CN112230275A - Seismic waveform identification method and device and electronic equipment - Google Patents
Seismic waveform identification method and device and electronic equipment Download PDFInfo
- Publication number
- CN112230275A CN112230275A CN202010961181.XA CN202010961181A CN112230275A CN 112230275 A CN112230275 A CN 112230275A CN 202010961181 A CN202010961181 A CN 202010961181A CN 112230275 A CN112230275 A CN 112230275A
- Authority
- CN
- China
- Prior art keywords
- seismic
- data set
- neural network
- network model
- data
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 56
- 238000003062 neural network model Methods 0.000 claims abstract description 88
- 238000012549 training Methods 0.000 claims abstract description 17
- 238000007781 pre-processing Methods 0.000 claims abstract description 14
- 238000012545 processing Methods 0.000 claims description 41
- 238000004422 calculation algorithm Methods 0.000 claims description 19
- 238000010606 normalization Methods 0.000 claims description 17
- 238000000605 extraction Methods 0.000 claims description 14
- 238000004458 analytical method Methods 0.000 claims description 13
- 230000008569 process Effects 0.000 claims description 13
- 238000006243 chemical reaction Methods 0.000 claims description 12
- 238000013528 artificial neural network Methods 0.000 claims description 11
- 238000012935 Averaging Methods 0.000 claims description 10
- 238000001914 filtration Methods 0.000 claims description 10
- 238000005070 sampling Methods 0.000 claims description 10
- 238000003860 storage Methods 0.000 claims description 9
- 238000013519 translation Methods 0.000 claims description 8
- 238000013507 mapping Methods 0.000 claims description 6
- 230000008447 perception Effects 0.000 claims description 6
- 238000004806 packaging method and process Methods 0.000 claims description 5
- 238000011176 pooling Methods 0.000 claims description 5
- 238000012856 packing Methods 0.000 claims 1
- 230000009467 reduction Effects 0.000 abstract description 5
- 230000002265 prevention Effects 0.000 abstract description 2
- 238000005422 blasting Methods 0.000 description 26
- 239000011435 rock Substances 0.000 description 13
- 230000033001 locomotion Effects 0.000 description 12
- 238000009826 distribution Methods 0.000 description 8
- 230000006870 function Effects 0.000 description 8
- 230000000694 effects Effects 0.000 description 7
- 238000004364 calculation method Methods 0.000 description 6
- 230000005284 excitation Effects 0.000 description 6
- 238000001228 spectrum Methods 0.000 description 6
- 230000008901 benefit Effects 0.000 description 5
- 238000011161 development Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000004880 explosion Methods 0.000 description 4
- 238000012544 monitoring process Methods 0.000 description 4
- 208000035126 Facies Diseases 0.000 description 3
- 230000004913 activation Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 238000010276 construction Methods 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 238000005065 mining Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000006835 compression Effects 0.000 description 2
- 238000007906 compression Methods 0.000 description 2
- 238000004590 computer program Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 239000003673 groundwater Substances 0.000 description 2
- 239000002689 soil Substances 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 239000002352 surface water Substances 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 208000007356 Fracture Dislocation Diseases 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000009395 breeding Methods 0.000 description 1
- 230000001488 breeding effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000002247 constant time method Methods 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 239000002360 explosive Substances 0.000 description 1
- 229910052602 gypsum Inorganic materials 0.000 description 1
- 239000010440 gypsum Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 230000008595 infiltration Effects 0.000 description 1
- 238000001764 infiltration Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 239000002243 precursor Substances 0.000 description 1
- 230000010349 pulsation Effects 0.000 description 1
- 238000012797 qualification Methods 0.000 description 1
- 238000012857 repacking Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 238000010008 shearing Methods 0.000 description 1
- 230000035939 shock Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 230000001960 triggered effect Effects 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
- G01V1/28—Processing seismic data, e.g. analysis, for interpretation, for correction
- G01V1/288—Event detection in seismic signals, e.g. microseismics
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Abstract
The embodiment of the invention discloses a method and a device for identifying seismic waveforms and electronic equipment, wherein the identification method comprises the following steps: obtaining a seismic waveform data set according to historical seismic observation data, and preprocessing the seismic waveform data set to obtain a first data set, a second data set, a third data set and a fourth data set; training according to the first data set, the second data set, the third data set and the fourth data set to obtain a first neural network model, a second neural network model, a third neural network model and a fourth neural network model; obtaining a seismic waveform to be identified according to the data of the target seismic event; and identifying the seismic waveform to be identified according to the first neural network model, the second neural network model, the third neural network model and the fourth neural network model to obtain the final comprehensive qualitative probability of the seismic event. The method can quickly and accurately identify the seismic waveform property, and plays an important role in guaranteeing disaster prevention and reduction early warning.
Description
Technical Field
The embodiment of the invention relates to the fields of seismic monitoring, geological exploration, mine safety monitoring, industrial production safety monitoring and the like, in particular to a method and a device for identifying seismic waveforms and electronic equipment.
Background
In recent years, with the development of digital earthquake observation technology, the wide application of a broadband digital seismograph greatly improves the recording quality of earthquake waveforms, and the modern digital seismograph provides digital earthquake waveform data which has higher sampling rate and resolution, wide recording frequency band and large dynamic range and is convenient for data processing by a computer, thereby expanding the research range of scientific researchers.
Many researchers have conducted extensive research into the mechanisms of mining-induced surface subsidence and deformation, and have achieved some beneficial results. The roof surrounding rock is intermittently and upwards caved along with the underground mining, and the surrounding rock deformation presents different deformation characteristics in different areas; horizontal tectonic stress is the driving force for rock mass deformation, and structural weaknesses in the rock mass change the rock mass deformation direction and cause the earth surface deformation range to be expanded farther. However, surface subsidence is not only related to the above factors, but also closely linked to groundwater activity and roof lithology. Underground mining causes the drainage of underground water, and surface water infiltration not only weakens the rock mass strength, but also changes the underground water runoff condition. Karst development, gypsum softening, weathering zones, etc. control the deformation and destruction of the rock mass, and joint cracks in the rock mass make the rock mass the weakest part to be destroyed first. Statistics show that 30 major and middle cities and 420 counties and cities in China are in high-risk areas with ground collapse, and 40 mines, 25 railway lines and several hundred reservoirs suffer from karst land collapse for a long time. About 70% of karst ground collapse disasters which occur in China are induced by human activities. Excess production of groundwater and mine drainage is a major cause of karst surface subsidence. Other conditions such as surface water impoundment, geotechnical engineering construction, railway and highway construction, engineering blasting and the like can also induce karst ground collapse. Therefore, the karst ground collapse disasters often occur in cities, mines or traffic lines with dense population, which bring serious influence and threat to national economic construction and people's lives and properties.
