Neural Network-Driven Machine Vision Learning for Adaptive Monitoring of Additive Manufacturing Processes
With the rapid advancement of additive manufacturing(AM)technology,its quality control and process monito-ring have increasingly become focal points of research.This study is dedicated to addressing quality fluctuations and defect issues in the 3D printing process by leveraging a neural network-driven machine vision system for efficient and adaptive monitoring of the AM process.Convolutional neural networks(CNNs)were utilized to conduct an in-depth analysis of high-resolution melt pool im-age data,aiming at real-time detection and classification of defects during the AM process.Experimental data were collected through a laser metal deposition process,and analysis and annotation were conducted,resulting in a comprehensive and high-quality dataset.Based on this dataset,a CNN model specifically designed for melt pool image recognition was developed,effec-tively extracting key features from complex images for the AM process.The findings demonstrate that the developed model exhibits high accuracy,recall,and excellent F1 scores in detecting and classifying defects in the AM process,validating its effectiveness in enhancing the precision and reliability of the AM process.Through real-time monitoring and analysis of the AM process,this research showcases the potent capabilities of neural networks in machine vision applications,offering new avenues for quality con-trol and process optimization in additive manufacturing technology.
万方数据
维普
