Predicting porosity in wire arc additive manufacturing (WAAM) using wavelet scattering networks and sparse principal component analysis

Wire arc additive manufacturing (WAAM) is getting much research attention because of its cost-effectiveness in the metallic production of large and complex parts. In pursuit of best-quality products and minimizing material loss, multimodal process monitoring methods are key. This paper presents the use of acoustic and current signals in identifying one of the critical defects in WAAM, i.e., porosity. Aluminum and unalloyed steel were deposited in a controlled environment which developed different amounts of porosity alongside measurements from current and gas sensors. Feature reduction of the signals was carried out using a combination of wavelet scattering networks and sparse principal component analysis (sPCA). While the models predict porosity reasonably, the dominant features learned by the model are also investigated and reported.