In-situ qualification of lattice structures in L-PBF: A framework for robust geometrical deviation detection

Part qualification and process verification account for a substantial share of production costs in additive manufacturing, particularly for complex geometries that require expensive non-destructive evaluation. In-situ monitoring offers a promising strategy to mitigate these costs by enabling in-situ inspection and real-time anomaly detection. This study advances the development of industrial solutions for in-situ detection of geometrical deviations in laser powder bed fusion (L -PBF), with a focus on lattice structures as a representative class of complex geometries. From an industrial perspective, the literature reveals two main gaps: (i) the need for systematic calibration of image segmentation algorithms to improve in-situ reconstruction accuracy, and (ii) the lack of clear criteria for selecting deviation metrics that effectively capture actual geometrical defects. This work addresses both issues and provides practical guidelines for implementation. Specifically, we demonstrate that appropriate calibration settings make powder-bed image segmentation robust to variations in illumination and to changes in part size and geometry. In addition, we evaluate the influence of different deviation metrics on inspection performance, identifying those most suitable for reliable industrial use. The approach is validated through a case study on gyroid lattice structures manufactured via L -PBF, where controlled defects of varying severity were introduced to assess detection capability. Defects of varying severity were successfully detected through an appropriate selection of the deviation metric and calibration strategy, while maintaining a false alarm rate below 2%. These results highlight the feasibility of robust in-situ geometrical inspection for industrial applications.