CN112231974A - TBM rock breaking seismic source seismic wave field characteristic recovery method and system based on deep learning - Google Patents
TBM rock breaking seismic source seismic wave field characteristic recovery method and system based on deep learning Download PDFInfo
- Publication number
- CN112231974A CN112231974A CN202011061619.5A CN202011061619A CN112231974A CN 112231974 A CN112231974 A CN 112231974A CN 202011061619 A CN202011061619 A CN 202011061619A CN 112231974 A CN112231974 A CN 112231974A
- Authority
- CN
- China
- Prior art keywords
- seismic
- wave field
- rock breaking
- deep learning
- seismic source
- 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
- 238000011084 recovery Methods 0.000 title claims abstract description 95
- 239000011435 rock Substances 0.000 title claims abstract description 88
- 238000000034 method Methods 0.000 title claims abstract description 62
- 238000013135 deep learning Methods 0.000 title claims abstract description 56
- 239000011159 matrix material Substances 0.000 claims abstract description 54
- 238000013528 artificial neural network Methods 0.000 claims abstract description 32
- 238000004088 simulation Methods 0.000 claims abstract description 27
- 238000007781 pre-processing Methods 0.000 claims abstract description 20
- 230000006870 function Effects 0.000 claims description 19
- 238000012549 training Methods 0.000 claims description 16
- 238000004590 computer program Methods 0.000 claims description 14
- 230000008569 process Effects 0.000 claims description 12
- 230000006835 compression Effects 0.000 claims description 10
- 238000007906 compression Methods 0.000 claims description 10
- 238000003860 storage Methods 0.000 claims description 9
- 238000011176 pooling Methods 0.000 claims description 6
- 238000013508 migration Methods 0.000 claims description 5
- 230000005012 migration Effects 0.000 claims description 5
- 238000005457 optimization Methods 0.000 claims 1
- 239000010410 layer Substances 0.000 description 65
- 238000011156 evaluation Methods 0.000 description 18
- 238000010586 diagram Methods 0.000 description 14
- 238000012545 processing Methods 0.000 description 10
- 230000000007 visual effect Effects 0.000 description 9
- 238000004364 calculation method Methods 0.000 description 8
- 239000013598 vector Substances 0.000 description 8
- 239000002356 single layer Substances 0.000 description 7
- 238000004422 calculation algorithm Methods 0.000 description 6
- 238000009826 distribution Methods 0.000 description 6
- 238000012360 testing method Methods 0.000 description 6
- 238000013441 quality evaluation Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 238000003475 lamination Methods 0.000 description 4
- 238000005070 sampling Methods 0.000 description 4
- 238000010200 validation analysis Methods 0.000 description 4
- 230000009977 dual effect Effects 0.000 description 3
- 238000013507 mapping Methods 0.000 description 3
- 230000005641 tunneling Effects 0.000 description 3
- RTAQQCXQSZGOHL-UHFFFAOYSA-N Titanium Chemical compound [Ti] RTAQQCXQSZGOHL-UHFFFAOYSA-N 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 2
- 238000010521 absorption reaction Methods 0.000 description 2
- 230000004913 activation Effects 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 2
- 230000000903 blocking effect Effects 0.000 description 2
- 238000005520 cutting process Methods 0.000 description 2
- 238000009795 derivation Methods 0.000 description 2
- 230000004907 flux Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000005381 potential energy Methods 0.000 description 2
- 238000002203 pretreatment Methods 0.000 description 2
- 230000007480 spreading Effects 0.000 description 2
- 238000003892 spreading Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000013526 transfer learning Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- 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
- G01V1/282—Application of seismic models, synthetic seismograms
-
- 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
- G01V1/30—Analysis
- G01V1/307—Analysis for determining seismic attributes, e.g. amplitude, instantaneous phase or frequency, reflection strength or polarity
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- 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
- G06F2111/00—Details relating to CAD techniques
- G06F2111/10—Numerical modelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/10—Noise analysis or noise optimisation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Remote Sensing (AREA)
- Mathematical Analysis (AREA)
- Evolutionary Computation (AREA)
- Computational Mathematics (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Databases & Information Systems (AREA)
- Algebra (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Acoustics & Sound (AREA)
- Environmental & Geological Engineering (AREA)
- Geology (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Geophysics (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Operations Research (AREA)
- Probability & Statistics with Applications (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Medical Informatics (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention belongs to the field of geophysical exploration and provides a method and a system for recovering seismic wave field characteristics of a TBM rock breaking seismic source based on deep learning. The TBM rock breaking seismic source seismic wave field feature recovery method based on deep learning comprises the following steps: acquiring a rock breaking seismic source original signal and a pilot signal to obtain TBM rock breaking seismic source numerical simulation data; converting numerical simulation data of a TBM rock breaking seismic source into a pulse seismic source seismic record after wave field characteristic recovery through a wave field characteristic recovery network based on deep learning; the wave field characteristic recovery network based on deep learning comprises a preprocessing layer and a deep neural network layer; the pretreatment layer is used for performing channel-by-channel convolution on the original signal of the rock breaking seismic source and the pilot signal and performing segmented superposition on the convolution output; the output of the pre-processing layer, the time information constraint matrix and the same-phase axis information constraint matrix form three-channel data, and then the three-channel data are input into the deep neural network layer.
