Session: 02-01-03: 7th Annual Conference-Wide Symposium on Additive Manufacturing: Polymers I

Paper Number: 96111

96111 - Detecting Defects in Low-Cost 3D Printing 

An automated 3D printer defect detection system for low-cost 3D printers is described in this paper. One of the main issues with fused deposition modeling (FDM) based 3D printing is that defects are quite common. Currently, there is no widely adopted system to automatically monitor the 3D printing process for defects. Developing such a system would save time and money for both small-scale prototyping projects and large scale automated FDM production. The proposed system must be able to detect a number of defects, specifically slippage and filament runout. It should be able to recognize what defect is occurring and notify the user as early as possible in the printing process. Several systems have been previously proposed that perform similar functions. Many of these systems use sophisticated sensors for detecting defects in various metal printing processes, while many others use commercially accessible cameras to detect defects in FDM printing. Most of these systems use some form of machine learning model to detect and recognize defects. Hence, the  goal for this project was to create an automated detection system, with a hardware sub-goal of constructing a standardized environment for the printer and camera, and a software sub-goal of developing an algorithm that detects when the part has slipped off the print bed (slippage), and when the printer has stopped extruding filament (filament runout). For this project, several different algorithms were tested, and the environment surrounding the 3D printer was continuously improved to better assist the detection algorithms. The system’s sensor hardware consisted of a high-resolution webcam. This was affixed to one of several 3D printed mounts, which consistently positioned the camera relative to a given printer. Additionally, there was a 3D printer enclosure, and a ring light and LEDs in various locations to control the lighting around the printer. Two main algorithms were developed. The first algorithm extracted the edges of the 3D printed part on the print bed and then compared that to the STL file of the part. The second algorithm used statistical analysis on an image to determine how different it was from an image taken earlier in the 3D print. To test the detection algorithm, time lapse videos were generated by taking layer-by-layer pictures of numerous prints. Images from the videos were then run through the algorithm to determine its effectiveness in detecting defects. The system correctly detected and identified slippage in a number of trials. Further improvements are being carried out that includes developing a library of camera mount designs to work for various different cameras and printers. Future software work includes finer tuning of the algorithm to allow it to work in more general scenarios. This paper serves as a detailed description of what has been outlined above.

Presenting Author: Mark Forte Worcester Polytechnic Institute

Presenting Author Biography: Mark Forte is a student in Computer Science at WPI

Authors:

Mark Forte Worcester Polytechnic Institute
Madison Eisenhour Worcester Polytechnic Institute
Ryan Malkowski Worcester Polytechnic Institute
Pradeep Radhakrishnan Worcester Polytechnic Institute
David Brown Worcester Polytechnic Institute