中深孔爆破振动加速度峰值的遗传BP网络预测

Prediction on peak vibration acceleration value of mediumdeep-hole blasting using genetic BP network

为准确预测爆破振动加速度峰值,保证爆破安全,相对于考虑因素少的经验公式法以及存在收敛性差、易陷入局部极小和计算复杂等缺陷的BP算法,提出了遗传BP神经网络算法,该算法具有更高的预测精度。以田湾核电站船山二期工程的试验数据为背景,比较分析并选择最大段药量、水平距离、总药量、高程差、爆破台阶高度和段别规模等6个参数作为输入层因子,建立了相应的爆破振动加速度峰值预测模型。结果表明,预测精度达到96.97%,验证了方法的可行性和有效性。

To accurately predict the peak vibration acceleration value of blasting and ensure the blasting security,genetic BP algorithm was proposed.This algorithm was superior to the empirical formula method which considered fewer factors,and the BP algorithm with slow convergeve,local minimum and high computation complexity.Based on the test data in Tianwan Nuclear Power Station,a relevant forecasting model with the maximal charge,horizontal distance,total charge,height difference,blasting bench height and scale as its input elements,was established to predict the peak vibration acceleration value.The prediction accuracy achieves 96.97%,which verifies its feasibility and validity.