Description
Technical Field
The invention belongs to the field of geophysical exploration, and particularly relates to a TBM rock breaking seismic source seismic wave field characteristic recovery method and system based on deep learning.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
The advanced detection method for a rock-breaking seismic source is a prediction method for detecting a front abnormal body by using rock-breaking vibration generated during tunneling of a Tunnel Boring Machine (TBM) as a seismic source. The method can simultaneously detect and quickly obtain a data processing result in the TBM tunneling process, so that the method is very suitable for the requirement of TBM quick tunneling. Because the rock-breaking vibration wave field has the characteristics of time continuity, uncontrollable performance and the like, effective reflection information is difficult to directly identify, and therefore the real wave field characteristics need to be recovered through interference processing. According to the knowledge of the inventor, the existing rock breaking seismic source seismic wave field recovery mainly adopts a cross-correlation interference method to recover continuous and disordered rock breaking signals into pulse signals similar to a pulse seismic source, and then subsequent data processing is carried out on the basis. However, new noise is introduced in the cross-correlation process, so that the effective signal cannot be accurately identified, and the problems of false abnormality, strong interference and the like are brought to subsequent imaging, which has become a bottleneck for restricting further popularization and application of the rock breaking seismic source method. Therefore, a new method capable of effectively solving the problem of seismic wave field characteristic recovery of a rock breaking seismic source is urgently needed.
In recent years, artificial intelligence technology is continuously broken through, wherein the deep learning method has strong nonlinear function fitting capacity, and possibility is provided for corresponding relation mining of a model space and a data space in the geophysical field. At present, better application effects are obtained on the problems of data-model inversion, signal identification, data processing and the like in the fields of seismic and electrical prospecting.
The inventor finds that the method for realizing deep learning seismic wave field characteristic recovery of the rock breaking seismic source has the following two problems:
(1) the applicability of deep learning in solving the wave field feature recovery problem. The existing network architecture in the deep learning field is mostly used for solving the problems of computer vision, voice recognition and natural language processing, and the problem is greatly different from the problem of rock breaking seismic source wave field information characteristic recovery. The data volume required for the wave field feature recovery training is significantly larger than that of the conventional deep learning problem, and the difference in the meaning of the features represented by the input data is large. Meanwhile, incompatibility exists between the time and space variation characteristics of the seismic data and the weight sharing attributes of the deep neural network parameters.
(2) The transition problem from simulation data to data wavefield feature recovery. The final purpose of the wave field characteristic recovery method is actual landing and application in a TBM tunnel driving site, a large amount of random environmental noise and irregular mechanical noise exist in actual engineering application on the site, the complexity of actual data is far higher than that of simulation data, and the retraining of network parameters for carrying out wave field recovery on numerical simulation data is far from sufficient.
Disclosure of Invention
In order to solve at least one of the above problems, a first aspect of the present invention provides a method and a system for recovering a seismic wave field characteristic of a rock-breaking seismic source of a TBM based on deep learning, in which a wave field time information matrix and a homophase axis information weighting matrix established according to seismic data characteristics are added to a network, so as to obtain a better effect of recovering the wave field characteristic of the rock-breaking seismic source.
In order to achieve the purpose, the invention adopts the following technical scheme:
the invention provides a method for recovering seismic wave field characteristics of a TBM rock breaking seismic source based on deep learning.
In one or more embodiments, a method for recovering characteristics of a seismic wave field of a TBM rock breaking source based on deep learning comprises the following steps:
acquiring a rock breaking seismic source original signal and a pilot signal to obtain TBM rock breaking seismic source numerical simulation data;
converting numerical simulation data of a TBM rock breaking seismic source into a pulse seismic source seismic record after wave field characteristic recovery through a wave field characteristic recovery network based on deep learning;
the wave field characteristic recovery network based on deep learning comprises a preprocessing layer and a deep neural network layer; the pretreatment layer is used for performing channel-by-channel convolution on the original signal of the rock breaking seismic source and the pilot signal and performing segmented superposition on the convolution output; the output of the pre-processing layer, the time information constraint matrix and the same-phase axis information constraint matrix form three-channel data, and then the three-channel data are input into the deep neural network layer.
The invention provides a TBM rock breaking seismic source seismic wave field characteristic recovery system based on deep learning.
In one or more embodiments, a deep learning based TBM rock breaking source seismic wavefield feature recovery system includes:
the signal acquisition module is used for acquiring a rock breaking source original signal and a pilot signal to obtain TBM rock breaking source numerical simulation data;
the characteristic recovery module is used for converting the TBM rock breaking seismic source numerical simulation data into a pulse seismic source seismic record after wave field characteristic recovery through a wave field characteristic recovery network based on deep learning;
the wave field characteristic recovery network based on deep learning comprises a preprocessing layer and a deep neural network layer; the pretreatment layer is used for performing channel-by-channel convolution on the original signal of the rock breaking seismic source and the pilot signal and performing segmented superposition on the convolution output; the output of the pre-processing layer, the time information constraint matrix and the same-phase axis information constraint matrix form three-channel data, and then the three-channel data are input into the deep neural network layer.
A third aspect of the invention provides a computer-readable storage medium.
In one or more embodiments, a computer-readable storage medium has stored thereon a computer program which, when executed by a processor, implements the steps in the deep learning-based TBM lithotripsy source seismic wavefield feature recovery method as described above.
A fourth aspect of the invention provides a computer apparatus.
In one or more embodiments, a computer device includes a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor executes the program to implement the steps of the deep learning based TBM rock breaking source seismic wavefield feature recovery method as described above.
Compared with the prior art, the invention has the beneficial effects that:
the invention provides a TBM rock breaking seismic source seismic wave field characteristic recovery method based on deep learning, aiming at the problems that the traditional seismic wave field characteristic recovery cross-correlation interference method can introduce new noise, the result resolution ratio is low, and the abnormal body description precision is not enough. The method utilizes the advantages of big data, fully learns the distribution rule and the same-phase axis characteristics of seismic data by utilizing information provided by mass data from the angle of data driving, establishes the nonlinear mapping from TBM rock breaking seismic source numerical simulation data to pulse seismic source seismic records after wave field characteristic recovery, completes the conversion of input data from low frequency to high frequency and from low resolution to high resolution, and solves the problem that new noise is introduced in the traditional cross-correlation method.
