基于BP神经网络方法研究岩石节理形貌粗糙度系数

Roughness Coefficient of Rock Joints Based on BP Neural Network

岩石节理粗糙度系数(JRC)是研究岩石力学性质的重要参数。为了更准确地描述这一参数,本文基于人工神经网络的原理,提出一种研究JRC的新方法——BP神经网络预测法。选取节理表面最大峰高S_p、表面最大高度S_z、表面最大谷深S_v、峰度系数S_(ku)、偏斜度系数S_(sk)、均方根高度S_q、算术平均高度S_a7个表面形貌高度特征参数作为网络输入,剖面线分维值和JRC作为网络输出,以此为基础构建网络模型,并对10组实测数据进行了预测验证。结果表明:该方法误差很小,具有很高的预测精度,可为进一步的研究提供新的思路和方法。

Rock joint roughness coefficient（JRC） is an important parameter to study the mechanical properties of rock.In order to more accurately describe this parameter,a new method of rock joint roughness coefficient based on artificial neural network（BP） is proposed.Seven surface morphology feature height parameters,including maximum peak high S_p,surface and the maximum height S_z,maximum surface valley deep S_v,kurtosis coefficient S_（ku）,partial slope coefficient S_（sk）,square highly S_q,arithmetic mean height S_a are selected as network input,fractal dimension of joint section line values and JRC as the output of the network.Based on these,network model is constructed and 10 sets of measured data are predicted and verified.The results show that the method has the very small error and high accuracy,which can provide new ideas and methods for further research.