基于BP神经网络的Mathews图表法分区调整

发布时间：2018-03-17 08:08

本文选题：稳定图表法　切入点：稳定性评价　出处：《西安科技大学学报》2016年01期 　论文类型：期刊论文

【摘要】：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。
[Abstract]:In order to solve the problem that the partition of Mathews chart method modified by Mawdesley is limited to regression analysis, it is too simplified, and there is no definite stability judgment in the two areas of too small hydraulic radius and too large hydraulic radius, etc. Through the establishment of BP neural network rock mass stability prediction model and the training of historical sample data of Mawdesley modified chart method, the stability state of nodes in Mathews chart can be predicted under the constraint of satisfying the precision requirement. On the basis of drawing the contour map of stability, the optimal adjustment of the zone is carried out. For the two regions with small hydraulic radius with low stability coefficient and large hydraulic radius with high stability coefficient, the precise zone definition of stability is given. The adjusted partition basically coincides with the original map in the middle area, which can fit the historical data sample very well, and proves the rationality of using neural network to partition the Mathews chart method. The industrial test was carried out in Jiaojia Gold Mine. The structural parameters are optimized by using the adjusted zoning and critical span design method. The approach span is increased to 7.9 m, and the stope stability is good.
【作者单位】：东北大学资源与土木工程学院;
【分类号】：TP183;TD76

【相似文献】