Abstract

In the field of additive manufacturing (AM), mid-print failure is exceedingly common due to user error, bad design, or environmental factors that cannot be readily prepared for. This holds for most if not all types of AM, but perhaps none more so than the popular Filament Deposition Modeling (FDM) method machines. Absent total power failure, the bulk of the common modes of failure in FDM can be expressed as having an immediate impact on the mechanical system, whether that be a head collision due to warping, increased pressure on the stepper as it tries to push jammed filament, etc. The open loop nature of FDM machines does nothing to help the high rate of failure that FDM printers are known for compared to traditional methods of manufacturing. In this paper, a method for predicting failure due to mechanical malfunction of an FDM 3D printer is presented. The method proposed seeks to close the loop on FDM machines by characterizing the vibrations of the stepper motors which comprise an FDM machine. Using the acoustic emissions, a classifier is trained in order to assess the state of a print based off of supervised learning of known modes of failure. The resulting model is able to successfully predict jamming or air printing during a print with 90% training accuracy.

This content is only available via PDF.
You do not currently have access to this content.