岩体节理粗糙度系数JRC与其统计参数之间具有复杂的非线性关系,单一统计参数因存在对结构面形貌描述的片面性,从而导致JRC计算结果的可靠性较低。从结构面起伏角、起伏高度及其分布特征的角度选取了平均起伏角avei、平均相对起伏度H/L、起伏角标准偏差iSD和起伏高度标准偏差hSD 4个参数共同反映结构面形貌。以已知JRC试验反算值的102条结构面剖面线作为样本数据对支持向量机进行训练,构建JRC与所选取的统计参数之间的非线性映射关系,建立了JRC支持向量回归(SVR)预测模型,并通过Barton标准剖面线的JRC预测值与试验反算值的对比证明了模型的可靠性。以三峡库区秭归县马家沟滑坡所处地层岩体结构面为例,基于三维激光扫描试验获取了其表面形貌数据并建立了三维形貌模型,开展室内直剪试验反算得到了其JRC。实例JRC的计算结果表明,SVR模型预测结果与试验反算值的相对误差仅为4.5%,不同统计参数回归关系式对于相同剖面线的估算结果存在较大差异,表明基于所选取的统计参数,采用SVR模型预测得到的JRC更加可靠。该方法为JRC的定量确定提供了新思路。 There is a complicated nonlinear relationship between the joint roughness coefficient JRC of rock mass and its statistical parameters. Because of the unilateral nature of the single statistical parameters, the reliability of the JRC results is low. The average undulation angle avei, the average relative undulation H / L, the undulating angle standard deviation iSD and the standard deviation of undulation height hSD are all selected to reflect the topography of structural plane from the perspective of relief, undulation height and distribution of structural plane. Based on the cross-sectional line of 102 structural planes with the known JRC test backscattering value, the SVM was trained to construct the nonlinear mapping relationship between JRC and the selected statistical parameters. The JRC SVR (Support Vector Regression) The model is verified by comparing the JRC predicted values ​​of the Barton standard profile with those of the test. Taking the formation face of the strata in Majiagou landslide of Zigui County in the Three Gorges Reservoir area as an example, the surface topography data and the three-dimensional topography model were obtained based on the three-dimensional laser scanning test. The JRC . The results of the example JRC show that the relative error between the SVR model prediction and the experimental backcalculation is only 4.5%. The regression results of different statistical parameters have great differences for the same section line estimation. Based on the selected statistical parameters, The JRC predicted by the SVR model is more reliable. This method provides a new idea for the quantitative determination of JRC.