CN114114383A - Earthquake activity prediction method and system based on multiple characteristics - Google Patents
Earthquake activity prediction method and system based on multiple characteristics Download PDFInfo
- Publication number
- CN114114383A CN114114383A CN202111504788.6A CN202111504788A CN114114383A CN 114114383 A CN114114383 A CN 114114383A CN 202111504788 A CN202111504788 A CN 202111504788A CN 114114383 A CN114114383 A CN 114114383A
- Authority
- CN
- China
- Prior art keywords
- signal data
- energy
- layer
- maximum
- seismic
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000000694 effects Effects 0.000 title claims abstract description 44
- 238000000034 method Methods 0.000 title claims abstract description 37
- 239000002243 precursor Substances 0.000 claims abstract description 95
- 238000000605 extraction Methods 0.000 claims abstract description 64
- 238000001228 spectrum Methods 0.000 claims description 26
- 238000007781 pre-processing Methods 0.000 claims description 21
- 238000001514 detection method Methods 0.000 claims description 11
- 230000005236 sound signal Effects 0.000 claims description 9
- 230000008439 repair process Effects 0.000 claims description 8
- 230000005484 gravity Effects 0.000 claims description 6
- 238000010801 machine learning Methods 0.000 claims description 3
- 230000002159 abnormal effect Effects 0.000 description 13
- 238000004364 calculation method Methods 0.000 description 13
- 239000000126 substance Substances 0.000 description 9
- 230000008859 change Effects 0.000 description 7
- 238000006243 chemical reaction Methods 0.000 description 5
- 230000009466 transformation Effects 0.000 description 5
- 230000003595 spectral effect Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000010276 construction Methods 0.000 description 2
- 238000010219 correlation analysis Methods 0.000 description 2
- 238000000354 decomposition reaction Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 230000002547 anomalous effect Effects 0.000 description 1
- 230000006735 deficit Effects 0.000 description 1
- 230000005672 electromagnetic field Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 239000012530 fluid Substances 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000000844 transformation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/01—Measuring or predicting earthquakes
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/28—Processing seismic data, e.g. for interpretation or for event detection
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V3/00—Electric or magnetic prospecting or detecting; Measuring magnetic field characteristics of the earth, e.g. declination, deviation
- G01V3/38—Processing data, e.g. for analysis, for interpretation, for correction
Landscapes
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Remote Sensing (AREA)
- Environmental & Geological Engineering (AREA)
- Geology (AREA)
- General Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Geophysics (AREA)
- Acoustics & Sound (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention discloses a seismic activity prediction method and a system based on multiple characteristics, which relate to the technical field of seismic activity prediction and comprise the following steps: acquiring original seismic precursor signal data of a target area at the current stage; the raw seismic precursor signal data comprises subsurface electromagnetic disturbance signal data or earth-sound signal data; respectively performing time domain feature extraction, frequency domain feature extraction and transform domain feature extraction on the original seismic precursor signal data to obtain a feature set of a target area at the current stage; the feature set comprises time domain features, frequency domain features and transform domain features; predicting seismic activity of a target area based on a feature set in a period of time; the period of time includes at least a current phase. According to the invention, the original seismic precursor signals can be respectively processed through three domains (time domain, frequency domain and transform domain) to obtain various characteristics related to seismic activity, so that the accuracy of seismic activity prediction is improved.
Description
Technical Field
The invention relates to the technical field of seismic activity prediction, in particular to a seismic activity prediction method and system based on multiple characteristics.
Background
During the seismic process, physical changes of the subsurface materials or chemical changes of the subsurface materials are involved. The original seismic precursor signals have complex and variable characteristics such as seismic activity, surface deformation, electromagnetic fields, earth temperature, earth sound, gravity, subsurface fluids, and geochemistry. The original earthquake precursor signal also has the characteristics of large data volume and large change rule, and when the signal is directly observed, information related to an earthquake is difficult to acquire. If it is desired to monitor or predict the earthquake by observing the change of the signal, it is necessary to extract relevant features from the physical variables or the chemical variables and then predict the earthquake activity by the abnormal change or change law of the extracted feature values.
With respect to the extraction of the features of the original seismic precursor signals, a more common method is to extract features from a seismic catalogue or the original seismic precursor signals. The seismic catalogue will typically extract features such as seismic elapsed time, mean magnitude, b-value (slope of the goodberg curve), mean square deviation, and magnitude deficit from frequency, energy, magnitude-frequency relationships, etc., according to the goodberg-riekt power law (G-R relationship). For feature extraction of an original seismic precursor signal, feature extraction is generally performed by using basic statistical values, such as a mean value, a central moment (variance, skewness and kurtosis), energy of a frequency band after fourier transform and a value of a percentile, and a magnitude of autocorrelation as parameters.
Although the method can perform feature extraction on the original seismic precursor signal, the acquired feature quantity is single, and the accuracy of predicting seismic activity is low.
Disclosure of Invention
The invention aims to provide a seismic activity prediction method and a system based on multiple characteristics, which can be used for respectively processing an original seismic precursor signal through three domains (a time domain, a frequency domain and a transform domain) to obtain multiple characteristics related to seismic activity, so that the accuracy of seismic activity prediction is improved.
