Experimental investigation and machine learning-based modelling of shrinkage evolution due to natural sea sand on fiber reinforced cementitious composites

This study investigates the shrinkage behavior of Engineered Cementitious Composites (ECCs) made with natural sea sand from Shenzhen Bay (SBS), focusing on the effects of ionic concentrations modified by rinsing. EDS and XRD analyses showed that rinsing significantly reduced chloride content by 80% after two rinses and nearly 100% after eight; sodium dropped to <0.01%, and sulfur trioxide decreased by 60%. Shrinkage testing revealed that two- and eight-times rinsed sand reduced shrinkage by 11.3% and 20.5%, respectively, at 28 days compared to unwashed sand. Increasing fiber content to ~2% and optimizing the water-binder ratio also contributed to reduced shrinkage. A machine learning model developed for temporal shrinkage prediction achieved ~0.9 overall accuracy and >0.85 temporal accuracy in early-age measurements. Principal Component Analysis showed that rinsing degree and fiber volume were the dominant factors (~40% influence each), while the water-binder ratio played a lesser role (~20%), with factor importance varying over time.

This is an original manuscript of an article published by Taylor & Francis in Journal of Sustainable Cement-Based Material on 06 Jul 2025, available at: https://doi.org/10.1080/21650373.2025.2524763.