基于极限学习机的隧洞岩体蠕变参数反分析方法

以某水电站坝区左岸导流洞工程为依托,将极限学习机(ELM)应用于隧洞岩体蠕变参数反演计算中。通过正交实验设计,确定导流洞出口段的16组岩体力学参数,选取其中的14组用FLAC3D中的Cvisc模型进行数值分析,根据各组蠕变参数和与其对应的各测点计算位移,对ELM网络进行训练,输入岩体中关键点实际监测的位移变化过程线,反演出相应的岩体蠕变参数确定二者之间的非线性关系,其余两组用于检验训练结果。将该模型应用于某水电站左岸隧洞岩体蠕变参数反演分析中,计算结果与实测位移值拟合较好,说明该模型简单、实用,具有良好的反演精度,可满足工程设计要求。

Back analysis of rock creep parameter based on extreme learning machine

On the left bank of a hydropower station dam project diversion tunnel based on the extreme learning machine (ELM) is applied to the tunnel rock creep parameter inversion calculation. Through the orthogonal experimental design to determine 16 groups of rock mechanics parameters of diversion tunnel exit section, select one of the 14 groups using Cvisc FLAC3D model in numerical analysis, the calculation of creep parameters for each group and the corresponding of the displacement, the training of ELM network, the key point of rock displacement monitoring actual input the process line, the nonlinear relationship between the inversion of rock creep parameters determined two, the remaining two groups to test the training results. The model is applied to a hydropower station on the left bank of tunnel rock creep parameters inversion analysis, calculation results and the measured displacement values fitted well, which shows that the model is simple and practical, has good retrieval accuracy, which can meet the requirement of engineering design.