Analysis of Diaphragm Wall Deflection Induced by Excavation Based on Machine Learning

For the concrete diaphragm wall (CDW) supported excavation, excessive deflection may pose a potential risk to adjacent structures and utilities in urban areas. Therefore, it is of significance predict CDW deformation with high accuracy efficiency. This paper investigates three machine learning algorithms, namely, back-propagation neural network (BPNN), long short-term memory (LSTM), gated recurrent unit (GRU), excavation-induced deflection. A database field measurement collected from an excavation project Suzhou, China, used verify proposed models. The results show that GRU exhibits lower prediction errors better robustness 10-fold cross validation than BPNN executes less computational time LSTM. most suitable algorithm for considering both effectiveness efficiency, predicted can provide reasonable assistance safety monitoring early warning strategies conducted on construction site.