Transverse wave velocity profile reconstruction method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a shear wave velocity profile reconstruction method and device, electronic equipment and a storage medium, and relates to the field of geophysics. The method comprises the following steps: acquiring a base-order modal dispersion curve and a target high-order modal dispersion curve from a Rayleigh surface wave dispersion energy diagram corresponding to an observation region; identifying the modal number of the target high-order modal frequency dispersion curve by using a pre-trained frequency dispersion curve modal number prediction model; and based on the modal number of the target high-order modal dispersion curve, performing joint inversion on the base-order modal dispersion curve and the target high-order modal dispersion curve, and reconstructing a shear wave velocity profile corresponding to the observation region. Since the training sample of the frequency dispersion curve modal number prediction model is generated based on the frequency point range of the sample base-order modal frequency dispersion curve and the sample high-order modal frequency dispersion curve, the modal number of the target high-order modal frequency dispersion curve can be accurately identified, the effect of joint inversion is improved, and the reconstruction precision of the shear wave velocity profile is improved.

Description

Technical Field

[0001] The present invention relates to the field of geophysics, and in particular to a shear wave velocity profile reconstruction method, device, electronic equipment and storage medium. Background Art

[0002] Shear wave velocity is a key parameter to characterize the stiffness of underground materials. The dispersion curve of Rayleigh surface waves is highly sensitive to the changes in shear wave velocity in various underground layers. Therefore, dispersion curve inversion is widely used to reconstruct shear wave velocity profiles.

[0003] Dispersion curve inversion requires prior identification of the dispersion curve's mode number. In practice, fundamental modal dispersion curves are generally stable and easy to identify in dispersion energy diagrams. Methods for identifying the mode number of higher-order modal dispersion curves can be divided into two categories: one relies on visual interpretation based on expert experience; the other follows a "preliminary inversion-fitting estimation" process. This involves first inverting the fundamental modal dispersion curve to construct a preliminary shear-wave velocity model. Theoretical multi-order modal dispersion curves are then generated based on this preliminary shear-wave velocity model. These are then compared with the actual extracted higher-order modal dispersion curves to estimate the mode number of the higher-order dispersion curves.

[0004] The former not only limits the degree of automation in data processing but can also lead to misidentification when high-order modal dispersion curves are missing. While the latter improves the accuracy of shear-wave velocity profile reconstruction to a certain extent, its effectiveness depends largely on the accuracy of the shear-wave velocity model and the stability of the inversion process. Summary of the Invention

[0005] The present invention provides a shear wave velocity profile reconstruction method, which can accurately identify the mode number of high-order modal dispersion curves, improve the effect of dispersion curve inversion, and thus enhance the reconstruction accuracy of shear wave velocity profiles.

[0006] The embodiments of the present invention can be implemented as follows: In a first aspect, the present invention provides a method for reconstructing a shear wave velocity profile, the method comprising: Obtain the fundamental mode dispersion curve and target high-order mode dispersion curve from the Rayleigh surface wave dispersion energy diagram corresponding to the observation area; Identifying the modal number of the target high-order modal dispersion curve using a pre-trained dispersion curve modal number prediction model, wherein training samples of the dispersion curve modal number prediction model are generated based on the frequency point range of the sample fundamental modal dispersion curve and the sample high-order modal dispersion curve; Based on the mode number of the target high-order modal dispersion curve, the fundamental modal dispersion curve and the target high-order modal dispersion curve are jointly inverted to reconstruct the shear wave velocity profile corresponding to the observation area.

[0007] Optionally, the method further includes a step of training a dispersion curve modal number prediction model, and the training step of the dispersion curve modal number prediction model includes: Extracting a first frequency point range of the sample fundamental modal dispersion curve and a second frequency point range of the sample high-order modal dispersion curve; For each preset earth model, extracting a reference multi-order modal dispersion curve corresponding to the preset earth model from the theoretical multi-order modal dispersion curve corresponding to the preset earth model according to the first frequency point range and the second frequency point range, to obtain a reference multi-order modal dispersion curve corresponding to each preset earth model; Generate a training sample set and a validation sample set using the reference multi-order modal dispersion curve corresponding to each of the preset earth models; Using the training sample set to train a plurality of pre-constructed width learning networks corresponding to different network complexities to obtain a plurality of trained width learning networks; The plurality of trained width learning networks are evaluated using the validation sample set, and the dispersion curve mode number prediction model is determined from the plurality of trained width learning networks.

