Abstract: In order to improve the accuracy and applicability of inversion analysis model of material parameters for rockfill dam, an adaptive model based on quantum genetic algorithm (QGA) and multi-output mixed kernel relevance vector machine (MMRVM) is established. By introducing mixed kernel function, the MMRVM can accurately simulate the nonlinear relationship between the material parameters and the settlement of rockfill dam. Therefore, the finite element method (FEM) can be replaced by the MMRVM to reduce time consumption. Then, the kernel parameters of the MMRVM is optimized by the QGA, thus the QGA-MMRVM is adaptable to different inversion analysis problems. The parameters of constitutive model of dam materials can be inverted by fully utilizing QGA's global searching ability. Finally, the influences of the signal-noise ratio and the number of measured points on the calculation result are analyzed. The examples of Gongboxia dam show that the parameters of constitutive model of material can quickly and accurately calculated by the QGA-MMRVM. With its adaptability, the QGA-MMRVM has good application prospect and popularization value in practical engineering.

Key words: rock-fill dam, parameter inversion analysis, multi-output mixed kernel relevance vector machine, quantum genetic algorithm, adaptivity