Because the vibration waves in the same or similar frequency ranges are excited by explosion, earthquake, even collapse and the like, the vibration waves picked and recorded by the actual earthquake observation station reflect various excitation source information of the explosion, the earthquake, the collapse and the like, the excitation sources are identified by recording the waves, the breeding source information is extracted, the types of the excitation sources of the earthquake, the explosion and the like are accurately distinguished and identified, the classification and analysis of earthquake emergency and earthquake data are carried out, regional earthquake activities are evaluated, and the reaction spectrum characteristics of regional earthquake engineering and earthquake prediction are very important to know.
In the aspect of earthquake prediction, because the occurrence of artificial blasting and collapse can generate earthquake waveforms recorded by a broadband digital seismograph, after the two earthquake waves are recorded by different stations, the events are often mistakenly judged as earthquake precursor abnormity due to imperfect quantitative indexes, the accuracy of earthquake prediction work is greatly influenced, and the accuracy and the qualification of the earthquake events, particularly tiny earthquake events are very important.
A plurality of seismologists use various effective earthquake mode identification and earthquake phase identification criteria to realize primary earthquake event judgment on a computer, but the final result of the judgment mostly depends on later manual intervention, the judgment depends on the experience and the intelligence of people to a great extent, a set of automatic identification system which is completely separated from the manual intervention is not formed, the earthquake events such as explosion, collapse, mine earthquake, landslide and the like are found and screened from a large number of earthquake events, and especially under the condition of small earthquake magnitude, the judgment of waveform characteristics is seriously influenced by the subjective error factors of people.
Disclosure of Invention
The embodiment of the invention aims to provide a method and a device for identifying seismic waveforms and electronic equipment, which are used for solving the problems that the existing method for judging the seismic property through manual intervention is influenced by subjective factors, and the existing method is low in efficiency and accurate.
In order to achieve the above object, the embodiments of the present invention mainly provide the following technical solutions:
in a first aspect, an embodiment of the present invention provides a method for identifying a seismic waveform, including: obtaining a seismic waveform data set according to historical seismic observation data; preprocessing the seismic waveform data set to obtain a first data set, a second data set, a third data set and a fourth data set, wherein the second data set is the mapping of the first data set in a frequency domain; training according to the first data set, the second data set, the third data set and the fourth data set to obtain a first neural network model, a second neural network model, a third neural network model and a fourth neural network model; the first neural network model and the second neural network model are respectively fitted models of a network structure in a sequence form extracted by using single convolution operation in time domain and frequency domain data, the third neural network model is a fitted model of a multilayer global pooled two-dimensional convolution structure in time domain data, and the fourth neural network model is a fitted model of a multilayer perception neural network based on characteristics in the time domain data; obtaining a seismic waveform to be identified according to data of a target seismic event, identifying the seismic waveform to be identified respectively through the first neural network model, the second neural network model, the third neural network model and the fourth neural network model to obtain four subentry qualitative probabilities of the target seismic event, and obtaining a final comprehensive qualitative probability of the target seismic event through an integrated judgment algorithm according to the four subentry qualitative probabilities.
According to an embodiment of the invention, the obtaining of the seismic waveform data set from the historical seismic observation data comprises: performing analog-to-digital conversion and sampling packaging on historical seismic observation data to obtain a plurality of continuous seismic waveforms; performing data processing according to the plurality of continuous seismic waveforms to obtain a plurality of waveform segments; and after event deduplication is carried out on the plurality of waveform segments, the seismic waveform data set is obtained by carrying out batch processing through a parallel processing algorithm.
According to one embodiment of the invention, the preprocessing includes data set splitting, normalization, filtering, data enhancement, feature extraction, and time-frequency analysis.
According to one embodiment of the invention, the normalization process includes averaging, derotation, detrending, extremum removal, averaging missing values, and whole value normalization.
According to one embodiment of the invention, the data enhancement process includes at least one of waveform rotation, P-wave perturbation, random noise, back-and-forth translation, and random window.
In a second aspect, an embodiment of the present invention further provides an apparatus for identifying a seismic waveform, including: the acquisition module is used for acquiring historical seismic observation data and acquiring data of a target seismic event; the control processing module is used for obtaining a seismic waveform data set according to historical seismic observation data, preprocessing the seismic waveform data set to obtain a first data set, a second data set, a third data set and a fourth data set, and then training according to the first data set, the second data set, the third data set and the fourth data set to obtain a first neural network model, a second neural network model, a third neural network model and a fourth neural network model; the control processing module is also used for obtaining a seismic waveform to be identified according to the seismic event data; respectively identifying the seismic waveform to be identified through the first neural network model, the second neural network model, the third neural network model and the fourth neural network model to obtain the subentry qualitative probability of the target seismic event, and outputting the final qualitative probability of the seismic event waveform to be identified through an integrated judgment algorithm according to the four subentry qualitative probabilities; an output module for passing the final integrated qualitative probability; the second data set is a mapping of the first data set in a frequency domain, the first neural network model and the second neural network model are models which are respectively fitted in time domain data and frequency domain data by using a network structure in a single convolution operation extraction sequence form, the third neural network model is a multilayer global pooling two-dimensional convolution structure, and the fourth neural network model is a feature-based multilayer perception neural network.
According to an embodiment of the invention, the control processing module is specifically configured to perform analog-to-digital conversion and sampling and packaging on historical seismic observation data to obtain a plurality of continuous seismic waveforms, perform data processing according to the plurality of continuous seismic waveforms to obtain a plurality of waveform segments, perform event deduplication on the plurality of waveform segments, and perform batch processing through a parallel processing algorithm to obtain the seismic waveform data set.
