Elasto-plastic parameter inversion of tunnel engineering based on genetic-Gaussian process regression algorithm

Performance of learning machines is the key to determine the effectiveness of intelligent displacement back analysis. The Gaussian process regression (GPR) algorithm is introduced into the field of parameter inversion to make up for the deficiency of the present intelligent inversion method. In addition, a combined kernel function of GPR(CKGPR) obtained by additive single standard isotropy covariance functions is put forward to improve the generalization ability of a single kernel function. At present, the hyper-parameters of GPR are achieved by maximizing likelihood function of training samples based on the conjugate gradient algorithm. The conjugate gradient algorithm is replaced by the genetic algorithm (GA) coded in decimal system to optimize the hyper-parameters of GPR with the combined kernel function, and the corresponding calculation code is programmed in Matlab. From the elasto-plastic parameter inversion results of Beikou tunnel, it can be concluded that the GA-CKGPR algorithm can obviously improve the inversion precision than the standard GA-GPR and GA-SVR algorithms, so it can be utilized in parameter inversion of geotechnical engineering and meanwhile can serve as a reference for similar projects.