Aiming at the problems of poor depiction effect, high output ambiguity and the like of an original U-Net neural network on deep seismic data, effective information constraint is constructed in the problem of wave field characteristic recovery of a rock breaking seismic source, and the effective information constraint is helpful for guiding the neural network to fully mine deep reflection information.
Aiming at the problem that tunnel noise influences the wave field recovery effect, the method further recovers the data containing real noise by adopting fine tuning operation based on transfer learning on the basis of network parameters trained by data without real noise, and efficiently realizes the dual purposes of denoising the data containing noise and recovering wave field information.
Aiming at the problem that the traditional wave field characteristic recovery quality evaluation indexes such as mean square error can not accurately reflect the characteristics of seismic data, the evaluation index design of the wave field characteristic recovery quality is carried out from two angles of error statistics and a human eye visual system, the evaluation index based on the error statistics between two groups of data emphasizes the evaluation of the amplitude distribution characteristics of the seismic data, the evaluation index based on the human eye visual system carries out weighting correction on different areas of an image by utilizing the real visual characteristics when a human observes an object, the special depth characteristics of the seismic data are introduced on the basis, the inherent characteristics of seismic observation data are considered, and the overall quality of wave field recovery of a rock-breaking seismic source can be effectively reflected.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the invention and together with the description serve to explain the invention and not to limit the invention.
FIG. 1 is a flow chart of a method for recovering seismic wave field characteristics of a TBM rock breaking seismic source based on deep learning according to an embodiment of the invention;
FIG. 2 is a schematic diagram of a wave field feature recovery network based on deep learning according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a three-channel information matrix according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of spherical wave dispersion according to an embodiment of the present invention;
FIG. 5(a) is a single-layer geological model of a database built in an embodiment of the present invention;
FIG. 5(b) is a diagram of a double-layer geological model in a database built according to an embodiment of the present invention;
FIG. 5(c) is a solution cavity geological model in the database established in the embodiment of the present invention;
FIG. 6(a) is a layout diagram of a rock breaking seismic source observation system according to an embodiment of the invention;
FIG. 6(b) is a diagram of a pulsed source observation system arrangement according to an embodiment of the invention;
FIG. 7 illustrates a source wavefield recovery result for a rock breaking source in an embodiment of the invention.
Detailed Description
The invention is further described with reference to the following figures and examples.
It is to be understood that the following detailed description is exemplary and is intended to provide further explanation of the invention as claimed. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
Example one
As shown in fig. 1, the embodiment provides a method for recovering characteristics of a seismic wave field of a TBM rock breaking source based on deep learning, including:
step 1: and acquiring an original signal and a pilot signal of a rock breaking seismic source to obtain TBM rock breaking seismic source numerical simulation data.
The example is mainly used for simulating the underground geological condition that a single-layer interface, a double-layer interface, a karst cave and/or a double-layer interface and the karst cave exist in front of the tunnel face of the tunnel, as shown in fig. 5(a), 5(b) and 5 (c).
The model size of this embodiment is 290m × 90m, the grid distance Δ x ═ Δ y ═ 1m, and PML absorption boundaries of 20 grids are set around the model. The arrangement form of the seismic source and the detectors is shown in fig. 6(a) and (b), 40 detectors are respectively arranged on the upper side wall and the lower side wall of the rock breaking seismic source observation system and the impulse seismic source observation system, the channel spacing is 1m, and a pilot detector is arranged in the middle of the tunnel face of the rock breaking seismic source observation system and used for observing a pilot signal. The seismic sources in the rock breaking seismic source observation system are arranged on all grid points of the tunnel face, and the seismic sources in the pulse seismic source observation system are arranged in the center of the tunnel face. The seismic source adopts 200Hz Rake wavelets, the sampling interval of the detector is 0.075ms, the recording time of forward simulation of the rock breaking seismic source is 75s, and the pulse seismic source is set to be 0.15 s.
Step 2: converting numerical simulation data of a TBM rock breaking seismic source into a pulse seismic source seismic record after wave field characteristic recovery through a wave field characteristic recovery network based on deep learning;
as shown in fig. 2, the wave field feature recovery network based on deep learning includes a preprocessing layer and a deep neural network layer; the pretreatment layer is used for performing channel-by-channel convolution on the original signal of the rock breaking seismic source and the pilot signal and performing segmented superposition on the convolution output; the output of the pre-processing layer, the time information constraint matrix and the same-phase axis information constraint matrix form three-channel data, and then the three-channel data are input into the deep neural network layer.
Specifically, a deep neural network layer in the wave field feature recovery network based on deep learning is constructed based on a U-Net network. The wave field characteristic recovery deep neural network comprises a preprocessing layer and a U-Net network, wherein the U-Net network comprises a compression channel and an expansion channel. The operation steps of the pre-treatment layer comprise reading original detector data and pilot signals of the rock breaking seismic source, performing channel-by-channel convolution on the original detector data and the pilot signals, and performing segmented superposition on convolution output.
In this embodiment, a time information constraint matrix and a phase axis information constraint matrix with the same dimension as the output of the pre-processing layer are introduced, and three channels of data are formed with the output of the pre-processing layer as shown in fig. 3 and then input to the compression channel. The time information constraint matrix form is derived from the propagation theory of seismic waves. The total energy of a wave in the formation for simple harmonic vibration can be represented by:
E＝Et+Ep∝ρ·A2·f2·W
wherein E istBeing kinetic energy of waves, EpIs the potential energy of the wave; w is the volume of the medium through which the wave passes, ρ is the density of the medium, a represents the amplitude of the wave, and f represents the frequency of the wave. The seismic wave energy passing through a unit area in unit time is energy flux density I, and then the expression is as follows:
FIG. 4 is an image of the spreading of a spherical wave whose wavefront starts at the center o and spreads out at r1And r2The partial spherical surface area of which is the radius is S1And S2. Then I1·S1＝I2·S2And then further on
Is inherently
It is proved that the amplitude of the seismic wave is inversely proportional to the first power of the wave propagation distance, i.e., the amplitude is also proportional to the time step length. According to the above derivation, the time information constraint matrix has a horizon information characteristic that varies with the time step, and the value of each row of elements of data in the time information constraint matrix of the embodiment is equal to the number of rows of the row. The in-phase axis information constraint matrix is realized based on an amplitude automatic gain control method AGC (automatic gain control)The amplitude of the data between the gathers can be made to equalize while destroying the relative nature of the amplitudes. And at the position corresponding to the event of the reflected wave in the seismic record, the absolute value of the event information constraint matrix is larger, and the absolute values of the rest positions are smaller. The two matrices are identical in spatial dimension to the input seismic data.
