Inclination prediction of a super-sized open caisson foundation during sinking process based on ensemble learning

Open caisson foundations are widely used in the construction of various large structures，and the inclination of an open caisson is one of the most important indexes of its sinking attitude. Accurate prediction of the inclination is conducive to ensuring the sinking safety and steady of the open caisson and preventing potential construction risks. Based on two ensemble learning techniques，bagging and boosting，the random forest algorithm and XGBoost framework are applied for the inclination prediction modeling. The monitoring data of the structural stress at the bottom of the open caisson are used to predict the longitudinal height difference and transverse height difference. The reliability of the prediction model was verified by applying it to the super-sized open caisson foundation of the main bridge pylon in the Changtai Yangtze River Bridge Project，and the proposed model was compared with the prediction models applying other single machine learning algorithms. Then，the important parameters of the ensemble learning model were analyzed to study their influence on prediction accuracy. The results show that the prediction model in this paper can accurately predict the longitudinal height difference and transverse height difference and reasonably determine the inclination of the open caisson foundation. With fast operating speed and strong practicability，the proposed model has higher prediction accuracy than other single machine learning models. In addition，the prediction accuracy increases with the number of base learners and the maximum tree depth. The research results achieve the real-time prediction of the inclination of the open caisson foundation during the sinking process，which can provide an important reference for the monitoring of similar foundations. © 2023 Academia Sinica. All rights reserved.