Detection of layer height–based defects in additively manufactured part using deep learning algorithm

Additive manufacturing (AM) is the most extensively researched area considering the advantages and impact that would be made when commercialized. The reason additive manufacturing is not being widely used in industrial applications is due to the defects they produce, causing dimensional inaccuracy. These problems could be rectified or minimized by optimizing the process parameters. A few parameters could be controlled during the printing process, while a few printing process needs to be corrected step by step. However, in this study, error is not rectified, though it has been prevented from growing into a bigger problem. Once the error occurs, the user can decide whether the error is within acceptable range or not. If the error is within the acceptable range, the user can continue the printing process. If not, the entire printing process could be stopped with only a minimal amount of time and material wastage. The defect being monitored in this study is based on the layer thickness of the part. Since the layer thickness of a part majorly affects the mechanical properties of the part, the primary objective of this study would be monitoring the layer height of the 3D-printed part. Three-layer thickness values 0.1 mm, 0.2 mm, and 0.3 mm have been considered for this study with acrylonitrile butadiene styrene (ABS) as the filament. A deep learning algorithm is used to develop a model for predicting the occurrence of layer defects during the printing of part.