In order to achieve the purpose, the invention provides the following scheme:
a method for seismic activity prediction based on multiple features, comprising:
acquiring original seismic precursor signal data of a target area at the current stage; the raw seismic precursor signal data comprises subsurface electromagnetic disturbance signal data or earth-sound signal data;
respectively performing time domain feature extraction, frequency domain feature extraction and transform domain feature extraction on the original seismic precursor signal data to obtain a feature set of a target area at the current stage; the feature set comprises time domain features, frequency domain features and transform domain features;
predicting seismic activity of a target area based on a feature set in a period of time; the period of time includes at least a current phase.
Optionally, the time domain feature extraction, the frequency domain feature extraction, and the transform domain feature extraction are respectively performed on the original seismic precursor signal data to obtain a feature set of a target area at a current stage, and the method specifically includes:
preprocessing the original earthquake precursor signal data to obtain preprocessed earthquake precursor signal data; the preprocessing comprises data missing exception processing and equal interval pulse exception processing; and respectively carrying out time domain feature extraction, frequency domain feature extraction and transform domain feature extraction on the preprocessed seismic precursor signal data to obtain a feature set of a target area at the current stage.
Optionally, the preprocessing the original seismic precursor signal data specifically includes:
performing data missing exception processing on the original seismic precursor signal data to obtain seismic precursor signal data subjected to data missing exception processing; processing the seismic precursor signal data subjected to data missing exception processing by adopting a time domain detection and repair pulse-based algorithm to obtain first data; processing the seismic precursor signal data subjected to data missing exception processing by adopting a frequency domain detection and repair pulse-based algorithm to obtain second data; wherein the preprocessed seismic precursor signal data includes first data and second data.
Optionally, the time domain feature extraction, the frequency domain feature extraction, and the transform domain feature extraction are respectively performed on the preprocessed seismic precursor signal data to obtain a feature set of a target area at a current stage, and the method specifically includes:
extracting time domain characteristics of the first data to obtain time domain characteristics of a target area at the current stage; extracting frequency domain characteristics of the second data to obtain frequency domain characteristics of a target area at the current stage; and extracting the transform domain characteristics of the second data to obtain the transform domain characteristics of the target area at the current stage.
Optionally, the predicting the seismic activity of the target area based on the feature set in a period of time specifically includes: and predicting the seismic activity of the target area based on the feature set in a period of time and a machine learning algorithm.
Optionally, the time domain features corresponding to the underground electromagnetic disturbance signal data include: mean absolute value, variance, maximum absolute value, power, skewness, kurtosis, position of 5% of maximum absolute value, position of 10% of maximum absolute value, short-time energy standard deviation and maximum short-time energy;
the time domain characteristics corresponding to the earth-sound signal data comprise: mean absolute value, variance, maximum absolute value, power, skewness, kurtosis, position of 5% of maximum absolute value, position of 10% of maximum absolute value, short-term energy standard deviation, maximum short-term energy, mean short-term zero-crossing rate, and maximum short-term zero-crossing rate.
Optionally, the frequency domain feature corresponding to the underground electromagnetic disturbance signal data and the frequency domain feature corresponding to the ground sound signal data both include: 0-5 Hz energy, 5-10Hz energy, 10-15Hz energy, 15-20Hz energy, 20-25Hz energy, 25-30Hz energy, 30-35Hz energy, 35-40Hz energy, 40-60Hz energy, 140-160Hz energy, energy proportion of other frequency bands, center of gravity frequency, mean square frequency, frequency variance and spectrum entropy.
Optionally, the transform domain is a wavelet transform domain; the transform domain features corresponding to the subsurface electromagnetic disturbance signal data comprise: mean of layer 4 detail absolute values, layer 4 detail energy value maximum, layer 4 detail energy value variance, mean of layer 5 detail absolute values, layer 5 detail energy value maximum, layer 5 detail energy value variance, mean of layer 6 detail absolute values, layer 6 detail energy value maximum, layer 6 detail energy value variance, mean of layer 6 approximate absolute values, the energy approximation of the 6 th layer, the maximum value of the energy approximation value of the 6 th layer, the variance of the energy approximation value of the 6 th layer, the mean value of the ultra-low frequency absolute value, the ultra-low frequency variance, the ultra-low frequency power, the ultra-low frequency skewness, the ultra-low frequency kurtosis, the maximum value of the ultra-low frequency absolute value, the maximum 5% position of the ultra-low frequency absolute value, the maximum 10% position of the ultra-low frequency absolute value, the ultra-low frequency short-time energy standard deviation and the ultra-low frequency short-time energy maximum value;
the transform domain features corresponding to the ground sound signal data comprise: mean of layer 4 detail absolute values, layer 4 detail energy value maximum, layer 4 detail energy value variance, mean of layer 5 detail absolute values, layer 5 detail energy value maximum, layer 5 detail energy value variance, mean of layer 6 detail absolute values, layer 6 detail energy value maximum, layer 6 detail energy value variance, mean of layer 6 approximate absolute values, layer 6 approximate energy value maximum, and layer 6 approximate energy value variance and mean.
