Oral Presentation Ninth International Symposium on Life-Cycle Civil Engineering 2025

Determination of surface chloride concentration accumulation of concrete in marine environment by machine learning (112162)

Zhi-Yu Hao 1 2 , Wei-Ping Zhang 1 2 , Chao Jiang 1 2
  1. Key Laboratory of Performance Evolution and Control for Engineering Structures (Tongji University), Ministry of Education, Shanghai, PR China
  2. Department of Structural Engineering, College of Civil Engineering, Tongji University, Shanghai, PR China

There are multiple aggressive agents like chloride or/and sulphate ions with high concentrations in typical saline soil, posing a serious threat to the durability of concrete structures. Similar to the marine atmospheric environment, the saline soil above the underground water level is unsaturated, and the moisture content in soil depends on the relative humidity of the surrounding atmosphere. The accumulation process of surface chloride concentration (Cs) of concrete is the main difference in chloride penetration within concrete under marine atmosphere and saline soil. Experimental study was conducted on chloride penetration within concrete in saline soil, considering the influence of saturation degree, salinity concentration, and compaction degree of soil. The spatial distribution of chloride concentration in concrete was measured by potentiometric titrator at different testing intervals. A machine learning (ML) model was developed to predict Cs based on a chloride penetration database of both field exposure and laboratory test. In order to ensure the model's generalization capability, the database covers Cs of concrete with environmental parameters and exposure time in both marine atmosphere and saline soil environments. Different machine learning algorithms including linear regression (LR), support vector machine (SVM), and extreme gradient boosting algorithm (XGBoost) were employed to train and validate the dataset. It was found that the ensemble learning method has a significant advantage in prediction accuracy. Thereafter, the machine learning model was used to predict the surface chloride concentration of concrete in saline soil. The results show that chloride concentration of concrete in unsaturated soil is higher than that in saturated soil, and increases with the increase of moisture content. Increased salt concentration in soil leads to higher Cs values, resulting in faster chloride transport and earlier corrosion initiation. Additionally, lower compaction degree of soil reduces the permeability of the soil matrix, thereby impeding the ingress of chloride transport.