基于支持向量机的振动加速度峰值预测模型

Forecasting Model of Blasting Vibration Acceleration Peak-value Based on SVM

针对传统爆破振动参数预测方法存在的泛化能力不强等缺陷,提出了基于灰色方法和支持向量机(SVM)相结合的爆破振动加速度峰值预测模型。该模型通过灰色关联度计算确定了影响爆破振动加速度峰值的主要因素,利用结构风险最小化代替传统的经验风险最小化原则,较好地解决了小样本、非线性和局部极小等实际问题,从而提高了预测精度。实际算例表明,SVM模型得到的预测结果与实际值的平均相对误差绝对值不足5%,证明了SVM在爆破振动参数预测中的可行性和有效性。

Aiming at the problem of low generalization capability in forecasting the parameters of blasting vibration,a new forecasting model based on combination of grey method and SVM was proposed.The model calculated and confirmed the main factors affecting the blasting vibration acceleration peak-value by grey method.By replacing the ERM rule with SRM rule,it solved the practical problems such as small sample,non-linear and partial infinitesimal,and improved the forecasting precision.The calculation instances showed that the average absolute relative error between the forecasting values（by SVM model） and real values was less than 5%,which proved that the SVM based forecasting model was feasible and effective.