Enhanced Surface Fault Detection in 3D Printed Objects: Leveraging Neural Networks and Transfer Learning

The concept of 3D printing, or additive manufacturing, has revolutionized the manufacturing sector by enabling the production of highly complex geometries with high precision. Variations in printing parameters or environmental conditions can lead to defects in 3D-printed products. Manually identifying these defective prints during the printing process is a difficult and time-consuming task, posing a significant obstacle to the final quality assurance of 3D-printed components. This research proposes a method to detect and classify defects in Additive Manufacturing (AM) by observing the surface of printed products using machine learning approaches such as Transfer Learning, Convolutional Neural Networks (CNN), Artificial Neural Networks (ANN), and Deep Neural Networks (DNN). Performing inference on a dataset of 1,288 images (based on OpenCV) using ConvNets has shown significant improvements in the accuracy of manufacturing fault detection. Using this method to identify surface defects on 3D-printed objects achieves detection accuracies of up to 97%. This approach has the potential to enhance quality control procedures in AM, thereby increasing the reliability of the final products.