论文部分内容阅读
通过RBF、BP神经网络及SVM算法3种预测方法,对爆破震动特征参量进行预测,并与传统萨道夫斯基公式进行对比分析研究。结果表明,3种方法预测精度均优于传统萨道夫斯基公式。当样本数有限时,BP、RBF神经网络在爆破振动峰值振动速度及主频率的预测中效果欠佳,SVM算法的预测精度优于RBF、BP神经网络,在实际工程应用中SVM算法对爆破振动特征参量的预测具有极强的适应性。
Through the three prediction methods of RBF, BP neural network and SVM algorithm, the blasting vibration characteristic parameters are predicted and compared with the traditional Sadowssky formula. The results show that the prediction accuracy of the three methods is better than the traditional Sadowssky formula. When the number of samples is limited, the prediction results of BP neural network and RBF neural network are not good enough in the prediction of peak vibration velocity and main frequency of blasting vibration. The prediction accuracy of SVM algorithm is better than that of RBF and BP neural network. In practical engineering application, The prediction of characteristic parameters is highly adaptable.