DRNN在多传感器泥石流监测系统中的应用  

Application of DRNN in multi-sensor debris flow monitoring system

作　　者：徐根祺 曹宁 谢国坤[1] 马婧 南江萍 张佳绮 XU Genqi;CAO Ning;XIE Guokun;MA Jing;NAN Jiangping;ZHANG Jiaqi(College of Mechanical and Electrical Engineering,Xi’an Traffic Enginering Institute,Xi’an 710030,China;College of Civil Engineering,Xi’an Traffic Enginering Institute,Xi’an 710030,China;China Gezhouba Group No 3 Engineering Co Ltd,Xi’an 710119,China)

机构地区：[1]西安交通工程学院机械与电气工程学院,陕西西安710030 [2]西安交通工程学院土木工程学院,陕西西安710030 [3]中国葛洲坝集团第三工程有限公司,陕西西安710119

出　　处：《传感器与微系统》2022年第11期152-155,160,共5页Transducer and Microsystem Technologies

基　　金：西安交通工程学院中青年基金资助项目(2022KY—48);国家自然科学基金资助项目(51578461);陕西省教育厅科研计划资助项目(2022JK0515)。

摘　　要：针对当前泥石流监测预警系统中预测准确性不高的问题,根据分布式计算和各传感器网络节点之间局部通信的方式,利用分布式递归神经网络(DRNN)预测故障传感器节点的数据,使用DRNN的预测值,对泥石流发生概率进行预测。实验结果表明:当传感器网络的链路质量较低或较多传感器发生故障的情况下,使用DRNN依然能够较准确地预测出泥石流发生的概率,该方法可靠性较高。Aiming at the problem of low prediction accuracy of current debris flow monitoring and early warning system, use distributed recurrent neural network(DRNN)to predict data of faulty sensor nodes, according to distributed computing and local communication between sensor network nodes, and the predicted value of DRNN is used to predict the occurrence probability of debris flow.The experimental results show that when the link quality of the sensor network is low or many sensors fail, the DRNN can still accurately predict the probability of debris flow, and the method has high reliability.

关 键 词：分布式递归神经网络 无线传感器网络 泥石流灾害 监测预警 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]