Continuous Data Collection of Under Extrusion in FDM 3D Printers for Deep-Learning Dataset

A shortcoming noted in fused deposition modelling (FDM) 3D printing technology refers to lack of intelligent monitoring and intervention during the printing process. Fail prints can still occur during the printing procedure even though the printer is of industrial grade and far more expensive than that of hobby grades. Under extrusion has been determined as one of the frequent failures in 3D printing. Such failure stems from insufficient extrusion rate and/or inadequate melting temperature of filament during the print. Under extrusion failure may result in undesired layer gaps, missing layers, unbalanced layers, and even holes in the printed models that would make the models completely unusable. Hence, an effective method that can reduce waste materials and overall costs is by integrating artificial intelligence (AI) into 3D printers. However, a large dataset is required prior to the training process of deep learning. Hence, this study proposes an automated and continuous data collection of under extrusion samples in FDM 3D printers using Raspberry Pi and webcam. As a result, adjustment of the G-code of the standard tessellation language (STL) models and repeated process of printing 3D models can effectively achieve the desired images.