According to one embodiment of the invention, the preprocessing includes data set splitting, normalization, filtering, data enhancement, feature extraction, and time-frequency analysis.
According to one embodiment of the invention, the normalization process includes averaging, derotation, detrending, extremum removal, averaging missing values, and whole value normalization.
According to one embodiment of the invention, the data enhancement process includes at least one of waveform rotation, P-wave perturbation, random noise, back-and-forth translation, and random window.
In a third aspect, an embodiment of the present invention further provides an electronic device, including: at least one processor and at least one memory; the memory is to store one or more program instructions; the processor is configured to execute one or more program instructions to perform the method of identifying seismic waveforms according to the first aspect.
In a fourth aspect, embodiments of the present invention also provide a computer-readable storage medium containing one or more program instructions for executing the method for identifying seismic waveforms according to the first aspect.
The technical scheme provided by the embodiment of the invention at least has the following advantages:
the seismic waveform identification method, the seismic waveform identification device and the electronic equipment provided by the embodiment of the invention can be used for extracting the seismic observation event waveform in real time, automatically preprocessing the seismic observation event waveform, quickly identifying the seismic event waveform property, judging the seismic event waveform property as natural earthquake, blasting or collapse, wherein the identification time is within 1 minute, the average accuracy is about 95%, the manual identification needs 30-60 minutes, and the average accuracy is about 70%.
The method has the advantages that the multistage neural network structure joint identification with different properties is adopted, the method is not common in the field, and the accuracy, reliability and interpretability of seismic event waveform identification are remarkably improved. In earthquake observation, the smaller the earthquake magnitude is, the harder the earthquake type property identification is, and generally, when the earthquake magnitude is below 0.6 level, human eyes are difficult to distinguish the earthquake type basically, but on the technical system applying the invention, when the earthquake magnitude is-0.2 level, the identification accuracy of the earthquake type can be up to more than 80%.
By applying the technical system of the invention, the conclusion of studying and judging the nature of the earthquake event can be quickly butted with other information issuing systems, basically has no time delay, and plays an important role in guaranteeing disaster prevention, reduction and early warning.
Drawings
Fig. 1 is a flowchart of a seismic waveform identification method according to an embodiment of the present invention.
Fig. 2 is a block diagram of a structure of a seismic waveform recognition apparatus according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention is provided for illustrative purposes, and other advantages and effects of the present invention will become apparent to those skilled in the art from the present disclosure.
In the following description, for purposes of explanation and not limitation, specific details are set forth such as particular system structures, interfaces, techniques, etc. in order to provide a thorough understanding of the present invention. It will be apparent, however, to one skilled in the art that the present invention may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, circuits, and methods are omitted so as not to obscure the description of the present invention with unnecessary detail.
In the description of the present invention, it is to be understood that the terms "first" and "second" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance.
In the description of the present invention, it is to be noted that, unless otherwise explicitly specified or limited, the terms "connected" and "connected" are to be interpreted broadly, e.g., as meaning directly connected or indirectly connected through an intermediate. The specific meanings of the above terms in the present invention can be understood in specific cases to those skilled in the art.
Fig. 1 is a flowchart of a seismic waveform identification method according to an embodiment of the present invention. As shown in fig. 1, the method for identifying a seismic waveform according to an embodiment of the present invention includes:
s1: and obtaining a seismic waveform data set according to the historical seismic observation data. Wherein the historical seismic observation data may be observation data of seismic events occurring in the last N years.
In one embodiment of the present invention, step S1 includes:
s1-1: and performing analog-to-digital conversion and sampling packaging on the historical seismic observation data to obtain a plurality of continuous seismic waveforms.
Specifically, the original data of the earthquake observation station generates basic data A/D conversion to the data collector to form an electric signal of earthquake waveform, and then the electric signal flows into the earthquake data flow server, and continuous earthquake waveform is formed after sampling and repacking service. In this embodiment, a data inflow process is monitored to obtain a seismic waveform data stream, and a long-short window comparison algorithm (STA/LTA) is used to detect a sudden change time probably triggered in a waveform, so as to obtain a continuous seismic waveform segment window. And (3) combining the automatic triggering service and the manual quick report cataloging result to obtain accurate P wave arrival time (the error is less than 1 second), and combining the seismic phase data and the cataloging data to obtain the accurate first arrival time of the seismic triggering.
S1-2: and performing data processing according to the plurality of continuous seismic waveforms to obtain a plurality of waveform segments.
Illustratively, a waveform segment of the precise seismic event of seismic event-2 seconds to +58 seconds of the plurality of consecutive seismic waveforms is truncated. For example, 100 data values per second (sampling rate 100Hz) are acquired, i.e. 60000 values per slice per pass.
S1-3: and after event deduplication is carried out on the plurality of waveform segments, batch processing is carried out through a parallel processing algorithm to obtain the seismic waveform data set.
Specifically, event deduplication is performed through manual cataloging data, the purpose of the step is to remove inaccurate training data in the initial stage by the most accurate means, and finally batch processing is performed through a parallel processing algorithm (multiprocessing), so that 107 ten thousand accurate seismic waveforms are obtained in total for training of the data recognition model.
S2: and preprocessing the seismic waveform data set to obtain a first data set, a second data set, a third data set and a fourth data set.
In an embodiment of the present invention, the preprocessing in step S2 includes data set splitting, normalization, filtering, data enhancement, feature extraction, and time-frequency analysis, and the specific steps are as follows:
s2-1: the data set splitting comprises: and splitting the seismic waveform data set according to the station, the channel code, the epicenter and the magnitude to obtain a plurality of seismic waveform data subsets so as to enhance the waveform characteristics.
Specifically, after obtaining the seismic waveform, the invention firstly splits the seismic waveform, and the technical parameters are as follows through splitting according to the station, the channel code, the epicenter distance and the seismic level:
and (3) grading the epicenter distance: 100KM, 300KM, 450KM, 500KM, 800KM or less;
grading the magnitude: mL1.0 or less, ML1.5 or less, ML2.5 or less, MLLess than 3.0, MLLess than 5.0, ML1.0-2.5，ML1.0-3.0, etc.
