Investigation on shear strength properties of water-bearing concrete-rock interface based on convolutional neural network recognition method.

Concrete has been commonly employed in building facilities such as dam foundations, but few scholars pay attention to the effects of high-humidity environments on the interfacial failure properties between rock and concrete. In this study, a new efficient convolutional neural network (CNN)-based method for the identification of concrete-rock interfacial transaction zone (ITZ) was proposed, and the shear strength variation of the concrete-rock ITZ affected by water content was systematically studied. Scanning electron microscope images were utilized for image recognition to elucidate the reasons behind the effect of moisture content on interfacial strength. The results show that the shear strength of the concrete-rock ITZ first increases and then decreases with prolonged soaking time. Moreover, at a microscopic level, it can be observed that the thickness of the concrete-rock ITZ decreases as the soaking time increases; however, over-soaking leads to the deterioration of the transition zone with powder agglomeration and increased cracking. The CNN image recognition has considerable prospects in the field of the concrete-rock ITZ, allowing for the incorporation of additional functionalities in the future. • A new efficient convolutional neural network based method for the identification of concrete-rock interfacial transaction zone was proposed. • The shear strength evolution law of concrete rock interface was explored by using the shear test. • The research confirms the capability and potential of CNN in concrete image recognition. [ABSTRACT FROM AUTHOR]