@article{01KDQSTRSE87WB6XA8PJTHKV69,
  abstract     = {{Porosity formation in polymer selective laser sintering (SLS) significantly impacts part quality, as unwanted pores compromise mechanical properties. In-process control of porosity is a promising solution, but existing in-situ monitoring systems lack the necessary speed to meaningfully intervene in small localized areas, and the specificity to guide targeted process adjustments. This study addresses these limitations by introducing an in-situ monitoring system that predicts control-actionable porosity targets at high speeds (2 kHz). The system combines multimodal visible and short-wave infrared imaging, complementing information about the powder bed condition and thermal behavior. A neural network then makes predictions based on learned correlations between the image data and the porosity targets. The neural network incorporates both contextual and temporal information through a specialized architecture that leverages adaptive weight networks and a temporal sliding window. The proposed targets are extracted post-print using x-ray computed tomography analysis and serve as ground truth targets to train the neural network. Based on pore concentrations and spatial uniformity measurements, the targets have a clear physical justification, and allow any printed part to be used as training data. Crucially, by distinguishing between insufficient melting and overheating, these targets resolve the directionality of the required laser power adjustment, enabling non-ambiguous control actions. Experimental results on cylindrical test parts demonstrate good correlations between the porosity targets and expected pore formations. Furthermore, the proposed in-situ system shows strong predictive performance, with normalized mean absolute errors of 14% for predicting relative pore concentrations, and a correlation coefficient of 0.68 for predicting the underlying cause. The combination of a high-speed system and the prediction of control-actionable porosity targets pave the way for effective process control in SLS, improving part quality.}},
  articleno    = {{025602}},
  author       = {{Vandecasteele, Mathieu and Santander, Ricardo and Iuso, Domenico and Nourazar, Mohsen and Philips, Wilfried and Cornelissen, Sven and Booth, Brian}},
  issn         = {{0957-0233}},
  journal      = {{MEASUREMENT SCIENCE AND TECHNOLOGY}},
  keywords     = {{powder bed fusion,porosity,defect detection,selective laser sintering,in-situ}},
  language     = {{eng}},
  number       = {{2}},
  pages        = {{23}},
  title        = {{High-speed multimodal actionable porosity prediction for enabling in-process control of selective laser sintering}},
  url          = {{http://doi.org/10.1088/1361-6501/ae2cbe}},
  volume       = {{37}},
  year         = {{2026}},
}

Altmetric

View in Altmetric

Web of Science

Times cited: