露天转地下边坡变形υ-SVR模型及参数敏感性分析

SENSITIVITY ANALYSIS OF MODEL PARAMETERS AND υ-SVR MODEL OF SLOPE DEFORMATION DUE TO EXCAVATING FROM OPEN-PIT TO UNDERGROUND MINING

基于改进支持向量回归(υ-SVR)原理,建立反映地下开采扰动影响下露天矿边坡变形非线性规律的υ-SVR模型。将该模型应用于露天转地下开采的杏山铁矿,结果表明变形预测精度高且计算方便,验证改进支持向量回归原理研究二次开采扰动下边坡变形规律的有效性。并分析自由因子C,υ的敏感性,直观反映C,υ对MAE和支持向量数目的影响程度,且确定参数C,υ的合理取值区间,即惩罚参数C应该在[1000,5000]内取值,取值范围较大,但应同时考虑惩罚因子C对泛化性能的影响,参数υ的取值区间为[0.028,0.125]。

Based on the improved support vector regression（v-SVR） principle, the open-pit mine slope deformation v-SVR model is established which represents the nonlinear variation rule due to the disturbance of excavation from open pit into underground mine. v-SVR model is applied to Xingshan iron mine excavating from open pit to underground mining. The results show that the forecast value which is close to true value and calculation is convenient, and the validity of the improved support vector regression principle is verified by studying the open-pit slope nonlinear deformation rule with virtue of the secondary mining perturbations. The sensitivity of the parameters is analyzed, intuitively reflecting the effect of C, v on mean absolute error（MAE） and the number of support vectors, and the reasonable ranges of the parameters C, vare determined, that is, [1 000, 5 000] for parameter C. The interval is larger, but the effect of parameter C on the generalization performance should be considered when selecting parameter C; for parameter v= [0.028, 0.125] is a rather reasonable range.