摘要
正确确定基本参数是提高有关岩体移动及变形预估精度的关键。通过对ＢＰ神经网络算法的改进，引进动量项系数及采用变步长方法，提高了收敛速度和稳定性。建立了地表移动基本参数的反分析方法，使用该方法及相应程序对大量观测结果进行反分析，结果表明效果良好。
Correct choice of parameters is critical to improve the precision of prediction for rockmass movement and deformation. In order to improve computing rate and stability of BP neural networks, the coefficient of dynamic-item and varied steps are adopted for back analysis of rockmass displacement. The comparison between calculated and measured displacements is made and the result is satisfactory. The proposed methods are useful for design of underground excavation or mining.
出处
《岩石力学与工程学报》
EI
CAS
CSCD
北大核心
2001年第S1期1762-1765,共4页
Chinese Journal of Rock Mechanics and Engineering
关键词
岩体移动
BP神经网络
反分析
rockmass movement
HP neural networks
back analysis