This example uses 3 methods of waveform feature enhancement, 1. waveform classification: i.e. the waveform of each channel is used as a unique sample for minimum granularity training. 2. Station lane dividing method: namely, the waveform of the same station and the waveform of the same seismic event are used as a superposition data set and used as a sample for training. 3. Waveform tandem method: namely three-element waveforms of the same station are connected in series to be used as a sample for training, and the waveform characteristics are enhanced.
S2-2: filtering the plurality of seismic waveform data subsets to remove interference and retain waveform data for characterizing earthquakes, industrial blasting and mine collapse.
Specifically, based on the characteristics of seismic waves, the frequency range of the seismic near shock is 1-20Hz, and for the minor seismic events less than 1.0 grade, the released energy is smaller, the frequency range is about 200-1500Hz, and the duration is less than 1 second. The vibration frequency of industrial blasting may vary between 0.5-200Hz, and studies have shown that the vibration frequency of the blast wave is mainly distributed between 1-13Hz, and the vibration frequency of the seismic wave is mainly concentrated between 1-6Hz, while other studies suggest that the peak frequency of blasting is distributed in the range of 5-7Hz, while the peak frequency of natural earthquakes is relatively large, in the range of 10-18 Hz. The peak frequency distribution range of the mine collapse is 2-4Hz, and a series of viewpoints that the average excellent period of blasting is 0.67s from different angles, which reflects a softer soil layer, the average excellent period of earthquake is 0.28s, which reflects a deeper rock layer and the like are provided. In general, for blasting, mine collapse and natural earthquakes, the energy of the high frequency components decays faster than the low frequencies. The natural earthquake has richer frequency and wider frequency domain than blasting and mine collapse.
The earthquake, industrial blasting, even mine collapse and the like excite vibration waves in the same or similar frequency ranges, the vibration waves picked up and recorded by the actual earthquake observation station reflect various excitation source information such as blasting, earthquake, collapse and the like, the excitation sources are identified through recording waves, the earthquake source information is extracted, the types of the excitation sources such as the earthquake and the blasting are accurately distinguished and identified, and the classification and analysis of earthquake emergency and earthquake data, the regional earthquake activity evaluation, the reaction spectrum characteristics of regional earthquake engineering and earthquake prediction are very important. In the embodiment, a combined filtering algorithm is adopted, and combined parameters such as 2Hz, 1-20Hz, 2-25Hz, 1-45Hz, no filtering and the like are used for filtering seismic waveforms, decomposing different data sets and providing calculation optimization for subsequent model training.
S2-3: and performing normalization processing on the filtered seismic waveform data subsets to enable seismic waveform data to have consistency.
Specifically, operations such as mean value removing, dip angle removing, trend removing, extreme value removing and the like are adopted, finally, the missing values are subjected to averaging processing, finally, full value normalization processing is carried out, and the data range is limited to be 0-1.
The method comprises the following steps of averaging to enable all dimensions of input data to be centered to be 0, and solving the problems that in subsequent neural network processing, under the condition that a characteristic value is large, weight is synchronously increased, the variation of an activation function is too small, gradient dissipation is caused during reverse propagation, the variation of parameters is small, and the fitting effect is poor.
The detrending and detrending enable the seismic waveform to eliminate the effects of the offset generated by the sensor in acquiring data on later calculations. In the operation of the step, pure seismic waveforms are obtained by response processing and suppression of instruments of different stations, and all seismic waveforms are compared under uniform migration.
The full-value normalization processing solves the problem that the distribution of each batch of training data of the neural network is different, the essence of the neural network is to find balance points in the distribution of a plurality of data sets and fit an optimal function, and if the distribution difference of each batch of data sets is too large, the convergence efficiency of the neural network function is greatly reduced.
S2-4: and performing data enhancement processing on the data seismic waveform data subsets to decompose more data subsets, and balancing the data subsets to enable all classification weights to be consistent.
In particular, because seismic grids have large deviations from the number of records for different seismic event types, the individual classification weights are made consistent in order to equalize the data set. The invention adopts a data enhancement technology to increase the number of data sets.
In one embodiment of the invention, the data enhancement process includes at least one of waveform rotation, P-wave perturbation, random noise, back-and-forth translation, and random window.
Wave-shaped rotation: the present embodiment adds a portion of the training data set by randomly tilting ± 5 degrees by rotating the original wave train by a certain angle so that the waveform looks like another new data set.
P-wave disturbance: the interference on the P wave train is realized by integrally circularly translating the arrival time position of the P wave of the original wave train. In this embodiment, the P-wave interference parameter is set to random 1-3 seconds.
Random noise: while the noise of the disturbance is randomly added to the training data set, thereby affecting the weight of the machine learning, the present embodiment adds the noise of ± 0.3 vertically to the array using the gaussian method.
Front-back translation: the series characteristic enhancement is carried out by a front-back translation integral wave train mode, and the translation parameter adopted by the embodiment is random for 1-3 seconds.
Random window: the data set is increased in a random card window mode, in order to ensure that key characteristics are omitted as little as possible, the window length set in the embodiment is 50 seconds, the length of the whole wave train is covered by 80%, and partial data sets are effectively increased.
S2-5: and performing feature extraction on the seismic waveform data subset after data enhancement to obtain potential waveform feature information in the seismic waveform.
Specifically, the present embodiment pre-extracts waveform features, and mainly extracts the following waveform feature information:
initial movement symbol: the initial motion direction of the natural earthquake P wave is uncertain, and the blasting recording waveform P wave is upward in initial motion, the initial motion waveform is strong and sharp, and the natural earthquake has long duration; the compression waves generated by artificial blasting are expansion waves without quadrant distribution, theoretically, P wave initial motion symbols are upward in the vertical direction of seismic records, the vertical components of the P waves of the compression waves generated by natural earthquakes are upward and downward in quadrant distribution, certain azimuth distribution characteristics exist, clear waveform records of 4 stations in different azimuths are used for identification, in an actual algorithm, extraction of medium and small seismic event initial motion is not easy, and the direction of 5 count values before P time after equalization is used as the initial motion direction in the embodiment.
