Next-generation net-zero composite for underwater 3D printing construction: Hybrid machine learning optimized LC3 with recycled rubber

This study introduces the first sustainable limestone-calcined clay cement (LC 3 ) composite specifically developed for underwater 3D printing applications. Kaolinite, illite, and montmorillonite identified as the dominant clay types in Oman’s indigenous deposits were selected as the basis for formulating scalable, printable, and low-carbon LC 3 binders. A systematic factorial mix design strategy was employed to generate 287 unique formulations, each experimentally evaluated for the most critical mechanical and rheological properties for 3D printing, including compressive strength, flexural strength, and yield stress. Multiple machine learning (ML) models were trained to predict these properties, with a hybrid Random Forest + ANN framework selected to optimize the mix design. The optimization process was guided by practical engineering constraints, including raw material availability, printability thresholds, rheological performance, and mechanical strength requirements, ensuring industrial feasibility. The final formulation was designed to be user-friendly and adaptable for use by construction professionals and field engineers. It was further enhanced with underwater-compatible additives—anti-washout admixture (AWA), water-repellent admixture (WRA), and accelerator (ACC)—and incorporated recycled waste rubber (WR) at various sand replacement levels. WR was introduced to enhance resistance against wave impact, cyclic fatigue, and hydrodynamic loads, while simultaneously improving crack-bridging and long-term chemical durability in aggressive submerged environments. The WR–LC 3 -3DPC composite demonstrated compressive and flexural strength of 30.5 MPa and 3.8 MPa, respectively, and a yield stress of approximately 700 Pa. XCT and TGA analyses revealed matrix densification and thermal stability. After 120 days of immersion in acid–chloride solution, WR–LC 3 -3DPC retained up to 87% of its strength. These findings serve as a stepping stone toward the development of resilient, eco-efficient, and digitally printable materials for underwater construction, establishing a foundation for future research and large-scale implementation. All Python codes, decision tree outputs, and the complete experimental dataset are provided in the Appendix to facilitate reproducibility and further research.