AI-Driven Time-Variant Reliability Assessment of an Elevated Water Tank Over Its Service Life Using Radial Basis Function and Artificial Neural Networks

Civil Engineering and Construction Laboratory (LGCC), Mohammadia School of Engineering, Mohammed V University, Rabat, Morocco

^* Najib ZEMED: najibzem@gmail.com

Abstract

This study introduces a time-dependent reliability assessment of elevated reinforced-concrete water towers, highlighting the combined influence of material behaviour and environmental actions. The proposed framework integrates Kriging and Artificial Neural Networks within an active-learning ensemble, enabling a substantial reduction in computational effort while preserving the accuracy of Monte Carlo simulations. The analysis is based on a concrete crack-width limit state and explicitly incorporates long-term material effects, including concrete creep, as well as realistic wind conditions whose variability is adjusted according to projected climate-change scenarios. Relative humidity is also considered to capture its influence on creep development and serviceability performance. Results show a marked reduction in reliability under low-humidity conditions due to increased deformation, and a significant contribution of climate-induced wind variability to structural demand. Sensitivity analyses indicate that geometric parameters, particularly tank height and radius, play a major role in failure probability, whereas improvements in material properties help to reduce crack widths and enhance durability. A maintenance strategy based on a target reliability index is finally examined, demonstrating its effectiveness in improving performance when re-evaluated over the service life.