How to Quickly Identify the Incorporation of Steel Slag in Manufactured Sand

Abstract

In recent years, the quality problem caused by the application of steel slag in concrete buildings have occurred one after another, which has been widely concerned and valued by the construction industry. How to effectively prevent steel slag from being incorporated into the raw materials of concrete has become a major problem. In this paper, the phase composition, chemical composition and macroscopic morphology of steel slags were analyzed in detail to clarify their characteristic identification elements. Subsequently, the rapid identification and detection methods of steel slag mixed in common fine aggregate of concrete were explored by means of grey value analysis, spectral analysis and deep learning. It can be found that the steel slag with high iron content makes the color of particles relatively dark. With the increase of steel slag content maxed in common fine aggregate of concrete, there is a linear positive correlation between the diffuse reflection absorption intensity and the steel slag content in visible light and near-infrared light range, while the grey value shows a decreasing trend with the steel slag content. Compared with the UNet model, the TransUNet model, which introduces a self-attention mechanism, can enhance the global perspective of recognition, thereby significantly improving the recognition accuracy for low content. Therefore, the use of grey value analysis, diffuse absorption spectrum and deep learning image recognition technology can effectively identify whether steel slag is mixed in manufactured sand, so as to ensure the safety and stability of concrete structures.