基于机器学习模型的泥石流地声识别方法

Debris flow geoacoustic identification method based on machine learning model

摘要

地声是泥石流运动过程中产生的震动信号，基于地声识别泥石流信号是检测泥石流事件最重要的方法之一。传统泥石流地声识别方法仅依靠地声时频域幅度阈值进行判定，未全面考虑泥石流和干扰地声特征，易导致泥石流地声识别高虚警率。以蒋家沟泥石流为研究对象，获取真实泥石流地声数据，并分析其与典型干扰数据的时频特征差异，构建泥石流地声时频多维特征向量数据集，结合支持向量机和随机森林模型构建泥石流地声信号识别模型。研究结果表明，基于支持向量机地声识别模型的准确率为92.4%，命中率为100%，误报率为12.2%。随机森林模型准确率为93.7%，命中率为100%，误报率为10.8%。相比阈值模型，该方法准确率至少提高21.1%，误报率降低幅度不低于29%。研究结果说明基于地声时频多维特征向量的信号识别模型可以高效地识别蒋家沟区域泥石流地声信号，可为不同地区泥石流事件检测提供参考。

Abstract

Ground sound is a vibration signal generated during the movement of debris flows, and identification based on geoacoustic signals is one of the most important methods for detecting debris flow events. Traditional ground sound identification methods for debris flows only rely on the amplitude threshold in the time-frequency domain for determination, failing to comprehensively consider the characteristics of debris flow and interfering ground sounds, easily leading to a high false alarm rate in geoacoustic recognition. Taking the debris flows in Jiangjia Gully as the research object, this paper obtained real debris flow geoacoustic data, analyzed the differences in time-frequency characteristics between it and typical interference data, constructed a time-frequency multi-dimensional feature vector dataset for debris flow ground sound, and established a geosound signal identification model for debris flows by combining the support vector machine (SVM) and random forest (RF) models. The results show that the accuracy of the geoacoustic identification model based on SVM is 92.4%, the hit rate is 100%, and the false alarm rate is 12.2%. The RF model has an accuracy of 93.7%, a hit rate of 100%, and a false alarm rate of 10.8%. Compared with the threshold model, this method improves the accuracy by at least 21.1% and reduces the false alarm rate by at least 29%. The findings indicate that the signal identification model based on time-frequency multi-dimensional feature vectors of ground sound can efficiently identify the geoacoustic signals of debris flows in the Jiangjia Gully area, providing a reference for debris flow event detection in different regions.

关键词

泥石流 / 地声识别 / 随机森林 / 支持向量机

Key words

debris flow / geoacoustic recognition / random forest / support vector machine

参考文献

[1] 韦方强. 城镇泥石流减灾决策支持系统[D]. 成都: 西南交通大学, 2002.

[2] WEI Fangqiang. Decision supporting system of debris flow mitigation for mountain towns[D]. Chengdu: Southwest Jiaotong University, 2002. (in Chinese)

[3] 钟敦伦, 张金山, 谢洪, 等 . 泥石流警报技术探索[J]. 山地学报, 2011, 29(2): 234-242.

[4] ZHONG Dunlun, ZHANG Jinshan, XIE Hong, et al. Techniques of debris flow alarm[J]. Journal of Mountain Science, 2011, 29(2): 234-242. (in Chinese)

[5] 熊江, 唐川, 陈明. 泥石流早期识别与监测预警研究进展探讨[J]. 自然灾害学报, 2021, 30(1): 165-173.

[6] XIONG Jiang, TANG Chuan, CHEN Ming. Discussion on the research progress of early identification, monitoring and early warning of debris flow[J]. Journal of Natural Disasters, 2021, 30(1): 165-173. (in Chinese)

[7] 卜峰, 许强, 陈琴, 等 . 地声波与次声波在泥石流监测预警中的应用[J]. 科技与创新, 2023(15): 179-181.

[8] BU Feng, XU Qiang, CHEN Qin, et al. Application of geosonic waves and infrasonic waves in monitoring and early warning of debris flows[J]. Science and Technology & Innovation, 2023(15): 179-181. (in Chinese)

[9] DONG H C, FANG L D, TAO Z G, et al. Geoacoustic monitoring technology of debris flow based on piezoelectric sensor[J]. Arabian Journal of Geosciences, 2022, 16(1): 2.

[10] 陈精日, 刘立秋. NJ-2A泥石流地声报警器研制与应用[J]. 山地学报, 2001, 19(5): 452-455.