And (3) period: the period of the seismic wave is one of the dynamic characteristics of the seismic wave, and the change of the dynamic characteristics of the seismic wave depends on a plurality of factors such as a seismic source mechanism of the earthquake and the structure and the property of a medium. Compared with the earthquake, the artificial blasting has the advantages of large earthquake period, shallow earthquake source, concentrated earthquake generating range, strong P wave group, strong and sharp initial motion of P waves and quick vibration attenuation; the natural earthquake has the characteristics of clear pulsation and high frequency, the dominant period generally increases gradually along with the time from the beginning to the beginning, and the S wave period is generally larger than the P wave period. The invention measures the period size by defining the extreme value distance of a detection value window.
Amplitude ratio: the artificial blasting mainly generates longitudinal waves, but shear waves may be generated due to the influence of a blasting mode, a complex propagation path and the like, so that the artificial blasting has a strong P-wave group, and the S-wave is relatively weak. In natural earthquakes, shearing and dislocation of rocks are mainly used, so that most earthquakes can generate strong S waves. Because the ground vertical vibration acceleration, the vibration speed and the like of the artificial blasting near-source region are larger than the horizontal motion, the S wave amplitude is smaller than the P wave amplitude, namely As/AP < l, the attenuation of the near-source region is very fast, and generally the P wave and the S wave almost arrive at the same time. The natural earthquake amplitude ratio in Fujian area is 2.8-4.8 on average, the blasting is close to 1, 0.8-1.1 on average, and the ratio is related to the area difference and the earthquake source mechanism. In the present embodiment, in the feature extraction, since the arrival time of the S-wave is already obtained from the cataloged data, it is only necessary to intercept the same-length wave band as the P-wave train in the S-wave train, and obtain the feature by calculating the absolute amplitude ratio of the P-wave train and the S-wave train.
Seismic phase characteristics: the natural earthquake is generated by rock fracture dislocation, the depth of a seismic source is generally much deeper than artificial blasting, and the propagation path of the earthquake waves is reflected and refracted by a multilayer medium, so that the components of the earthquake waves recorded by the station are rich. The artificial blasting source is small in size, can be regarded as an instant expansion source and is mostly generated on the earth surface, the wave propagation path is single, and the seismic facies are simple. In the aspect of extracting the seismic facies features, the complexity of the seismic facies is measured by setting the number of inflection points of the wave train, and generally, the wave train complexity of a natural earthquake is obviously higher than that of artificial blasting.
Attenuation characteristics: because the artificial blasting occurs on the ground surface, a relatively long path is in the shallow layer (soil layer) of the ground surface in the propagation process, the medium density is low, the energy loss is large, the attenuation is very fast along with the distance, the attenuation is in meters, and the blasting duration is about several seconds generally (the explosive quantity is small and the blasting center distance is close, and is generally not more than 0.55). While natural earthquakes generally occur from underground to tens of kilometers, the wave propagation path is mainly in the rock stratum, the energy loss is small, the attenuation is slow, the waveform duration is long, and the time duration is generally more than tens of seconds. In the invention, because a waveform segment with a fixed length is set, the wave train attenuation condition is measured by extracting the average value of the last 200 bits (2 seconds) of the wave train.
Surface wave development: from the surface wave development condition, the artificial blasting seismic source is shallow, the prominent characteristic is that the surface wave is developed in a short period, and obvious surface waves can be observed in the record. The natural earthquake has two direct longitudinal waves and transverse waves Pg and Sg within 200km of the seismic distance, the surface waves generally do not develop, and when the seismic distance is smaller, the surface waves cannot be distinguished. In the present invention, the feature is set to a boolean value, that is, whether the data is a plane wave sequence is determined by the difference of the last 5 seconds of the transverse quantity, if the data in the segment has obvious directivity, the existence of the plane wave band is considered (sequence angle >5 degrees), and the characteristic boolean value is set to true.
S2-6: and performing time-frequency analysis on the seismic waveform data subset according to the extracted waveform characteristic information to obtain a first data set, a second data set, a third data set and a fourth data set.
In recent years, with the richness and perfection of seismic motion records, methods such as a short-time technology, Cohen time-frequency analysis, wavelet analysis and the like are also applied to the aspect of seismic time-frequency analysis, but a plurality of defects also occur, such as window leakage, frequency shift, cross terms, noise and the like, which are inaccurate to remove information and mix in signals, and are not beneficial to knowing the real characteristics of frequency spectrums.
In the aspect of non-stationary research of a frequency domain, Fourier spectrums and power spectrums are adopted to describe the spectrum characteristics of earthquake motion, the distribution or energy proportion of different frequencies in the earthquake motion process can be directly reflected, the physical significance is clear, the application is wide, and the algorithm is convenient to realize. In order to identify more seismic waveform characteristics, the present embodiment performs frequency domain conversion on the waveform sequence observed in the time domain, and adopts a CT method to calculate one-dimensional discrete fourier transform in the frequency domain conversion process, so as to obtain a spectrogram with the same sequence as a new data set for analysis and calculation.
Through the above data conversion process, 4 data sets for training learning, i.e., a first data set, a second data set, a third data set, and a fourth data set, are finally determined. Wherein the second data set is a mapping of the first data set in the frequency domain.
S3: and training according to the first data set, the second data set, the third data set and the fourth data set to obtain a first neural network model, a second neural network model, a third neural network model and a fourth neural network model. The first neural network model and the second neural network model are models which are respectively fitted in time domain data and frequency domain data by using a network structure in a single convolution operation extraction sequence form, the third neural network model is a fitting model of a multilayer global pooling two-dimensional convolution structure in the time domain data, and the fourth neural network model is a fitting model of a multilayer perception neural network based on characteristics in the time domain data.
Specifically, in the seismic observation of the current seismic table network, a seismometer adopts a sampling rate of 100Hz, according to error evaluation, the error of the current seismic moment does not exceed 1 second, a time window is obtained by about 5 seconds before and 95 seconds after the theoretical seismic moment, and 10000-dimensional vector input is generated on the assumption that calculation is carried out according to the length of the 100 second window. In the embodiment, the calculation amount is reduced through a dimensionality reduction algorithm, and the purpose of qualifying the seismic event waveform is finally achieved through characteristic dimensionality reduction or multilayer convolution dimensionality reduction and softmax classification.
In the embodiment, three network structures are mainly adopted, if no special description is given, the loss function adopted by the network structure is a cross entropy method, and the activation function adopts a linear rectification method.
