CN117970479A - Surrounding rock grading method and grading system based on earthquake body waves and machine learning - Google Patents
Surrounding rock grading method and grading system based on earthquake body waves and machine learning Download PDFInfo
- Publication number
- CN117970479A CN117970479A CN202410111491.0A CN202410111491A CN117970479A CN 117970479 A CN117970479 A CN 117970479A CN 202410111491 A CN202410111491 A CN 202410111491A CN 117970479 A CN117970479 A CN 117970479A
- Authority
- CN
- China
- Prior art keywords
- surrounding rock
- seismic
- attribute
- machine learning
- frequency
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 239000011435 rock Substances 0.000 title claims abstract description 77
- 238000000034 method Methods 0.000 title claims abstract description 60
- 238000010801 machine learning Methods 0.000 title claims abstract description 37
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 76
- 238000001228 spectrum Methods 0.000 claims abstract description 64
- 238000012545 processing Methods 0.000 claims abstract description 17
- 238000006243 chemical reaction Methods 0.000 claims abstract description 12
- 238000003064 k means clustering Methods 0.000 claims abstract description 7
- 239000011159 matrix material Substances 0.000 claims description 12
- 238000010606 normalization Methods 0.000 claims description 7
- 230000003595 spectral effect Effects 0.000 claims description 7
- 238000000513 principal component analysis Methods 0.000 claims description 4
- 238000001125 extrusion Methods 0.000 claims description 3
- 238000001514 detection method Methods 0.000 abstract description 11
- 238000010276 construction Methods 0.000 abstract description 10
- 230000009286 beneficial effect Effects 0.000 abstract description 4
- 238000010183 spectrum analysis Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 4
- 230000004927 fusion Effects 0.000 description 3
- 238000005259 measurement Methods 0.000 description 3
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 2
- 238000002247 constant time method Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 229910000831 Steel Inorganic materials 0.000 description 1
- 238000007792 addition Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 238000005553 drilling Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 229910052742 iron Inorganic materials 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000149 penetrating effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000002689 soil Substances 0.000 description 1
- 239000010959 steel Substances 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
Landscapes
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention relates to the technical field of surrounding rock classification, in particular to a surrounding rock classification method and system based on earthquake body waves and machine learning, comprising the following steps: collecting shallow seismic body wave data at a position to be detected; establishing a deep conversion relation; obtaining a common offset profile, extracting spectrum decomposition information corresponding to the same phase axis of the target layer in different offset profiles, and converting the spectrum decomposition information into a spectrum decomposition attribute profile; the PCA machine learning method is used for respectively processing the spectrum decomposition attribute sections of the high frequency band, the middle frequency band and the low frequency band to obtain a seismic attribute main component section; and (5) quantitatively predicting the surrounding rock grade by using a K-means clustering method. The scheme establishes the connection between the surrounding rock grade and the elastic wave detection result, simplifies the field data acquisition difficulty, gets rid of the dependence on the working experience of field operators, realizes the quantitative prediction of the high-resolution, continuous and accurate space surrounding rock grade, is beneficial to reducing the construction cost, reduces the working period and ensures the construction safety of highway tunnels.
Description
Technical Field
The invention relates to the technical field of surrounding rock classification, in particular to a surrounding rock classification method and system based on earthquake body waves and machine learning.
Background
In the process of excavating a tunnel, the instability of surrounding rock mass or soil mass increases the difficulty for excavating, the classification of surrounding rock of the tunnel is an important basis for the design and construction of a tunnel supporting structure, the present research mainly takes an intelligent algorithm for classifying the surrounding rock as a main basis, and the core thought of the intelligent algorithm is to take classification indexes such as rock mass integrity, rock mass strength, structural surface parameters, hydraulic conditions, the state of a ground stress field and the like as inputs, and after various intelligent algorithms are used, the classification of the surrounding rock is taken as output, and the intelligent algorithms comprise BP neural network algorithm, support vector machine, genetic algorithm and the like.
