Multiple Nonlinear Regression Models for Predicting Deformation Behavior of Concrete-Face Rockfill Dams

Abstract

Deformation assessment and control are important issues in the construction of concrete face rockfill dams (CFRDs). The design and construction of CFRDs require deformation behavior that can be estimated rapidly to support engineering optimization and safety assessment. This study aims to develop robust empirical prediction models with physical meaning for predicting key indices of CFRD deformation behavior based on in-service case history data. A database of 87 case histories of in-service CFRD constructed over the past 50 years was compiled. A multiple regression method is adopted to develop empirical relationships between three key indices (crest settlement, internal settlement, and face slab deflection) and six dam construction-related control variables (dam height, void ratio, foundation condition, intact rockfill strength, valley shape, and operation time). The internal correlation between the key indices and control variables is discussed. Dam height, intact rockfill strength, and foundation condition are found to be the important factors influencing the three key indices. The developed models are compared with some published methods to discuss model rationality and accuracy. The feasibility and application of the models are further validated considering one case study.