A seismic activity prediction system based on multiple features, comprising:
the data acquisition module is used for acquiring original earthquake precursor signal data of a target area at the current stage; the raw seismic precursor signal data comprises subsurface electromagnetic disturbance signal data or earth-sound signal data;
the characteristic set determining module is used for respectively carrying out time domain characteristic extraction, frequency domain characteristic extraction and transform domain characteristic extraction on the original seismic precursor signal data to obtain a characteristic set of a target area at the current stage; the feature set comprises time domain features, frequency domain features and transform domain features;
the earthquake activity prediction module is used for predicting earthquake activity of the target area based on the feature set in a period of time; the period of time includes at least a current phase.
Optionally, the feature set determining module specifically includes:
the preprocessing unit is used for preprocessing the original earthquake precursor signal data to obtain preprocessed earthquake precursor signal data; the preprocessing comprises data missing exception processing and equal interval pulse exception processing; and the feature set determining unit is used for respectively performing time domain feature extraction, frequency domain feature extraction and transform domain feature extraction on the preprocessed seismic precursor signal data to obtain a feature set of a target area at the current stage.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the method creatively processes the characteristics of the earthquake precursor signal from three angles, and can observe the trend or change of the precursor signal in a time domain, a frequency domain and a transformation domain simultaneously, thereby improving the accuracy of earthquake activity prediction.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a block diagram of a feature extraction framework of the present invention;
FIG. 2 is a pulse-type abnormal spectrum diagram of the present invention; FIG. 2 (a) is a time domain spectrogram; FIG. 2 (b) is a frequency domain spectrogram;
FIG. 3 is a schematic flow chart of a seismic activity prediction method based on multiple features of the present invention;
FIG. 4 is a schematic diagram of a seismic activity prediction system based on multiple features according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
For many years, integrated microsystems laboratories at the university of Beijing have been dedicated to observing or capturing anomalous signature signals in the original seismic precursor signals, further to analyzing and attempting to predict upcoming seismic activity. Acoustics & electromagnetic to AI (AETA) is a system invented by the integrated microsystem laboratories of the university of beijing and used to collect raw seismic precursor signals, including subsurface electromagnetic disturbance signals or earth-sound signals. Currently, the AETA system has collected 38TB of data, and has discovered some signal features related to the impending earthquake and made an imminent earthquake prediction. While some progress has been made, the solution of the distance seismic prediction problem requires more analysis and research.
In view of the above, the invention firstly performs time domain/frequency domain/transform domain feature extraction based on the original seismic precursor signals (including underground electromagnetic disturbance signals or earth sound signals) acquired by the AETA system, and then performs seismic activity prediction based on the extracted features.
The original seismic precursor signals acquired by the AETA system have the characteristics of large data volume and low single-point information density, and particularly, the original underground electromagnetic disturbance signals have obvious periodicity, have single or double peaks in a frequency domain, have relatively independent frequencies of all wave bands, and need to more mine the characteristics of ultra-low frequency bands. For the original earth sound signal, the amplitude is mainly distributed around zero and the fluctuation is discontinuous, similar to the voice signal. Meanwhile, in addition to signals related to original seismic precursors, signals with change time related to human activity time exist, and the signals are different from seismic wave signals in frequency spectrum, and the characteristics of the signals need to be further mined from the frequency spectrum. Based on the above characteristics, the features extracted from the original signal are divided into three major categories: time domain related features, frequency domain related features, and wavelet transform domain related features. The former two are mainly from time domain and frequency domain, extract statistical characteristics, the latter adopts the method based on transform, uses db4 wavelet decomposition reconstruction, can refine the characteristics of analysis signal on time frequency from multi-scale angle. Therefore, the method can comprehensively describe the earthquake precursor signals and provide comprehensive and rich data input for further performing earthquake correlation analysis, abnormal extraction and prediction model construction.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Example one
The embodiment provides a method for extracting characteristics of an original seismic precursor signal, and with reference to fig. 1, the method includes:
the method comprises the following steps: preprocessing original seismic precursor signal data acquired by an AETA system; the raw seismic precursor signal data includes subsurface electromagnetic disturbance signals or earth-sound signals.
At present, data missing abnormity caused by network difference of the AETA system location and equal-interval pulse abnormity which continuously appears in a period of time exist in original earthquake precursor signal data. The preprocessing comprises data missing exception processing and equal interval pulse exception processing. The data missing exception handling comprises: whether a plurality of continuous 0 data points exist at the tail of the original earthquake precursor signal data or not is judged, and if yes, the data are judged to be abnormal in missing. For the data with data missing abnormality, firstly, a binary search mode is used for finding out the first abnormal point which becomes 0, and then, the missing value is used for replacing the catastrophe point and the following data. The algorithm adopted by the equal interval pulse exception handling comprises the following steps: the pulse type algorithm is repaired based on time domain detection and the pulse type algorithm is repaired based on frequency domain detection.
The pulse type repairing algorithm based on time domain detection comprises the following steps: firstly, judging abnormal points of first-order difference through an IQR algorithm, and then connecting adjacent abnormal points by using an open operation to obtain an abnormal interval. And detecting the equispaced pulse abnormal data according to the equispaced judgment of whether the middle point of the abnormal section is mostly. The detected abnormal interval is filled up by a linear interpolation mode.