Along with the increasing complexity of tunnel construction conditions, the requirement on the definition degree of surrounding rock classification is continuously increased, and geophysical prospecting means gradually play a role in the surrounding rock classification technology. The current common shallow layer investigation technology mainly comprises an electric method and an earthquake, wherein the electric method comprises a high-density resistivity method, a transient electromagnetic method and a ground penetrating radar method, and the earthquake comprises a shallow layer earthquake reflection wave method, a refraction method and a surface wave method; in combination with drilling, there is also a cross-hole resistivity CT method and a cross-hole seismic wave CT method. The earthquake body wave detection method is low in noise, high in resolution and low in cost, and can be used for grading detection of tunnel surrounding rocks.
The patent publication number is: the patent document of CN115062375A discloses a grading method and a grading system for tunnel surrounding rocks, wherein the grading method comprises the steps of collecting surrounding rock characteristic data and corresponding surrounding rock grades; constructing a preset fusion model, training the preset fusion model according to surrounding rock characteristic data and corresponding surrounding rock levels to form an intelligent surrounding rock classification model, wherein the preset fusion model comprises a characteristic recognition part and a target learning part, and the characteristic recognition part is provided with characteristic learning networks respectively built for different types of measurement data; and carrying out grading evaluation on the target surrounding rock by using the intelligent surrounding rock grading model. The invention digs the internal relation between various measurement and surrounding rock grading targets, ensures the data basis, builds the intelligent grading model with good flexibility and wide applicability, and can effectively realize the feature learning and target learning of the surrounding rock intelligent grading task.
Compared with the traditional method, the method for detecting the earthquake body waves has the advantages of high resolution, continuous space, quantitative measurement results and the like, but the test results of the earthquake body waves do not directly reflect the grades of the surrounding rocks, so that a set of surrounding rock grading method suitable for detecting the earthquake body waves needs to be established by means of a machine learning method, and the purpose of quantitatively evaluating the grades of the surrounding rocks is achieved by processing the detection results of the earthquake body waves.
Disclosure of Invention
In order to overcome the technical problems, the invention aims to provide a surrounding rock grading method and a surrounding rock grading system based on earthquake body waves and machine learning, which are used for processing earthquake body wave detection results so as to achieve the purpose of quantitatively evaluating surrounding rock grades.
The aim of the invention can be achieved by the following technical scheme:
a method for classification of surrounding rock based on seismic body wave and machine learning, comprising the steps of:
S1, collecting shallow seismic body wave data at a position to be detected;
S2, establishing a time depth conversion relation;
s3, obtaining a common offset profile, extracting spectrum decomposition information of the target layer corresponding to the phase axis in the profiles with different offset distances, and converting the spectrum decomposition information into a spectrum decomposition attribute profile;
s4, respectively processing the spectrum decomposition attribute sections of the high frequency band, the middle frequency band and the low frequency band by using a PCA machine learning method to obtain a seismic attribute main component section;
S5, performing quantitative prediction on the surrounding rock grade by using a K-means clustering method.
Further, the step S2 includes the steps of:
s21, obtaining a single shot record based on body wave exploration, obtaining a same phase axis corresponding to a target layer, and counting the same phase axis travel time corresponding to different offset distances;
s22, counting dominant frequencies of the phase axes of the target layer at different offset distances, and calculating wavelength, wherein the wavelength can be obtained by the following formula:
Wherein lambda is the wavelength of the bulk wave, v is the wave velocity of the bulk wave, and f is the frequency of the bulk wave;
S23, based on the body wave wavelength, calculating the depth corresponding to the body waves with different offset distances, wherein the depth is about 1/2 to 1/3 of the body wave wavelength.
Further, the step S3 includes the steps of:
S31, extracting common offset seismic traces in each single shot record to form a common offset seismic section;
s32, performing spectrum decomposition on the co-offset profile according to the dominant frequency distribution range of the seismic profile obtained in the step S22 to obtain seismic body wave information corresponding to different frequencies;
And S33, extracting the layer-following spectrum decomposition attribute corresponding to the target layer phase axis in each spectrum decomposition profile according to the travel of the phase axis obtained in the step S21 at different offset distances, carrying out normalization processing on the layer-following attribute value, and drawing the layer-following spectrum decomposition attribute of different offset distances in the same spectrum decomposition attribute profile.
