Sensory Data Fusion Using Machine Learning Methods For In-Situ Defect Registration In Additive Manufacturing: A Review

Abstract

In-situ control to predict and mitigate defects in Additive Manufacturing (AM) could significantly increase these technologies’ quality and reliability. Thorough knowledge of the AM processes is needed to develop such a controller. Recent studies utilized various methods to acquire data from the process, build insight into the process, and detect anomalies within the process. However, each sensory method has its unique limitations and capabilities. Sensor fusion techniques based on Machine Learning (ML) methods can combine all the data acquisition sources to form a holistic monitoring system for better data aggregation and enhanced detection. This holistic approach could also be used to train a controller on top of the fusion system to master the AM production and increase its reliance. This article summarizes recent studies on sensor utilization, followed by ML-based sensor fusion and control strategies.