A Bibliometric Study on the Applications of Neural Networks in Metal Surface Defect Inspection

The transformation industries are tasked with the process of converting vast amounts of chemical compounds into metal sheets. By applying artificial intelligence to detect anomalies and enhance the scope of visual inspection on metal surfaces, these industries have significantly advanced their digital transformation. In this research, we explore and highlight extensively used neural network architectures, since, despite the existing literature, recent bibliometric studies on this important topic remain scarce. Keeping this in mind, this work presents an analysis of a quantitative approach based on bibliometric metrics to evaluate the application of neural network architectures aimed at inspecting defects on metal alloy surfaces, using data from the Scopus and Web of Science databases as references. It primarily focuses on identifying aspects related to the research in this field, such as the distribution of scientific publications, the most relevant sources, the interaction dynamics among these sources, the most prominent subjects in recent research (trend topics), and the application of machine learning techniques to the detection of metal surfaces' failures. Finally, our results show a lack of relevant public datasets of adequate size, as well as structured methods to apply the combined neural network architectures and techniques of machine learning in real-world applications within the scope of this research.