基于改进蝙蝠算法优化广义回归神经网络的岩质边坡稳定性预测

在对边坡进行稳定性评价时,传统的数值分析法计算量大,对经验的依赖性强,无法很好地反映边坡动态开放和非线性的特征.针对岩质边坡的上述特点,采用广义Hoek-Brown非线性破坏准则力学参数作为边坡稳定性的影响因素.利用改进后的蝙蝠算法(bat algorithm,BA)搜寻最优解来更新广义回归神经网络(generalized regression neural network,GRNN)的光滑因子,建立改进的BA-GRNN边坡稳定性预测网络.针对蝙蝠算法种群个体缺乏变异机制,在迭代过程中寻优能力下降的问题,引入交叉变异算子改进蝙蝠种群的多样性,使其保持持续优化能力.将改进BA-GRNN网络、BA-GRNN和GRNN3种网络得到预测结果进行对比,发现改进后的BA-GRNN预测网络对于边坡状态和安全系数预测精度更高,在边坡稳定性的预测方面有更好的适用性.

Stability Prediction of Rock Slope Based on Improved BA-GRNN Neural Network

When evaluating the stability of a slope, the traditional numerical analysis method has a large amount of calculation and a strong dependence on experience, and cannot reflect the dynamic open and nonlinear characteristics of the slope well. Aiming at the above characteristics of rock slopes, the mechanical parameters of the generalized Hoek-Brown nonlinear failure criterion was introduced as the influencing factors of slope stability in this paper. The improved bat algorithm (BA) is used to search for the optimal solution to update the smoothing factor of the generalized regression neural network (GRNN), and an improved BA-GRNN slope stability prediction network is established. Otherwise, aiming at the problem that the individual bat algorithm population lacks a mutation mechanism and the ability to search for optimization declines in the iterative process, a cross mutation operator is introduced to improve the diversity of the bat population, so that it can maintain continuous optimization capabilities. Comparing the prediction results of the improved BA-GRNN network, BA-GRNN and GRNN three networks, it is found that the improved BA-GRNN prediction network has higher prediction accuracy for the slope status and safety factor, and has better applicability in the prediction of slope stability.