[0008] Optionally, the theoretical multi-order modal dispersion curve includes a theoretical fundamental-order modal dispersion curve and a plurality of different theoretical higher-order modal dispersion curves, and the reference multi-order modal dispersion curve includes a reference fundamental-order modal dispersion curve and a plurality of different reference higher-order modal dispersion curves; The step of extracting the reference multi-order modal dispersion curve corresponding to the preset earth model from the theoretical multi-order modal dispersion curve corresponding to the preset earth model according to the first frequency point range and the second frequency point range comprises: Using the portion of the theoretical fundamental-order modal dispersion curve within the first frequency point range as the reference fundamental-order modal dispersion curve; The portion of each theoretical high-order modal dispersion curve that is within the second frequency point range is used as a reference high-order modal dispersion curve to obtain a plurality of reference high-order modal dispersion curves.

[0009] Optionally, the step of generating a training sample set and a validation sample set using a reference multi-order modal dispersion curve corresponding to each of the preset earth models includes: For each of the preset earth models, using a reference fundamental modal dispersion curve corresponding to the preset earth model and each of the reference higher-order modal dispersion curves corresponding to the preset earth model, a plurality of sample vectors corresponding to the preset earth model are generated; All the sample vectors are divided into a plurality of training sample vectors and a plurality of validation sample vectors, the training sample set includes the plurality of training sample vectors, and the validation sample set includes the plurality of validation sample vectors.

[0010] Optionally, the step of evaluating the plurality of trained width learning networks using the validation sample set and determining the dispersion curve mode number prediction model from the plurality of trained width learning networks includes: Inputting the verification sample set into each of the trained width learning networks respectively to obtain a cross entropy loss value of each of the trained width learning networks; The trained width learning network with the smallest cross entropy loss value is used as the dispersion curve mode number prediction model.

[0011] In a second aspect, the present invention provides a shear wave velocity profile reconstruction device, comprising: An acquisition module is used to obtain the fundamental mode dispersion curve and the target high-order mode dispersion curve from the Rayleigh surface wave dispersion energy map corresponding to the observation area; an identification module, configured to identify the modal number of the target high-order modal dispersion curve using a pre-trained dispersion curve modal number prediction model, wherein the training samples of the dispersion curve modal number prediction model are generated based on the frequency point range of the sample fundamental modal dispersion curve and the sample high-order modal dispersion curve; An inversion module is used to jointly invert the fundamental modal dispersion curve and the target higher-order modal dispersion curve based on the mode number of the target higher-order modal dispersion curve, and reconstruct the shear wave velocity profile corresponding to the observation area.

[0012] Optionally, the device further comprises a training module; The training module is used to extract the first frequency point range of the sample fundamental modal dispersion curve and the second frequency point range of the sample high-order modal dispersion curve; for each preset earth model, the reference multi-order modal dispersion curve corresponding to the preset earth model is extracted from the theoretical multi-order modal dispersion curve corresponding to the preset earth model according to the first frequency point range and the second frequency point range, so as to obtain the reference multi-order modal dispersion curve corresponding to each preset earth model; using the reference multi-order modal dispersion curve corresponding to each preset earth model, a training sample set and a verification sample set are generated; using the training sample set, a plurality of pre-constructed width learning networks corresponding to different network complexities are trained to obtain a plurality of trained width learning networks; using the verification sample set, the plurality of trained width learning networks are evaluated, and the dispersion curve modal number prediction model is determined from the plurality of trained width learning networks.

[0013] Optionally, when the training module is used to evaluate the multiple trained width learning networks using the verification sample set and determine the dispersion curve modal number prediction model from the multiple trained width learning networks, it is specifically used to input the verification sample set into each of the trained width learning networks respectively to obtain the cross-entropy loss value of each of the trained width learning networks; and use the trained width learning network with the smallest cross-entropy loss value as the dispersion curve modal number prediction model.

[0014] In a third aspect, the present invention provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the shear wave velocity profile reconstruction method as described in the first aspect is implemented.

[0015] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the shear wave velocity profile reconstruction method as described in the first aspect.

