@Article{sdhm.2026.084515,
AUTHOR = {Zeyu Wang, Ying Yu, Xin Huang, Guangwen Guo, Degao Zou, Kaiyan Tan, Jiayu Long, Song Gao},
TITLE = {AI-Enabled Prediction and Control of Settlement during Construction of Concrete-Faced Rockfill Dams: Method and Engineering Application},
JOURNAL = {Structural Durability \& Health Monitoring},
VOLUME = {},
YEAR = {},
NUMBER = {},
PAGES = {{pages}},
URL = {http://www.techscience.com/sdhm/online/detail/28219},
ISSN = {1930-2991},
ABSTRACT = {(1) Background: During the construction of the concrete-faced sand-gravel dam with a height of 247 m at the Dashixia Water Conservancy Project in Xinjiang, China, pronounced stage-related changes in dam settlement and marked settlement differences among dam zones were observed. Under these conditions, conventional settlement monitoring and prediction methods were insufficient to support the real-time decision-making required for on-site construction. (2) Methods: To address this practical engineering problem, this study developed an AI-enabled method for settlement prediction and control during construction. Settlement monitoring data collected throughout construction were combined with key dam filling parameters to establish a multi-source dataset for the construction period. On this basis, a settlement prediction model with rolling updates based on random forest regression was established. The model can predict the settlement development trend, support the identification of potential nonuniform settlement zones, and provide references for construction control based on actual construction conditions. (3) Results: The results of field application and quantitative error evaluation show that the predicted settlement is generally consistent with the measured settlement trend. The model effectively captures settlement responses during construction stoppage and the subsequent resumption of filling. (4) Conclusions: The proposed method provides practical quantitative support for optimizing construction organization and improving the precision of settlement risk control.},
DOI = {10.32604/sdhm.2026.084515}
}