Zoning design of cemented sand and gravel dam materials based on DeepFit–memory gradient search–particle swarm optimization fitting optimization

Abstract

Listen

Translate article

Cemented sand and gravel (CSG) dams are preferred for their cost-effectiveness and adaptability. This study introduces a zoning fitting optimization design method aimed at improving material strength utilization and reducing construction costs. Using Latin sampling and reduced strength safety calculations, an implicit function relationship is established between zoning parameters and safety through an improved deep neural network (DNN). Subsequent dimensionality reduction facilitates optimization using the memory gradient search–particle swarm optimization (MGS-PSO) algorithm. The method's efficacy is demonstrated by optimizing zoning parameters for Japan's Tobetsu Dam, achieving significant improvements in safety and material use. This approach increases safety by 22% and reduces cementing dosage by 36% compared to traditional methods, underscoring its potential in enhancing dam stability and efficiency.