A Combination of Vision- and Sensor-Based Defect Classifications in Extrusion-Based Additive Manufacturing

Additive manufacturing, also known as 3D printing, has been facing the problem of inconsistent processing defects and product quality as a transformative technology, thus hindering its wide application in industry and other fields. In this context, machine learning (ML) algorithms are increasingly u...

Full description

Saved in:
Bibliographic Details
Published in:Journal of sensors 2023-03, Vol.2023 (1)
Main Authors: Li, Xiao-Yu, Liu, Fu-Long, Zhang, Meng-Na, Zhou, Ming-Xia, Wu, Chuan, Zhang, Xiao
Format: Article
Language:English
Subjects:
Accuracy
Acoustics
Additive manufacturing
Algorithms
Artificial intelligence
Cameras
Classification
Data processing
Datasets
Defects
Experiments
Fused deposition modeling
Geometry
Industrial applications
Interlayers
Lasers
Machine learning
Manufacturing
Mechanical properties
Methods
Neural networks
Prediction models
Printing
Quality assurance
Sensors
Signal monitoring
Three dimensional printing
Vibration monitoring
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Additive manufacturing, also known as 3D printing, has been facing the problem of inconsistent processing defects and product quality as a transformative technology, thus hindering its wide application in industry and other fields. In this context, machine learning (ML) algorithms are increasingly used for automatic classification of process data to achieve computer-aided defect detection. Specifically, in this paper, two data-driven classification prediction models are built by monitoring the sensing signals (temperature and vibration data) and interlayer images during the printing process, using the fused deposition model (FDM) as the base case, and the prediction results of the two machine learning models are fused for prediction. The experimental results show that by fusing the prediction results of the two models, the classification accuracy is significantly higher than the prediction results of a single model. These findings can benefit researchers studying FDM with the goal of producing systems that self-adjust online for quality assurance.
ISSN:1687-725X
1687-7268
DOI:10.1155/2023/1441936