Machine learning methods for predicting the durability behavior of earth mortars with date palm ash

This study examines the use of machine learning models to predict the durability characteristics of earth mortars enhanced with date palm ash (DPA), a crucial factor in ensuring the long-term performance and sustainability of earthen construction. A comprehensive dataset derived from experimental investigations was used to train and validate two models: An Artificial Neural Network hybridised by Neural Architecture Search (NAS-ANN) and a Random Forest with Neural Architecture Search (NAS-RF). Five key durability parameters, initial durability, capillary absorption, abrasion resistance, mass loss due to abrasion, and swelling behavior, were selected as outputs based on their relevance to structural integrity and longevity. The K-fold cross-validation technique rigorously assessed each model's predictive capabilities. Results indicate that the NAS-ANN model consistently outperforms the other models across all durability parameters, demonstrating superior accuracy and robustness. Across all parameters, the NAS-ANN model exhibits superior predictive performance compared to the NAS-RF model, accurately capturing complex relationships between material composition and long-term performance. These findings highlight the efficacy of DPA as a sustainable additive for enhancing the mechanical and physical properties of earth mortars, offering a promising avenue for environmentally responsible construction practices. The NAS-ANN model's accurate predictive capabilities provide a valuable tool for optimizing material design, creating durable and sustainable earth-based structures that withstand diverse environmental conditions. This research supports the broader adoption of DPA-modified earth mortars as a viable alternative to conventional building materials, promoting resource efficiency and reducing environmental impact within the construction industry.