In Situ Process Monitoring and Machine Learning Based Modeling of Defects and Anomalies in Wire-Arc Additive Manufacturing

Wire Arc Additive Manufacturing (WAAM) has made great strides in recent years however, there remain numerous persistent challenges still hindering more widespread adoption. Defects in the parts produced degrade their mechanical performance. Inconsistency in the geometry of the weld beads or undesirable anomalies such as waviness, or humps can lead to loss of geometric accuracy and in extreme cases, when anomalies propagate to subsequent layers, build failure. Such defects can be mitigated by a controls framework, which would require a model that maps undesirable outcomes to information about the process that can be obtained in real time. This thesis explores the development of a multi-sensor framework for real time data acquisition and several approaches for arriving at such a model, employing well known machine learning methodologies including Random Forests, Artificial Neural Networks and Long Short Term Memory. The merits and drawbacks of these methods is discussed, and a physics based approach intended to mitigate some of the drawbacks is explored. The models are trained first on data obtained on a single build layer, and subsequently on a multi-layer wall.