基于BP神经网络的岩土体细观力学参数研究

目前颗粒流计算方法中所采用的细观力学参数，均需根据岩土体宏观力学参数试验结果反复调试获取，调试效率很低且有较大的盲目性，因此，急需引入一种新方法，建立岩土体宏观力学参数与细观力学参数的关系。基于PFC3D程序，采用BP(back propagation)神经网络方法，建立宏观力学参数与细观力学参数的非线性模型，通过输入宏观力学参数，即可快速、准确地反演岩土体细观力学参数。研究结果表明：(1)根据BP神经网络模型反演确定的细观力学参数，输入数值模型计算其宏观力学参数，结果与试验值相比，精度一般高于90%；(2)当模型最小尺度上的颗粒数RES = 10、隐含层含6个神经元时，BP神经网络模型的反演性能最佳。实例计算表明，BP神经网络模型对于岩土体细观力学参数的确定具有良好的反演性能，该方法为颗粒流理论的推广应用提供了新的技术手段。

Research on mesomechanical parameters of rock and soil mass based on BP neural network

At present,in particle flow theory,mesomechanical parameters can only be obtained by varying them until the macro mechanical parameters of the numerical sample match that of the laboratory rock-soil mass sample.The adjustment process is inefficient with some blindness;so a new method should be introduced to establish the relationship between macro mechanical parameters and mesomechanical parameters.Based on PFC3D program,a nonlinear network model linked macro mechanical parameters and mesomechanical paramet...