论文部分内容阅读
为获得可靠的开采沉陷数值模拟岩体力学参数,将覆岩按照不同的移动破坏特征分为冒落带、裂隙带、弯曲带和松散层带4部分。采用大量数值模拟结果进行BP神经网络训练,建立模拟地表移动与覆岩力学参数间的关系。以淮北某矿为例,根据正交设计思想进行了50次FLAC3D数值模拟实验。采用实验结果对BP神经网络模型进行训练,获得以地表下沉值为输入、以各层力学参数为输出的BP神经网络模型。采用实测地表下沉值进行力学参数反分析。以反演得到的力学参数进行数值计算,地表下沉数值模拟结果与实测结果吻合,表明所获得的力学参数正确可靠。采用BP神经网络和正交实验方法进行力学参数的正演反分析,可以用较少的实验次数建立地表移动变形量和岩层力学参数之间的非线性映射,获得与实测结果吻合的数值模拟结果,为开采沉陷数值分析中力学参数的选取提供了依据。
In order to obtain reliable mining subsidence numerical simulation of mechanical parameters of rock mass, the overlying strata are divided into 4 parts of caving zone, fracturing zone, bending zone and loose belt according to different moving failure characteristics. A large number of numerical simulation results were used to train BP neural network to establish the relationship between simulated ground surface movement and overburden mechanical parameters. Taking a mine in Huaibei as an example, FLAC3D numerical simulation experiment was carried out 50 times according to the orthogonal design idea. The BP neural network model is trained by the experimental results, and the BP neural network model which takes the surface subsidence value as input and the mechanical parameters of each layer as output is obtained. The measured surface subsidence values were used to conduct back analysis of mechanical parameters. The mechanical parameters obtained by inversion are used for numerical calculation. The numerical simulation results of surface subsidence agree well with the measured ones, which shows that the obtained mechanical parameters are correct and reliable. BP neural network and orthogonal experimental method for forward analysis of mechanical parameters, the number of experiments can be used to establish non-linear mapping between the amount of surface deformation and rock mechanics parameters to obtain the numerical simulation results in line with the measured results , Which provides the basis for the selection of mechanical parameters in the numerical analysis of mining subsidence.