| On the basis of summing up the tunnel engineering monitoring, parameter back analysis of surrounding rock, the tunnel of time series prediction, the focus of this paper include the following aspects:(1) The Gaussian process-differential evolution algorithm is proposed, and developing the GP-DE program based on matlab. Using the non-linear mapping relations between processing ability of Gaussian process. Difference evolution algorithm was used to optimize the GP-DE contains super parameters, effectively improve the precision of the nonlinear mapping relation model, parameters of intelligent optimization back analysis for underground engineering rock mass and rock mass changes in time series prediction provides a new method.(2) In combination with the practical engineering of dalian subway, the multivariate information automation monitoring system is decorated in the engineering field, in order to get more timely abundantly and accuratly information of the surrounding rock, and carring on the comprehensive analysis of the collected information. Otherwise by using numerical calculation, constructing orthogonal experiment scheme for parameter sensitivity analysis of surrounding rock, selected the obvious parameters for surrounding rock. Using field monitoring results of displacement of surrounding rock stress information as the control values, surrounding rock parameters GP-DE stress joint displacement back analysis, and compared with DE displacement back analysis of surrounding rock parameters, verify the superiority of the method.(3) In combination with the practical engineering of Chen village shop mountain tunnel, consider the effect of seepage. By using numerical calculation method, the parameter sensitivity analysis of surrounding rock and GP-DE back analysis of surrounding rock parameters. So we can obtain the parameters of surrounding rock, and comparing the results of numerical calculation with the actual monitoring value, Verifing the reliability of the results. Using the back analysis parameters to simulation,compare with the result of different construction methods, the presence of seepage effect of construction.(4) Using the GP-DE algorithm prediction the time series of Dalian metro tunnel vault subsidence. The principal component analysis of the training sample optimization method, to achieve the two variable time series distance value and the tunnel vault subsidence monitoring section and tunnel face prediction, prediction and comparison of different sample composition method, the single variable time series and multivariate time series. |
