Fine-Tuned CNN with Data Augmentation for 3D Printer Fault Detection

Massive implementation of additive manufacturing is facing a challenge. A defect in the surface of the goods often causes an increase in product prices. To overcome this, effective anomaly detection is needed. Machine learning comes to help develop automatic detection to reduce the helps production efficiency. This paper proposed vision-based anomaly detection for surface defect detection of FDM printers. The model mainly employs CNN as the deep learning method, which outperformed image classification because of the ability to reduce the high dimensionality of an image without losing the information. Some additional layers grouped in the data augmentation layer are added to the trained model to improve the system's efficiency. It helps vary the number of input data and avoid overfitting during training. Public and private datasets are trained and validated to check the effectiveness of the proposed model. The result shows that the application of CNN with data augmentation could achieve better accuracy, loss, and classification time.