基于BP网络的采矿巷道围岩力学参数反分析

为较理想地处理巷道的收敛变形与围岩力学间的复杂非线性关系,应用反分析方法研究了大冶铁矿龙洞-74 m水平采矿巷道的围岩力学参数.首先应用正交试验理论确定有限元正分析的围岩力学参数取值样本,运用2D-Sigma有限元软件计算巷道的收敛变形值;然后将有限元计算位移值作为输入样本,相应的围岩力学参数作为输出样本训练BP神经网络;再将实测位移带入训练好的神经网络反分析得到相应围岩力学参数,以此围岩力学参数作为二次正分析的计算参数进行有限元计算,并比较分析位移计算值与实测值的误差.结果表明,采用该方法获取围岩力学参数是可行的,其结果符合工程实际要求.

BP Network-Based Back Analysis of Wall-Rock Mechanics Parameters in Mining Tunnel

To ideally treat the complex nonlinear relationship between the convergent deformation of mining tunnel and the wall-rock mechanics,back analysis method was used to study the wall-rock mechanics parameters of Longdong's mining tunnel at -74 m level.The value-taking samples for wall-rock mechanics parameters in the EFM normal analysis were ascertained by orthogonality theorem and the value of the tunnel convergent deformation was calculated by 2D-Sigma finite element software,which was then used as the input sample.The correspondent wall-rock mechanics parameters were used as the output samples for training the BP neural network.The practically measured displacement was used in the back analysis of the trained BP network to obtain the correspondent wall-rock mechanics parameters,which were then used in the secondary normal analysis by EPM.The comparison between the calculated displacement value in the analysis and that practically measured shows that the mechanics parameters obtained by this method are feasible with the results can meet the practical requirements of engineering.