This paper presents a multiobjective expectation–value optimization (MEVO)‐guided dynamic safety‐threshold framework for deep excavations in soft soils, integrating soft‐soil‐oriented indicator construction, drift‐aware rolling updates, and full‐process field validation. This study applies and extends the MEVO model to the Hangzhou Fengshouhu Metro Station excavation, a 28 m‐deep cut located in layered marine–lacustrine clays. A comprehensive database, including wall deflection, strut axial force, and pore‐water pressure was established. Region‐specific adaptations were introduced by (i) redefining the search domain through a combined expert judgement/historical‐case procedure and (ii) embedding a displacement–slope dual‐trigger rolling‐update mechanism to mitigate threshold drift. The optimized dynamic thresholds were implemented and validated over an 81‐day observation window. Compared with the original static design limits, the calibrated MEVO thresholds reduced the average alarm frequency by 41% and eliminated missed alarms (missed alarm rate = 0), while ensuring all key deformation indices remained within code‐prescribed envelopes. These results confirm the transferability of the MEVO framework to soft‐soil environments and provide a replicable workflow for deriving engineering‐ready dynamic safety thresholds in Hangzhou and similar coastal regions. The proposed methodology underpins proactive risk management for deep excavations.
Zheng et al. (Thu,) studied this question.