Stratum anisotropy cause classification method and equipment based on neural network

Abstract

The invention provides a stratum anisotropy cause classification method and equipment based on a neural network, and relates to the field of applied geophysics and oil exploration and development, and the method comprises the steps: taking a stratum parameter combination as a sample feature of a frequency dispersion data set; a frequency dispersion equation is calculated through a numerical algorithm, and a theoretical frequency dispersion curve is solved to serve as a label of a frequency dispersion data set; constructing a frequency dispersion calculation forward model, and correcting individual error samples in the frequency dispersion data set; using a stratum parameter combination as a sample feature, using the corrected frequency dispersion data set as a label, and updating a frequency dispersion calculation forward model; extracting a frequency dispersion curve of the actually measured sound wave data; obtaining an initial prediction frequency dispersion curve by using a frequency dispersion calculation forward model, establishing an objective function, and updating a frequency dispersion data fitting optimal solution based on iteration of a leapfrog algorithm; and classifying the formation anisotropy causes by using the fitted optimal solution. According to the technical scheme, the problems of long calculation time and poor algorithm stability of a frequency dispersion curve classification method in the prior art are solved.

Description

Technical Field

[0001] The invention relates to the field of applied geophysics and oil exploration and development, and in particular to a neural network-based formation anisotropy genesis classification method and equipment. Background Art

[0002] The identification of wellbore rupture, gas leakage, instrument eccentricity and other phenomena, as well as the determination of anisotropic areas caused by inherent anisotropy, stress or cracks can all be distinguished and evaluated through dispersion curves, realizing downstream applications such as wellbore stability analysis, horizontal stress determination, and geological model calibration. Winkler et al. determined the cause of formation anisotropy through the intersection of fast and slow bending wave dispersion curves; Plona et al. combined field orthogonal dipole acoustic data with the dispersion characteristics of bending waves to identify mechanical damage and ground stress. Therefore, correctly distinguishing the cause of anisotropy of induced dispersion curves is a key step in guiding the application of rock physics and geomechanics. However, the existing traditional dispersion curve classification method has defects such as long calculation time and poor algorithm stability, especially when processing field noisy data, which seriously restricts its application and development.

[0003] The development of machine learning technology has facilitated efficient dispersion data processing. Its good adaptability and generalization ability have laid a solid foundation for self-adjustment of model performance and accurate prediction of new data. Assous (2018) and Liang (2021) studied the method of calculating dispersion curves using neural networks. Extending this method to the processing of cross-dipole flexural wave dispersion data is of great significance for the practical application of this technology.

[0004] Therefore, there is a need for a neural network-based formation anisotropy cause classification method and device that can achieve accurate classification of the causes of formation anisotropy and improve the processing speed of cross-dipole bending wave dispersion. Summary of the invention

[0005] The main purpose of the present invention is to provide a method and device for classifying the causes of formation anisotropy based on a neural network, so as to solve the problems of long calculation time and poor algorithm stability in the classification method of dispersion curves in the prior art.

[0006] To achieve the above object, the present invention provides a neural network-based formation anisotropy genesis classification method, which specifically includes the following steps: S1, based on the dipole flexural wave logging model in transversely isotropic formations and the equivalent instrument theory, the sensitivity analysis of different formation parameters is performed, the parameter range and step size are divided, and the formation parameter combination is used as the sample feature of the dispersion data set.

[0007] S2, calculating the dispersion equation by numerical algorithm, solving the theoretical dispersion curve of the dipole flexural wave logging model in the VTI formation defined by each formation parameter combination obtained in step S1 as the label of the dispersion data set.

[0008] S3, uses a fully connected neural network model multi-layer perceptron MLP to build a dispersion calculation forward model, sets the maximum threshold of the difference between the dispersion data set calculated by the numerical algorithm and the dispersion calculation forward model, compares the error size of the training set and the test set, and corrects individual erroneous samples in the dispersion data set.

[0009] S4, using the formation parameter combination in step S1 as a sample feature and the corrected dispersion data set in step S3 as a label, updating the dispersion calculation forward model.

[0010] S5, using the Prony method or the weighted spectrum WSS method to extract the dispersion curve of the measured acoustic wave data, and marking and extracting the dispersion data through the density-based clustering method.

[0011] S6, based on the known formation parameters of the current formation depth point and the range of formation parameters that need to be inverted, different groups of formation parameter combinations are formed, the dispersion calculation forward model trained in step S4 is used to obtain the initial predicted dispersion curve, the objective function is established, and based on the leapfrog algorithm iteration, the optimal solution fitting the dispersion data processed in step S5 is updated.

