Classification of Aging Grade of Heat-resistant Steel based on Improved Shufflenet Network

Heat-resistant steel is important material for manufacture of power generation equipment. The aging grade of heat-resistant steel has an important impact on whether the equipment can operate safely. The aging grade of heat-resistant steel needs to be judged according to metallographic images of material during on-site inspection. However, traditional manual classification method has high requirements for experience, an improved shuffle net network is proposed in this paper, which can automatically and accurately classify heat-resistant steel aging grade based on metallographic images. Due to the fact that the background of metallographic images is complex and deep image features are difficult to extract, the dual attention module is introduced into the improved network to realize accurate extraction of image feature information. In order to meet the requirements of both high accuracy and lightweight, h-swish activation function is introduced into the improved network. Metallographic images of heat-resistant steel of different aging grades are used for model training and verification. Experimental results demonstrate the feasibility and accuracy of this method, which is based on metallographic images to realize automatic classification of aging grade of heat-resistant steel.