基于BP神经网络的Mathews图表法分区调整

东北大学 资源与土木工程学院,辽宁 沈阳 110819

稳定图表法; 稳定性评价; BP神经网络

Adjustment of Mathews' stability graph method based on BP neural network
WANG Yun-sen,Zheng Gui-ping,CAO Wei-dong,LI Yun-hui

(College of Resources and Civil Engineering,Northeastern University,Shenyang 110819,China)

stability graph method; stability assessment; BP neural network

DOI: 10.13800/j.cnki.xakjdxxb.2016.0120

备注

为解决经过Mawdesley修正的Mathews图表法的分区划分限于回归分析,过于简化,在水力半径过小及过大2个区域没有给出明确的稳定性判断等问题,通过建立BP神经网络岩体稳定性预测模型,利用Mawdesley修正的图表法历史样本数据进行训练,在满足精度要求的约束下对Mathews图表中节点的稳定性状态进行预测,绘制稳定性等值线图,在此基础上进行分区的优化调整,对于低稳定性系数小水力半径和高稳定性系数大水力半径的2个区域,给出了稳定性确切的分区定义,调整后的分区在中间区域和原图基本重合,能很好的符合历史数据样本,证明了应用神经网络对Mathews图表法进行分区的合理性。在焦家金矿进行了工业试验,利用调整后的分区及临界跨度设计法进行结构参数优化,进路跨度提高到7.9 m,试验结果采场稳定性良好。

The partition of Mathews' stability graph modified using regression analysis by Mawdesley is too simplified,and do not give a clear judgment of stability in the two areas in small or large hydraulic radius.BP neural network prediction model of rock mass stability was established,using the original Mathews chart's historical data for training samples,and predicting the stability of the nodes in Mathews' stability graph under the accuracy constraint.By drawing the contour map of stability,the partition of the graph was optimized that in the low stability coefficient of small hydraulic radius and high stability coefficient of large hydraulic radius of two regions,the precise definition of stability zoning was given.The adjusted partitions basically coincide in the middle region with the original historical data,which proves the rationality of the application of neural network to partition of the Mathews chart.In Jiaojia Gold Mine the structure parameter optimization was carried out by using modified Mathews stability graph method.The research result can ensure stability field at less than 7.9 m of stoping span with a good industrial test effect.