CN116299665A - LSTM surface wave inversion method, device and medium - Google Patents
LSTM surface wave inversion method, device and medium Download PDFInfo
- Publication number
- CN116299665A CN116299665A CN202211090913.8A CN202211090913A CN116299665A CN 116299665 A CN116299665 A CN 116299665A CN 202211090913 A CN202211090913 A CN 202211090913A CN 116299665 A CN116299665 A CN 116299665A
- Authority
- CN
- China
- Prior art keywords
- stratum
- dispersion
- lstm
- fuzzy
- formation
- 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
- 238000000034 method Methods 0.000 title claims abstract description 71
- 239000006185 dispersion Substances 0.000 claims abstract description 87
- 238000012549 training Methods 0.000 claims abstract description 47
- 230000006870 function Effects 0.000 claims abstract description 38
- 238000003384 imaging method Methods 0.000 claims abstract description 12
- 238000007781 pre-processing Methods 0.000 claims abstract description 8
- 230000015572 biosynthetic process Effects 0.000 claims description 31
- 239000011159 matrix material Substances 0.000 claims description 7
- 239000013598 vector Substances 0.000 claims description 6
- 238000012216 screening Methods 0.000 claims description 4
- 238000012545 processing Methods 0.000 abstract description 9
- 238000010586 diagram Methods 0.000 description 10
- 230000000694 effects Effects 0.000 description 9
- 238000013135 deep learning Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000007796 conventional method Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000013016 damping Methods 0.000 description 2
- 238000005553 drilling Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000009191 jumping Effects 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012797 qualification Methods 0.000 description 1
- 238000002922 simulated annealing Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/28—Processing seismic data, e.g. analysis, for interpretation, for correction
- 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. analysis, for interpretation, for correction
- G01V1/30—Analysis
- G01V1/307—Analysis for determining seismic attributes, e.g. amplitude, instantaneous phase or frequency, reflection strength or polarity
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Abstract
The invention discloses an LSTM surface wave inversion method, a device and a medium, wherein the method comprises the steps of determining a fuzzy stratum parameter interval of a work area to be inverted; according to the fuzzy stratum parameter interval, different models are randomly generated, theoretical plane wave dispersion curves of the models are calculated by using a generalized reflection-transmission coefficient method, and training sample data pairs are constructed; preprocessing the training sample data pair; constructing an LSTM network based on the FHLV loss function, training the LSTM network based on the preprocessed training sample data pair, and storing a trained model; and (3) automatically picking up a dispersion curve from actual dispersion imaging data by using unsupervised learning, and predicting the dispersion curve with uniform dimension by using the trained model to obtain a near-surface transverse wave velocity model. The invention improves the processing efficiency and inversion precision, and is suitable for large-scale data processing.
Description
Technical Field
The invention belongs to the field of near-surface exploration seismic data processing, and particularly relates to an LSTM surface wave inversion method, an LSTM surface wave inversion device and a medium.
Background
Surface wave analysis methods are widely used to build near-surface shear wave velocity structures. The near-surface ground transverse wave speed structure is obtained through inversion of a surface wave dispersion curve regardless of an active source or a passive source surface wave analysis method. Inversion of the surface wave dispersion curve is a typical highly nonlinear, multiparameter geophysical inversion problem. The conventional inversion method of the surface wave dispersion curve is divided into two major types, wherein one type is a local linearization method, such as least square method, damping least square inversion and the like; because the forward dispersion equation of the rayleigh wave in the layered medium is a nonlinear function, when the selected initial model is improper, the local linearization method is difficult to find the global optimal solution of the objective function. The other type is a global nonlinear optimization method, and genetic algorithm, simulated annealing inversion and the like are commonly used. Such algorithms can avoid to some extent the dependence of the local linearization inversion method on the initial model, however in practical applications the local search capability is not strong and the time is long. Therefore, the face inversion method needs to be explored in a new field.
At present, various traditional surface wave dispersion curve inversion methods have certain use limitations. Deep learning has the ability to solve many non-linear problems, and in recent years, machine learning and deep learning have shown great potential when applied to various geophysical research problems, providing automated performance in some tasks. Therefore, in order to solve the problems of low inversion efficiency, poor inversion effect and the like in the traditional surface wave exploration, the LSTM surface wave inversion method based on the FHLV loss function is provided, the processing efficiency and the inversion precision are improved, and the method is suitable for large-scale data processing.