Wherein, the network structure 1 (the structure of the first neural network model and the second neural network model): the waveform characteristics based on the sequence form are extracted through single convolution operation, the whole structure only uses one convolution operation, the calculation consumption is low, the convolution operation is rapidly pooled, the data dimensionality is further reduced, the learning robustness of the network structure is increased through flattening and discarding, finally, classification output is carried out on the softmax function through two full-connection hidden layers, and the whole network structure is simple. Experiments show that the network structure 1 has strong stability and recognition accuracy.
Network structure 2 (structure of third neural network model): the network structure uses a VGG16 model as a reference, and is mainly characterized in that the fact that a very small convolution kernel is used is proved, the network depth is increased, meanwhile, the efficiency of the model is effectively improved, and the structure has strong generalization capability. The invention greatly reduces the number of the model convolution operation layers, reduces the operation node parameters and obtains better experiment results.
Network structure 3 (structure of the fourth neural network model): the network adopts a multilayer sensing structure, adopts more fully-connected hidden layers, has higher computational efficiency, adopts a mode of combining seismic wave train characteristics and extraction characteristics on a data set, reduces dimensions through the multiple fully-connected hidden layers, adopts ReLu as an activation function, and can achieve more than 90% of precision on data set identification.
S4: obtaining a seismic waveform to be recognized according to data of a target seismic event, respectively recognizing the seismic waveform to be recognized through a first neural network model, a second neural network model, a third neural network model and a fourth neural network model to obtain four subentry qualitative probabilities of the target seismic event, and obtaining a final qualitative probability of the target seismic event through an integrated judgment method according to the four subentry qualitative probabilities.
Specifically, the corresponding seismic waveform is automatically intercepted by monitoring data transmission of a seismic measurement technology system data stream server and automatic triggering of a seismic event or manual seismic triggering alarm.
After the seismic waveform to be recognized is obtained, after the waveform processing of the seismic event, the first neural network model, the second neural network model, the third neural network model and the fourth neural network model are called to recognize and judge the seismic waveform, the four seismic recognition models all output the qualitative probability of the event according to rules, and the final probability of the seismic event is judged through a designed integrated judgment algorithm.
An integrated judgment algorithm:
where P is the final comprehensive qualitative probability, PiFor fractional qualitative probability, diIs a fractional weight.
Fig. 2 is a block diagram of a structure of a seismic waveform recognition apparatus according to an embodiment of the present invention. As shown in fig. 2, an apparatus for identifying a seismic waveform according to an embodiment of the present invention includes: an acquisition module 100, a control processing module 200 and an output module 300.
The obtaining module 100 is configured to obtain historical seismic observation data and obtain data of a target seismic event. The control processing module 200 is configured to obtain a seismic waveform data set according to historical seismic observation data, perform preprocessing on the seismic waveform data set to obtain a first data set, a second data set, a third data set, and a fourth data set, and then perform training according to the first data set, the second data set, the third data set, and the fourth data set to obtain a first neural network model, a second neural network model, a third neural network model, and a fourth neural network model. The control processing module 200 is further configured to obtain a seismic waveform to be identified according to the data of the target seismic event; the seismic waveform to be recognized is recognized through the first neural network model, the second neural network model, the third neural network model and the fourth neural network model respectively to obtain four subentry qualitative probabilities of the target seismic event, and the final qualitative probability of the target seismic event is obtained through an integrated judgment algorithm according to the four subentry qualitative probabilities. The output module 300 is used for outputting the final comprehensive qualitative probability of the target seismic event. The second data set is the mapping of the first data set in a frequency domain, the first neural network model and the second neural network model are models which are respectively fitted in time domain and frequency domain data by using a network structure of a single convolution operation extraction sequence form, the third neural network model is a multilayer global pooling two-dimensional convolution structure, and the fourth neural network model is a multilayer perception neural network based on characteristics.
In an embodiment of the present invention, the control processing module 200 is specifically configured to perform analog-to-digital conversion and sampling and packaging on historical seismic observation data to obtain a plurality of continuous seismic waveforms, perform data processing according to the plurality of continuous seismic waveforms to obtain a plurality of waveform segments, further perform event deduplication on the plurality of waveform segments, and perform batch processing through a parallel processing algorithm to obtain a seismic waveform data set.
In one embodiment of the invention, the preprocessing includes data set splitting, normalization, filtering, data enhancement, feature extraction, and time-frequency analysis.
In one embodiment of the invention, the normalization process includes averaging, derotation, detrending, extremum removal, averaging missing values, and whole value normalization.
In one embodiment of the invention, the data enhancement process includes at least one of waveform rotation, P-wave perturbation, random noise, back-and-forth translation, and random window.
It should be noted that the specific implementation of the seismic waveform identification device in the embodiment of the present invention is similar to the specific implementation of the seismic waveform identification method in the embodiment of the present invention, and specific reference is specifically made to the description of the seismic waveform identification method, and details are not repeated for reducing redundancy.
In addition, other configurations and functions of the seismic waveform recognition device according to the embodiment of the present invention are known to those skilled in the art, and are not described in detail to reduce redundancy.
An embodiment of the present invention further provides an electronic device, including: at least one processor and at least one memory; the memory is to store one or more program instructions; the processor is configured to execute one or more program instructions to perform the method of identifying seismic waveforms according to the first aspect.
The disclosed embodiments of the present invention provide a computer-readable storage medium having stored therein computer program instructions that, when run on a computer, cause the computer to perform the above-described method of identification of seismic waveforms.
In an embodiment of the invention, the processor may be an integrated circuit chip having signal processing capability. The Processor may be a general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic device, discrete hardware component.
The various methods, steps and logic blocks disclosed in the embodiments of the present invention may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of the method disclosed in connection with the embodiments of the present invention may be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software module may be located in ram, flash memory, rom, prom, or eprom, registers, etc. storage media as is well known in the art. The processor reads the information in the storage medium and completes the steps of the method in combination with the hardware.
The storage medium may be a memory, for example, which may be volatile memory or nonvolatile memory, or which may include both volatile and nonvolatile memory.
The nonvolatile Memory may be a Read-Only Memory (ROM), a Programmable ROM (PROM), an Erasable PROM (EPROM), an Electrically Erasable PROM (EEPROM), or a flash Memory.
