泥石流次声预警机制及降低误报率的算法研究

Infrasound early-warning for debris flows can be 10 minutes to 40 minutes in advance. This advance is very precious for disaster prevention and mitigation. This device is simple to use and easy to maintain. But in the practical application, it has a serious problem of false alarm. Aiming at the two main causes of false alarm: there are different warning critical values for different types of debris flows and there is so much environment infrasound interference, we focus on the research of quantitative infrasound characteristics and signal feature extraction and classification algorithms to reduce the false-alarm rate of debris flows early-warning. By using theoretical research, numerical simulation, we research on the characteristics of time domain, frequency domain and propagation characteristics. Based on this, we will obtain the characteristic parameters and their threshold values by which we can discriminate the infrasound of debris flows from other kind of infrasound effectively and determine the type and scale of debris flows. On this basis, we also research on the most appropriate signal feature extraction and intelligent recognition algorithms which reduce the false-alarm rate. The indoor experiment and on-site observation are used to verify our research on which a practical and high efficient method and procedure of debris flow early-warning are established. These research findings make a new step of debris flow warning technology, protect the safety of people's life and property, and can also be used in other geological disasters, such as landslide, avalanche after some variations.

泥石流次声预警可以实现10到40分钟的提前量，为防灾减灾赢得了宝贵的时间，而且使用简单、维护方便。但是在实际应用中发现，该方法存在较严重的误报问题。针对产生误报的两个主要原因：不同类型泥石流的预警临界值不同以及环境次声信号的干扰问题，本项目将对泥石流的次声特性进行定量研究：采用理论研究和数值模拟的方法研究泥石流次声信号的时域、频域和传播特性，在此基础上建立各类泥石流区别于其他类型次声信号的特征参数和临界值；根据采用的特征参数研究相应的信号特征提取算法和智能分类聚类算法，以有效地区分和判别泥石流次声信号以及泥石流的类型和规模，达到降低误报率的目的；通过室内模拟和现场实验验证等手段，对上述方法进行验证和完善，形成一套实用高效的泥石流次声预警技术方案，使泥石流预警技术上一个新的台阶，保护人民生命财产安全。本项目的研究成果对于滑坡、雪崩等地质灾害的监测预警具有借鉴意义。

本课题解决泥石流次声预警的误报问题。围绕这个研究目标，我们针对产生误报的两个主要原因：不同类型泥石流的预警临界值不同以及环境次声信号的干扰问题，对泥石流的次声特性进行定量研究：①采用理论研究和数值模拟的方法研究泥石流次声信号的时域、频域和传播特性；②各类泥石流的特征参数和临界值；③次声信号的特征提取算法和基于机器学习的分类聚类算法；在此基础上④完成了实验用泥石流次声预警仪器的设计和制作（包括第一版和改进版）以及⑤泥石流声波信号处理分析软件的开发。通过泥石流模型实验和野外采集的数据对上述方法进行了验证和完善，形成一套实用高效的泥石流次声预警技术方案。.本项目达到并超过了预期目标，使泥石流预警技术上了一个新的台阶。本项目的研究成果对于滑坡、雪崩等地质灾害的监测预警具有借鉴意义。同时，在项目执行过程中，培养了多名动手能力强、综合素质高的研究生。