Failure Detection of Fused Filament Fabrication via Deep Learning

Zhang, Zhicheng
Fidan, Ismail
Journal Title
Journal ISSN
Volume Title
University of Texas at Austin

Additive Manufacturing (AM) is used in several fields and its utilization is growing sharply in almost every aspect of daily life. The focus of the current studies in the AM field is generally focused on the development of new technologies and materials. In addition, there is a limited number of research studies on the troubleshooting aspects of the AM processes. For the most commonly used Fused Filament Fabrication (FFF) process, the waste of material and time due to the printing errors are still an unsolved problem. The typical errors such as nozzle jamming and layer mis-alignment are inevitable during the printing process, and thus cause the failure of printing. It is a challenging task to clearly understand the physical behavior of FFF process with uncertainty, due to the phase transition and heterogeneity of the materials. Therefore, to detect the printing error, this research proposes a deep learning (DL) based printing failure detection technique. In this study, DL is utilized to monitor the printing process, and detect its failures. This newly developed DL framework was beta-tested with a commercially available FFF setup. The beta testing results showed that this technique could effectively detect printing failures with high accuracy.