ABSTRACT
In this study, the origins and development of caissons are briefly reviewed, CiteSpace bibliometric software is used to summarize the traditional research focuses of open caissons, and the rise of artificial intelligence in the field is identified. Therefore, this paper reviews representative papers employing AI algorithms in caissons, including 30 journal articles published since 2014. Many scholars have conducted data-driven intelligent analyses based on various machine learning methods using numerous measured data such as inclination, stress, strain, soil settlement, etc., or combining the results of refined finite element model calculations aiming to sense the caisson states, provide feedbacks on construction risks, and provide scientific decisions. However, there are research shortcomings in the existing literature, such as high dependence on data and weak interpretability of models, which need to be addressed in the future to achieve more accurate, fast and efficient analysis and prediction of caissons.
