A novel method for detecting the type of surface defects of hot rolled steel strip using the convolutional neural network

Steel production is essential in today's world. The classification of surface defects of steel strips in the steel industry is essential for their diagnosis because it is closely related to the quality of the final product. In this study, classification is considered to identify six types of defects from the North Eastern University dataset on hot rolled steel strip surfaces using artificial intelligence (AI). The proposed method is a kind of architecture based on a convolutional neural network. 200 × 200 images enter the convolutional neural network, changing to 32 × 32 in the first layer, 64 × 64 in the second layer, and 128 × 128 in the third layer. The test results show that this architecture achieves 93.54% accuracy in the test set, which is much more than comparable architectures. To evaluate the results of the proposed architecture, the criteria of accuracy, precision, and recall have been used.

A novel method for detecting the type of surface defects of hot rolled steel strip using the convolutional neural network Keywords:

Surface defects , Hot-rolled steel strip , Convolutional neural network , Deep Learning , Artificial Intelligence