TABLE 1 Impulse type algorithm execution flow chart based on time domain detection and repair
The pulse type algorithm for detecting and repairing based on the frequency domain comprises the step of judging whether the frequency exceeding a threshold value is equal interval or not, so as to judge whether the earthquake precursor signal data subjected to data missing abnormity processing is abnormal by equal interval pulse within one minute or not. First determining the fundamental frequencyf s Then the frequency spectrum around the fundamental frequency and the frequency multiplication is set to 0, and finally the frequency spectrum is converted into the time domain through the inverse fourier transform.
There is a spectrum aliasing phenomenon in the frequency spectrum larger than 250Hz, and as shown in FIG. 2, in the frequency spectrum of 100 to 200Hz, the small peak beside each high peak is actually the frequency spectrum corresponding to 300 to 400 Hz. It is found through research that the amplitude of the frequency multiplication larger than 400Hz is very small, and the spectrum energy of the original signal is mainly concentrated in the low frequency part, and only the frequency multiplication not exceeding 430Hz is removed in order not to influence the spectrum of the low frequency part.
Table 2 repair pulse type algorithm execution flow chart based on frequency domain detection
| Pulse type algorithm for detecting and repairing algorithm based on frequency domain |
| Inputting: |
| v i : seismic precursor signal data after data missing exception processing |
| T H : maximum percentage threshold, default 1% |
| T e : percentage of fold error, default 5% |
| T n : minimum number of elements in the set, default 5 |
| And (3) outputting: |
| v o : repaired frequency seismic precursor signal data |
| The method comprises the following steps: |
| 1. to pairv i Fast Fourier transform is carried out to obtain frequency spectrumf i (ii) a Based on symmetry, taking frequency spectrumf i First half sectionf ir 。 |
| 2. To pairf i In the method, the values except the value near 50-150 Hz are sequenced to obtain the maximum valueT H Spectral amplitude of a location as a spectral amplitude thresholdf T 。 |
| 3. Finding a frequency spectrumf ir Greater than spectral amplitude thresholdf T And in the maximum value in each intervalf max Representing the interval, forming a setF。 |
| 4. When in useFNon-empty time, newly-built setDAnd will beFMinimum value off 0 Put into a collectionD。 |
5. To pairFElement (1) off j Go through from small to large, letFOf which the current maximum isf m If, if |
6. If it isFMiddle element is greater thanT n For in the frequency spectrum |
| 7. Repeating the steps 4-6 to finally obtain a new frequency spectrumf ol Will bef ol The first term to the second last term are overturned and conjugated to obtain the right half section of the frequency spectrumf or Spliced to obtain a converted spectrumf o 。 |
| 8. If all sets areFAll are less thanT n Direct return ofv i Otherwise, it willf o Performing inverse Fourier transform to obtainv o And output. |
Step two: and performing time domain conversion on the preprocessed seismic precursor signal data.
The pre-processed seismic precursor signal data are electromagnetic disturbance signals and earth-sound signals at 500Hz, and the pre-processed seismic precursor signal data at 1 minute contains 30000 values.
The time domain conversion is to extract the electromagnetic disturbance signal or the earth sound signal (30000 points) corresponding to each minute into individual time domain characteristic values.
The time domain features include: mean, variance, absolute value, power, skewness, kurtosis, position of 5% of the maximum absolute value, position of 10% of the maximum absolute value, short-term energy standard deviation, maximum short-term energy, short-term zero-crossing rate.
The calculation formula of the mean value is as follows:
wherein the content of the first and second substances,
to sample the mean of one minute data, T is the data length of one minute, 30000, s (T), T =1,2, 3.
The calculation formula of the variance is as follows:
wherein the content of the first and second substances,
is the variance of the sampled one minute data.
The calculation formula of the power is as follows:
wherein the content of the first and second substances,
is the power of the sampled one minute data.
The calculation formula of the skewness is as follows:
wherein the content of the first and second substances,
is the power of the sampled one minute data.
The kurtosis calculation formula is as follows:
wherein the content of the first and second substances,
is the power of the sampled one minute data.
The calculation formula of the short-time energy is as follows:
wherein the content of the first and second substances,E n is the short-time average energy of the signal at time N, N being the window length, here taken to be 500,x(m)m =1,2, 3.. for pre-processed seismic precursor signal data,
representing a window function.
The position of the maximum value x% is the value of the x% position selected after 30000 points are sorted from large to small according to absolute values.
The short-time zero crossing rate is used for counting the times of changing the signal sampled every second from positive to negative and from negative to positive, namely the times of zero crossing point of the signal every second.
The ultralow frequency band data of the electromagnetic disturbance signal needs to be analyzed, so that the high-frequency electromagnetic disturbance signal is subjected to low-pass filtering before time domain conversion, and the ground sound signal with main energy concentrated in the ultralow frequency band does not need to be specially extracted for the ultralow frequency band feature. In addition, the short-time zero crossing rate has a large correlation with frequency and is a classic feature in speech signal processing. Because the original signal of the electromagnetic disturbance is mainly 50Hz or 150Hz, the signal has a long fixed period, and the short-time zero-crossing rate is basically invalid, the extraction of the short-time energy standard deviation, the short-time energy maximum value and the short-time zero-crossing rate is only carried out on the earth sound signal similar to the voice signal.