In this embodiment, the compression channel includes four convolution layers and four maximum pooling layers, and performs a ReLU operation on the output of each convolution layer, where the four convolution layers and the four maximum pooling layers are arranged at intervals.
In this embodiment, the extended channel includes four convolution layers, four deconvolution layers, and two full-connection layers, and performs a ReLU operation on the output of each convolution layer, where the four convolution layers and the four deconvolution layers are arranged at intervals, and the two full-connection layers are spliced at the end, and the feature map output by each deconvolution layer in the extended channel is spliced with the feature map output by one volume of lamination in the compressed channel and then input to the next volume of lamination, and the two feature maps subjected to splicing should have the same size.
The wavefield feature recovery deep neural network finally converts the input matrix of size 40 × 1000000 into a column vector of size 1 × 1000000.
Optimizing a deep neural network by adopting a mean square error loss function weighted by prior information, and the specific process comprises the following steps:
calculating a loss function of the seismic record after wave field characteristic recovery and the seismic record obtained by pulse seismic source numerical simulation, wherein the calculation formula is as follows:
Wi,j＝Di,j·Ai,j
where w represents the width of the seismic record, h represents the height of the seismic record,
values, x, representing i-row and j-column positions of seismic records obtained by numerical simulation of the impulsive seismic sourcei,jValues, W, representing i-row and j-column positions of seismic records after wavefield feature recoveryi,jValues representing i rows and j columns positions in the dual information matrix, Di,jValues representing i rows and j columns of positions in the depth information matrix, both matrices being formally the same as described above, Ai,jValues representing the i row and j column positions in the in-phase axis weighting information matrix. The calculated loss function is subjected to gradient return through a back propagation algorithm in the neural network, and is used for updating parameters of the network.
The valley process of the training wave field characteristic recovery network is as follows:
the database of the embodiment comprises 1100 geological models, wherein a forward model 200 of a single-layer geological interface, a karst cave geological model 200, a forward model 500 of a double-layer geological interface, and a combined geological model 200 of a karst cave and a single-layer interface. The method comprises the steps of obtaining 4400 pairs of different TBM rock breaking seismic source data numbers (40 multiplied by 1000000 two-dimensional matrixes) and pulse seismic source data (1 multiplied by 1000000 column vectors) of corresponding models through forward modeling, then longitudinally cutting each data into four parts, conducting bilinear interpolation to restore the data to the original size, and further expanding a data set to 17600 pairs of data. These data were randomly divided into training, validation and test sets (training 14080, validation 1760, test 1760) at an 8:1:1 ratio, and the resulting data were normalized to a [0,1] range based on the magnitude of the seismic signal.
The main network parameters and hardware conditions in this embodiment are: the calculation is implemented using a single chip NVIDIA TITAN Xp. A network is built based on a PyTorch platform, a ReLU function is used as an activation function of network training, 7 x 3 convolution kernels are adopted for up-down sampling in a neural network, the batch processing quantity (batch size) of an SGD optimizer is 16, the learning rate (learning rate) is 0.1, and the working frequency (epoch) of a learning algorithm in the whole training data set is 250.
And adding the noise data actually acquired in the field into a TBM rock breaking seismic source numerical simulation data set. Due to the similarity of the migration learning problem of the noise-containing data wave field feature recovery and the previous problem, after the wave field feature recovery deep neural network finishes the training on the noise-free data, the network parameters of the compression channel are fixed, the network parameters of the expansion channel are trained only in a network fine tuning mode, the migration learning is carried out, and the noise removal and the wave field information recovery are carried out on the noise data.
Wave field characteristic recovery quality evaluation indexes aiming at seismic data characteristics are adopted, the evaluation indexes comprise evaluation indexes based on error statistics and evaluation indexes based on a human visual system, and a Q1 evaluation index calculation formula based on the error statistics is as follows:
let I be the output data matrix, size w h (width and height, respectively), Ii,jWhich represents the small block data matrix formed by division, i and j represent the block number indices in the width and height directions, respectively. Ki,j＝(k1,k2kn) An amplitude distribution vector, k, representing the block matrixnRepresenting the number of pixels contained in the nth amplitude level.
i,j＝(
1,
2 
n) An amplitude distribution vector representing the corresponding desired output blocking matrix,
nrepresenting the number of pixels contained in the nth amplitude level of the desired output block matrix, a and b are the block numbers in the width and height directions, respectively.
The Q2 index is calculated by adopting an evaluation index calculation formula based on a human visual system:
PCm(X)＝max(PC1(x),PC2(x))
the formula is derived from an FSIM algorithm based on a human eye vision system (HVS) in the existing image quality evaluation method, wherein rich local texture structure characteristics of an image are obtained by a Phase Consistency (PC) theory, and contrast information is extracted by a gradient amplitude value (GM). After the PC and GM characteristics of the two images are obtained, the PC and wave field time matrix combination is introduced as a weighting function to obtain an image structure similarity score Q2 considering the visual characteristics of human eyes. In the formula, subscripts 1 and 2 represent two image data to be compared, respectively. D (x) is a seismic wavefield time information matrix.