Disclosure of Invention
The invention aims to solve the technical problems of the background technology and provides an LSTM surface wave inversion method, an LSTM surface wave inversion device and a LSTM surface wave inversion medium. In particular, where a single data consumer generates a perceived data demand, multiple data owners compete for participation in a shared task qualification. In the method, a blockchain technology is adopted to solve the trust problem brought by a trusted third party. And an incentive mechanism is designed based on the reverse auction model, so that miners are helped to screen out data owners of irrational quotations, the workload of subsequently verifying the quality level of the data is reduced, and the performance of the incentive model is improved. The quality level of the perceived data is calculated using a softmax regression algorithm. And finally calculating the value of the data through the quotation of the data owner and the quality grade of the data, and distributing rewards according to different data values to encourage the data owner to upload the data with reasonable price and high quality and reliability.
The specific technical scheme of the invention is as follows:
according to a first aspect of the present invention there is provided a method of LSTM surface wave inversion, the method comprising:
determining a fuzzy stratum parameter interval of a work area to be inverted;
according to the fuzzy stratum parameter interval, different models are randomly generated, theoretical plane wave dispersion curves of the models are calculated by using a generalized reflection-transmission coefficient method, and training sample data pairs are constructed;
preprocessing the training sample data pair;
constructing an LSTM network based on the FHLV loss function, training the LSTM network based on the preprocessed training sample data pair, and storing a trained model;
and (3) automatically picking up a dispersion curve from actual dispersion imaging data by using unsupervised learning, and predicting the dispersion curve with uniform dimension by using the trained model to obtain a near-surface transverse wave velocity model.
Further, the fuzzy stratum parameter interval of the work area to be inverted is determined by using known logging information or according to an existing inversion method.
Further, the formation parameters include formation shear wave velocity for each layerThe degree, the stratum longitudinal wave speed, the stratum thickness and the stratum density are 2.4-3 times of the stratum transverse wave speed of the same stratum, and the stratum density is 1.5-2Kg/m 3 The fuzzy stratum parameter interval is a fuzzy interval range of stratum parameters of each layer, and is determined by the minimum value and the maximum value of the parameters of each layer in the extracted one-dimensional speed structures at different positions, the initial range of the fuzzy stratum parameter interval is 40-80% of the minimum value, the end range is 120-140% of the maximum value, the larger the numerical value of the stratum parameters is, the smaller the proportion of the corresponding initial range and the end range is, and the one-dimensional speed structures at different positions are obtained according to known logging information or the existing inversion method.
Further, the preprocessing the training sample data pair includes:
normalizing the training sample data pair, and vectorizing the stratum transverse wave speed and stratum thickness column in the training sample data pair.
Further, the LSTM network based on the FHLV loss function includes three hidden layers, each hidden layer corresponds to 256 LSTM units, each LSTM unit is connected in a jumping manner, and an output end of a last LSTM unit is connected to a full connection layer to control an output dimension.
Further, the training the LSTM network and saving the trained model based on the preprocessed training sample data pair includes:
inputting a theoretical plane wave dispersion curve into the LSTM network, outputting a corresponding stratum transverse wave speed and stratum thickness, and finishing training of the LSTM network;
the loss function of the LSTM network is:
Loss1＝mean(abs(v_label-v))
Loss2＝mean(abs(h_label-h))
where los 1 is the velocity Loss, los 2 is the thickness Loss, FHLV is the actual Loss function during training, label is the label, mean is the average operator, abs is the absolute operator, v is the formation shear wave velocity, h is the formation thickness, e is the ideal velocity Loss, set to 15-20% of the normalized minimum layer velocity, wh is the weight attenuation coefficient, set to 0.01-0.1.
Further, the automatic picking up of the dispersion curve for the actual dispersion imaging data by using the unsupervised learning includes:
determining a surface wave dispersion energy matrix according to actual dispersion imaging data;
normalizing the surface wave dispersion energy matrix, and taking the coordinates of points larger than a first dispersion energy threshold value as first dispersion energy points;
clustering the first scattered energy points;
and comparing the frequency dispersion energy corresponding to the clustered points, and screening the phase velocity of the frequency vector corresponding to the local peak value to obtain a final frequency dispersion curve.
