Characteristics And Type Identification Of Debris Flow Acoustic Signal

This research relies on the National Natural Science Foundation of China(41572347)"Research on Infrasound Warning Mechanism of Debris Flows and Algorithms to Reduce False Alarm Rate" project.Based on indoor debris flow acoustic simulation experiments and field observations of synchronous observation data,effective waveform feature extraction,first-arrival identification,and machine learning classification algorithms for multiple sets of different types of debris flow infrasound signals are studied.This thesis mainly conducted research and exploration in the following aspects:(1)In view of the non-stationary and non-stationary characteristics of debris flow acoustic wave signals,a generalized S transform is proposed by introducing window function parameters.The transform can flexibly adjust the width of the Gaussian window function according to the signal difference,and effectively improve the resolution of the time-frequency distribution.Finally,MATLAB was used to compile a set of sonic signal processing software,which realized the rapid standardization of sonic signals,which laid the foundation for the subsequent processing of massive debris flow sonic signals.(2)The original acoustic wave signal is pre-processed by the integrated empirical mode decomposition method(EEMD),the dominant component of the signal is extracted and reconstructed using the threshold rule,and the reconstructed signal is analyzed using a dual-scale box-dimensional fractal algorithm to obtain The box dimensions of the original infrasound signal of typical dilute,transitional,and viscous debris flows are obviously different,that is,as the bulk density increases,the box dimension values show a downward trend,which can distinguish the types of debris flows.(3)The principal component IMF box dimension value decomposed by EEMD is used as the eigenvalue input to the least square support vector machine(LS-SVM)classifier for training and classification.Through the model training test,the correct recognition rate of debris flow type reached 87%,of which the rare and transitional debris flow reached 80%,and the recognition rate of viscous debris flow reached 100%.(4)The wavelet packet transform method is used to extract the energy distribution characteristics of the frequency band of the acoustic wave signal,so as to comprehensively identify different types of debris flows based on the frequency interval of the acoustic wave signal and the energy of the frequency band.The results show that as the bulk density of the debris flow increases,the peak frequency of its acoustic signal shifts to low frequencies.The energy distribution of different types of debris flow sound wave signal bands is obviously different.The dilute debris flow is mainly distributed in the higher frequency band above 30 Hz,and the transitional and viscous debris flow is distributed in the low frequency band.(5)The analysis of the acoustic wave signal of the debris flow in Jiangjia gully on August 9,2000 found that the peak frequency of the acoustic wave signal of this section of viscous debris flow was 10-15 Hz,and the energy of the frequency band was mainly concentrated in 7.5-15 Hz.The measured signals effectively verified the results of the wavelet packet frequency band classification test,that is,with the increase of the bulk density,the peak frequency and energy of the debris flow acoustic wave signal are mainly distributed in the low and middle frequency bands.