[0016] The shear wave velocity profile reconstruction method, device, electronic device, and storage medium provided by the present invention obtain a fundamental modal dispersion curve and a target higher-order modal dispersion curve from a Rayleigh surface wave dispersion energy map corresponding to an observation area; utilize a pre-trained dispersion curve modal number prediction model to identify the modal number of the target higher-order modal dispersion curve; and based on the modal number of the target higher-order modal dispersion curve, jointly invert the fundamental modal dispersion curve and the target higher-order modal dispersion curve to reconstruct the shear wave velocity profile corresponding to the observation area. Because the training samples of the dispersion curve modal number prediction model of the present invention are generated based on the frequency point range of the sample fundamental modal dispersion curve and the sample higher-order modal dispersion curve, it can accurately identify the modal number of the target higher-order modal dispersion curve, improve the effect of the joint inversion of the fundamental modal dispersion curve and the target higher-order modal dispersion curve, and thus improve the reconstruction accuracy of the shear wave velocity profile. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 A schematic flow chart of a shear wave velocity profile reconstruction method provided by an embodiment of the present invention; Figure 2 A schematic diagram of a process for training a dispersion curve modal number prediction model provided by an embodiment of the present invention; Figure 3 A schematic diagram of a sample vector generation process provided by an embodiment of the present invention; Figure 4 A Rayleigh wave dispersion energy diagram based on numerical simulation provided by an embodiment of the present invention; Figure 5 A schematic diagram of high-order modal number identification results of a numerical experiment provided in an embodiment of the present invention; Figure 6 A dispersion energy diagram of active source survey and passive source survey drawn from field measured data provided by an embodiment of the present invention; Figure 7 A schematic diagram of high-order modal number recognition results of field observation data provided by an embodiment of the present invention; Figure 8 A functional unit block diagram of a shear wave velocity profile reconstruction device provided by an embodiment of the present invention; Figure 9 A schematic block diagram of the structure of an electronic device provided by an embodiment of the present invention.

[0019] Icons: 100 - shear wave velocity profile reconstruction device; 101 - acquisition module; 102 - identification module; 103 - inversion module; 104 - training module; 200 - electronic equipment; 210 - memory; 220 - processor. DETAILED DESCRIPTION

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0021] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort shall fall within the scope of protection of the present invention.

[0022] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings.

[0023] In addition, the terms "first", "second", etc., if used, are merely used to distinguish and describe, and should not be understood as indicating or implying relative importance.

[0024] It should be noted that, in the absence of conflict, the features in the embodiments of the present invention may be combined with each other.

[0025] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

[0026] In order to overcome the deficiencies of the prior art, an embodiment of the present invention provides a shear wave velocity profile reconstruction method, which will be described in detail below.

[0027] Please refer to Figure 1 The shear wave velocity profile reconstruction method includes steps S101 to S103.

[0028] S101, obtaining a fundamental mode dispersion curve and a target high-order mode dispersion curve from a Rayleigh surface wave dispersion energy map corresponding to the observation area.

[0029] The observation area refers to the area where the shear wave velocity profile needs to be reconstructed. The fundamental modal dispersion curve and the target high-order modal dispersion curve are both extracted from the Rayleigh surface wave dispersion energy map corresponding to the observation area.

[0030] It can be understood that the target high-order modal dispersion curve is the dispersion curve of the mode number to be determined.

[0031] S102 , using a pre-trained dispersion curve modal number prediction model to identify the modal number of a target high-order modal dispersion curve.

[0032] The training samples of the dispersion curve modal number prediction model are generated based on the frequency point range of the sample fundamental modal dispersion curve and the sample high-order modal dispersion curve.

[0033] By inputting the fundamental modal dispersion curve and the target high-order modal dispersion curve into a pre-trained dispersion curve modal number prediction model, the modal number of the target high-order modal dispersion curve is obtained.

[0034] S103 , based on the mode number of the target high-order modal dispersion curve, jointly invert the fundamental modal dispersion curve and the target high-order modal dispersion curve to reconstruct the shear wave velocity profile corresponding to the observation area.

[0035] Since the modal number of the target high-order modal dispersion curve is identified by using the pre-trained dispersion curve modal number prediction model, the joint inversion effect of the fundamental modal dispersion curve and the target high-order modal dispersion curve is improved, thereby improving the reconstruction accuracy of the shear wave velocity profile.

[0036] The following introduces the training process of the dispersion curve mode number prediction model.

[0037] Please refer to Figure 2 The training process of the dispersion curve modal number prediction model includes steps S201 to S205.

[0038] S201 , extracting a first frequency point range of a sample fundamental-order modal dispersion curve and a second frequency point range of a sample high-order modal dispersion curve.

[0039] For example, the following is drawn based on the field measured data: Figure 3 The Rayleigh surface wave dispersion energy diagram is shown in , and the fundamental modal dispersion curve determined from the Rayleigh surface wave dispersion energy diagram is used as the sample fundamental modal dispersion curve, and any one of the determined high-order modal dispersion curves is selected as the sample high-order modal dispersion curve.

