Smart Additive Manufacturing: Sensing, Monitoring, and Machine Learning for Quality Assurance in Metal Additive Manufacturing

The long-term goal of this research is to advance in-situ sensing and monitoring approaches to mitigate flaw formation in metal additive manufacturing (AM) processes. Despite the considerable cost and time savings, the lack of process consistency has deterred the application of AM in safety-critical production environments. Due to the layer-wise addition of material in AM, a flaw, e.g., porosity, in a layer may get permanently sealed in by subsequent layers thereby adversely affecting part performance, such as fatigue life and strength. These flaws tend to occur notwithstanding extensive a priori materials and process optimization. Thus, to ensure the commercial viability of AM processes it is necessary to develop efficient sensing, diagnosis, and flaw correction approaches. The goal of this dissertation is to detect flaw formation using data from heterogeneous sensing arrays embedded inside AM machines – a Big Data Analytics and Sensor Fusion problem. As a step towards this goal, the objective of this dissertation is to develop and apply advanced data analytics algorithms that can fuse process signatures acquired from a heterogeneous in-situ sensor array, and subsequently identify the nature (type) and severity of an evolving flaw. Two types of metal-based AM processes are specifically studied in this work: laser powder bed fusion and droplet-on-demand liquid metal jetting. The approaches developed in this work are capable of detecting the type and severity of flaw formation with a statistical fidelity exceeding 95%