Predicting the deformation behaviour of concrete face rockfill dams by combining support vector machine and AdaBoost ensemble algorithm

Abstract

The accurate evaluation of deformation behaviour is very important to the optimal design and safety evaluation of concrete face rockfill dams (CFRDs). Deformation prediction is a key problem in the CFRD construction. Based on statistical review of typical deformation behaviour of 94 CFRD cases, the relationship amongst three typical deformation indexes and six influencing factors of CFRDs is analysed using multiple linear regression method. The main influencing factors of the dam behaviour are discussed. An improved support vector machine (SVM) dam deformation prediction model is proposed based on the construction of an adaptive hybrid kernel function. The particle swarm intelligent optimisation algorithm is used to determine the main parameters. To solve the problem of nonlinear mutation and discreteness of case data, the SVM-AdaBoost prediction model of typical deformation indexes is established by iterative optimisation of the basic learner composed of the improved SVM model through the AdaBoost strategy. The comparative analysis of the proposed model and existing models shows that the proposed model has good prediction accuracy and can be used to predict the typical deformation behaviour of CFRDs accurately.