In this embodiment, the wave field characteristic recovery network constructs a mapping relationship between a TBM rock-breaking seismic source signal and a pulse seismic source signal, which may represent a rock-breaking seismic source wave field characteristic recovery process, and partial results of the substitution into the test set are shown in fig. 7. In the evaluation indexes corresponding to the double-layer model data in fig. 7, the record Q1 after the correlation of the rock breaking seismic source is 0.297, the record Q2 is 0.284, the record Q1 after the recovery of the rock breaking seismic source is 0.680, and the record Q2 is 0.569; and according to evaluation indexes corresponding to the karst cave model data, the record Q1 after the correlation of the rock breaking seismic source is 0.385, the record Q2 is 0.514, the record Q1 after the recovery of the rock breaking seismic source is 0.771, and the record Q2 is 0.684. The above results can show that the embodiment can effectively complete the process of recovering the wave field characteristics of the rock breaking seismic source.
Example two
The embodiment provides a system for recovering seismic wave field characteristics of a TBM rock breaking seismic source based on deep learning, which comprises:
(1) and the signal acquisition module is used for acquiring the original signal and the pilot signal of the rock breaking seismic source to obtain the numerical simulation data of the TBM rock breaking seismic source.
The example is mainly used for simulating the underground geological condition that a single-layer interface, a double-layer interface, a karst cave and/or a double-layer interface and the karst cave exist in front of the tunnel face of the tunnel, as shown in fig. 5(a), 5(b) and 5 (c).
The model size of this embodiment is 290m × 90m, the grid distance Δ x ═ Δ y ═ 1m, and PML absorption boundaries of 20 grids are set around the model. The arrangement form of the seismic source and the detectors is shown in fig. 6(a) and fig. 6(b), 40 detectors are respectively arranged on the upper side wall and the lower side wall of the rock breaking seismic source observation system and the pulse seismic source observation system, the channel spacing is 1m, and a pilot detector is arranged in the middle of the tunnel face of the rock breaking seismic source observation system and used for observing a pilot signal. The seismic sources in the rock breaking seismic source observation system are arranged on all grid points of the tunnel face, and the seismic sources in the pulse seismic source observation system are arranged in the center of the tunnel face. The seismic source adopts 200Hz Rake wavelets, the sampling interval of the detector is 0.075ms, the recording time of forward simulation of the rock breaking seismic source is 75s, and the pulse seismic source is set to be 0.15 s.
(2) The characteristic recovery module is used for converting the TBM rock breaking seismic source numerical simulation data into a pulse seismic source seismic record after wave field characteristic recovery through a wave field characteristic recovery network based on deep learning;
the wave field characteristic recovery network based on deep learning comprises a preprocessing layer and a deep neural network layer; the pretreatment layer is used for performing channel-by-channel convolution on the original signal of the rock breaking seismic source and the pilot signal and performing segmented superposition on the convolution output; the output of the pre-processing layer, the time information constraint matrix and the same-phase axis information constraint matrix form three-channel data, and then the three-channel data are input into the deep neural network layer.
Specifically, a deep neural network layer in the wave field feature recovery network based on deep learning is constructed based on a U-Net network. The wave field characteristic recovery deep neural network comprises a preprocessing layer and a U-Net network, wherein the U-Net network comprises a compression channel and an expansion channel. The operation steps of the pre-treatment layer comprise reading original detector data and pilot signals of the rock breaking seismic source, performing channel-by-channel convolution on the original detector data and the pilot signals, and performing segmented superposition on convolution output.
In this embodiment, a time information constraint matrix and a phase axis information constraint matrix with the same dimension as the output of the pre-processing layer are introduced, and three channels of data are formed with the output of the pre-processing layer as shown in fig. 3 and then input to the compression channel. The time information constraint matrix form is derived from the propagation theory of seismic waves. The total energy of a wave in the formation for simple harmonic vibration can be represented by:
E＝Et+Ep∝ρ·A2·f2·W
wherein E istBeing kinetic energy of waves, EpIs the potential energy of the wave; w is the volume of the medium through which the wave passes, ρ is the density of the medium, a represents the amplitude of the wave, and f represents the frequency of the wave. The seismic wave energy passing through a unit area in unit time is energy flux density I, and then the expression is as follows:
FIG. 4 is an image of the spreading of a spherical wave whose wavefront starts at the center o and spreads out at r1And r2The partial spherical surface area of which is the radius is S1And S2. Then I1·S1＝I2·S2And then further on
Is inherently
It is proved that the amplitude of the seismic wave is inversely proportional to the first power of the wave propagation distance, i.e., the amplitude is also proportional to the time step length. According to the above derivation, the time information constraint matrix has a horizon information characteristic that varies with the time step, and the value of each row of elements of data in the time information constraint matrix of the embodiment is equal to the number of rows of the row. The in-phase axis information constraint matrix is realized based on an amplitude automatic gain control method AGC, and the method can enable the amplitude of data among all the gathers to tend to be balanced and simultaneously destroy the relative attribute of the amplitude. In correspondence with the event of reflection in seismic recordingAt the position, the absolute value of the in-phase axis information constraint matrix is larger, and the absolute values of the rest positions are smaller. The two matrices are identical in spatial dimension to the input seismic data.
In this embodiment, the compression channel includes four convolution layers and four maximum pooling layers, and performs a ReLU operation on the output of each convolution layer, where the four convolution layers and the four maximum pooling layers are arranged at intervals.
In this embodiment, the extended channel includes four convolution layers, four deconvolution layers, and two full-connection layers, and performs a ReLU operation on the output of each convolution layer, where the four convolution layers and the four deconvolution layers are arranged at intervals, and the two full-connection layers are spliced at the end, and the feature map output by each deconvolution layer in the extended channel is spliced with the feature map output by one volume of lamination in the compressed channel and then input to the next volume of lamination, and the two feature maps subjected to splicing should have the same size.