Step three: and performing frequency domain conversion on the preprocessed seismic precursor signal data.
The frequency domain features include: the power, the center of gravity frequency, the mean square frequency, the spectrum variance and the spectrum entropy of each spectrum band after Fourier transformation.
The frequency spectrum band after Fourier transformation comprises 0-5 Hz, 5-10Hz, 10-15Hz, 15-20Hz, 20-25Hz, 25-30Hz, 30-35Hz, 35-40Hz, 40-60Hz, 140-160Hz and other frequencies.
The formula for calculating the energy of the a-bHz frequency band is as follows:
wherein abs is an absolute value function, FFT is fast Fourier transform, and v is seismic precursor signal data after 1 minute of preprocessing.
The gravity center frequency calculation formula is as follows:
where f (v) n represents the nth value of the data column.
The mean square frequency calculation formula is as follows:
the spectral variance calculation formula is as follows:
the spectrum entropy calculation formula is as follows:
step four: and performing transform domain feature conversion on the preprocessed seismic precursor signal data.
The transform domain features include: and (5) adopting db4 as a wavelet base to carry out 6-layer wavelet decomposition, and extracting the characteristic values of the reconstructed detail parts of the 4 th-6 th layers and the approximate part of the 6 th layer.
The reconstruction detail parts of the 4 th to 6 th layers respectively comprise the following frequency bands: 15.63Hz to 31.25Hz, 7.81Hz to 15.63Hz, 3.91Hz to 7.81 Hz.
The wavelet reconstruction detail partial frequency band calculation formula is as follows:
wherein the content of the first and second substances,f s in order to be able to sample the frequency,jis the number of layers.
The approximate portion of layer 6 includes frequency bands of: 0 to 3.91 Hz.
The wavelet reconstruction approximate partial band calculation formula is as follows:
the characteristic values comprise an absolute value mean value, energy, an energy variance and an absolute value maximum value.
As mentioned above, 51 electromagnetic features and 44 earth sound features are extracted, and the specific features are shown in tables 3-4.
TABLE 3 electromagnetic disturbance characteristics List
TABLE 4 land Sound feature List
Example two
Referring to fig. 3, the method for predicting seismic activity based on multiple features according to this embodiment includes:
step 301: acquiring original seismic precursor signal data of a target area at the current stage; the raw seismic precursor signal data includes subsurface electromagnetic disturbance signal data or earth-sound signal data.
Step 302: respectively performing time domain feature extraction, frequency domain feature extraction and transform domain feature extraction on the original seismic precursor signal data to obtain a feature set of a target area at the current stage; the feature set includes time domain features, frequency domain features, and transform domain features.
Step 303: predicting seismic activity of a target area based on a feature set in a period of time; the period of time includes at least a current phase.
Further, the preprocessing the original seismic precursor signal data specifically includes:
and carrying out data missing exception processing on the original seismic precursor signal data to obtain the seismic precursor signal data after the data missing exception processing. And processing the seismic precursor signal data subjected to data missing exception processing by adopting a time domain detection and repair pulse type algorithm to obtain first data. And processing the seismic precursor signal data subjected to data missing exception processing by adopting a frequency domain detection and repair pulse type algorithm to obtain second data. Wherein the preprocessed seismic precursor signal data includes first data and second data.
Further, the time domain feature extraction, the frequency domain feature extraction and the transform domain feature extraction are respectively performed on the preprocessed seismic precursor signal data to obtain a feature set of a target area at the current stage, and the method specifically includes: and extracting time domain characteristics of the first data to obtain the time domain characteristics of the target area at the current stage. And extracting frequency domain characteristics of the second data to obtain the frequency domain characteristics of the target area at the current stage. And extracting the transform domain characteristics of the second data to obtain the transform domain characteristics of the target area at the current stage.
Step 303 specifically includes: and predicting the seismic activity of the target area based on the feature set in a period of time and a machine learning algorithm.
The time domain features corresponding to the underground electromagnetic disturbance signal data comprise: mean absolute value, variance, maximum absolute value, power, skewness, kurtosis, position of maximum absolute value of 5%, position of maximum absolute value of 10%, short-term energy standard deviation, and short-term energy maximum.
The time domain characteristics corresponding to the earth-sound signal data comprise: mean absolute value, variance, maximum absolute value, power, skewness, kurtosis, position of 5% of maximum absolute value, position of 10% of maximum absolute value, short-term energy standard deviation, maximum short-term energy, mean short-term zero-crossing rate, and maximum short-term zero-crossing rate.
The frequency domain characteristics corresponding to the underground electromagnetic disturbance signal data and the frequency domain characteristics corresponding to the ground sound signal data both comprise: 0-5 Hz energy, 5-10Hz energy, 10-15Hz energy, 15-20Hz energy, 20-25Hz energy, 25-30Hz energy, 30-35Hz energy, 35-40Hz energy, 40-60Hz energy, 140-160Hz energy, energy proportion of other frequency bands, center of gravity frequency, mean square frequency, frequency variance and spectrum entropy.
