An intelligent method for temperature load of arch dams.

Temperature load significantly affects the deformation and stress of arch dams. Traditional methods for determining temperature loads may not adapt to complex environments, such as high-altitude regions with large temperature variations. This study proposes an intelligent method for temperature load of arch dams. Based on the measured data from existing dams, a physics-informed convolutional neural network is developed, enabling direct application to newly designed dams. The intelligent method delivers annual temperature boundaries on the dam surface, aiding the finite element model in predicting deformation and stress during operational phases. Validation demonstrates accurate results for dam body temperature and deformation, aligning closely with measured data, while also pinpointing critical stress areas within the dam. This method advances the AI-assisted design by predicting temperature loads during dam operation, providing strong support for determining the most unfavorable loading without measured data and ensuring the safety of the structure. [Display omitted] • An intelligent physics-informed model for temperature load of dams is proposed. • Proposed model is transferable to the design phase where measured data is limited. • The frequency of the model outputs can be adjusted on actual demand. • Continuous temperature loads enable detection of the most unfavorable conditions. • Validation shows a relative temperature error of only 4.3 % without measured data. [ABSTRACT FROM AUTHOR]