The wavefield feature recovery deep neural network finally converts the input matrix of size 40 × 1000000 into a column vector of size 1 × 1000000.
Optimizing a deep neural network by adopting a mean square error loss function weighted by prior information, and the specific process comprises the following steps:
calculating a loss function of the seismic record after wave field characteristic recovery and the seismic record obtained by pulse seismic source numerical simulation, wherein the calculation formula is as follows:
Wi,j＝Di,j·Ai,j
where w represents the width of the seismic record, h represents the height of the seismic record,
values, x, representing i-row and j-column positions of seismic records obtained by numerical simulation of the impulsive seismic sourcei,jValues, W, representing i-row and j-column positions of seismic records after wavefield feature recoveryi,jRepresenting i rows and j columns in a dual information matrixValue of position, Di,jValues representing i rows and j columns of positions in the depth information matrix, both matrices being formally the same as described above, Ai,jValues representing the i row and j column positions in the in-phase axis weighting information matrix. The calculated loss function is subjected to gradient return through a back propagation algorithm in the neural network, and is used for updating parameters of the network.
The valley process of the training wave field characteristic recovery network is as follows:
the database of the embodiment comprises 1100 geological models, wherein a forward model 200 of a single-layer geological interface, a karst cave geological model 200, a forward model 500 of a double-layer geological interface, and a combined geological model 200 of a karst cave and a single-layer interface. The method comprises the steps of obtaining 4400 pairs of different TBM rock breaking seismic source data numbers (40 multiplied by 1000000 two-dimensional matrixes) and pulse seismic source data (1 multiplied by 1000000 column vectors) of corresponding models through forward modeling, then longitudinally cutting each data into four parts, conducting bilinear interpolation to restore the data to the original size, and further expanding a data set to 17600 pairs of data. These data were randomly divided into training, validation and test sets (training 14080, validation 1760, test 1760) at an 8:1:1 ratio, and the resulting data were normalized to a [0,1] range based on the magnitude of the seismic signal.
The main network parameters and hardware conditions in this embodiment are: the calculation is implemented using a single chip NVIDIA TITAN Xp. A network is built based on a PyTorch platform, a ReLU function is used as an activation function of network training, 7 x 3 convolution kernels are adopted for up-down sampling in a neural network, the batch processing quantity (batch size) of an SGD optimizer is 16, the learning rate (learning rate) is 0.1, and the working frequency (epoch) of a learning algorithm in the whole training data set is 250.
And adding the noise data actually acquired in the field into a TBM rock breaking seismic source numerical simulation data set. Due to the similarity of the migration learning problem of the noise-containing data wave field feature recovery and the previous problem, after the wave field feature recovery deep neural network finishes the training on the noise-free data, the network parameters of the compression channel are fixed, the network parameters of the expansion channel are trained only in a network fine tuning mode, the migration learning is carried out, and the noise removal and the wave field information recovery are carried out on the noise data.
Wave field characteristic recovery quality evaluation indexes aiming at seismic data characteristics are adopted, the evaluation indexes comprise evaluation indexes based on error statistics and evaluation indexes based on a human visual system, and a Q1 evaluation index calculation formula based on the error statistics is as follows:
let I be the output data matrix, size w h (width and height, respectively), Ii,jWhich represents the small block data matrix formed by division, i and j represent the block number indices in the width and height directions, respectively. Ki,j＝(k1,k2kn) An amplitude distribution vector, k, representing the block matrixnRepresenting the number of pixels contained in the nth amplitude level.
i,j＝(
1,
2 
n) An amplitude distribution vector representing the corresponding desired output blocking matrix,
nrepresenting the number of pixels contained in the nth amplitude level of the desired output block matrix, a and b are the block numbers in the width and height directions, respectively.
The Q2 index is calculated by adopting an evaluation index calculation formula based on a human visual system:
PCm(X)＝max(PC1(x),PC2(x))
the formula is derived from an FSIM algorithm based on a human eye vision system (HVS) in the existing image quality evaluation method, wherein rich local texture structure characteristics of an image are obtained by a Phase Consistency (PC) theory, and contrast information is extracted by a gradient amplitude value (GM). After the PC and GM characteristics of the two images are obtained, the PC and wave field time matrix combination is introduced as a weighting function to obtain an image structure similarity score Q2 considering the visual characteristics of human eyes. In the formula, subscripts 1 and 2 represent two image data to be compared, respectively. D (x) is a seismic wavefield time information matrix.
In this embodiment, the wave field characteristic recovery network constructs a mapping relationship between a TBM rock-breaking seismic source signal and a pulse seismic source signal, which may represent a rock-breaking seismic source wave field characteristic recovery process, and partial results of the substitution into the test set are shown in fig. 7. In the evaluation indexes corresponding to the double-layer model data in fig. 7, the record Q1 after the correlation of the rock breaking seismic source is 0.297, the record Q2 is 0.284, the record Q1 after the recovery of the rock breaking seismic source is 0.680, and the record Q2 is 0.569; and according to evaluation indexes corresponding to the karst cave model data, the record Q1 after the correlation of the rock breaking seismic source is 0.385, the record Q2 is 0.514, the record Q1 after the recovery of the rock breaking seismic source is 0.771, and the record Q2 is 0.684. The above results can show that the embodiment can effectively complete the process of recovering the wave field characteristics of the rock breaking seismic source.
EXAMPLE III
The present embodiment provides a computer-readable storage medium, on which a computer program is stored, where the program is executed by a processor to implement the steps in the deep learning-based TBM lithotripsy source seismic wave field feature recovery method as described above.
