CN-115456028-B - Automatic identification method for first arrival of vibration event based on multi-module filter integration

CN115456028BCN 115456028 BCN115456028 BCN 115456028BCN-115456028-B

Abstract

The invention discloses an automatic identification method of a first arrival of a vibration event based on multi-module filter integration, which mainly comprises the following steps of preprocessing a continuous recording original signal, cutting, and judging vibration energy effective frequency distribution section by section aiming at a cutting signal; then sequentially performing operations such as embedding an anti-aliasing low-pass filter, signal downsampling, embedding a self-adaptive trending term filter, embedding median filtering, embedding an edge-free delay filter and the like, calculating signal envelope energy based on the edge-free delay filter, finally embedding a discrimination filter, normalizing the signal, and further obtaining a final recognition characteristic function. The invention establishes a set of automatic picking algorithm integrated by multiple filters, and builds the characteristic function while reducing noise, so that on one hand, only the most basic signal parameters are needed, on the other hand, the noise interference is effectively reduced, the first arrival time of the disaster event is extracted, and the occurrence accuracy of the automatic picking disaster event is improved.

Inventors

SUN DONG
MA NING
LUO BING
ZHANG ZHIHOU
WANG QIANG
ZHAO MINGHAO
LIU WEIZU
ZHANG TIANYI

Assignees

四川省华地建设工程有限责任公司
西南交通大学

Dates

Publication Date: 20260505
Application Date: 20220921

Claims (6)

1. An automatic identification method of a first arrival of a vibration event based on multi-module filter integration is characterized by comprising the following steps: s1, preprocessing a continuous recording original signal, then cutting, and judging vibration energy effective frequency distribution section by section aiming at a cutting signal; s2, embedding an anti-aliasing low-pass filter; S3, signal downsampling, namely downsampling on the basis of the low-pass filtering in the previous step, and introducing a downsampled scaling factor R, wherein the downsampled scaling factor R is expressed as: fsp=2f high wherein fs is the original sampling frequency of the signal, fsp is the sampling frequency after downsampling, and f high is the preset high-frequency cut-off frequency; s4, embedding an adaptive trending term filter, wherein the filter is expressed as: B=FIR(M,ω c ) M=round(d max ×fsp) trend=filtfit(B,1,x r ) x rd ＝x r -trned Wherein B is a filter, M is a cutoff angular frequency order, omega c is a cutoff angular frequency, d max is a preset maximum duration of disasters, f low is a preset low-frequency cutoff frequency, trend is trend-removing filtering, x r is a downsampled signal, and x rd is a trend-term-filtered signal; s5, embedding median filtering, wherein the median filtering is expressed as: y=medfilt(x rd ,round(fsp×d min /10)) Wherein d min is the shortest duration of a disaster set in advance; S6, embedding edge-free delay filtering, wherein the filter is expressed as: y′=filter(H,1,y) Wherein y' is a signal after non-edge delay filtering, H is a filter parameter, y is a signal after median filtering; s7, calculating signal envelope energy; s8, firstly embedding a discrimination filter, and then establishing a characteristic function, wherein the expression of the characteristic function is as follows: CF=10×(log 10 (1+D/V)/log 10 (2)) wherein CF is the first arrival identification characteristic function, V is the maximum amplitude of the signal, and D is the signal after discrimination and filtering; according to the method, stability of a characteristic function under interference of different noise intensities is identified, verification tests of the same signal and different noise levels are conducted, signal-to-noise ratio SNR is set to be 5, 3, 2 and 1 respectively through mixing Gaussian random white noise of an original signal, when the signal-to-noise ratio is 5, the noise intensity has small influence on the signal, vibration characteristics of the signal are clear, first arrival time of four main events can be picked up based on the characteristic function and corresponds to signal time sequence, when the signal-to-noise ratio is reduced to 3, noise content in the signal is increased, the form of the characteristic function is basically unchanged, the first arrival time of the four main events is still obvious, when the signal-to-noise ratio is further reduced to 2, the characteristic curve is only slightly changed, when the signal-to-noise ratio is 1, namely, the amplitude of the signal is equal to the amplitude of the noise, and the shape of the characteristic curve is obviously changed compared with the previous first arrival time of the first event.
2. The method for automatically identifying the first arrival of a vibration event based on multi-module filter integration as set forth in claim 1, wherein in step S6, the filter parameters H are set as follows: H=acos(2π×fc×th) a=1:length(th):0 k=1,2,......,N f Where f c is the center frequency of [ f low ,f high ], th is a time series, a is an amplitude conversion coefficient, and represents the number from 0 to 1, th, k is the number of frequency points, and N f is the number of frequency-decomposable frequency band windows defined in advance.
3. The method for automatically recognizing the first arrival of a vibration event based on multi-module filter integration as claimed in claim 1, wherein in the step S8, the expression of the discrimination filter is as follows: D(m)=filter(T,1,E k ) m=1,2,......,N d k=1,2,......,N f Wherein D is the intermediate variable of penalty factor beta containing non-pulse signal, E k is the signal envelope energy, m is the natural number sequence, and N d is the time window number which can be decomposed in holding time from the 1 st to the N d th; the intermediate variable T is expressed as: T=β×((1:L)-(L+1)) L=round(D(m)×fsp) D=[d min ,N d ,d max ] Where β is a penalty factor for the non-pulse signal and L is a data length representing D.
4. The method for automatically identifying the first arrival of a vibration event based on multi-module filter integration according to claim 1, wherein the step S1 is specifically as follows: S11, preprocessing the continuous recording original signals, wherein the preprocessing comprises the steps of baseline leveling, mean removal and trend removal; s12, cutting the preprocessed original signals, and cutting by adopting an overlap joint mode; s13, judging effective influence time of vibration energy segment by segment aiming at cutting signals, and setting an influence duration range of time: d min , the shortest duration of the disaster, d max , the longest duration of the disaster, N d , the number of time windows which can be decomposed when held; s14, judging vibration energy effective frequency distribution segment by segment according to the cutting signals, selecting a low-frequency band, and setting a frequency band range: f low a low frequency cut-off frequency, f high a high frequency cut-off frequency, and N f a frequency resolvable frequency band window number.
5. The method for automatically identifying a first arrival of a vibration event based on multi-module filter integration according to claim 4, wherein in the step S2, the anti-aliasing low pass filter is expressed as: Where n is the filter order, ω c is the cut-off angular frequency, and ω p is the passband edge frequency.
6. The method for automatically identifying a first arrival of a vibration event based on multi-module filter integration as set forth in claim 5, wherein the step S7 is specifically to divide the signal into sub-signals of different frequency bands according to different center frequency signals based on non-edge delay filtering, and further calculate envelope energy E k of each signal according to the following calculation formula: k=1,2,......,N f where H is the Hilbert transform of the signal and y' is the signal after non-edge delay filtering.