According to a second aspect of the present invention there is provided an LSTM surface wave inversion apparatus, the apparatus comprising a processor configured to:
determining a fuzzy stratum parameter interval of a work area to be inverted;
according to the fuzzy stratum parameter interval, different models are randomly generated, theoretical plane wave dispersion curves of the models are calculated by using a generalized reflection-transmission coefficient method, and training sample data pairs are constructed;
preprocessing the training sample data pair;
constructing an LSTM network based on the FHLV loss function, training the LSTM network based on the preprocessed training sample data pair, and storing a trained model;
and (3) automatically picking up a dispersion curve from actual dispersion imaging data by using unsupervised learning, and predicting the dispersion curve with uniform dimension by using the trained model to obtain a near-surface transverse wave velocity model.
Further, the formation parameters include formation shear wave velocity, formation longitudinal wave velocity, formation thickness, and formation density for each layer, wherein the formationThe longitudinal wave velocity of the stratum is 2.4-3 times of the transverse wave velocity of the stratum at the same stratum, and the stratum density is 1.5-2Kg/m 3 The fuzzy stratum parameter interval is a fuzzy interval range of stratum parameters of each layer, and is determined by the minimum value and the maximum value of the parameters of each layer in the extracted one-dimensional speed structures at different positions, the initial range of the fuzzy stratum parameter interval is 40-80% of the minimum value, the end range is 120-140% of the maximum value, the larger the numerical value of the stratum parameters is, the smaller the proportion of the corresponding initial range and the end range is, and the one-dimensional speed structures at different positions are obtained according to known logging information or the existing inversion method.
Further, the LSTM network based on the FHLV loss function includes three hidden layers, each hidden layer corresponds to 256 LSTM units, each LSTM unit is connected in a jumping manner, and an output end of a last LSTM unit is connected to a full connection layer to control an output dimension.
Further, the processor is further configured to:
inputting a theoretical plane wave dispersion curve into the LSTM network, outputting a corresponding stratum transverse wave speed and stratum thickness, and finishing training of the LSTM network;
the loss function of the LSTM network is:
Loss1＝mean(abs(v_label-v))
Loss2＝mean(abs(h_label-h))
where los 1 is the velocity Loss, los 2 is the thickness Loss, FHLV is the actual Loss function during training, label is the label, mean is the average operator, abs is the absolute operator, v is the formation shear wave velocity, h is the formation thickness, e is the ideal velocity Loss, set to 15-20% of the normalized minimum layer velocity, wh is the weight attenuation coefficient, set to 0.01-0.1.
Further, the processor is further configured to:
determining a surface wave dispersion energy matrix according to actual dispersion imaging data;
normalizing the surface wave dispersion energy matrix, and taking the coordinates of points larger than a first dispersion energy threshold value as first dispersion energy points;
clustering the first scattered energy points;
and comparing the frequency dispersion energy corresponding to the clustered points, and screening the phase velocity of the frequency vector corresponding to the local peak value to obtain a final frequency dispersion curve.
According to a third aspect of the present invention there is provided a computer readable storage medium having stored thereon computer readable instructions which, when executed by a processor of a computer, cause the computer to perform an LSTM face wave inversion method as described in the various embodiments of the present invention.
According to the LSTM plane wave inversion method, the LSTM plane wave inversion device and the LSTM plane wave inversion medium provided by the embodiments of the invention, the processing efficiency and the inversion precision are improved, and the LSTM plane wave inversion method, the LSTM plane wave inversion device and the LSTM plane wave inversion medium are suitable for large-scale data processing. The stratum in the same area has continuity, the inversion of the current dispersion curve can provide reference for the inversion of the subsequent adjacent dispersion curve, and the LSTM network has memory characteristics and is suitable for the feature extraction of one-dimensional data, so that the LSTM network is used for solving the problem of inversion of the surface wave dispersion curve, and has good applicability. The LSTM network can learn the nonlinear mapping between the surface wave dispersion curve and the near-surface transverse wave velocity structure by excavating a large number of data features, but because the transverse wave velocity and the stratum thickness have large difference between the value ranges of two data volumes, the dispersion curve is insensitive to the thickness, and the network based on the conventional loss function often has difficulty in accurately predicting the transverse wave velocity and the stratum thickness at the same time. The FHLV loss function is designed to improve the learning ability of the network to the thickness parameter, the loss function can keep the optimized balance of the stratum speed and the stratum thickness, the network convergence is better, the inversion precision is higher, the LSTM surface wave inversion method based on the FHLV loss function is formed, and finally the high-precision near-surface transverse wave speed structure is obtained.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below. Like elements or portions are generally identified by like reference numerals throughout the several figures. In the drawings, elements or portions thereof are not necessarily drawn to scale.
Fig. 1 is a process flow diagram of an LSTM surface wave inversion method based on FHLV loss function.
Fig. 2 is a schematic diagram of a theoretical two-dimensional velocity model composed of 2000 one-dimensional velocity structures.
Fig. 3 is a schematic diagram of a model, wherein (a) - (c) respectively represent a real two-dimensional shear wave velocity model schematic diagram, a two-dimensional shear wave velocity model schematic diagram predicted by MAE, and a two-dimensional shear wave velocity model schematic diagram predicted by the present invention.
FIG. 4 is a schematic diagram of predicted absolute differences, wherein (a) represents MAE predicted absolute differences and (b) represents predicted absolute differences according to the present invention.
FIG. 5 is a graph of MAE versus one-dimensional velocity structure predicted by the present invention versus true velocity structure, (a) is a 420 th pass inversion result versus graph, (b) is a 1000 th pass inversion result versus graph, and (c) is a 1000 th pass inversion result versus graph.
Fig. 6 is a schematic diagram of actual data, in which (a) represents a positive offset portion of a single shot and (b) represents a corresponding high resolution dispersion image.
FIG. 7 is a graph of the inversion result forward dispersion curve versus the observation dispersion curve of the present invention, wherein (a) is the 60 th inversion result comparison graph, (b) is the 100 th inversion result comparison graph, and (c) is the 260 th inversion result comparison graph.
FIG. 8 is a graph showing the verification of inversion effects, wherein (a) is a schematic diagram of a two-dimensional shear wave velocity model obtained by inversion according to the invention, (b) is a graph comparing the effects of the invention with conventional methods and well logging data, and (c) is a schematic diagram of drilling layering in region A.
FIG. 9 shows the LSTM deep neural network structure for the inversion of the surface wave dispersion curve, wherein X0, … Xn are dispersion curve points, H0, … Hn are output results of the LSTM network, P1, … Pn are network outputs after dimension control, pn represents predicted model parameters, and dimension change is explained on the left side.
FIG. 10 is a flow chart of an unsupervised learning automatic pickup dispersion curve according to the present invention, wherein (a) represents a normalized dispersion energy map, (b) represents a larger energy point of interest, (c) represents a clustered point, (d) represents a peak energy point of a picked frequency vector, and (e) represents a final pickup effect map.
Detailed Description
The following description of the technical solutions in the embodiments of the present invention will be clear and complete, and it is obvious that the described embodiments are only some embodiments of the present invention, but 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.
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
The invention will now be further described with reference to the accompanying drawings.
Because of the low cost and no damage, the surface wave imaging method has great potential in urban near-surface exploration and is widely focused by geophysicists. The core is to construct an accurate near-surface transverse wave velocity model, and the accurate inversion of the surface wave dispersion curve is definitely the foundation of the construction of the near-surface transverse wave velocity model. The invention provides a surface wave inversion method through deep learning network and improved loss function fusion, in particular to an LSTM surface wave inversion method.
The basic technical principle of the invention is as follows: determining a fuzzy stratum parameter interval of a work area to be inverted by using known logging information or a conventional inversion method, randomly generating a large number of different models in the stratum model parameter fuzzy interval, and calculating a corresponding theoretical surface wave dispersion curve by a generalized reflection-transmission coefficient method so as to construct a large number of sample data pairs. And carrying out nonlinear mapping between the training learning surface wave dispersion curve and the near-surface transverse wave velocity structure on the sample data pair through an LSTM network based on the FHLV loss function, and predicting the data to be inverted to obtain an accurate near-surface transverse wave velocity model. Networks based on conventional loss functions have difficulty learning the characteristics of thickness parameters to fully converge, resulting in inaccurate inversion of thickness parameters for multi-layer models and fine-layer models. The FHLV loss function is the core of the proposed method, and consists of two parts, namely speed loss and thickness loss, and optimizes the learning of the network on the thickness parameters, thereby improving the overall prediction precision and better solving the problem of poor thickness prediction effect. And finally, an LSTM surface wave inversion method based on the FHLV loss function is formed, and a high-precision near-surface transverse wave speed structure is obtained.
Referring to fig. 1, a flowchart of an LSTM surface wave inversion method is shown, and the specific processing steps of the method are as follows:
1) The fuzzy formation parameter interval of the work area to be inverted is determined by using known logging information or by a conventional inversion method.
The number of stratum model layers is set, the determined stratum parameters mainly comprise stratum transverse wave speed VS and stratum thickness H of each stratum, the corresponding stratum longitudinal wave speed VP is set to be 2.4-3 times VS, and the stratum density is set to be 1.5-2Kg/m3. The range of the fuzzy interval of each layer of stratum parameters is determined by the minimum value and the maximum value of each layer of parameters in the extracted one-dimensional speed structures at different positions, the initial range is 40% -80% of the minimum value, the end range is 120% -140% of the maximum value, the larger the stratum parameter value is, the smaller the proportion is, and the one-dimensional speed structures at different positions correspond to the inversion result of known logging information or other methods.
2) Randomly generating a large number of different models in a stratum model parameter fuzzy interval, and calculating a corresponding theoretical surface wave dispersion curve by a generalized reflection-transmission coefficient method, so as to construct a large number of sample data pairs;
3) Constructing an LSTM network based on the FHLV loss function, preprocessing a large amount of generated sample data, sending the preprocessed sample data into the network for training, and storing a trained model;
the LSTM based on the FHLV loss function takes an LSTM unit module as a main body, wherein three hidden layers are provided, each layer corresponds to 256 LSTM units, jump connection is adopted between the layers, and the output of the last LSTM unit is connected with a full connection layer to control the output dimension; inputting a surface wave dispersion curve into the LSTM, outputting a corresponding stratum model transverse wave speed and stratum thickness, and finishing training of the LSTM; in the training process, the error of the output stratum model parameters and the sample labels is defined by FHLV Loss functions, wherein the FHLV Loss functions are the core of the proposed method, and consist of two parts, namely speed Loss1 and thickness Loss2, so that the learning of the network on the thickness parameters is optimized, and the overall prediction accuracy is improved, and the method is characterized by the following steps:
Loss1＝mean(abs(v_label-v))
Loss2＝mean(abs(h_label-h))
label is the label, mean is the average operator, abs is the absolute operator, v is the formation shear wave velocity, h is the formation thickness, e is the ideal velocity loss, typically set to 15-20% of the normalized minimum layer velocity, wh is the weight decay coefficient, typically set to 0.01-0.1.
4) The method comprises the steps of obtaining an actual surface wave dispersion curve by using a dispersion imaging method, automatically picking up the dispersion curve by using an unsupervised clustering method, and further obtaining a near-surface transverse wave velocity model by inversion of a trained neural network.
The method comprises the steps of extracting energy peaks of corresponding frequency vectors from the thought angle of people, automatically picking up the frequency dispersion curves mainly by adopting a DBSCAN clustering method based on density, normalizing a surface wave frequency dispersion energy matrix, selecting coordinates of points with larger energy, clustering the coordinates of the selected frequency dispersion energy points by adopting the DBSCAN clustering method based on density, comparing the frequency dispersion energy corresponding to the clustered points, screening the phase velocity of the frequency vectors corresponding to local peaks, and obtaining the final frequency dispersion curve.
Through the processing of the specific steps, the difficult problem of carrying out high-precision face wave inversion by utilizing deep learning is solved.
In order to verify the LSTM surface wave inversion method based on the FHLV loss function and the inversion effect thereof, the theoretical two-dimensional speed model and the training process and inversion result of near-surface exploration seismic surface wave data of cities in a western area A are taken as examples respectively for analysis.
FIG. 2 is a 6-layer theoretical two-dimensional transverse wave velocity model consisting of 2000 channels of one-dimensional velocity structures. And 5 one-dimensional speed structures (corresponding to logging data or inversion results of a conventional method in an actual work area) are extracted from the model, and a model parameter range is determined. The training set and test set model parameter ranges are shown in table 1. 60000 groups of model parameters are randomly generated in the interval range, and a corresponding base-order dispersion curve is calculated to obtain 60000 pairs of samples. Training and predicting by using the MAE loss function and the FHLV loss function respectively under the same iteration times. 60000 is used for training all samples, and a dispersion curve corresponding to 2000 one-dimensional speed structures in a two-dimensional model is used for testing.
TABLE 1 Low speed thin layer theoretical model test set and training set parameter Range
The following fig. 3 shows that 2000 channel prediction results of the two networks are respectively interpolated to form a two-dimensional near-surface transverse wave velocity model for network inversion, and the two network inversion results basically reflect the change trend of the model. To compare the inversion effects, it can be observed from the inversion error map of fig. 4 (a), 4 (b) that the model based on the net inversion of the MAE loss function differs significantly from the original model in some local places (especially at the red boxes). On the integral inversion effect, the absolute difference between the inversion result based on the FHLV loss function and the real model is smaller, the inversion result is more consistent with the original model, and the inversion precision is higher. For better comparison of the details of the low-speed thin-layer model, three inverse one-dimensional velocity structures were extracted from the test set and compared with the model velocity structures, which were located at the red lines i, ii, iii in fig. 3 (a). From fig. 5 (a) - (c), it can be found that the network inversion result based on FHLV loss function according to the present invention achieves better effect on the whole, especially the accuracy of thickness parameter prediction is higher.
To further illustrate the effectiveness of the method, a near-surface survey seismic surface data of a city in a western region A is selected, and FIG. 6 (q) shows a typical single shot record in which significant surface wave information is visible. Fig. 6 (b) shows a corresponding dispersion image from which significant base-order dispersion energy can be identified.
Inversion is carried out on the near-surface transverse wave speed structure within the range of 1.2km of the acquired data, and 300 groups of dispersion curves are picked up. The initial model was set to invert the dispersion curves at 5 different positions according to the layering ratio method to determine the range of formation parameters, the specific parameter ranges of which are shown in table 2. And randomly generating 50000 groups of model parameter structures in a given interval range, and calculating a corresponding theoretical dispersion curve to form a large amount of training data. As shown in FIG. 7, the picked 300 groups of dispersion curves are predicted, the inversion result forward data and the observation data are well matched, and the correlation degree is greater than 92%. Since each inverted 1D shear wave velocity structure reflects the subsurface structure below the receiver, the 1D inversion results of all the 300 picked up dispersion curves are arranged on the corresponding coordinates and interpolated smoothly to obtain the final 2D shear wave velocity model. As shown in FIG. 8, the inversion result of 20m is shown, and by combining the existing drilling data (the well position is positioned at the red line position in FIG. 8 (a)) of the work area, the network inversion result can be found to be basically consistent with the actual near-surface situation compared with the damping least square inversion result, so that the accuracy and the practicability of the automatic inversion method are further demonstrated.
Table 2A actual data training set parameter ranges
Through theoretical data and two data experiments in area A, an LSTM surface wave inversion method based on FHLV loss function can be used to obtain a relatively accurate near-surface transverse wave velocity structure. Through comparison with the conventional method and logging data, the research method can effectively improve inversion accuracy.
The above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the invention, and are intended to be included within the scope of the appended claims and description.
Claims (10)
1. A method of LSTM face wave inversion, the method comprising:
determining a fuzzy stratum parameter interval of a work area to be inverted;
according to the fuzzy stratum parameter interval, different models are randomly generated, theoretical plane wave dispersion curves of the models are calculated by using a generalized reflection-transmission coefficient method, and training sample data pairs are constructed;
preprocessing the training sample data pair;
constructing an LSTM network based on the FHLV loss function, training the LSTM network based on the preprocessed training sample data pair, and storing a trained model;
and (3) automatically picking up a dispersion curve from actual dispersion imaging data by using unsupervised learning, and predicting the dispersion curve with uniform dimension by using the trained model to obtain a near-surface transverse wave velocity model.
2. The method of claim 1, wherein the section of the fuzzy formation parameters of the work area to be inverted is determined using known logging information or according to existing inversion methods.
3. The method of claim 1, wherein the formation parameters include formation shear wave velocity, formation longitudinal wave velocity, formation thickness, and formation density for each layer, wherein theThe longitudinal wave velocity of the stratum is 2.4-3 times of the transverse wave velocity of the stratum at the same stratum, and the density of the stratum is 1.5-2Kg/m 3 The fuzzy stratum parameter interval is a fuzzy interval range of stratum parameters of each layer, and is determined by the minimum value and the maximum value of the parameters of each layer in the extracted one-dimensional speed structures at different positions, the initial range of the fuzzy stratum parameter interval is 40-80% of the minimum value, the end range is 120-140% of the maximum value, the larger the numerical value of the stratum parameters is, the smaller the proportion of the corresponding initial range and the end range is, and the one-dimensional speed structures at different positions are obtained according to known logging information or the existing inversion method.
4. A method according to claim 3, wherein said pre-processing said training sample data pairs comprises:
normalizing the training sample data pair, and vectorizing the stratum transverse wave speed and stratum thickness column in the training sample data pair.
5. The method of claim 1, wherein the LSTM network based on FHLV loss function comprises three hidden layers, each hidden layer corresponds to 256 LSTM units, each LSTM unit uses a jump connection, and a full connection layer is connected to the output terminal of the last LSTM unit to control the output dimension.
6. The method of claim 5, wherein training the LSTM network and saving the trained model based on the preprocessed training sample data pair, comprises:
inputting a theoretical plane wave dispersion curve into the LSTM network, outputting a corresponding stratum transverse wave speed and stratum thickness, and finishing training of the LSTM network;
the loss function of the LSTM network is:
Loss1＝mean(abs(v_label-v))
Loss2＝mean(abs(h_label-h))
where los 1 is the velocity Loss, los 2 is the thickness Loss, FHLV is the actual Loss function during training, label is the label, mean is the average operator, abs is the absolute operator, v is the formation shear wave velocity, h is the formation thickness, e is the ideal velocity Loss, set to 15-20% of the normalized minimum layer velocity, wh is the weight attenuation coefficient, set to 0.01-0.1.
7. The method of claim 1, wherein automatically picking up the dispersion curve for the actual dispersion imaging data using unsupervised learning comprises:
determining a surface wave dispersion energy matrix according to actual dispersion imaging data;
normalizing the surface wave dispersion energy matrix, and taking the coordinates of points larger than a first dispersion energy threshold value as first dispersion energy points;
clustering the first scattered energy points;
and comparing the frequency dispersion energy corresponding to the clustered points, and screening the phase velocity of the frequency vector corresponding to the local peak value to obtain a final frequency dispersion curve.
8. An LSTM surface wave inversion apparatus, the apparatus comprising a processor configured to:
determining a fuzzy stratum parameter interval of a work area to be inverted;
according to the fuzzy stratum parameter interval, different models are randomly generated, theoretical plane wave dispersion curves of the models are calculated by using a generalized reflection-transmission coefficient method, and training sample data pairs are constructed;
preprocessing the training sample data pair;
constructing an LSTM network based on the FHLV loss function, training the LSTM network based on the preprocessed training sample data pair, and storing a trained model;
and (3) automatically picking up a dispersion curve from actual dispersion imaging data by using unsupervised learning, and predicting the dispersion curve with uniform dimension by using the trained model to obtain a near-surface transverse wave velocity model.
9. The apparatus of claim 8, wherein the formation parameters include formation shear wave velocity, formation longitudinal wave velocity, formation thickness, and formation density for each layer, wherein the formation longitudinal wave velocity is 2.4-3 times the formation shear wave velocity at the same layer, and the formation density is 1.5-2Kg/m 3 The fuzzy stratum parameter interval is a fuzzy interval range of stratum parameters of each layer, and is determined by the minimum value and the maximum value of the parameters of each layer in the extracted one-dimensional speed structures at different positions, the initial range of the fuzzy stratum parameter interval is 40-80% of the minimum value, the end range is 120-140% of the maximum value, the larger the numerical value of the stratum parameters is, the smaller the proportion of the corresponding initial range and the end range is, and the one-dimensional speed structures at different positions are obtained according to known logging information or the existing inversion method.
10. A computer readable storage medium having stored thereon computer readable instructions which, when executed by a processor of a computer, cause the computer to perform the method of any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211090913.8A CN116299665A (en) | 2022-09-07 | 2022-09-07 | LSTM surface wave inversion method, device and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211090913.8A CN116299665A (en) | 2022-09-07 | 2022-09-07 | LSTM surface wave inversion method, device and medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116299665A true CN116299665A (en) | 2023-06-23 |
Family
ID=86794750
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211090913.8A Pending CN116299665A (en) | 2022-09-07 | 2022-09-07 | LSTM surface wave inversion method, device and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116299665A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116819622A (en) * | 2023-08-30 | 2023-09-29 | 北京工业大学 | Background noise level vertical spectrum ratio joint inversion method for soil layer three-dimensional speed structure |
| CN117631029A (en) * | 2024-01-26 | 2024-03-01 | 中国铁路设计集团有限公司 | Rayleigh surface wave dispersion curve inversion method based on joint algorithm |
-
2022
- 2022-09-07 CN CN202211090913.8A patent/CN116299665A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116819622A (en) * | 2023-08-30 | 2023-09-29 | 北京工业大学 | Background noise level vertical spectrum ratio joint inversion method for soil layer three-dimensional speed structure |
| CN116819622B (en) * | 2023-08-30 | 2023-11-21 | 北京工业大学 | Background noise level vertical spectrum ratio joint inversion method for soil layer three-dimensional speed structure |
| CN117631029A (en) * | 2024-01-26 | 2024-03-01 | 中国铁路设计集团有限公司 | Rayleigh surface wave dispersion curve inversion method based on joint algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wu et al. | Seismic impedance inversion using fully convolutional residual network and transfer learning | |
| CN116299665A (en) | LSTM surface wave inversion method, device and medium | |
| CN1239919C (en) | Neural net prediction of seismic streamer shape | |
| CN109709603A (en) | Seismic horizon identification and method for tracing, system | |
| CN105512635A (en) | Category attribute fused deep network underground target identification method and system | |
| CN113359212B (en) | Reservoir characteristic prediction method and model based on deep learning | |
| CN105319589B (en) | A kind of fully automatic stereo chromatography conversion method using local lineups slope | |
| Zong et al. | A deep learning approach for urban underground objects detection from vehicle-borne ground penetrating radar data in real-time | |
| CN116665067B (en) | Ore finding target area optimization system and method based on graph neural network | |
| CN109858523A (en) | A kind of shallow sea velocity of sound profile inversion method of neural network and ray theory | |
| CN110879412A (en) | Underground transverse wave velocity inversion method, device, computing equipment and storage medium | |
| CN111381275A (en) | First arrival picking method and device for seismic data | |
| CN109113729A (en) | Lithology Identification Methods and device based on log | |
| CN115841076A (en) | Shallow sea layered seabed ground sound parameter inversion method based on BP neural network model | |
| CN114743010B (en) | Ultrahigh voltage power transmission line point cloud data semantic segmentation method based on deep learning | |
| CN110954950A (en) | Underground transverse wave velocity inversion method, device, computing equipment and storage medium | |
| CN105044775A (en) | Seismic fluid inversion identification method and apparatus | |
| Wang et al. | Seismic velocity inversion transformer | |
| Liu et al. | Inversion of vehicle-induced signals based on seismic interferometry and recurrent neural networks | |
| Liu et al. | Physics-driven self-supervised learning system for seismic velocity inversion | |
| CN114566052A (en) | Method for judging rotation of highway traffic flow monitoring equipment based on traffic flow direction | |
| Nie et al. | An effective recognition of moving target seismic anomaly for security region based on deep bidirectional LSTM combined CNN | |
| CN104849751A (en) | Prestack seismic data imaging method | |
| Zhao et al. | Multidimensional scaling-based data dimension reduction method for application in short-term traffic flow prediction for urban road network | |
| Hao et al. | AFSs-RBF neural network for predicting earthquake-induced liquefaction of light loam |
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 |