Further, the step S4 includes the steps of:
S41, according to the dominant frequency distribution range of the seismic section obtained in the step S22, dividing the spectrum decomposition seismic attribute into three types of low frequency, medium frequency and high frequency;
S42, performing principal component analysis on the three types of spectrum decomposition profiles of low frequency, medium frequency and high frequency respectively, and calculating the contribution rate corresponding to each principal component;
s43, selecting principal component information with highest contribution rate as principal component profiles corresponding to low-frequency, medium-frequency and high-frequency information.
Further, the step S5 includes the steps of:
S51, converting the principal component profile into a quantitative depth domain principal component profile with the ordinate being depth and the abscissa being the length of the measuring line according to the time-depth conversion relation obtained in the step S2;
S52, judging the classification grade and the quantity of surrounding rocks of the area to be predicted by combining the local geological profile;
S53, quantitatively dividing the surrounding rock grade by using a K-means clustering method by using the main component section corresponding to the low-frequency, medium-frequency and high-frequency information.
The PCA machine learning method in the step S4 is characterized by comprising the following core steps:
U=ZTb,
where U is the principal component matrix of the attribute, Z T is the transpose matrix of the normalized input attribute matrix, and b is the unit eigenvector of the input attribute dataset correlation coefficient matrix.
Further, the K-means machine learning core step in the step S5 is as follows:
And randomly extracting k initial clustering centers C i (i is more than or equal to 1 is less than or equal to k) from an input data set, calculating the distance between the rest data objects and the clustering center C i, finding out the clustering center C i closest to the target data object, distributing the data objects into clusters corresponding to the clustering center C i, forming a new clustering center according to the value of the data object in each cluster, and performing the next iteration until the clustering center is not changed or the maximum iteration number is reached.
Further, in the step S32, the seismic data is decomposed into a series of time domain discrete frequency amplitude data volumes using a spectral decomposition method, including but not limited to a short time fourier transform, a continuous wavelet transform, an S-transform, or a synchronous extrusion transform method.
Further, in the step S33, the normalization method is as follows:
Wherein A is the value of the decomposition attribute along the target layer spectrum, A max is the maximum value of the decomposition attribute along the target layer spectrum, and A min is the minimum value of the decomposition attribute along the target layer spectrum.
A surrounding rock grading system based on seismic body waves and machine learning comprises a data input module, a data processing module and an output module;
The data input module is used for inputting seismic body wave data;
The data processing module stores a surrounding rock grading method based on earthquake body waves and machine learning, and the surrounding rock is graded by calling the grading method;
the output module comprises a terminal or a cloud server.
The invention has the beneficial effects that:
The scheme is based on a shallow seismic body wave detection section, adopts Matlab programming language as technical support, provides a surrounding rock grading method and grading system based on seismic body waves and machine learning through a series of technical means of time-depth conversion, spectrum decomposition, PCA and K-means machine learning, establishes connection between surrounding rock grades and elastic wave detection results, simplifies field data acquisition difficulty, gets rid of dependence on working experience of field operators, realizes high-resolution, spatially continuous and accurate quantitative prediction of the surrounding rock grades, is beneficial to reducing construction cost, reduces working period and ensures construction safety of highway tunnels.
Drawings
The invention is further described below with reference to the accompanying drawings.
FIG. 1 is a flow chart of a method of classification of surrounding rock in the present invention;
FIG. 2 is a cross-sectional view of shallow seismic single shot data acquired by the present invention;
FIG. 3 is a graph of velocity pickup results for determining the phase axis of a destination layer in accordance with the present invention;
FIG. 4 is a diagram of time-frequency analysis in the process of establishing a time-depth conversion relationship in the present invention;
FIG. 5 is a cross-sectional view of a common offset taken in accordance with the present invention;
FIG. 6 is a 45Hz spectrally exploded cross-sectional view of the co-offset cross-section of the present invention;
FIG. 7 is a characteristic cross-sectional view of 20Hz, 40Hz, 55Hz, 70Hz spectral decomposition attribute values as a function of offset in the present invention;
FIG. 8 is a cross-sectional view of principal components of PCA corresponding to spectral decomposition attribute values of low, medium and high frequency bands according to the present invention;
FIG. 9 is a diagram of the grading result of the road surrounding rock in the present invention.