The transform domain is a wavelet transform domain; the transform domain features corresponding to the subsurface electromagnetic disturbance signal data comprise: mean of layer 4 detail absolute values, layer 4 detail energy value maximum, layer 4 detail energy value variance, mean of layer 5 detail absolute values, layer 5 detail energy value maximum, layer 5 detail energy value variance, mean of layer 6 detail absolute values, layer 6 detail energy value maximum, layer 6 detail energy value variance, mean of layer 6 approximate absolute values, the energy approximation of the layer 6, the maximum value of the energy approximation of the layer 6, the variance of the energy approximation of the layer 6, the mean value of the absolute value of the ultralow frequency, the variance of the ultralow frequency, the power of the ultralow frequency, the skewness of the ultralow frequency, the kurtosis of the ultralow frequency, the maximum value of the absolute value of the ultralow frequency, the maximum position of the absolute value of the ultralow frequency of 5 percent, the maximum position of the absolute value of the ultralow frequency of 10 percent, the standard deviation of the energy of the ultralow frequency and the maximum value of the energy of the ultralow frequency and the short time.
The transform domain features corresponding to the earth-sound signal data include: mean of layer 4 detail absolute values, layer 4 detail energy value maximum, layer 4 detail energy value variance, mean of layer 5 detail absolute values, layer 5 detail energy value maximum, layer 5 detail energy value variance, mean of layer 6 detail absolute values, layer 6 detail energy value maximum, layer 6 detail energy value variance, mean of layer 6 approximate absolute values, layer 6 approximate energy value maximum, and layer 6 approximate energy value variance and mean.
EXAMPLE III
Referring to fig. 4, the present embodiment provides a seismic activity prediction system based on multiple features, including:
the data acquisition module 401: the method comprises the steps of acquiring original seismic precursor signal data of a target area at the current stage; the raw seismic precursor signal data includes subsurface electromagnetic disturbance signal data or earth-sound signal data.
The feature set determination module 402: the system is used for respectively carrying out time domain feature extraction, frequency domain feature extraction and transform domain feature extraction on the original seismic precursor signal data to obtain a feature set of a target area at the current stage; the feature set includes time domain features, frequency domain features, and transform domain features.
The seismic activity prediction module 403: predicting seismic activity of a target area based on a feature set over a period of time; the period of time includes at least a current phase.
The feature set determining module 402 specifically includes:
the preprocessing unit is used for preprocessing the original earthquake precursor signal data to obtain preprocessed earthquake precursor signal data; the preprocessing comprises data missing exception processing and equal interval pulse exception processing. And the feature set determining unit is used for respectively performing time domain feature extraction, frequency domain feature extraction and transform domain feature extraction on the preprocessed seismic precursor signal data to obtain a feature set of a target area at the current stage.
Compared with the prior art, the invention has the following effects: the method creatively processes the characteristics of the earthquake precursor signal from three angles, and can observe the trend or change of the precursor signal in a time domain, a frequency domain and a transformation domain simultaneously, thereby improving the accuracy of earthquake activity prediction.
The invention has high description accuracy, the characteristic value has wide covering information, and the 51 characteristic transformations do not have redundant characteristics, thereby providing comprehensive and rich data input for further performing seismic correlation analysis, abnormal extraction and prediction model construction. The complexity of the algorithm used by the invention is O (n) as much as possible and O (logn) at most, thereby greatly facilitating the feature calculation and extraction of large data volume.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.
Claims (10)
1. A seismic activity prediction method based on multiple features, comprising:
acquiring original seismic precursor signal data of a target area at the current stage; the raw seismic precursor signal data comprises subsurface electromagnetic disturbance signal data or earth-sound signal data;
respectively performing time domain feature extraction, frequency domain feature extraction and transform domain feature extraction on the original seismic precursor signal data to obtain a feature set of a target area at the current stage; the feature set comprises time domain features, frequency domain features and transform domain features;
predicting seismic activity of a target area based on a feature set in a period of time; the period of time includes at least a current phase.
2. The method according to claim 1, wherein the time domain feature extraction, the frequency domain feature extraction and the transform domain feature extraction are respectively performed on the original seismic precursor signal data to obtain a feature set of a target area at a current stage, and specifically comprises:
preprocessing the original earthquake precursor signal data to obtain preprocessed earthquake precursor signal data; the preprocessing comprises data missing exception processing and equal interval pulse exception processing;
and respectively carrying out time domain feature extraction, frequency domain feature extraction and transform domain feature extraction on the preprocessed seismic precursor signal data to obtain a feature set of a target area at the current stage.
3. The method of claim 2, wherein the preprocessing of the raw seismic precursor signal data comprises:
performing data missing exception processing on the original seismic precursor signal data to obtain seismic precursor signal data subjected to data missing exception processing;
processing the seismic precursor signal data subjected to data missing exception processing by adopting a time domain detection and repair pulse-based algorithm to obtain first data;
processing the seismic precursor signal data subjected to data missing exception processing by adopting a frequency domain detection and repair pulse-based algorithm to obtain second data;
wherein the preprocessed seismic precursor signal data includes first data and second data.
