Analysis and prediction of diaphragm wall deflection induced by deep braced excavations using finite element method and artificial neural network optimized by metaheuristic algorithms. (May 2022)

• Record Type: Journal Article
• Authors: Yong, Weixun; Zhang, Wengang; Nguyen, Hoang; Bui, Xuan-Nam; Choi, Yosoon; Nguyen-Thoi, Trung; Zhou, Jian; Tran, Trung Tin
• Is Part Of: Reliability engineering & system safety. Volume 221(2022)
• Publication Date: 2022-05
• Subjects: System safety -- Diaphragm wall -- Soft clay -- Braced excavation -- Artificial neural network -- Metaheuristic algorithms; Reliability (Engineering) -- Periodicals; System safety -- Periodicals; Industrial safety -- Periodicals; Fiabilité -- Périodiques; Sécurité des systèmes -- Périodiques; Sécurité du travail -- Périodiques; 620.00452
• DOI: 10.1016/j.ress.2022.108335
• Languages: English
• ISSNs: 0951-8320

Abstract

Highlights: MLP was applied to predict diaphragm walls deflection induced by braced excavations. HHO and WOA were applied to optimize the MLP model. HHO-MLP and WOA-MLP were proposed as the novel and robust models for this aim. HHO-MLP model yielded the most dominant performance in this study. SSR, STR, ED, and WS are the most important input parameters in this study. Abstract: The construction of metropolises in smart cities is the trend of developed countries. However, it may cause damages to the surrounding structures. For this reason, the diaphragm wall has been applied to prevent the deformation or collapse of the surrounding structures. Diaphragm walls can be deflected due to the swelling pressure or other geotechnical properties during construction. Therefore, the accurate prediction of diaphragm wall deflection (DWD) is challenging in construction aiming to ensure the safety of the surrounding structures. This study is, therefore, to propose two intelligent models for predicting DWD induced by deep braced excavations based on finite element method (FEM) and metaheuristic algorithms. Accordingly, a total of 1120 finite elements were analyzed to investigate the behaviors of DWD. Finally, a neural network with multiple layer perceptron (MLP) was optimized by two metaheuristic algorithms for predicting DWD, including whale optimization (WO) and Harris hawks optimization (HHO), called MLP-HHO and MLP-WO, respectively. The results indicated that the proposed MLP-HHO and MLP-WO provided high accuracy in predicting DWD. A comparison of the proposed models in this study and previous studies was also discussed to highlight the superiority of the proposed MLP-HHO and MLP-WO models. Graphical abstract: Image, graphical abstract