Detailed Description
The technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
As shown in fig. 1,2,3, 4, 5, 6, 7, 8 and 9, the surrounding rock grading method based on the earthquake body wave and the machine learning is used for processing the earthquake body wave detection result so as to achieve the purpose of quantitatively evaluating the surrounding rock grade.
As shown in fig. 1, the surrounding rock classification method comprises the following steps:
S1, collecting shallow seismic body wave data at a position to be detected;
In this embodiment, the hammer and the steel plate are used as the manually excited source to collect the seismic body wave data, and are connected with the host computer through the lead, when the hammer is knocked on the iron plate placed at the shot point, the electric signal is conducted, the host computer starts to collect the seismic data from the detector on the seismic big line from the moment, in the process, the energy excited by the manually knocked source is limited, so that the data of one shot point need to be overlapped for a plurality of times, the operation can reduce the interference data in each excitation, meanwhile, the effective signal is enhanced, and finally, the single shot record shown in fig. 2 is acquired.
S2, establishing a time depth conversion relation, wherein the method comprises the following steps of:
s21, obtaining a single shot record based on body wave exploration, obtaining a same phase axis corresponding to a target layer, and counting the same phase axis travel time corresponding to different offset distances;
In the embodiment, in the seismic data provided by the scheme, due to the influence of the dispersion effect, the multiple interference and the like of waves, more than one in-phase axis is concentrated in the middle-far offset distance channel, and the corresponding sampling time difference is far, so that the in-phase axis which truly belongs to the body wave is difficult to recognize on the common offset distance channel set; in order to avoid using the wrong in-phase axis to obtain the wrong result, the in-phase axis needs to be screened by combining the local rock wave speed test result, and the in-phase axis of the real record body wave is screened.
As shown in FIG. 3, the speed pickup result on the single shot record is compared and analyzed to find that the same phase axis with the speed of 1883.99m/s is most consistent with the rock speed in the construction area, so that the same phase axis corresponding to the speed is the same phase axis corresponding to the target layer, and the corresponding same phase axis of the target layer can be found on each common offset seismic section by counting travel time corresponding to different offset distances on the same phase axis.
S22, counting dominant frequencies of the phase axes of the target layer at different offset distances, and calculating wavelength, wherein the wavelength can be obtained by the following formula:
wherein lambda is the wavelength of the bulk wave, v is the wave velocity of the bulk wave, and f is the frequency of the bulk wave.
S23, based on the body wave wavelength, calculating the depth corresponding to the body waves with different offset distances, wherein the depth is about 1/2 to 1/3 of the body wave wavelength.
In this embodiment, as shown in fig. 4, by performing spectrum analysis on the seismic traces with different offset distances of the same phase axis of the target layer, the dominant frequency f of the same phase axis on each trace can be obtained. FIG. 4 is a graph showing the extraction of spectral analysis data of three near (16 m), middle (42 m) and far (66 m) seismic traces with significant dominant frequencies in spectral analysis, and it can be seen that the dominant frequency of bulk waves on the 14m offset seismic traces is 80Hz; the dominant frequency of the bulk wave on the 42m offset seismic trace is 50Hz; the dominant frequency of the bulk wave on the 66m offset seismic trace is 40Hz, and the frequency band range of the acquired data is about 40-80 Hz. And counting the longitudinal wave speed v of the selected homophase shaft in the single shot data and the dominant frequency f in the homophase shaft spectrum analysis, so that the following corresponding relation from the time domain to the depth domain can be obtained, and lambda/2 is selected as a time-depth conversion coefficient after the corresponding relation is compared with a construction site.
The following table one shows the time-depth conversion relationship in this embodiment:
Track offset | Wave speed m/s | Dominant frequency Hz | λ/2m | Time ms |
14m | 847.79 | 80 | 5.30 | 21.2 |
42m | 1883.99 | 50 | 18.84 | 49.2 |
66m | 1883.99 | 40 | 23.55 | 62.2 |
List one
S3, obtaining a common offset profile, extracting spectrum decomposition information of a target layer corresponding to the phase axis in different offset profiles, and converting the spectrum decomposition information into a spectrum decomposition attribute profile, wherein the method specifically comprises the following steps of:
S31, extracting common offset seismic traces in each single shot record to form a common offset seismic section.
