针对滚石运动特征参数间关系模糊等问题,在开展其敏感性分析的基础上,构建准确、快捷的滚石运动特征参数预测系统.首先引入灰色关联法(grey relation analysis,GRA),以水平运动距离、弹跳高度和碰撞能量等滚石运动特征为目标参数,研究了初始速度、边坡摩擦角、滚石质量、法向恢复系数和切向恢复系数等滚石运动影响因素的敏感性和显著性.随后,使用高斯过程回归模型,建立水平运动距离、弹跳高度和碰撞能量等滚石运动特征参数的预测模型,避免了传统的滚石运动数值模拟软件繁琐的建模计算过程.研究表明,高斯过程回归(Gaussian process regression,GPR)模型对滚石的水平距离和总动能的预测精度最高.因此,文中所构建的滚石运动特征参数敏感性分析及预测模型能够快速得出滚石运动特征参数,对支挡结构的设计和施工有重要的参考价值.
Sensitivity analysis and prediction model of rolling stone motion characteristic parameters based on GRA-GPR
In view of the fuzzy relationship between the characteristic parameters of rock rolling motion,an accurate and fast prediction system for the characteristic parameters of rock rolling motion is constructed on the basis of its sensitivity analysis.In this paper,the grey correlation method(GRA)is introduced firstly,and the sensitivity and significance of the influencing factors of rockfall motion,such as initial velocity,slope friction angle,rockfall mass,normal recovery coefficient and tangential recovery coefficient,are studied by taking the horizontal motion distance,bounce height,collision energy and other rockfall motion characteristics as the target parameters.Then,the Gaussian process regression model is used to establish the prediction model of the rolling stone movement characteristic parameters,such as horizontal movement distance,bounce height,collision energy,which avoids the tedious modeling and calculation process of traditional numerical simulation software for rolling stone movement.The engineering case proves that the Gaussian process regression(GPR)model has the highest accuracy in predicting the horizontal distance and total kinetic energy of rolling stones.Therefore,the sensitivity analysis and prediction model of rolling stone movement characteristic parameters constructed in this paper can quickly obtain the rolling stone movement characteristic parameters,which has important practical value for the design and construction of retaining structures.
slope rollingsensitivity analysisgrey correlationmodel predictioncharacteristic parameters of rolling
NETL
NSTL
万方数据
