基于遗传BP神经网络的煤矿爆破振动特征参量预测  

Prediction on characteristic parameters of blasting vibration based genetic BP neural network in coal mine

作　　者：张士科[1,2] 方宏远[2] 耿勇强 ZHANG Shike;FANG Hongyuan;GENG Yongqiang(School of Civil Engineering and Architecture,Anyang Normal University,Anyang 455000,China;School of Water Conservancy and Environment,Zhengzhou University,Zhengzhou 450001,China)

机构地区：[1]安阳师范学院建筑工程学院,河南安阳455000 [2]郑州大学水利与环境学院,河南郑州450001

出　　处：《煤炭科学技术》2018年第9期133-139,共7页Coal Science and Technology

基　　金：国家自然科学基金资助项目(51678536;41404096);中国博士后科学基金资助项目(2016M602259);河南省科技攻关计划资助项目(182102310804);河南省自然科学基金资助项目(182300410160);安阳师范学院大学生创新基金资助项目(ASCX/2017-Z141)

摘　　要：为了解决矿区爆破振动产生的危害大、影响因素多、特征参量监测结果离散和计算非线性的问题,通过建立基于遗传算法优化BP神经网络预测模型来拟合煤矿爆破振动参数与特征参量之间的非线性关系,并采用该模型对煤矿爆破振动特征参量进行了准确的预测。研究结果表明:在实际工程中,GA-BP神经网络模型对确定不容易测量的爆破振动特征参量能有效预测,同时又能节约大量人力和财力;GA-BP神经网络预测模型较经验公式、BP神经网络预测模型有更强的解决复杂非线性问题能力,其预测值与实际值的相对误差在10%以内,不易陷入局部极小值,稳定性更好,平均预测精度较高;从使用的样本数量可知该模型可为小样本、多因素影响参数预测问题提供一种切实有效的方法,且工作量小而灵活,适用性广泛。In order to solve a high danger,many influence factors,discrete measured results of the characteristic parameters and calculation nonlinear problems occurred by the blasting vibrations in the mining area,the establishment on the optimized BP neural network prediction model based on the genetic algorithm（ GA） was applied to fit the nonlinear relationship between the mine blasting vibration parameters and characteristic parameters. The model was also applied to the accurate prediction on the characteristic parameters of the mine blasting vibration. The study results showed that in the actual engineering,a GA-BP neural network model could have an important applied value to determine the blasting vibration characteristic parameters and also could save great labor and financial resources. In comparison with the experience formula and BP neural network prediction model,the GA-BP neural network prediction model could have the more strong capacity to solve the complicated nonlinear problem. The predicted value would have a relative error within 10% to the actual value and would not be a minimum value locally. The stability would be better and the average predicted accuracy would be high. From the applied sample number,the model would provide a practical and effective method to the small sample and multi factor parameter prediction and the work load would be small,flexible and wide suitable.

关 键 词：遗传算法 BP神经网络 爆破振动 特征参量预测 煤矿爆破参数 

分 类 号：TD235.4[矿业工程—矿井建设]