As shown in fig. 5, in this embodiment, the seismic traces with offset of 6m, 16m, 26m, 36m, 46m, 56m, 66m in each single shot record are extracted to form a common offset profile.
S32, according to the dominant frequency distribution range of the seismic profile obtained in the step S22, performing spectrum decomposition on the common offset profile to obtain seismic body wave information corresponding to different frequencies, and decomposing the seismic data into a series of time domain discrete frequency amplitude data volumes by using a spectrum decomposition method, wherein the spectrum decomposition method comprises but is not limited to a short-time Fourier transform, a continuous wavelet transform, an S transform or a synchronous extrusion transform method.
In this embodiment, according to the single shot time-frequency analysis result, the common offset section is decomposed into 20-70 Hz (one spectrum decomposition section is extracted every 5 Hz), the partial spectrum decomposition attribute (middle frequency 45 Hz) section is shown in fig. 6, the seismic body wave reflects the integral property of lithology in the detection range, the larger the offset, the deeper the lithology reflected by the body wave, and the surrounding rock property in the exploration range can be qualitatively analyzed through the strong and weak changes of the spectrum decomposition section energy of different offset.
S33, extracting the layer-along spectrum decomposition attribute corresponding to the target layer phase axis in each spectrum decomposition profile according to the travel of the phase axis corresponding to the different offset distances obtained in the S21, carrying out normalization processing on the layer-along attribute value, and drawing the layer-along spectrum decomposition attribute of the different offset distances in the same spectrum decomposition attribute profile.
The normalization method used above is:
Wherein A is the value of the decomposition attribute along the target layer spectrum, A max is the maximum value of the decomposition attribute along the target layer spectrum, and A min is the minimum value of the decomposition attribute along the target layer spectrum.
In the embodiment, the along-layer spectrum decomposition attribute of the same phase axis of the target layer is extracted, a section with the offset distance as an ordinate and the track number as an abscissa is constructed, and the deeper the stratum corresponding to the lithology condition is reflected along with the increase of the offset distance, so that the section with the offset distance as the ordinate can qualitatively reflect the quality of surrounding rocks at different depths in the underground; as the spectral decomposition energy becomes weaker with the increase of the offset distance, the spectral decomposition properties of the layers of different offset distances are not comparable, so that normalization processing is carried out on the spectral decomposition properties of the layers of layers before the cross section is drawn.
The spectrum decomposition attribute profile obtained after normalizing the edge layer spectrum decomposition attribute is shown in fig. 7, the spectrum decomposition attributes with different offset distances are drawn on the same spectrum decomposition profile through the operation, and only the edge layer attribute of the same phase axis of the target layer is reserved, so that the change characteristic of the spectrum decomposition attribute characteristic along with the depth is more obvious.
S4, respectively processing the spectrum decomposition attribute sections of the high frequency band, the middle frequency band and the low frequency band by using a PCA machine learning method to obtain a main component section of the seismic attribute, wherein the method specifically comprises the following steps:
s41, according to the dominant frequency distribution range of the seismic section obtained in the step S22, the spectrum decomposition seismic attribute is divided into three types of low frequency, medium frequency and high frequency.
In this embodiment, spectrum analysis is performed on the phase axis of the target layer, and it is found that the frequency band range of the phase axis of the target layer is between 20Hz and 80Hz, and the main frequency of the whole seismic wave is about 45Hz, so that the spectrum decomposition attribute smaller than 40Hz is divided into low frequency attribute, the spectrum decomposition attribute between 40Hz and 50Hz is divided into medium frequency attribute, and the spectrum decomposition attribute larger than 50Hz is divided into high frequency attribute.
S42, PCA principal component analysis is carried out on the three kinds of spectrum decomposition profiles of the low frequency, the medium frequency and the high frequency respectively, and the contribution rate corresponding to each principal component is calculated.
