基于BP神经网络的多台阶地形爆破振速预测对比分析

Comparative Analysis of Blasting Vibration Velocity Prediction Based on BP Neural Network in Multi-level Terrain

应用萨道夫斯基公式、考虑高程的萨道夫斯基修正公式及BP网络模型对多台阶地形的爆破振速进行预测。得出以下结论:在测点正对爆区的情况下,考虑高程情况下的萨道夫斯基公式的准确度要高于未修正的萨道夫斯基公式;在爆区的侧向,考虑高程情况下的萨道夫斯基修正公式和萨道夫斯基公式的准确度均较低;BP神经网络能很好地对上述两种情况下的爆破振速进行预测,且准确度较高。

In order to predict the blasting vibration velocity in the multi-level terrain, Sadaovsk formula and BP neural network model were adopted, as well as Sadaovsk correction formula considering the elevation. The results were concluded as follows. Firstly, when the measuring points were on the direction of the blasting area, the accuracy of the Sadaovsk correction formula considering the elevation was higher than that of the original Sadaovsk formula. Secondly, the Sadaovsk correction formula and the original Sadaovsk formula both had a lower accuracy in the sides of the blasting area. Thirdly, the BP neural network was flexible and accurate to predict blasting vibration velocity in above cases.