Example four
The embodiment provides a computer device, which comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the program to realize the steps in the deep learning-based TBM rock breaking source seismic wave field characteristic recovery method.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A TBM rock breaking seismic source seismic wave field feature recovery method based on deep learning is characterized by comprising the following steps:
acquiring a rock breaking seismic source original signal and a pilot signal to obtain TBM rock breaking seismic source numerical simulation data;
converting numerical simulation data of a TBM rock breaking seismic source into a pulse seismic source seismic record after wave field characteristic recovery through a wave field characteristic recovery network based on deep learning;
the wave field characteristic recovery network based on deep learning comprises a preprocessing layer and a deep neural network layer; the pretreatment layer is used for performing channel-by-channel convolution on the original signal of the rock breaking seismic source and the pilot signal and performing segmented superposition on the convolution output; the output of the pre-processing layer, the time information constraint matrix and the same-phase axis information constraint matrix form three-channel data, and then the three-channel data are input into the deep neural network layer.
2. The deep learning-based method for recovering the seismic wavefield characteristics of a rock breaking source of a TBM (tunnel boring machine) based on deep learning as claimed in claim 1, wherein the value of each row of elements of the data in the time information constraint matrix is equal to the number of rows of the row.
3. The deep learning-based TBM rock breaking source seismic wavefield feature recovery method of claim 1, wherein the in-phase axis information constraint matrix is implemented based on an amplitude Automatic Gain Control (AGC) method.
4. The deep learning-based method for recovering the seismic wave field characteristics of the TBM rock breaking source as claimed in claim 1, wherein the deep neural network layer comprises a compression channel and an expansion channel.
5. The deep learning-based TBM rock breaking source seismic wave field characteristic recovery method as claimed in claim 4, wherein the compression channel comprises four convolutional layers and four maximum pooling layers, and ReLU operation is performed on the output of each convolutional layer, wherein the four convolutional layers and the four maximum pooling layers are arranged at intervals;
or the extended channel comprises four convolution layers, four deconvolution layers and two full-connection layers, ReLU operation is carried out on the output of each convolution layer, wherein the four convolution layers and the four deconvolution layers are arranged at intervals, and the two full-connection layers are spliced at the end.
6. The deep learning-based method for recovering the seismic wavefield characteristics of the TBM rock breaking source based on the deep learning as claimed in claim 1, wherein the deep learning-based method for training the wavefield characteristic recovery network comprises the following steps:
constructing a noiseless rock breaking seismic source wave field recovery database, and carrying out noiseless training on a wave field characteristic recovery network based on deep learning;
and introducing real noise, establishing a rock breaking seismic source wave field recovery database containing the real noise, finely adjusting a wave field characteristic recovery network based on deep learning after noise-free training, and performing migration learning on noise data.
7. The deep learning-based seismic wave field feature recovery method for the TBM rock breaking source as claimed in claim 1, wherein the optimization process of the deep learning-based wave field feature recovery network is as follows:
and optimizing by adopting a dual-weighted loss function of the time information and the in-phase axis information, wherein the loss function is used for calculating the error between the seismic record subjected to wave field characteristic recovery and the seismic record obtained by pulse seismic source numerical simulation.
8. The utility model provides a broken rock seismic source seismic wave field characteristic recovery system of TBM based on deep learning which characterized in that includes:
the signal acquisition module is used for acquiring a rock breaking source original signal and a pilot signal to obtain TBM rock breaking source numerical simulation data;
the characteristic recovery module is used for converting the TBM rock breaking seismic source numerical simulation data into a pulse seismic source seismic record after wave field characteristic recovery through a wave field characteristic recovery network based on deep learning;
the wave field characteristic recovery network based on deep learning comprises a preprocessing layer and a deep neural network layer; the pretreatment layer is used for performing channel-by-channel convolution on the original signal of the rock breaking seismic source and the pilot signal and performing segmented superposition on the convolution output; the output of the pre-processing layer, the time information constraint matrix and the same-phase axis information constraint matrix form three-channel data, and then the three-channel data are input into the deep neural network layer.
9. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements the steps in the method for deep learning-based feature recovery of a seismic field of a TBM rock breaking source according to any one of claims 1 to 7.
10. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor when executing the program implements the steps in the deep learning based TBM lithotripsy source seismic wavefield feature recovery method of any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011061619.5A CN112231974B (en) | 2020-09-30 | 2020-09-30 | Deep learning-based method and system for recovering seismic wave field characteristics of rock breaking seismic source of TBM (Tunnel boring machine) |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011061619.5A CN112231974B (en) | 2020-09-30 | 2020-09-30 | Deep learning-based method and system for recovering seismic wave field characteristics of rock breaking seismic source of TBM (Tunnel boring machine) |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112231974A true CN112231974A (en) | 2021-01-15 |
| CN112231974B CN112231974B (en) | 2022-11-04 |
Family
ID=74119871
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011061619.5A Expired - Fee Related CN112231974B (en) | 2020-09-30 | 2020-09-30 | Deep learning-based method and system for recovering seismic wave field characteristics of rock breaking seismic source of TBM (Tunnel boring machine) |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112231974B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114419558A (en) * | 2022-03-31 | 2022-04-29 | 华南理工大学 | Fire video image identification method, fire video image identification system, computer equipment and storage medium |
