BP神经网络修正卡尔曼滤波在边坡监测中的应用
发布时间：2019-05-28 15:39
【摘要】：露天煤矿的开采是一种极其危险的工程,因为露天矿中高边坡会随着施工的进行越来越高,内部受力越来越不平衡,这样就造成了高边坡处于极不稳定状态。再经过降雨,暴晒,风化等不利因素,在某一时刻就有可能会发生滑坡。为了对滑坡进行预测,有很多学者对高边坡监测进行了研究,也产生了很多预测模型。但是如果预测模型的输入值误差较大,这样就会使得预测的效果不太好。为了解决此类问题,需要对数据进行滤波操作。由于卡尔曼滤波对于统计特征有着不稳定性,可能会导致离散现象。为了解决这个问题本文提出了使用BP神经网络修正卡尔曼滤波的改进算法BPKF对数据进行滤波处理。将训练好的BP神经网络运用到卡尔曼滤波中对数据进行平滑处理,最后进行预测。针对山西某矿边坡监测项目的特殊地理环境,本文进行了针对该矿高边坡和其余相似环境的施工组织方案的设计。并且对系统的运行、监测点的埋设,数据的报送等相关内容做了介绍。最后通过测试数据,运用均方根对BPKF算法和标准卡尔曼滤波进行了对比评估。结果显示BPKF算法的滤波结果更平滑,更有利于预测。同时通过结合设计好的监测方案、BPKF和预测模型,在山西某矿监测中一共成功预测了5次滑坡,其中较大滑坡1次,小范围滑坡4次。
[Abstract]:The mining of open pit coal mine is an extremely dangerous project, because the middle and high slope of open pit mine will become higher and higher with the construction, and the internal force will become more and more unbalanced, which results in the high slope in a very unstable state. After rainfall, sun exposure, weathering and other adverse factors, there may be landslides at some point. In order to predict landslide, many scholars have studied the monitoring of high slope and produced a lot of prediction models. However, if the input error of the prediction model is large, the prediction effect will not be very good. In order to solve this kind of problem, it is necessary to filter the data. Because Kalman filter is unstable to statistical characteristics, it may lead to discrete phenomenon. In order to solve this problem, an improved algorithm BPKF, which uses BP neural network to modify Kalman filter, is proposed to filter the data. The trained BP neural network is applied to Kalman filter to smooth the data, and finally, the prediction is carried out. In view of the special geographical environment of a mine slope monitoring project in Shanxi Province, this paper designs the construction organization scheme for the high slope and other similar environments of the mine. And the operation of the system, the embedding of monitoring points, the submission of data and other related contents are introduced. Finally, the BPKF algorithm and the standard Kalman filter are compared and evaluated by root mean square (root mean square) through the test data. The results show that the filtering results of BPKF algorithm are smoother and more conducive to prediction. At the same time, through the combination of the designed monitoring scheme, BPKF and prediction model, a total of five landslides have been successfully predicted in a mine monitoring in Shanxi Province, including 1 large landslide and 4 small scale landslides.
【学位授予单位】：郑州大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TD824.7;TP183
本文编号：2487170
[Abstract]:The mining of open pit coal mine is an extremely dangerous project, because the middle and high slope of open pit mine will become higher and higher with the construction, and the internal force will become more and more unbalanced, which results in the high slope in a very unstable state. After rainfall, sun exposure, weathering and other adverse factors, there may be landslides at some point. In order to predict landslide, many scholars have studied the monitoring of high slope and produced a lot of prediction models. However, if the input error of the prediction model is large, the prediction effect will not be very good. In order to solve this kind of problem, it is necessary to filter the data. Because Kalman filter is unstable to statistical characteristics, it may lead to discrete phenomenon. In order to solve this problem, an improved algorithm BPKF, which uses BP neural network to modify Kalman filter, is proposed to filter the data. The trained BP neural network is applied to Kalman filter to smooth the data, and finally, the prediction is carried out. In view of the special geographical environment of a mine slope monitoring project in Shanxi Province, this paper designs the construction organization scheme for the high slope and other similar environments of the mine. And the operation of the system, the embedding of monitoring points, the submission of data and other related contents are introduced. Finally, the BPKF algorithm and the standard Kalman filter are compared and evaluated by root mean square (root mean square) through the test data. The results show that the filtering results of BPKF algorithm are smoother and more conducive to prediction. At the same time, through the combination of the designed monitoring scheme, BPKF and prediction model, a total of five landslides have been successfully predicted in a mine monitoring in Shanxi Province, including 1 large landslide and 4 small scale landslides.
【学位授予单位】：郑州大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TD824.7;TP183
【参考文献】
相关期刊论文 前4条
1 胡丛玮,刘大杰;基于方差分量估计原理的自适应卡尔曼滤波及其应用[J];测绘学院学报;2002年01期
2 杨茂兴;小样本容量测量数据中粗差的剔除[J];计量与测试技术;2005年01期
3 赵琳;王小旭;孙明;丁继成;闫超;;基于极大后验估计和指数加权的自适应UKF滤波算法[J];自动化学报;2010年07期
4 胡大贺;吴侃;陈冉丽;;三维激光扫描用于开采沉陷监测研究[J];煤矿开采;2013年01期
相关博士学位论文 前1条
1 刘超;矿区GPS变形监测及其伪卫星增强技术[D];中国矿业大学;2011年
，本文编号：2487170
本文链接：https://www.wllwen.com/kejilunwen/kuangye/2487170.html
