Experimental and data-driven study of crack self-healing in lime mortar using Bacillus subtilis

Lime mortar is a sustainable and historically significant construction material widely used in heritage structures. However, its low initial mechanical strength and high susceptibility to cracking limit its use in modern construction. This study experimentally explores the crack self-healing in lime mortar by adding Bacillus subtilis and develops data-driven predictive models to evaluate its self-healing performance. Mortar samples were prepared with bacterial volume fractions of 10%, 15%, and 20% and tested for compressive strength over curing periods varying from 7 to 28 days. The self-healing process relies on microbially induced calcium carbonate precipitation, which helps seal cracks, refine pores, and densify the microstructure. Results showed that incorporating bacteria into lime mortar improved compressive strength and durability compared with traditional lime mortar, with optimal results at a 20% bacterial concentration after 28 days, achieving a maximum compressive strength of 1.55 MPa. To predict self-healing performance, machine learning methods, including Artificial Neural Networks, Random Forests, Gaussian Process Regression, and XGBoost, were applied to a smaller dataset of 120 experimental data points. The XGBoost model achieved the highest accuracy (R² = 0.984), followed by the Random Forest model (R² = 0.976). However, machine learning models need to be tested with larger datasets to provide concrete results. The findings suggest that bio-mediated self-healing is effective in reducing microcrack growth and enhancing fracture resistance in lime mortars. The combined experimental and computational approach provides a powerful tool for designing durable, self-healing plastering materials, offering essential benefits for sustainable construction, architecture, and heritage preservation.