Description

Automatic identification method for first arrival of vibration event based on multi-module filter integration Technical Field The invention relates to the technical field of disaster early warning prediction, in particular to an automatic identification method for a first arrival of a vibration event based on multi-module filter integration. Background At present, early warning prediction of disaster events is increasingly dependent on dynamic feature observation. Natural disasters in the nature such as earthquakes, volcanic, landslides, collapse stones and debris flows or artificial disasters such as mine rock burst, pipeline breakage and high-pressure gas leakage are accompanied with vibration propagation of energy, so that the vibration time sequence characteristics of the natural disasters are recorded through instruments, and the explosion characteristics of disaster sources can be analyzed based on a mathematical physical method. The most basic analysis is the determination of the moment of occurrence of an event, i.e. the pick-up of the first arrival. In theory, the observation platform array is composed of a sufficient number of monitoring instruments (wide/narrow frequency seismometers, long period seismometers, accelerometers, acoustic wave sensors and the like), so that the feature information such as the first arrival time of an event, the energy gauge model of the event, the energy source azimuth angle of time and the like can be obtained. However, in actual monitoring, the number of detection instruments is often small, so how to obtain the initial value of an event through a small number of monitoring instruments is always a research hot spot in each field, and meanwhile, a great number of technical difficulties exist, and how to find out the source energy carried by a disaster event from complicated random vibration records is significant for the early warning of disasters so as to form multi-scale disasters. The method for revealing disaster events of different scales by utilizing dynamic vibration signals is mainly aimed at accurately monitoring the time-space distribution characteristics of the disaster events, and is based on the detection of the first arrival of the disaster events, namely the occurrence time of the disaster events. The current pickup aiming at the first arrival time is mainly based on analysis of dynamic time sequence signals, usually, the time sequence information recorded by a detection sensor is basically background noise in the earlier stage of disaster event occurrence, the disaster event occurrence time is usually accompanied by stronger pulse signals, and then becomes a source to propagate in a medium until being received by sensors distributed at different distances, and the signal strength gradually decays along with the time until the vibration amplitude is reduced to the background noise scale. The first-arrival pick-up is based on the envelope intensity, absolute amplitude, or power spectral characteristics of the signal in the time or frequency domain as a whole. The algorithms that are currently popular are the short-time average/long-time average ratio algorithm (STA/LTA) and the power spectrum thresholding method (PSD). The short-time average/long-time average ratio algorithm, although it can pick up the first arrival of an event occurrence, is too dependent on preset parameters, although some studies have corrected the sensitivity of the algorithm by changing the feature function. In addition, the short and long time window lengths affect the threshold, low thresholds may lead to many false first-arrival triggers (false positives), while high thresholds may lead to missing weak events (false negatives). In order to improve the shortfall that the short-time average/long-time average ratio algorithm is too dependent on the set parameters, a power spectrum threshold method is generated. The spectrum content of background noise is generally uniform in each frequency band, but the spectrum content of disaster events is very large in certain specific frequency bands, so that the core of the power spectrum threshold algorithm is to find the moment when the frequency content is abnormally large, and pick up the first arrival of the disaster event. For the power spectrum threshold algorithm, although the early parameter setting is less, the requirement of the algorithm on the signal-to-noise ratio of the signal is higher. The two methods have good recognition capability on signals with high signal-to-noise ratio, such as strong vibration energy, but the algorithm has great error recognition on the disaster events such as microseismic, volcanic micro-motion, pipeline piping and the like, wherein the signal-to-noise ratio is low, even the vibration energy of signal energy annihilation and background noise is high. Therefore, a relatively stable algorithm comprising the noise reduction module and the characteristic pickup module is established, and the method has