Gaussian Process Monitoring of Layerwise-Dependent Imaging Data

Additive Manufacturing (AM) enables the direct production of complex geometries from computer-aided designs (CAD). The AM fabrication process is often executed in a layer-by-layer manner, whereby minute printing errors one layer can manifest significant defects final part. In-situ quality monitoring and control are currently limited for processes cause low repeatability. Recently. advanced imaging increasingly invested leads to proliferation layerwise data, which provides an opportunity transform post-build inspection in-situ monitoring. However, existing methodologies primarily focus on key characteristics image profiles that tend be ability analyze variance components, as well root causes failure patterns critical improvement. This letter presents Gaussian Process with dependent correlation (AGP-D) model spatio-temporal data AGP-D consists three independent GP modules. first approximates base profile, whereas second third capture within same among layers, respectively. Based posterior predictions new Hotelling T 2 generalized likelihood ratio (GLR) tests formulated detect shifts newly fabricated causes. proposed methodology evaluated validated using both simulation real-world case study cylinder build by laser powder bed fusion (LPBF) machine. Experimental results show effective real-time modeling layerwise-correlated data.