The volatile Memory may be a Random Access Memory (RAM) which serves as an external cache. By way of example and not limitation, many forms of RAM are available, such as Static random access memory (Static RAM, SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (ddr Data Rate SDRAM), Enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM), and Direct Rambus RAM (DRRAM).
The storage media described in connection with the embodiments of the invention are intended to comprise, without being limited to, these and any other suitable types of memory.
Those skilled in the art will appreciate that the functionality described in the present invention may be implemented in a combination of hardware and software in one or more of the examples described above. When software is applied, the corresponding functionality may be stored on or transmitted over as one or more instructions or code on a computer-readable medium. Computer-readable media includes both computer storage media and communication media including any medium that facilitates transfer of a computer program from one place to another. A storage media may be any available media that can be accessed by a general purpose or special purpose computer.
The above-mentioned embodiments, objects, technical solutions and advantages of the present invention are further described in detail, it should be understood that the above-mentioned embodiments are only exemplary embodiments of the present invention, and are not intended to limit the scope of the present invention, and any modifications, equivalent substitutions, improvements and the like made on the basis of the technical solutions of the present invention should be included in the scope of the present invention.
Claims (10)
1. A method of identifying seismic waveforms, comprising:
obtaining a seismic waveform data set according to historical seismic observation data;
preprocessing the seismic waveform data set to obtain a first data set, a second data set, a third data set and a fourth data set, wherein the second data set is the mapping of the first data set in a frequency domain;
training according to the first data set, the second data set, the third data set and the fourth data set to obtain a first neural network model, a second neural network model, a third neural network model and a fourth neural network model; the first neural network model and the second neural network model are respectively fitted models of network structures in a time domain and frequency domain data, wherein the network structures are extracted in a sequence form by using single convolution operation; the third neural network model is a fitting model of a multilayer global pooling two-dimensional convolution structure in time domain data, and the fourth neural network model is a fitting model of a multilayer perception neural network based on characteristics in the time domain data;
obtaining a seismic waveform to be identified according to data of a target seismic event, identifying the seismic waveform to be identified respectively through the first neural network model, the second neural network model, the third neural network model and the fourth neural network model to obtain four subentry qualitative probabilities of the target seismic event, and obtaining a final comprehensive qualitative probability of the target seismic event through an integrated judgment algorithm according to the four subentry qualitative probabilities.
2. A method of identifying seismic waveforms as claimed in claim 1, wherein said deriving a seismic waveform data set from historical seismic observation data comprises:
performing analog-to-digital conversion and sampling packaging on historical seismic observation data to obtain a plurality of continuous seismic waveforms;
performing data processing according to the plurality of continuous seismic waveforms to obtain a plurality of waveform segments;
and after event deduplication is carried out on the plurality of waveform segments, the seismic waveform data set is obtained by carrying out batch processing through a parallel processing algorithm.
3. The method of identifying seismic waveforms of claim 1, wherein said preprocessing comprises data set splitting, normalization, filtering, data enhancement, feature extraction, and time-frequency analysis.
4. The method of identifying seismic waveforms of claim 3, wherein said normalization process comprises de-averaging, de-dip, de-trending, de-extremum, averaging missing values, and full-value normalization.
5. The method of identifying seismic waveforms of claim 3, wherein said data enhancement processing includes at least one of waveform rotation, P-wave perturbation, random noise, back-and-forth translation, and random windowing.
6. An apparatus for identifying seismic waveforms, comprising:
the acquisition module is used for acquiring historical seismic observation data and acquiring data of a target seismic event;
the control processing module is used for obtaining a seismic waveform data set according to historical seismic observation data, preprocessing the seismic waveform data set to obtain a first data set, a second data set, a third data set and a fourth data set, and then training according to the first data set, the second data set, the third data set and the fourth data set to obtain a first neural network model, a second neural network model, a third neural network model and a fourth neural network model; the control processing module is also used for obtaining a seismic waveform to be identified according to the seismic event data; respectively identifying the seismic waveform to be identified through the first neural network model, the second neural network model, the third neural network model and the fourth neural network model to obtain four subentry qualitative probabilities of the target seismic event, and obtaining a final qualitative probability of the target seismic event through an integrated judgment algorithm according to the four subentry qualitative probabilities;
an output module for passing the final integrated qualitative probability;
the second data set is a mapping of the first data set in a frequency domain, the first neural network model and the second neural network model are models which are respectively fitted in time domain and frequency domain data by using a network structure of a single convolution operation extraction sequence form, the third neural network model is a multilayer global pooling two-dimensional convolution structure, and the fourth neural network model is a feature-based multilayer perception neural network.
7. The seismic waveform recognition device of claim 6, wherein the control processing module is specifically configured to perform analog-to-digital conversion and sampling packing on historical seismic observation data to obtain a plurality of continuous seismic waveforms, perform data processing according to the plurality of continuous seismic waveforms to obtain a plurality of waveform segments, perform event deduplication on the plurality of waveform segments, and perform batch processing through a parallel processing algorithm to obtain the seismic waveform data set.
8. The apparatus for seismic waveform identification as claimed in claim 6 wherein said preprocessing comprises data set splitting, normalization, filtering, data enhancement, feature extraction and time-frequency analysis.
9. An electronic device, characterized in that the electronic device comprises: at least one processor and at least one memory;
the memory is to store one or more program instructions;
the processor, operable to execute one or more program instructions to perform the method of identifying seismic waveforms of any of claims 1-5.