4. The method according to claim 3, wherein the time domain feature extraction, the frequency domain feature extraction and the transform domain feature extraction are respectively performed on the preprocessed seismic precursor signal data to obtain a feature set of a target area at a current stage, and specifically comprises:
extracting time domain characteristics of the first data to obtain time domain characteristics of a target area at the current stage;
extracting frequency domain characteristics of the second data to obtain frequency domain characteristics of a target area at the current stage;
and extracting the transform domain characteristics of the second data to obtain the transform domain characteristics of the target area at the current stage.
5. The method according to claim 1, wherein the predicting the seismic activity of the target area based on the feature set over a period of time specifically comprises:
and predicting the seismic activity of the target area based on the feature set in a period of time and a machine learning algorithm.
6. The method of claim 1, wherein the time domain features corresponding to the subsurface electromagnetic perturbation signal data comprise: mean absolute value, variance, maximum absolute value, power, skewness, kurtosis, position of 5% of maximum absolute value, position of 10% of maximum absolute value, short-time energy standard deviation and maximum short-time energy;
the time domain characteristics corresponding to the earth-sound signal data comprise: mean absolute value, variance, maximum absolute value, power, skewness, kurtosis, position of 5% of maximum absolute value, position of 10% of maximum absolute value, short-term energy standard deviation, maximum short-term energy, mean short-term zero-crossing rate, and maximum short-term zero-crossing rate.
7. The method of claim 1, wherein the frequency domain features corresponding to the subsurface electromagnetic perturbation signal data and the frequency domain features corresponding to the geophone signal data each comprise: 0-5 Hz energy, 5-10Hz energy, 10-15Hz energy, 15-20Hz energy, 20-25Hz energy, 25-30Hz energy, 30-35Hz energy, 35-40Hz energy, 40-60Hz energy, 140-160Hz energy, energy proportion of other frequency bands, center of gravity frequency, mean square frequency, frequency variance and spectrum entropy.
8. The method of claim 1, wherein the transform domain is a wavelet transform domain; the transform domain features corresponding to the subsurface electromagnetic disturbance signal data comprise: mean of layer 4 detail absolute values, layer 4 detail energy value maximum, layer 4 detail energy value variance, mean of layer 5 detail absolute values, layer 5 detail energy value maximum, layer 5 detail energy value variance, mean of layer 6 detail absolute values, layer 6 detail energy value maximum, layer 6 detail energy value variance, mean of layer 6 approximate absolute values, the energy approximation of the 6 th layer, the maximum value of the energy approximation value of the 6 th layer, the variance of the energy approximation value of the 6 th layer, the mean value of the ultra-low frequency absolute value, the ultra-low frequency variance, the ultra-low frequency power, the ultra-low frequency skewness, the ultra-low frequency kurtosis, the maximum value of the ultra-low frequency absolute value, the maximum 5% position of the ultra-low frequency absolute value, the maximum 10% position of the ultra-low frequency absolute value, the ultra-low frequency short-time energy standard deviation and the ultra-low frequency short-time energy maximum value;
the transform domain features corresponding to the ground sound signal data comprise: mean of layer 4 detail absolute values, layer 4 detail energy value maximum, layer 4 detail energy value variance, mean of layer 5 detail absolute values, layer 5 detail energy value maximum, layer 5 detail energy value variance, mean of layer 6 detail absolute values, layer 6 detail energy value maximum, layer 6 detail energy value variance, mean of layer 6 approximate absolute values, layer 6 approximate energy value maximum, and layer 6 approximate energy value variance and mean.
9. A seismic activity prediction system based on a plurality of features, comprising:
the data acquisition module is used for acquiring original earthquake precursor signal data of a target area at the current stage; the raw seismic precursor signal data comprises subsurface electromagnetic disturbance signal data or earth-sound signal data;
the characteristic set determining module is used for respectively carrying out time domain characteristic extraction, frequency domain characteristic extraction and transform domain characteristic extraction on the original seismic precursor signal data to obtain a characteristic set of a target area at the current stage; the feature set comprises time domain features, frequency domain features and transform domain features;
the earthquake activity prediction module is used for predicting earthquake activity of the target area based on the feature set in a period of time; the period of time includes at least a current phase.
