泥石流次声特征分析及流量智能反演方法  

Analysis of infrasound characteristics and intelligent inversion method for debris flow discharge

作　　者：胡耕源 刘敦龙 桑学佳 张少杰[3] 陈乔[4] HU Gengyuan;LIU Dunlong;SANG Xuejia;ZHANG Shaojie;CHEN Qiao(School of Software Engineering,Chengdu University of Information Technoloy,Chengdu 610225,China;Sichuan Provincial Information Application Support Software Engineering Technology Research Center,Chengdu 636499,China;Institute of Mountain Hazards and Environment,Chinese Academy of Sciences and Ministry of Water Resources,Chengdu 610044,China;Chongqing Institute of Green and Intelligent Technology,Chinese Academy of Sciences,Chongqing 400722,China)

机构地区：[1]成都信息工程大学软件工程学院,四川成都610225 [2]四川省信息化应用支撑软件工程技术研究中心,四川成都636499 [3]中国科学院、水利部成都山地灾害与环境研究所,四川成都610044 [4]中国科学院重庆绿色智能技术研究院,重庆400722

出　　处：《自然灾害学报》2024年第2期54-64,共11页Journal of Natural Disasters

基　　金：国家自然科学基金青年项目(42001100);四川省自然科学基金项目(2023NSFSC0751);四川省信息化应用支撑软件工程技术研究中心开放课题(760115027)。

摘　　要：次声技术已被广泛应用于泥石流监测预警中,泥石流次声的某些特征可从一定程度上反映泥石流的流量。泥石流流量是评估泥石流规模的重要参数,准确预测泥石流流量对泥石流监测预警具有重要意义。围绕影响泥石流次声特性的关键物理参量,通过定量化配比水槽实验模拟泥石流产生次声的物理过程,采集次声信号并测算流量。通过分析泥石流流量与次声之间的关联,揭示泥石流流量对次声特性的影响规律。经特征提取和特征选择提炼出可表征泥石流流量的次声特征因子并构建特征向量集。通过对比分析k近邻算法(k-nearest neighbor,KNN)、神经网络、随机森林和梯度提升决策树(gradient boosting decision tree,GBDT)算法在流量预测方面的表现性能,构建了具有较高预测准确率的泥石流流量智能反演模型。通过该智能反演模型可实现泥石流流量的有效预测,从而为泥石流次声监测提供更丰富的报警信息。Infrasound technology has been widely applied in debris flow monitoring and early warning.Certain characteristics of debris flow infrasound can reflect the debris flow discharge to a certain extent.Debris flow discharge is an important parameter for evaluating the scale of debris flow,and the accurate prediction of debris flow discharge is of great significance for debris flow monitoring and early warning.Based on the key physical parameters that affect the infrasound characteristics of debris flow,a quantitative ratio water tank experiment is conducted to simulate the physical process of infrasound generated by debris flow.Infrasound signals are collected and the debris flow discharge is measured.By analyzing the correlation between the debris flow discharge and the infrasound of debris flow,the influence pattern of flow discharge on the infrasound characteristics is revealed.After feature extraction and selection,infrasound characteristics that can characterize the flow discharge of debris flow are refined and a feature vector set is constructed.By comparing the performance of KNN,neural networks,random forests,and GBDT algorithms in flow discharge prediction,an intelligent inversion model for debris flow discharge with high prediction accuracy is constructed.This intelligent inversion model can achieve effective prediction of debris flow discharge,providing richer alarm information for debris flow infrasound monitoring.

关 键 词：泥石流 次声 特征分析 流量 智能反演 

分 类 号：U418.56[交通运输工程—道路与铁道工程] X43[环境科学与工程—灾害防治]