10. A computer readable storage medium having one or more program instructions embodied therein for performing the method of identifying seismic waveforms of any of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010961181.XA CN112230275B (en) | 2020-09-14 | 2020-09-14 | Method and device for identifying seismic waveform and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010961181.XA CN112230275B (en) | 2020-09-14 | 2020-09-14 | Method and device for identifying seismic waveform and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112230275A true CN112230275A (en) | 2021-01-15 |
| CN112230275B CN112230275B (en) | 2024-01-26 |
Family
ID=74116231
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010961181.XA Active CN112230275B (en) | 2020-09-14 | 2020-09-14 | Method and device for identifying seismic waveform and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112230275B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112526606A (en) * | 2021-02-08 | 2021-03-19 | 南京云创大数据科技股份有限公司 | Seismic source type prediction method and system based on heterogeneous multi-classification model |
| CN112884134A (en) * | 2021-04-01 | 2021-06-01 | 山西云芯新一代信息技术研究院有限公司 | Time domain-based convolutional neural network model for seismic facies identification and application |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109086872A (en) * | 2018-07-30 | 2018-12-25 | 东北大学 | Seismic wave recognizer based on convolutional neural networks |
| WO2019040288A1 (en) * | 2017-08-25 | 2019-02-28 | Exxonmobil Upstream Researchcompany | Automated seismic interpretation using fully convolutional neural networks |
| CN109917457A (en) * | 2019-04-10 | 2019-06-21 | 苏州浪潮智能科技有限公司 | A kind of pick-up method and equipment of seismic first breaks point |
| CN110082822A (en) * | 2019-04-09 | 2019-08-02 | 中国科学技术大学 | The method for carrying out earthquake detection using convolutional neural networks |
| CN110308485A (en) * | 2019-07-05 | 2019-10-08 | 中南大学 | Microseismic signals classification method, device and storage medium based on deep learning |
| CN110488351A (en) * | 2019-08-15 | 2019-11-22 | 东北大学 | Seismic wave based on machine learning shakes property recognition methods |
-
2020
- 2020-09-14 CN CN202010961181.XA patent/CN112230275B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019040288A1 (en) * | 2017-08-25 | 2019-02-28 | Exxonmobil Upstream Researchcompany | Automated seismic interpretation using fully convolutional neural networks |
| CN109086872A (en) * | 2018-07-30 | 2018-12-25 | 东北大学 | Seismic wave recognizer based on convolutional neural networks |
| CN110082822A (en) * | 2019-04-09 | 2019-08-02 | 中国科学技术大学 | The method for carrying out earthquake detection using convolutional neural networks |
| CN109917457A (en) * | 2019-04-10 | 2019-06-21 | 苏州浪潮智能科技有限公司 | A kind of pick-up method and equipment of seismic first breaks point |
| CN110308485A (en) * | 2019-07-05 | 2019-10-08 | 中南大学 | Microseismic signals classification method, device and storage medium based on deep learning |
| CN110488351A (en) * | 2019-08-15 | 2019-11-22 | 东北大学 | Seismic wave based on machine learning shakes property recognition methods |
Non-Patent Citations (2)
| Title |
|---|
| LOMAX ANTHONY 等: "An investigation of rapid earthquake characterization using single-station waveforms and a convolutional neural network", SEISMOLOGICAL RESEARCH LETTERS, vol. 90, no. 2, pages 517 - 529 * |
| 周本伟 等: "利用卷积神经网络检测地震的方法与优化", 地震学报, pages 1 - 20 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112526606A (en) * | 2021-02-08 | 2021-03-19 | 南京云创大数据科技股份有限公司 | Seismic source type prediction method and system based on heterogeneous multi-classification model |
| CN112884134A (en) * | 2021-04-01 | 2021-06-01 | 山西云芯新一代信息技术研究院有限公司 | Time domain-based convolutional neural network model for seismic facies identification and application |
| CN112884134B (en) * | 2021-04-01 | 2022-04-26 | 山西云芯新一代信息技术研究院有限公司 | Time domain-based convolutional neural network model for seismic facies identification and application |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112230275B (en) | 2024-01-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Varotsos et al. | Natural time analysis: the new view of time: precursory seismic electric signals, earthquakes and other complex time series | |
| CA2743611C (en) | Methods and systems for monitoring and modeling hydraulic fracturing of a reservoir field | |
| Zeng et al. | Surface microseismic monitoring of hydraulic fracturing of a shale‐gas reservoir using short‐period and broadband seismic sensors | |
| CN112230275B (en) | Method and device for identifying seismic waveform and electronic equipment | |
| D’hour et al. | Detection of subtle hydromechanical medium changes caused by a small-magnitude earthquake swarm in NE Brazil | |
| Maity | Microseismicity analysis for HFTS pad and correlation with completion parameters | |
| Hafez et al. | Earthquake onset detection using spectro-ratio on multi-threshold time–frequency sub-band | |
| Rodríguez‐Pradilla et al. | Automated microseismic processing and integrated interpretation of induced seismicity during a multistage hydraulic‐fracturing stimulation, Alberta, Canada | |
| Petrosino et al. | Low frequency seismic source investigation in volcanic environment: The Mt. Vesuvius atypical case | |
| McNamara et al. | Site response in the eastern United States: A comparison of VS30 measurements with estimates from horizontal: vertical spectral ratios | |
| Brown et al. | Automatic infrasonic signal detection using the Hough transform | |
| Sović et al. | High frequency microseismic noise as possible earthquake precursor | |
| Mendecki et al. | Determination of the resonance frequency-thickness relation based on the ambient seismic noise records from Upper Silesia Coal Basin | |
| Lacazette et al. | Geomechanical and flow simulation of hydraulic fractures using high-resolution passive seismic images | |
| CN113075656A (en) | Method for comprehensively predicting coal mine working face fracture through ground penetrating radar and three-dimensional earthquake | |
| Rohnacher et al. | Source Mechanism of Seismic Explosion Signals at Santiaguito Volcano, Guatemala: New Insights From Seismic Analysis and Numerical Modeling | |
| Rabinovich et al. | Modeling of a reservoir fracture zone formed by hydraulic fracturing | |
| Shang et al. | Cluster analysis of the domain of microseismic event attributes for floor water inrush warning in the working face | |
| CN111915844B (en) | Method and device for evaluating coal rock stability by analyzing vibration signal through cepstrum coefficient | |
| Hu et al. | Discriminator of mining blasts and microseismic events based on multi-scale discrete wavelet transform | |
| CN105929445A (en) | Micro-fracture vector scanning method | |
| Núñez Escribano et al. | Recent seismicity at Ceboruco Volcano (Mexico) | |
| Scala et al. | Monitoring the Microseismicity through a Dense Seismic Array and a Similarity Search Detection Technique: Application to the Gas-Storage Site of Collalto, North Italy | |
| Mukuhira et al. | Introduction of natural fracture information to induced seismicity analysis | |
| CN114966845A (en) | Unsupervised clustering seismic waveform real-time classification detection method |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |