Search

In North America, where winter temperatures frequently dip below freezing, deicing salt is a common solution for melting snow and ice on roadways. However, this practice can result in the top layer of concrete flaking off, a phenomenon known as salt frost scaling. This problem not only leads to costly damage to infrastructure such as roads and bridges but also poses safety hazards for drivers and pedestrians. Therefore, it is crucial to understand the mechanisms underlying salt frost scaling and to develop effective strategies to mitigate its impacts. This thesis proposes a three-pronged approach to investigate this issue, including laboratory studies, mechanistic analyses, and predictive modeling. The experimental results have shown the low concentration of salt solution causes the most severe scaling. The previously established cryogenic suction mechanism combined with the micro-ice-lens pumping provides a sound explanation for the phenomenon. Pore drying and ice expansion due to freezing likely creates additional space for moisture uptake during freeze-thaw (F-T) cycles by enlarging the capillary pores and opening the channel to connect previously unconnected pores. This study also has revealed that hydrophobic impregnations fail to prevent the pumping effect associated with the F-T cycles and a marked negative effect of hydrophobic impregnation on the F-T resistance has been observed. These results shed light on the link between salt frost scaling and internal cracking, both of which depend on the degree of saturation, with the former associated with local moisture conditions and the latter with global moisture conditions. Large moisture uptake and poor air void system increase the likelihood of reaching the critical degree of saturation which leads to severe damage. These findings pave the way for modeling scaling. The polynomial regression model is proposed to quantify the scaling based on sorptivity, which is a measure of capillary suction and freezing water susceptibility, and the ability of the air void system to provide pressure relief. The results show that more permeable concrete should be equipped with a better-quality air void system. Finally, a machine learning model implemented by XGBoost algorithms was proposed to distinguish between scaling-resistant and non-scaling-resistant concrete based on previously identified important parameters. The results demonstrate the importance of sorptivity and spacing factor in scaling. This model, combined with the Shapley value, provides valuable insights into the scaling resistance of concrete and can be used as a reliable and efficient assessment tool for concrete quality.