S43, selecting principal component information with highest contribution rate as principal component profiles corresponding to low-frequency, medium-frequency and high-frequency information.
The PCA machine learning method comprises the following core steps:
U=ZTb,
where U is the principal component matrix of the attribute, Z T is the transpose matrix of the normalized input attribute matrix, and b is the unit eigenvector of the input attribute dataset correlation coefficient matrix.
In the present embodiment, taking the hyperspectral decomposition attribute as an example, the characteristic value and the contribution rate of the principal component of the hyperspectral decomposition are shown in the following table, it can be seen that the principal component "PC1" with the highest contribution rate represents that "PC1" contributes 92.1% of the information amount to the hyperspectral decomposition, and "PC1" can represent most of the hyperspectral decomposition information, so "PC1" is selected as the principal component profile of all the hyperspectral decomposition attribute profiles; the principal component analysis process of the low-frequency and medium-frequency spectrum decomposition attribute is similar, and the principal component profile of the spectrum decomposition information of the three groups of frequencies is shown in fig. 8.
The second table below shows the principal component profile of PC1, PC2, PC3, and PC4 in this example:
Main component | Eigenvalues | Contribution rate | Cumulative contribution rate |
PC1 | 0.309 | 0.921 | 0.921 |
PC2 | 0.025 | 0.076 | 0.997 |
PC3 | 0.001 | 0.003 | 1 |
PC4 | 0.0001 | 0 | 1 |
Watch II
S5, quantitatively predicting the surrounding rock grade by using a K-means clustering method, and specifically comprising the following steps of:
S51, converting the principal component profile into a quantitative depth domain principal component profile with the ordinate being the depth and the abscissa being the length of the measuring line according to the time-depth conversion relation obtained in the step S2.
S52, judging the classification grade and the quantity of surrounding rocks of the area to be predicted by combining the local geological profile.
S53, quantitatively dividing the surrounding rock grade by using a K-means clustering method by using the main component section corresponding to the low-frequency, medium-frequency and high-frequency information.
The K-means machine learning core steps in the step are as follows:
And randomly extracting k initial clustering centers C i (i is more than or equal to 1 is less than or equal to k) from an input data set, calculating the distance between the rest data objects and the clustering center C i, finding out the clustering center C i closest to the target data object, distributing the data objects into clusters corresponding to the clustering center C i, forming a new clustering center according to the value of the data object in each cluster, and performing the next iteration until the clustering center is not changed or the maximum iteration number is reached.
The invention also provides a surrounding rock grading system based on the earthquake body wave and the machine learning, which comprises a data input module, a data processing module and an output module;
The data input module is used for inputting seismic body wave data;
the data processing module stores a surrounding rock grading method based on earthquake body waves and machine learning, and the surrounding rock is graded by calling the grading method;
the output module comprises a terminal or a cloud server.
The scheme is based on a shallow seismic body wave detection section, adopts Matlab programming language as technical support, provides a surrounding rock grading method and grading system based on seismic body waves and machine learning through a series of technical means of time-depth conversion, spectrum decomposition, PCA and K-means machine learning, establishes connection between surrounding rock grades and elastic wave detection results, simplifies field data acquisition difficulty, gets rid of dependence on working experience of field operators, realizes high-resolution, spatially continuous and accurate quantitative prediction of the surrounding rock grades, is beneficial to reducing construction cost, reduces working period and ensures construction safety of highway tunnels.
In the description of the present specification, the descriptions of the terms "one embodiment," "example," "specific example," and the like, mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiments or examples. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
The foregoing is merely illustrative and explanatory of the invention, as various modifications and additions may be made to the particular embodiments described, or in a similar manner, by those skilled in the art, without departing from the scope of the invention or exceeding the scope of the invention as defined in the claims.
