Multi-objective optimization-based updating of predictions during excavation

In this paper, an efficient multi-objective optimization (MOOP)-based updating framework is established, which involves (1) the development of an enhanced multi-objective differential evolution algorithm with good searching ability and high convergence speed, (2) the development of an enhanced anisotropic elastoplastic model considering small-strain stiffness with its implementation into a finite element code, and (3) the proposal of an identification procedure for parameters using field measurements followed by an updating procedure. The proposed updating framework is verified with a well-documented excavation case where the small-strain stiffness, the anisotropy of elasticity, the anisotropy of yield surface for natural clays, and the parameters of the supporting structures and diaphragm wall are consecutively updated during the staged excavation process. The advantages of the proposed updating framework compared to the Bayesian updating on the same case are also illustrated.