Metal additive manufacturing (MAM) offers a larger design space with greater manufacturability than traditional manufacturing has offered. Despite continued advances, MAM processes still face huge uncertainty, resulting in variable part quality. Real-time sensing for MAM processing helps quantify uncertainty by detecting build failure and process anomalies. While the high volume of multidimensional sensor data—such as melt pool geometries and temperature gradients—is beginning to be explored, sensor selection does not yet effectively link sensor data to part quality. To begin investigating such connections, we propose network-based models that capture in real time 1) sensor data's association with process variables and 2) as-built part qualities' association with related physical phenomena. These sensor models and networks lay the foundation for a comprehensive framework to monitor and manage the quality of MAM process outcomes.