Identifying process interruption regions in laser powder bed fusion through mechanical property classification

Process interruptions in Laser Powder Bed Fusion (L-PBF) compromise part integrity by disturbing thermal cycles. This study presents a machine learning approach to identify interruption-affected regions using mechanical properties. Fifteen stainless steel 316L samples with controlled interruptions were fabricated varying laser power (200–250 W), scan speed (800–1200 mm/s), and interruption time (0.5–12 h). Nanoindentation provided 600 measurements of hardness and elastic modulus. Random forest, gradient boosting, and K-nearest neighbor algorithms were compared for classifying regions as before, during, or after interruption. Random forest achieved optimal performance with 97.31% accuracy, demonstrating that mechanical property measurements reliably identify interruption-affected regions for quality assessment.