[11] CHEN Jingri, LIU Liqiu. Research and application on NJ-2A ground sound warning device for debris flow[J]. Journal of Mountain Science, 2001, 19(5): 452-455. (in Chinese)

[12] 周铭. 不同形态泥石流地声与次声特性比较研究[D]. 南宁: 广西大学, 2014.

[13] ZHOU Ming. Research on the characteristic of ground vibrations and infrasound produced by different types of debris flows[D]. Nanning: Guangxi University, 2014. (in Chinese)

[14] 吴积善, 康志成, 田连权, 等 . 云南蒋家沟泥石流观测研究[M]. 北京: 科学出版社, 1990.

[15] WU Jishan, KANG Zhicheng, TIAN Lianquan, et al. Observation and research on debris flows in Jiangjiagou, Yunnan[M]. Beijing: Science Press, 1990. (in Chinese)

[16] 陈精日, 刘立秋, 叶明富. 泥石流地声参数传播特征值的测试与分析[J]. 山地学报, 1999, 17(4): 349-352.

[17] CHEN Jingri, LIU Liqiu, YE Mingfu. Test and analysis for debris flow ground sound interfex value[J]. Mountain Research, 1999, 17(4): 349-352. (in Chinese)

[18] 黃清哲, 朱崇锐, 田宗谟, 等 . 土石流光纤感测系统之整合与应用[J]. 中华水土保持学报, 2013, 44(3): 191-201.

[19] HUANG Qingzhe, ZHU Chongrui, TIAN Zongmo, et al. Integration and application of a fiber-optic sensing system for monitoring debris flows[J]. Journal of Chinese Soil and Water Conservation, 2013, 44(3): 191-201. (in Chinese).

[20] 陈志波, 李鼎兴, 陈澄, 等 . 基于机器学习和InSAR技术的滑坡易发性动态评价[J]. 自然灾害学报, 2025, 34(2): 56-65.

[21] CHEN Zhibo, LI Dingxing, CHEN Cheng, et al. Dynamic evaluation of landslide susceptibility based on machine learning and InSAR technology[J]. Journal of Natural Disasters, 2025, 34(2): 56-65. (in Chinese)

[22] 翟文华, 王小东, 吴明堂, 等 . 基于频率比模型和随机森林模型耦合的地质灾害易发性评价[J]. 自然灾害学报, 2023, 32(6): 74-82.

[23] ZHAI Wenhua, WANG Xiaodong, WU Mingtang, et al. Geological disaster susceptibility evaluation based on coupling of frequency ratio model and random forest model[J]. Journal of Natural Disasters, 2023, 32(6): 74-82. (in Chinese)

[24] 胡耕源, 刘敦龙, 桑学佳, 等 . 泥石流次声特征分析及流量智能反演方法[J]. 自然灾害学报, 2024, 33(2): 54-64.

[25] HU Gengyuan, LIU Dunlong, SANG Xuejia, et al. Analysis of infrasound characteristics and intelligent inversion method for debris flow discharge[J]. Journal of Natural Disasters, 2024, 33(2): 54-64. (in Chinese)

[26] 刘艳辉, 黄俊宝, 肖锐铧, 等 . 基于随机森林的福建省区域滑坡灾害预警模型研究[J]. 工程地质学报, 2022, 30(3): 944-955.

[27] LIU Yanhui, HUANG Junbao, XIAO Ruihua, et al. Study on early warning model for regional landslides based on random forest in Fujian Province[J]. Journal of Engineering Geology, 2022, 30(3): 944-955. (in Chinese)

[28] LEI Y G, HE Z J, ZI Y Y, et al. Fault diagnosis of rotating machinery based on multiple ANFIS combination with GAs[J]. Mechanical Systems and Signal Processing, 2007, 21(5): 2280-2294.

[29] 方迎潮, 王道杰, 何松膛, 等 . 云南东川蒋家沟泥石流2003-2014年冲淤演变特征[J]. 山地学报, 2018, 36(6): 907-916.

[30] FANG Yingchao, WANG Daojie, HE Songtang, et al. Characteristics of debris flow erosion and deposition at Jiangjia Gully, Dongchuan, Yunnan Province, China for 2003-2014[J]. Mountain Research, 2018, 36(6): 907-916. (in Chinese)

[31] 王源程, 苏鹏程, 陈聪, 等 . 地基GPS反演泥石流流域局地降水水汽特征: 以云南东川蒋家沟为例[J]. 山地学报, 2016, 34(6): 749-756.

[32] WANG Yuancheng, SU Pengcheng, CHEN Cong, et al. Precipitable water vapor in the debris-flow watershed inversed by ground-based GPS: A case study of the Jiangjia Valley in Yunnan Province[J]. Mountain Research, 2016, 34(6): 749-756. (in Chinese)