Digital twins for rapid in-situ qualification of part quality in laser powder bed fusion additive manufacturing

Abstract

This work concerns the laser powder bed fusion (LPBF) additive manufacturing process. Currently, LPBF parts are inspected post-process using such techniques as X-ray computed tomography, optical and scanning electron microscopy, among others. This empirical build-and-test approach for qualification of part quality is prohibitively expensive and cumbersome. To enable rapid and accurate in-situ qualification of LPBF part quality, in this work, we developed a physics and data-integrated digital twin approach. To demonstrate the approach, Inconel 718 parts of various shapes were manufactured under differing LPBF processing conditions. The process was continuously monitored using in-situ thermal and optical tomography imaging cameras. The part-scale thermal history was predicted using an experimentally validated computational thermal simulation. The simulation-derived thermal history and sensor signatures were used as inputs to a k-nearest neighbor machine learning model. The machine learning model was trained with ground truth porosity and microstructure data obtained from post-process characterization. The approach predicted the onset of porosity, meltpool depth, grain size, and microhardness with an accuracy exceeding 90 % (R2). This work thus takes a critical step towards realizing an in-situ Born Qualified part quality assessment paradigm in LPBF.