A Hierarchical Gated Recurrent Unit Network Model for Settlement Prediction

초록

Serious issues regarding ground settlement can arise when cities or roads are constructed on soft soil. Soft soil experiences gradual consolidation over time, requiring accurate prediction of settlement through ground surveys and monitoring during construction. However, uncertainties in ground information and time-consuming factors pose challenges to precise settlement prediction. To address this, we propose a hierarchical network based on GRU (Gated Recurrent Unit) for predicting settlement in soft soil. Our network model involves analyzing patterns of preloading for consolidation and subsequent settlement changes, applying hierarchical models based on soft soil layer characteristics. The experimental implementation of the proposed network shows that choosing a suitable network based on the soil characteristics of the target region can lead to improved accuracy.1 © 2023 Owner/Author.

키워드