10. The system of claim 9, wherein the feature set determination module comprises:
the preprocessing unit is used for preprocessing the original earthquake precursor signal data to obtain preprocessed earthquake precursor signal data; the preprocessing comprises data missing exception processing and equal interval pulse exception processing;
and the feature set determining unit is used for respectively performing time domain feature extraction, frequency domain feature extraction and transform domain feature extraction on the preprocessed seismic precursor signal data to obtain a feature set of a target area at the current stage.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111504788.6A CN114114383A (en) | 2021-12-10 | 2021-12-10 | Earthquake activity prediction method and system based on multiple characteristics |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111504788.6A CN114114383A (en) | 2021-12-10 | 2021-12-10 | Earthquake activity prediction method and system based on multiple characteristics |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114114383A true CN114114383A (en) | 2022-03-01 |
Family
ID=80364288
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111504788.6A Pending CN114114383A (en) | 2021-12-10 | 2021-12-10 | Earthquake activity prediction method and system based on multiple characteristics |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114114383A (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO1996018119A1 (en) * | 1994-12-06 | 1996-06-13 | Farnsworth David F | Method for forecasting an earthquake from precusor signals |
| KR101028779B1 (en) * | 2011-01-19 | 2011-04-14 | (주)에이케이지씨 | A seismic wave auto-detection system using time-frequency domain changes and variable threshold values and the method thereof |
| CN104570093A (en) * | 2014-12-26 | 2015-04-29 | 深圳市微纳集成电路与系统应用研究院 | Earthquake precursor signal processing system and method |
| WO2019232791A1 (en) * | 2018-06-08 | 2019-12-12 | 北京大学深圳研究生院 | Monitoring data processing method for earthquake prediction, and earthquake prediction method and system |
| US20200271804A1 (en) * | 2019-02-21 | 2020-08-27 | Emerson Paradigm Holding Llc | Accelerated seismic interpolation of measured data collected by a set of geophones arranged in an irregularly spaced grid |
| CN111626153A (en) * | 2020-05-13 | 2020-09-04 | 电子科技大学 | Integrated learning-based partial discharge fault state identification method |
-
2021
- 2021-12-10 CN CN202111504788.6A patent/CN114114383A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO1996018119A1 (en) * | 1994-12-06 | 1996-06-13 | Farnsworth David F | Method for forecasting an earthquake from precusor signals |
| KR101028779B1 (en) * | 2011-01-19 | 2011-04-14 | (주)에이케이지씨 | A seismic wave auto-detection system using time-frequency domain changes and variable threshold values and the method thereof |
| CN104570093A (en) * | 2014-12-26 | 2015-04-29 | 深圳市微纳集成电路与系统应用研究院 | Earthquake precursor signal processing system and method |
| WO2019232791A1 (en) * | 2018-06-08 | 2019-12-12 | 北京大学深圳研究生院 | Monitoring data processing method for earthquake prediction, and earthquake prediction method and system |
| US20200271804A1 (en) * | 2019-02-21 | 2020-08-27 | Emerson Paradigm Holding Llc | Accelerated seismic interpolation of measured data collected by a set of geophones arranged in an irregularly spaced grid |
| CN111626153A (en) * | 2020-05-13 | 2020-09-04 | 电子科技大学 | Integrated learning-based partial discharge fault state identification method |
Non-Patent Citations (3)
| Title |
|---|
| 杨超 等: "基于张衡一号电磁卫星数据对印尼Ms_7.4地震的研究", 北京大学学报（自然科学版）, vol. 57, no. 6, pages 997 - 1005 * |
| 耿杰 等: "郯庐断裂带中南段前兆综合信息量地震预测方法研究", 地震研究, vol. 38, no. 4, pages 517 - 524 * |
| 薛玮: "震前大地电磁信号分析与研究", 中国优秀硕士学位论文全文数据库-基础科学辑, vol. 2019, no. 5, pages 184 - 186 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Tary et al. | Analysis of time-varying signals using continuous wavelet and synchrosqueezed transforms | |
| Yiou et al. | Spectral analysis of climate data | |
| CN110454687A (en) | A kind of pipeline multipoint leakage localization method based on improvement VMD | |
| CN107991706B (en) | Coal seam hydraulic fracturing microseismic signal combined noise reduction method based on wavelet packet multiple threshold and improved empirical mode decomposition | |
| CN109101910B (en) | A kind of Magnetotelluric signal denoising method screened based on noise | |
| Ma et al. | A novel blind source separation method for single-channel signal | |
| CN107957566A (en) | Magnetic resonance depth measurement method for extracting signal based on frequency selection singular spectrum analysis | |
| CN107274911A (en) | A kind of similarity analysis method based on sound characteristic | |
| CN104089699B (en) | A kind of substation equipment speech reconstruction algorithm | |
| CN108196164B (en) | Method for extracting cable fault point discharge sound signal under strong background noise | |
| Li et al. | A novel feature extraction method for roller bearing using sparse decomposition based on self-Adaptive complete dictionary | |
| Astuti et al. | Adaptive Short Time Fourier Transform (STFT) Analysis of seismic electric signal (SES): A comparison of Hamming and rectangular window | |
| Giudicepietro et al. | Fast discrimination of local earthquakes using a neural approach | |
| CN110609088B (en) | Post insulator fault identification method and device | |
| Yang et al. | Fast nonlinear Hoyergram for bearings fault diagnosis under random impact interference | |
| Melgoza et al. | Comparing radar receiver pulse deinterleaving performance of differing window functions for bandpass FIR filter design | |
| Li | New sparse regularization approach for extracting transient impulses from fault vibration signal of rotating machinery | |
| Li et al. | Magnetotelluric data denoising method combining two deep-learning-based models | |
| CN109507292A (en) | A kind of method for extracting signal | |
| Li et al. | Magnetotelluric signal-noise separation method based on SVM–CEEMDWT | |
| Wang et al. | The diagnosis of rolling bearing based on the parameters of pulse atoms and degree of cyclostationarity | |
| CN114114383A (en) | Earthquake activity prediction method and system based on multiple characteristics | |
| Biao et al. | Rotating machine fault diagnosis by a novel fast sparsity-enabled feature-energy-ratio method | |
| CN104778342A (en) | Heart sound characteristic extraction method based on wavelet singularity entropy | |
| Wu et al. | Statistical significance test of intrinsic mode functions |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |