Process Monitoring in Friction Stir Welding Using Convolutional Neural Networks

Preliminary studies have shown the superiority of convolutional neural networks (CNNs) compared to other network architectures for determining surface quality friction stir welds. In this paper, CNNs were employed detect cavities inside welds by evaluating inline measured process data. The aim was determine whether are suitable identifying defects exclusively, or if approach is transferable internal weld defects. For purpose, 120 produced and examined ultrasonic testing, which basis labeling data as “good” “defective.” Different types artificial tested predicting placement into defined classes. It found that way significant accuracy achievable. When complete uniformly labeled “defective,” an 98.5% achieved a CNN, improvement state art. segment-wise, 79.2% obtained using showing segment-wise prediction also possible. results confirm well suited monitoring in welding their application enables identification various defect types.