Claims (10)
1. The surrounding rock grading method based on the earthquake body wave and the machine learning is characterized by comprising the following steps of:
S1, collecting shallow seismic body wave data at a position to be detected;
S2, establishing a time depth conversion relation;
s3, obtaining a common offset profile, extracting spectrum decomposition information of the target layer corresponding to the phase axis in the profiles with different offset distances, and converting the spectrum decomposition information into a spectrum decomposition attribute profile;
s4, respectively processing the spectrum decomposition attribute sections of the high frequency band, the middle frequency band and the low frequency band by using a PCA machine learning method to obtain a seismic attribute main component section;
S5, performing quantitative prediction on the surrounding rock grade by using a K-means clustering method.
2. The surrounding rock grading method based on seismic body waves and machine learning according to claim 1, wherein the step S2 comprises the steps of:
s21, obtaining a single shot record based on body wave exploration, obtaining a same phase axis corresponding to a target layer, and counting the same phase axis travel time corresponding to different offset distances;
s22, counting dominant frequencies of the phase axes of the target layer at different offset distances, and calculating wavelength, wherein the wavelength can be obtained by the following formula:
Wherein lambda is the wavelength of the bulk wave, v is the wave velocity of the bulk wave, and f is the frequency of the bulk wave;
S23, based on the body wave wavelength, calculating the depth corresponding to the body waves with different offset distances, wherein the depth is about 1/2 to 1/3 of the body wave wavelength.
3. The surrounding rock grading method based on seismic body waves and machine learning according to claim 1, wherein the step S3 comprises the steps of:
S31, extracting common offset seismic traces in each single shot record to form a common offset seismic section;
s32, performing spectrum decomposition on the co-offset profile according to the dominant frequency distribution range of the seismic profile obtained in the step S22 to obtain seismic body wave information corresponding to different frequencies;
And S33, extracting the layer-following spectrum decomposition attribute corresponding to the target layer phase axis in each spectrum decomposition profile according to the travel of the phase axis obtained in the step S21 at different offset distances, carrying out normalization processing on the layer-following attribute value, and drawing the layer-following spectrum decomposition attribute of different offset distances in the same spectrum decomposition attribute profile.
4. The surrounding rock grading method based on seismic body waves and machine learning according to claim 1, wherein the step S4 comprises the steps of:
S41, according to the dominant frequency distribution range of the seismic section obtained in the step S22, dividing the spectrum decomposition seismic attribute into three types of low frequency, medium frequency and high frequency;
S42, performing principal component analysis on the three types of spectrum decomposition profiles of low frequency, medium frequency and high frequency respectively, and calculating the contribution rate corresponding to each principal component;
s43, selecting principal component information with highest contribution rate as principal component profiles corresponding to low-frequency, medium-frequency and high-frequency information.
5. The surrounding rock grading method based on seismic body waves and machine learning according to claim 1, wherein the step S5 comprises the steps of:
S51, converting the principal component profile into a quantitative depth domain principal component profile with the ordinate being depth and the abscissa being the length of the measuring line according to the time-depth conversion relation obtained in the step S2;
S52, judging the classification grade and the quantity of surrounding rocks of the area to be predicted by combining the local geological profile;
S53, quantitatively dividing the surrounding rock grade by using a K-means clustering method by using the main component section corresponding to the low-frequency, medium-frequency and high-frequency information.
6. The surrounding rock grading method based on seismic body waves and machine learning according to claim 1, wherein the PCA machine learning method in step S4 comprises the following core steps:
U=ZTb,
where U is the principal component matrix of the attribute, Z T is the transpose matrix of the normalized input attribute matrix, and b is the unit eigenvector of the input attribute dataset correlation coefficient matrix.
7. The surrounding rock grading method based on seismic body waves and machine learning according to claim 1, wherein the K-means machine learning core step in the step S5 is as follows:
And randomly extracting k initial clustering centers C i (i is more than or equal to 1 is less than or equal to k) from an input data set, calculating the distance between the rest data objects and the clustering center C i, finding out the clustering center C i closest to the target data object, distributing the data objects into clusters corresponding to the clustering center C i, forming a new clustering center according to the value of the data object in each cluster, and performing the next iteration until the clustering center is not changed or the maximum iteration number is reached.
8. A method of classification of surrounding rock based on seismic body waves and machine learning according to claim 3, wherein in step S32, the seismic data is decomposed into a series of time domain discrete frequency amplitude data volumes using a spectral decomposition method including but not limited to short time fourier transform, continuous wavelet transform, S transform or synchro extrusion transform methods.
