Design of process monitoring system for extrusion based 3D printers

• Main Author: Chin, Ryan Jun Xiang
• Other Authors: Erdal Kayacan
• Format: Final Year Project
• Language: English
• Published: 2018
• Subjects: DRNTU::Engineering; DRNTU::Engineering::Aeronautical engineering
• Online Access: http://hdl.handle.net/10356/75603
• Institution: Nanyang Technological University

Additive manufacturing (AM) processes are capable of producing objects of high precision. There are many AM methods and Fused Deposition Modelling (FDM) will be the method that will be focused on in this report. The precision of printing is dependent on many parameters that have to be controlled well during printing. These parameters are subjected to errors happening during the print and may deviate from their original values as a result. Typically, a printing process will take many hours and will unlikely be under manual observation. As a result, these errors go undetected and the printer ultimately produces a defective product, resulting in material and time wastage. A solution to reducing material and time wastage is to develop a process monitoring system which is capable of detecting errors and taking counter measures in real time. In this project, the concept of machine learning and visual detection is used to train neural networks to identify and classify errors that happen during a print job. Data sets used to train the neural networks are generated using image editing software. These data sets consist of different errors that are digitally generated. Three data sets representing different errors of stringing, under extrusion and misalignment are generated. They are subsequently used to train the neural networks and results show that the training with the stringing data set is effective. The neural networks are able to correctly detect and identify the type of error that occur on a layer of material. This project proves that visual detection can indeed be effective in detecting errors that may occur during a print job. This will greatly reduce material and time wastage as well as eliminate the need for manual observation of a print process.