| CN115290761A (en) * | 2022-09-27 | 2022-11-04 | 中国科学院地质与地球物理研究所 | Acoustic emission data reconstruction method and device based on deep learning |
| CN115453616A (en) * | 2022-08-04 | 2022-12-09 | 山东大学 | Tunnel seismic exploration data enhancement method and system based on deep learning |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104678427A (en) * | 2015-03-11 | 2015-06-03 | 山东大学 | Tunnel boring machine rock breaking seismic source three-dimensional seismic advanced detection device and method |
| CN204496014U (en) * | 2015-03-11 | 2015-07-22 | 山东大学 | Rock tunnel(ling) machine breaks rock focus and active source 3-D seismics associating forward probe system |
| CN108798690A (en) * | 2018-06-01 | 2018-11-13 | 中国科学院武汉岩土力学研究所 | Realize the combination type T BM and geology detecting driving method of geology detecting |
| CN109993280A (en) * | 2019-03-27 | 2019-07-09 | 东南大学 | A kind of underwater sound source localization method based on deep learning |
| CN110414675A (en) * | 2019-09-02 | 2019-11-05 | 中北大学 | A kind of underground shallow layer seismic source location method based on deep learning |
| CN111562611A (en) * | 2020-04-08 | 2020-08-21 | 山东大学 | Semi-supervised depth learning seismic data inversion method based on wave equation drive |
| CN111722279A (en) * | 2020-05-12 | 2020-09-29 | 山东大学 | TBM rock breaking seismic source seismic detection device and method based on ground-tunnel combination |
-
2020
- 2020-09-30 CN CN202011061619.5A patent/CN112231974B/en not_active Expired - Fee Related
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104678427A (en) * | 2015-03-11 | 2015-06-03 | 山东大学 | Tunnel boring machine rock breaking seismic source three-dimensional seismic advanced detection device and method |
| CN204496014U (en) * | 2015-03-11 | 2015-07-22 | 山东大学 | Rock tunnel(ling) machine breaks rock focus and active source 3-D seismics associating forward probe system |
| CN108798690A (en) * | 2018-06-01 | 2018-11-13 | 中国科学院武汉岩土力学研究所 | Realize the combination type T BM and geology detecting driving method of geology detecting |
| CN109993280A (en) * | 2019-03-27 | 2019-07-09 | 东南大学 | A kind of underwater sound source localization method based on deep learning |
| CN110414675A (en) * | 2019-09-02 | 2019-11-05 | 中北大学 | A kind of underground shallow layer seismic source location method based on deep learning |
| CN111562611A (en) * | 2020-04-08 | 2020-08-21 | 山东大学 | Semi-supervised depth learning seismic data inversion method based on wave equation drive |
| CN111722279A (en) * | 2020-05-12 | 2020-09-29 | 山东大学 | TBM rock breaking seismic source seismic detection device and method based on ground-tunnel combination |
Non-Patent Citations (2)
| Title |
|---|
| YUXIAO REN等: "《A Physics-Based Neural-Network Way to Perform Seismic Full Waveform Inversion》", 《IEEE ACCESS》 * |
| 张攀龙等: "《基于U-Net深度神经网络的地震数据去噪研究》", 《金属矿山》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114419558A (en) * | 2022-03-31 | 2022-04-29 | 华南理工大学 | Fire video image identification method, fire video image identification system, computer equipment and storage medium |
| CN114419558B (en) * | 2022-03-31 | 2022-07-05 | 华南理工大学 | Fire video image identification method, fire video image identification system, computer equipment and storage medium |
| CN115453616A (en) * | 2022-08-04 | 2022-12-09 | 山东大学 | Tunnel seismic exploration data enhancement method and system based on deep learning |
| CN115290761A (en) * | 2022-09-27 | 2022-11-04 | 中国科学院地质与地球物理研究所 | Acoustic emission data reconstruction method and device based on deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112231974B (en) | 2022-11-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112231974B (en) | Deep learning-based method and system for recovering seismic wave field characteristics of rock breaking seismic source of TBM (Tunnel boring machine) | |
| US11828894B2 (en) | Multi-scale unsupervised seismic velocity inversion method based on autoencoder for observation data | |
| CN106772583B (en) | A kind of earthquake diffracted wave separation method and device | |
| CN109709603B (en) | Seismic horizon identification and tracking method and system | |
| CN110031896B (en) | Seismic random inversion method and device based on multi-point geostatistics prior information | |
| CN107479092B (en) | A kind of frequency domain high order ACOUSTIC WAVE EQUATION the Forward Modeling based on directional derivative | |
| CN108181653B (en) | For VTI medium reverse-time migration method, equipment and medium | |
| CN110954950B (en) | Underground transverse wave velocity inversion method, device, computing equipment and storage medium | |
| US11397273B2 (en) | Full waveform inversion in the midpoint-offset domain | |
| CN117546051A (en) | Method and system for seismic imaging using an S-wave velocity model and machine learning | |
| CN103913768A (en) | Method and device for modeling superficial layer in earth surface based on seismic wave data | |
| CN110927793A (en) | Reservoir prediction method and system based on sequential random fuzzy simulation | |
| US20240134080A1 (en) | Method and System for Real-Time Calculating a Microseismic Focal Mechanism Based on Deep Learning | |
| US20160034612A1 (en) | Re-ordered Interpolation and Convolution for Faster Staggered-Grid Processing | |
| CN111123361B (en) | Method and device for regularly reconstructing seismic data of vertical seismic profile and storage medium | |
| CN115639605B (en) | Automatic identification method and device for high-resolution fault based on deep learning | |
| CN110687598A (en) | Method and device for accelerating microseismic numerical simulation | |
| CN117933319B (en) | Modeling method and system for deep random mixed medium non-stationary model and electronic equipment | |
| CN113219534B (en) | Pre-stack depth migration speed quality control method and device, medium and electronic equipment | |
| CN118364227A (en) | Method and system for suppressing near-surface noise of earthquake based on multi-modal feature constraint DNN | |
| US20230074047A1 (en) | Method and Apparatus for Performing Wavefield Predictions By Using Wavefront Estimations | |
| CN117706623A (en) | Microseism speed inversion method based on deep learning and wave equation combined driving | |
| Grubas | Seismic-wave traveltime computation by supervised and unsupervised training of artificial neural networks | |
| Feng et al. | Calculation method of formation dip angle based on deep learning | |
| CN117665911A (en) | Surface wave signal windowing analysis-based earth-rock interface identification method and related device |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20221104 |