基于前馈网络的岩体爆破效应预测研究

将神经网络理论知识和爆破专业知识有机地结合在一起，提出了一种新的岩体爆破效应预测的前馈网络理论方法。该方法适合于不同的爆破参数和不同的岩体条件，是一种普遍适用的方法，同时也是一种“面向数据”的方法。通过对三峡工程左岸坝区岩体爆破效应预测的研究表明，本文方法与通常的经验公式法、回归分析法以及ＢＰ网络方法相比，具有较高的预报精度

Forecast Research on Effect of Vibration and Damage in Rock Mass Blasting

With the neural network theory applied to the forecast of blasting effect and the excavation of the bed rock in Three Gorges project , a new feedforward neural network model based on the prior knowledge is advanced in this paper. The precise prediction of both the vibration speed of seismic wave and the damage scope in rock mass blasting can be accomplished. On the basis of the practical data measured from 6 blasting experiments in Three Gorges dam regeion(the number of the samples of vibration speed is 36 and that of the damage scope is 6), the results of the forecast and analyses show that the feedforward neural network can predict the vibration speed and damage scope accurately. The average value of relative error for the forecast vibration is smaller than 8%, the forecast values of damage scope are quite close to the observed ones and its maximum value of relative error does not exceed 15%. In addition, the measured data of different explosions within the same region, can be used for training the neural network proposed, which not only increases the accuracy of prediction with the increasing data, but also makes up for the deficiency of the existed empiric formulas.