[0012] S7, classifying the causes of formation anisotropy using the optimal solution fitted in step S6.

[0013] Furthermore, the sensitivity calculation formula in step S1 is: (1); in, represents parameter sensitivity, represents the formation parameters, represents the phase velocity of the wave at different frequencies, is the angular frequency.

[0014] Specifically, in step S2, the specific expression of the theoretical dispersion curve of the dipole flexural wave logging model in the VTI formation defined by each formation parameter combination obtained in step S1 is solved as follows: (2); in, is the dispersion matrix using equivalent instrument theory, is the wave number, is the well diameter, is the formation density, and are the fluid density and fluid velocity, respectively, is the Thomsen parameter, is Poisson's ratio, is the equivalent instrument radius.

[0015] Furthermore, step S3 specifically includes the following steps: S3.1, a fully connected neural network model multi-layer perceptron MLP is used to construct a dispersion calculation forward model. The dispersion calculation forward model includes: an input layer, three hidden layers and an output layer connected in sequence; the dispersion calculation forward model is expressed as: (3); in, represents the dispersion curve calculated by the dispersion calculation forward model, represents the formation parameters, , Respectively represent The weights and biases of the layer network, where , represents the activation function, The specific expression is: (4).

[0016] S3.2, subtract the dispersion data set label obtained in step S2 from the data set label obtained by the dispersion calculation forward model trained by inputting the formation parameter combination, and determine whether it is an erroneous sample by setting the maximum difference threshold.

[0017] S3.3, Correction of erroneous samples.

[0018] S3.4, input the corrected samples into the dispersion calculation forward model for training.

[0019] S3.5, divide the training set and the test set, and determine whether the training set error is smaller than the test set error. If it is larger than the test set error, continue to step S3.2. If it is smaller than the test set error, the correction is completed.

[0020] Furthermore, in step S4, the cosine annealing algorithm is used to iteratively calculate the forward model parameters: (5); in, is the current batch learning rate, is the number of annealing cycles, is the annealing rate, and are the maximum and minimum values ​​of the learning rate, respectively. , They are the training round after the last restart and the round with a given restart interval, respectively.

[0021] Furthermore, step S5 specifically includes the following steps: S5.1, use Prony or weighted spectrum WSS method to extract the dispersion data of measured sound wave data. The expression of Prony is as follows: (6); in, Indicates The spectrum prediction value of each receiver is is a complex exponential term, is an imaginary number, is the spacing between receivers, is the slowness spectrum, For the The amplitude of the vibration mode, is the total number of vibration modes.

[0022] The specific expression of the weighted spectrum method is as follows: (7); in, is the spectral coherence function, For slowness, For the The spectrum of the receiver, for The conjugate of is the number of receivers, and .

[0023] S5.2, use the DBSCAN method to mark and extract dispersion data. The specific expression is as follows: (8); in, is the extracted dispersion data, is the number of clusters, and are samples of different clusters, and is the average distance from the samples in the cluster to the cluster center, is the distance between cluster centers, and is the cluster sample set.

[0024] Furthermore, step S6 specifically includes the following steps: S6.1, Objective Function for: (9); in, is the dispersion data after processing in step S5, is the dispersion curve value predicted by the neural network, Count the number of frequencies, is the frequency calculation range, To calculate the frequency.

[0025] S6.2, the specific calculation formula of the frog leaping algorithm update strategy is: (10); (11); in, and Respectively represent The best and worst solutions in a population, is the updated optimal solution, is the jump step length, and are the minimum and maximum values ​​of the jump step length, For updated new individuals.

[0026] The present invention also provides a formation anisotropy genesis classification device based on a neural network, comprising: at least one processor and a memory; Memory stores computer-executable instructions; At least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor performs a formation anisotropy genesis classification method based on a neural network.

[0027] The present invention has the following beneficial effects: The present invention uses eight related parameters such as well diameter and instrument as sample features, and dispersion curves as sample labels. A dispersion data set is produced based on the equivalent instrument theoretical model, and a dispersion calculation forward model based on machine learning is obtained after forward training. The target parameters are inverted using the dispersion calculation forward model and the leapfrog algorithm, and the dispersion curves extracted from the actual waveform are fitted to achieve accurate classification of the causes of formation anisotropy. Compared with traditional dispersion processing and classification methods, the present invention greatly improves the processing speed of cross-dipole flexural wave dispersion, and the feasibility and effectiveness of the present invention are proved by combining field data processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the specific implementation of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the specific implementation or the prior art description. Obviously, the drawings described below are some implementations of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings: Figure 1 A flowchart of a neural network-based formation anisotropy genesis classification method of the present invention is shown.

[0029] Figure 2 A dispersion sensitivity analysis diagram of a dipole flexural wave fast formation beneath a VTI formation is shown.

[0030] Figure 3 A dispersion sensitivity analysis diagram of dipole flexural wave slow formations beneath the VTI formation is shown.

[0031] Figure 4 The dipole fast and slow shear wave waveforms are shown.

[0032] Figure 5 The fitting diagram of the actual dispersion data of dipole fast and slow shear waves is shown.

[0033] Figure 6 The diagram shows the classification results of the causes of stratigraphic anisotropy. DETAILED DESCRIPTION

[0034] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0035] Embodiment 1 like Figure 1 A neural network-based formation anisotropy genesis classification method is shown, which specifically includes the following steps: S1, based on the dipole flexural wave logging model in transversely isotropic formations and the equivalent instrument theory, sensitivity analysis of different formation parameters is performed, parameter ranges and step sizes are divided, and formation parameter combinations are used as sample features of the dispersion data set. Among them, the dipole flexural wave logging model refers to a logging model constructed by studying the propagation characteristics of flexural waves under the conditions of formation parameters, which is excited by a dipole sound source; the equivalent instrument theory considers the elastic effect of the logging instrument and introduces instrument model parameters to simulate its influence on the logging data.

[0036] The parameter range and step size are divided according to the principle that "the greater the parameter sensitivity, the denser the value", such as Figure 2 and Figure 3 As shown in the sensitivity analysis diagram, the shear wave velocity shows high sensitivity in both fast and slow formations, so a smaller step size is used to divide more values ​​within the shear wave velocity range. The eight formation parameters in the dipole flexural wave logging model are valued at appropriate step sizes and then arranged and combined to determine the formation parameter combination as the sample feature of the dispersion data set.

[0037] S2, calculating the dispersion equation by numerical algorithm, solving the theoretical dispersion curve of the dipole flexural wave logging model in the VTI formation defined by each formation parameter combination obtained in step S1 as the label of the dispersion data set.

[0038] S3, uses a fully connected neural network model multi-layer perceptron MLP to build a dispersion calculation forward model, sets the maximum threshold of the difference between the dispersion data set calculated by the numerical algorithm and the dispersion calculation forward model, compares the error size of the training set and the test set, and corrects individual erroneous samples in the dispersion data set.

[0039] S4, using the formation parameter combination in step S1 as a sample feature and the corrected dispersion data set in step S3 as a label, updating the dispersion calculation forward model.

[0040] S5, using the Prony method or the weighted spectrum WSS method to extract the dispersion curve of the measured acoustic wave data, and marking and extracting the dispersion data through the density-based clustering method.

[0041] S6, according to the known formation parameters of the current formation depth point and the range of formation parameters to be inverted, including: Thomsen parameters and Poisson's ratio, different groups of formation parameter combinations are formed, and the dispersion calculation forward model trained in step S4 is used to obtain the initial predicted dispersion curve, establish the objective function, and update the optimal solution for fitting the dispersion data processed in step S5 based on the iteration of the frog leaping algorithm, that is, the current formation depth point is obtained. value, values ​​and fitted dispersion curves.

[0042] S7, classifying the causes of formation anisotropy using the optimal solution fitted in step S6.

[0043] The present invention processes time-domain acoustic waveforms, effectively extracts and marks dispersion data, trains a dispersion calculation forward model to quickly calculate the theoretical dispersion curve of the dipole bending wave logging model in the VTI formation, combines the frog leaping algorithm to invert the target function, fits the actual dispersion data, and completes the classification of anisotropic formation genesis. Compared with traditional dispersion extraction and classification methods, the present invention significantly improves the processing efficiency of acoustic logging data and can better meet the timeliness requirements of field production.

[0044] Specifically, the sensitivity analysis of different formation parameters according to the dipole acoustic logging model in the transversely isotropic formation in step S1 is as follows:

[0045] The influence of different formation parameters on the dispersion curve is analyzed according to the parameter sensitivity. Generally, the step size and parameter range are defined in accordance with the principle that “the more sensitive the formation parameter, the denser the value”.

[0046] Sensitivity analysis diagram Figure 2 and Figure 3 As shown, in Figure 2 In the fast formation shown in Figure 1, the sensitivity of dipole flexural waves to VTI media is mainly controlled by the shear wave velocity at low frequencies, and as the frequency increases, the fluid velocity becomes dominant. Figure 3 In the slow formation shown, the sensitivity of dipole bending waves to the VTI formation is high in the entire frequency range.

[0047] The sensitivity analysis is performed on the eight input formation parameters, and the sensitivity calculation formula is: (1); in, represents parameter sensitivity, represents the formation parameters, represents the phase velocity of the wave at different frequencies, is the angular frequency.

[0048] As shown in Table 1, based on sensitivity analysis, appropriate step sizes and parameter ranges are divided, and the sample characteristic combination of the dipole bending wave dispersion data set is obtained through permutations and combinations. The formation shear wave velocity shows high sensitivity in both fast and slow formations, so a smaller step size is used to divide more values ​​within the formation shear wave velocity range; while the equivalent instrument model modulus is at a low value in the entire frequency range, whether in fast or slow formations, and does not change with frequency, so it is set to a constant value and not used as a sample characteristic parameter.

[0049] Table 1. Sample characteristics of dipole bending wave dispersion data set

[0050] Specifically, in step S2, the borehole bending wave excited by the dipole acoustic logging tool is a dispersive guided wave propagating along the wellbore wall. The specific expression for solving the theoretical dispersion curve of the dipole bending wave logging model in the VTI formation defined by each formation parameter combination obtained in step S1 is: (2); in, is the dispersion matrix using equivalent instrument theory, is the wave number, is the well diameter, is the formation density, and are the fluid density and fluid velocity, respectively, is the Thomsen parameter, is Poisson's ratio, is the equivalent instrument radius.

[0051] A total of 8 parameters are required to solve the dispersion equation required for equation (2). In step 1, the present invention obtains dispersion equations defined by 725,760 parameter combinations. The theoretical dispersion curve of the dipole flexural wave logging model in the VTI formation defined by each formation parameter combination corresponding to the dispersion equation is solved by step S2. Each dispersion curve consists of flexural wave velocities corresponding to 76 frequency points equidistantly distributed between 0.1 kHz and 15.1 kHz.

[0052] Specifically, since the dispersion equation is complex and has multiple solutions, the algorithm needs to be constantly adjusted during the process of calculating the dispersion equation using the numerical algorithm in step S2 to ensure the accuracy of the result.

[0053] Step S3 specifically includes the following steps: S3.1, the present invention uses 725760 sample features and labels in the data set to train a fully connected neural network model with three hidden layers, namely, a dispersion calculation forward model. The trained dispersion calculation forward model is used to replace the time-consuming and unstable dispersion search mode in step S2.

[0054] Multilayer Perceptron (MLP) is a classic fully connected neural network model, which consists of an input layer, a hidden layer and an output layer, and uses an activation function in each hidden layer for nonlinear transformation. The present invention sets the neural network model as a fully connected neural network with three hidden layers, and the hidden layer contains 1200 neurons. The model can be regarded as a function with eight formation parameters as independent variables and the dispersion curve of VTI formation bending waves as dependent variables.

[0055] The dispersion calculation forward model is constructed using a fully connected neural network model multi-layer perceptron MLP. The dispersion calculation forward model includes: an input layer, three hidden layers and an output layer connected in sequence; the dispersion calculation forward model is expressed as: (3); in, represents the dispersion curve calculated by the dispersion calculation forward model, represents the formation parameters, , Respectively represent The weights and biases of the layer network, where , represents the activation function, The specific expression is: (4).

[0056] S3.2, subtract the dispersion data set label obtained in step S2 from the data set label obtained by the dispersion calculation forward model trained by inputting the formation parameter combination, and determine whether it is an erroneous sample by setting the maximum difference threshold.

[0057] S3.3, Correction of erroneous samples.

[0058] S3.4, input the corrected samples into the neural network for training.

[0059] S3.5, divide the training set and the test set, and determine whether the training set error is smaller than the test set error. If it is larger than the test set error, continue to step S3.2. If it is smaller than the test set error, the correction is completed.

[0060] In step S4, the cosine annealing algorithm is used to iterate the dispersion calculation of the forward model parameters. The training set and the validation set are divided into a ratio of 95:5. The cosine annealing algorithm gradually reduces the learning rate by using the cosine function to smooth the learning rate change and prevent gradient oscillation. The specific expression of the algorithm principle is: (5); in, For the current moment The batch learning rate is is the number of annealing cycles, is the annealing rate, and are the maximum and minimum values ​​of the learning rate, respectively. , They are the training round after the last restart and the round with a given restart interval, respectively.

[0061] Since the target optimization function may have multiple local optimal solutions, this hot restart method makes it possible to jump out of the local optimal solution, thereby improving the global search capability of network training convergence.

[0062] Specifically, step S5 includes the following steps: S5.1, use Prony or weighted spectrum WSS method to extract the dispersion data of measured sound wave data. The expression of Prony is as follows: (6); in, Indicates The spectrum prediction value of each receiver is is a complex exponential term, is an imaginary number, is the spacing between receivers, is the slowness spectrum, For the The amplitude of the vibration mode, is the total number of vibration modes.

[0063] The specific expression of the weighted spectrum method is as follows: (7); in, is the spectral coherence function, For slowness, For the The spectrum of the receiver, for The conjugate of is the number of receivers, and .

[0064] S5.2, using the DBSCAN method to mark and extract dispersion data, and then remove noise, the specific expression is as follows: (8); in, is the extracted dispersion data, is the number of clusters, and are samples of different clusters, and is the average distance from the samples in the cluster to the cluster center, is the distance between cluster centers, and is the cluster sample set.

[0065] Specifically, step S6 includes the following steps: S6.1, Objective Function for: (9); in, is the dispersion data after processing in step S5, is the dispersion curve value predicted by the neural network, Count the number of frequencies, is the frequency calculation range, To calculate the frequency.

[0066] S6.2, the specific calculation formula of the frog leaping algorithm update strategy is: (10); (11); in, and Respectively represent The best and worst solutions in a population, is the updated optimal solution, is the jump step length, and are the minimum and maximum values ​​of the jump step length, For updated new individuals.

[0067] This heuristic optimization algorithm combines local search and global search to update the optimal individual according to the fitness value of the solution. When the local search reaches a certain threshold, it integrates and updates the information within the group and iterates the global optimal solution.

[0068] Figure 4 To extract the fast and slow shear wave waveforms from the orthogonal dipole four-component data, the processing step S5 is used to obtain Figure 5 The fast and slow shear wave field dispersion data in are scattered points, and the fast and slow shear wave dispersion curves are obtained by fitting the field dispersion data in step S6, which are dotted lines. This field example is used to verify the feasibility of the data processing method of extracting waveforms from full wave train signals to effectively mark dipole bending wave dispersion data and automatically obtain fitting dispersion curves.

[0069] Based on the dispersion characteristics of dipole flexural waves, anisotropic formation classification is performed at each depth point based on different phenomena such as dispersion curve overlap, crossover, and mid-frequency splitting. Figure 6 As shown in the figure, at different depths, the fast and slow shear wave dispersion curves after field data processing and fitting are used to accurately and efficiently classify the dispersion types with different colors. This neural network-based formation anisotropy genesis classification method has the characteristics of high-efficiency dispersion processing and high-precision parameter inversion.

[0070] Embodiment 2 A neural network-based formation anisotropy genesis classification device comprises: at least one processor and a memory.

[0071] The memory stores computer-executable instructions.

[0072] At least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes a neural network-based formation anisotropy genesis classification method of the present invention.

[0073] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by technicians in this technical field within the essential scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A neural network-based classification method for formation anisotropy, characterized in that: The specific steps include: S1, based on the dipole flexural wave logging model in transversely isotropic formations and the equivalent instrument theory, sensitivity analysis of different formation parameters is performed, parameter ranges and step sizes are divided, and formation parameter combinations are used as sample features of dispersion data sets; S2, calculating the dispersion equation by numerical algorithm, solving the theoretical dispersion curve of the dipole bending wave logging model in the VTI formation defined by each formation parameter combination obtained in step S1 as a label of the dispersion data set; S3, using a fully connected neural network model multi-layer perceptron MLP to build a dispersion calculation forward model, set the maximum threshold of the difference between the dispersion data set calculated by the numerical algorithm and the dispersion calculation forward model, compare the error size of the training set and the test set, and correct individual error samples in the dispersion data set; S4, using the formation parameter combination in step S1 as a sample feature and the corrected dispersion data set in step S3 as a label, updating the dispersion calculation forward model; S5, extracting the dispersion curve of the measured acoustic wave data using the Prony method or the weighted spectrum WSS method, and marking and extracting the dispersion data using a density-based clustering method; S6, forming different groups of formation parameter combinations according to the known formation parameters at the current formation depth point and the range of formation parameters that need to be inverted, using the dispersion calculation forward model trained in step S4 to obtain an initial predicted dispersion curve, establishing an objective function, and updating the optimal solution for fitting the dispersion data processed in step S5 based on the iteration of the leapfrog algorithm; S7, classifying the causes of formation anisotropy using the optimal solution fitted in step S6.

2. The neural network-based formation anisotropy genesis classification method according to claim 1, characterized in that: The sensitivity calculation formula in step S1 is: （1）； in, represents parameter sensitivity, represents the formation parameters, represents the phase velocity of the wave at different frequencies, is the angular frequency.

3. The neural network-based formation anisotropy genesis classification method according to claim 1, characterized in that: In step S2, the specific expression of the theoretical dispersion curve of the dipole flexural wave logging model in the VTI formation defined by each formation parameter combination obtained in step S1 is solved as follows: （2）； in, is the dispersion matrix using equivalent instrument theory, is the wave number, is the well diameter, is the formation density, and are the fluid density and fluid velocity, respectively, is the Thomsen parameter, is Poisson's ratio, is the equivalent instrument radius.

4. The neural network-based formation anisotropy genesis classification method according to claim 1, characterized in that: Step S3 specifically includes the following steps: S3.1, a fully connected neural network model multi-layer perceptron MLP is used to construct a dispersion calculation forward model. The dispersion calculation forward model includes: an input layer, three hidden layers and an output layer connected in sequence; the dispersion calculation forward model is expressed as: （3）； in, represents the dispersion curve calculated by the dispersion calculation forward model, represents the formation parameters, , Respectively represent The weights and biases of the layer network, where , represents the activation function, The specific expression is: （4）； S3.2, subtract the dispersion data set label obtained in step S2 from the data set label obtained by the dispersion calculation forward model trained by inputting the formation parameter combination, and determine whether it is an erroneous sample by setting the maximum difference threshold; S3.3, correcting erroneous samples; S3.4, inputting the corrected samples into the dispersion calculation forward model for training; S3.5, divide the training set and the test set, and determine whether the training set error is smaller than the test set error. If it is larger than the test set error, continue to step S3.

2. If it is smaller than the test set error, the correction is completed.

5. The neural network-based formation anisotropy genesis classification method according to claim 1, characterized in that: In step S4, the cosine annealing algorithm is used to iteratively calculate the forward model parameters: （5）； in, is the current batch learning rate, is the number of annealing cycles, is the annealing rate, and are the maximum and minimum values ​​of the learning rate, respectively. , They are the training round after the last restart and the round with a given restart interval, respectively.

6. The neural network-based formation anisotropy genesis classification method according to claim 1, characterized in that: Step S5 specifically includes the following steps: S5.1, use Prony or weighted spectrum WSS method to extract the dispersion data of measured sound wave data. The expression of Prony is as follows: （6）； in, Indicates The spectrum prediction value of each receiver is is a complex exponential term, is an imaginary number, is the spacing between receivers, is the slowness spectrum, For the The amplitude of the vibration mode, is the total number of vibration modes; The specific expression of the weighted spectrum method is as follows: （7）； in, is the spectral coherence function, For slowness, For the The spectrum of the receiver, for The conjugate of is the number of receivers, and ; S5.2, use the DBSCAN method to mark and extract dispersion data. The specific expression is as follows: （8）； in, is the extracted dispersion data, is the number of clusters, and are samples of different clusters, and is the average distance from the samples in the cluster to the cluster center, is the distance between cluster centers, and is the cluster sample set.

7. The neural network-based formation anisotropy genesis classification method according to claim 1, characterized in that: Step S6 specifically includes the following steps: S6.1, Objective Function for: （9）； in, is the dispersion data after processing in step S5, is the dispersion curve value predicted by the neural network, Count the number of frequencies, is the frequency calculation range, To calculate the frequency; S6.2, the specific calculation formula of the frog leaping algorithm update strategy is: （10）； （11）； in, and Respectively represent The best and worst solutions in a population, is the updated optimal solution, is the jump step length, and are the minimum and maximum values ​​of the jump step length, For updated new individuals.

8. A neural network-based formation anisotropy genesis classification device, characterized in that: include: at least one processor and memory; The memory stores computer-executable instructions; The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor performs the neural network-based formation anisotropy genesis classification method as described in any one of claims 1 to 7.

Images