9. A surrounding rock grading method based on seismic body waves and machine learning according to claim 3, wherein in the step S33, the normalization method is as follows:
Wherein A is the value of the decomposition attribute along the target layer spectrum, A max is the maximum value of the decomposition attribute along the target layer spectrum, and A min is the minimum value of the decomposition attribute along the target layer spectrum.
10. The surrounding rock grading system based on earthquake body waves and machine learning is characterized by comprising a data input module, a data processing module and an output module;
The data input module is used for inputting seismic body wave data;
The data processing module stores a surrounding rock grading method based on earthquake body waves and machine learning, and the surrounding rock is graded by calling the grading method;
the output module comprises a terminal or a cloud server.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410111491.0A CN117970479A (en) | 2024-01-26 | 2024-01-26 | Surrounding rock grading method and grading system based on earthquake body waves and machine learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410111491.0A CN117970479A (en) | 2024-01-26 | 2024-01-26 | Surrounding rock grading method and grading system based on earthquake body waves and machine learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117970479A true CN117970479A (en) | 2024-05-03 |
Family
ID=90854858
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202410111491.0A Pending CN117970479A (en) | 2024-01-26 | 2024-01-26 | Surrounding rock grading method and grading system based on earthquake body waves and machine learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117970479A (en) |
-
2024
- 2024-01-26 CN CN202410111491.0A patent/CN117970479A/en active Pending
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105221133B (en) | A kind of method and apparatus that content of organic carbon of hydrocarbon source rock is determined based on well logging multi-parameter | |
CN106154323B (en) | The thin method for predicting reservoir of phased stochastic inverse of frequency processing is opened up based on earthquake | |
CN103728659B (en) | A kind of method improving detection of karst cave precision | |
CN105137498A (en) | Underground target detection and recognition system and method based on feature fusion | |
CN107748399B (en) | Method for identifying deep tectonic layer of mountain front zone by utilizing gravity interface inversion | |
Bonal et al. | Wavelet analysis of bender element signals | |
CN105445800A (en) | Thick sand body top differentiation lithologic reservoir identification method | |
Galvis et al. | Seismic attribute selection and clustering to detect and classify surface waves in multicomponent seismic data by using k-means algorithm | |
Shi et al. | Brittleness index prediction in shale gas reservoirs based on efficient network models | |
CN105044793A (en) | Inversion method and apparatus for multi-channel transient electromagnetic detection data | |
CN111622201B (en) | Soil layer equivalent shear wave velocity direct calculation method based on surface wave frequency dispersion characteristics | |
Farrugia et al. | Noninvasive techniques for site characterization of Alberta seismic stations based on shear‐wave velocity | |
Mola-Abasi et al. | Shear wave velocity by polynomial neural networks and genetic algorithms based on geotechnical soil properties | |
CN102253414B (en) | Reservoir detecting method based on analysis of earthquake lines | |
CN112946752B (en) | Method for predicting fracture probability body based on fracture density model | |
CN112835124B (en) | Crack effectiveness evaluation method based on imaging logging and array acoustic logging data | |
CN117970479A (en) | Surrounding rock grading method and grading system based on earthquake body waves and machine learning | |
CN106934725A (en) | Rock formation median radius forecast model method for building up, apparatus and system | |
Young et al. | Seismic attenuation spectra in rock mass characterization; a case study in open-pit mining | |
CN114482995A (en) | Fine determination method for argillaceous content of fine-grain sediment | |
Rong et al. | Improvement of the objective function in the velocity structure inversion based on horizontal-to-vertical spectral ratio of earthquake ground motions | |
Huang et al. | New seismic attribute: Fractal scaling exponent based on gray detrended fluctuation analysis | |
CN115266914B (en) | Pile sinking quality monitoring system and method based on acoustic signal processing | |
RU2205434C1 (en) | Method of geophysical prospecting to establish oil-and-gas productive types of geological section of variable thickness | |
CN111679326B (en) | Transient broadband electromagnetic detection signal processing method based on geological detection |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination |