data on internal defects, obtained through X-ray computed tomography. Particularly, a machine learning workflow including feature extraction and logistic regression was designed and implemented, with a specific focus on predicting lack-of-fusion porosity formation directly from in-process monitoring data.
INVESTIGATION ON THE USE OF MACHINE LEARNING AND X-RAY COMPUTED TOMOGRAPHY FOR LACK-OF-FUSION POROSITY PREDICTION
Bonato N.
;Zanini F.;Carmignato S.
2024
Abstract
While laser powder bed fusion technology has gained widespread adoption across various industries, its susceptibility to low repeatability and manufacturing challenges often results in defect formation. Consequently, there is a growing interest in the development of in-process monitoring systems to detect defects as fabrication progresses. This necessitates robust correlations between process events and actual defects, based on an accurate comparison between datasets acquired during in-process monitoring and post-process measurements. This work explores the correlation between off-axis long-exposure process monitoring and aligned reference data on internal defects, obtained through X-ray computed tomography. Particularly, a machine learning workflow including feature extraction and logistic regression was designed and implemented, with a specific focus on predicting lack-of-fusion porosity formation directly from in-process monitoring data.Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.