IDE-MLSSVR-Based Back Analysis Method for Multiple Mechanical Parameters of Concrete Dams
Tengfei Bao, Jianming Li, Lu Yuanfu, Chongshi Gu
Abstract
A back analysis method based on multioutput least-squares support vector regression machine (MLSSVR) and improved differential evolution algorithm (IDE) is proposed to estimate multiple mechanical parameters of concrete dams. Based on the uniform design method, representative combinations of mechanical parameters are generated. Using these combinations, calculated hydrostatic displacement component differences are obtained through the finite-element method (FEM). The calculated hydrostatic component differences and mechanical parameters are then used to train MLSSVR models, with the model parameters selected by IDE. This allows for establishing the mapping relationship between dam displacements and corresponding mechanical parameters. The hydrostatic component differences separated from prototype observed data are then substituted into the model to obtain back analyzed mechanical parameters. If predefined terminating conditions are not satisfied, new combinations are added to the training set, and the training process continues until satisfied. The proposed back analysis method was successfully implemented to the Jinping I arch dam. Results indicate that the proposed method has high precision and strong generation ability.