[0040] The first frequency point range of the sample fundamental modal dispersion curve and the second frequency point range of the sample high-order modal dispersion curve are extracted respectively, such as Figure 3 As shown, the number of frequency points within the first frequency point range of the sample fundamental modal dispersion curve is , the number of frequency points within the second frequency point range of the sample high-order modal dispersion curve is .

[0041] S202, for each preset earth model, extracting a reference multi-order modal dispersion curve corresponding to the preset earth model from the theoretical multi-order modal dispersion curve corresponding to the preset earth model according to the first frequency point range and the second frequency point range, to obtain a reference multi-order modal dispersion curve corresponding to each preset earth model.

[0042] In the embodiment of the present invention, the parameter search space of the earth model is set to generate randomly N The theoretical multi-order modal dispersion curves corresponding to the respective earth models are respectively obtained by using the first frequency point range and the second frequency point range to extract the reference multi-order modal dispersion curves corresponding to the respective preset earth models from the theoretical multi-order modal dispersion curves corresponding to the respective preset earth models.

[0043] The theoretical multi-order modal dispersion curve includes a theoretical fundamental-order modal dispersion curve and a plurality of different theoretical higher-order modal dispersion curves; the reference multi-order modal dispersion curve includes a reference fundamental-order modal dispersion curve and a plurality of different reference higher-order modal dispersion curves.

[0044] In a possible implementation, for each preset earth model, the portion of the theoretical fundamental-order modal dispersion curve corresponding to the preset earth model that is within the first frequency point range can be used as a reference fundamental-order modal dispersion curve corresponding to the preset earth model; the portion of each theoretical high-order modal dispersion curve corresponding to the preset earth model that is within the second frequency point range can be used as a reference high-order modal dispersion curve corresponding to the preset earth model, thereby obtaining multiple reference high-order modal dispersion curves corresponding to the preset earth model.

[0045] For example, Figure 3 As shown, a reference fundamental-order modal dispersion curve is extracted from the theoretical fundamental-order modal dispersion curve according to the first frequency point range, and a reference high-order modal dispersion curve is extracted from the theoretical high-order modal dispersion curve according to the second frequency point range.

[0046] S203 : Generate a training sample set and a verification sample set using the reference multi-order modal dispersion curve corresponding to each preset earth model.

[0047] In a possible implementation, for each preset earth model, a reference fundamental modal dispersion curve corresponding to the preset earth model and each reference higher-order modal dispersion curve corresponding to the preset earth model can be used to generate multiple sample vectors corresponding to the preset earth model.

[0048] Assume that the maximum number of modes is K , it can be understood that each preset earth model has its own corresponding K For any preset earth model, the reference fundamental modal dispersion curve and K Reference high-order modal dispersion curves can be composed of K sample vectors. Accordingly, N Preset Earth models are available NN = sample vectors.

[0049] like Figure 3 As shown, each sample vector includes input samples and output samples. The input sample can be expressed as:

[0050] Where, , .

[0051] The output sample can be expressed as:

[0052] Where, .

[0053] Can be obtained from Sample vectors are randomly selected Sample vectors are used as training sample vectors, each training sample vector includes the input training sample And output training samples .in, , , the output training samples are generated by one-hot encoding, so if k indivual is equal to 1, then this k is the modal number. The remaining The sample vectors are used as validation sample vectors. It can be understood that the training sample set includes multiple training sample vectors, and the validation sample set includes multiple validation sample vectors.

[0054] S204 , using the training sample set to train a plurality of pre-constructed width learning networks corresponding to different network complexities, to obtain a plurality of trained width learning networks.

[0055] The complexity of the width learning network is determined by the number of neurons in each mapping feature layer of the width learning network. , the number of mapping feature layers And the number of enhanced nodes Decision, which can form a hyperparameter space During the network training process, the grid search method performs calculations in the above parameter space, allowing the broadband learning network to learn the mapping relationship between high-order modal dispersion curves and mode numbers.

[0056] S205 , evaluating the multiple trained width learning networks using the validation sample set, and determining a dispersion curve mode number prediction model from the multiple trained width learning networks.

[0057] In a possible implementation, the validation sample set can be input into each trained width learning network respectively to obtain the cross entropy loss value of each trained width learning network; the trained width learning network with the smallest cross entropy loss value is used as the dispersion curve mode number prediction model.

[0058] Cross entropy loss value ,in, For the validation dataset i Earth model k The first sample m The one-hot encoded true value of each entry; is the corresponding softmax probability of the output of the trained network with the specified complexity.

[0059] The cross entropy loss value is used to evaluate the network complexity of each trained width learning network. The optimal network complexity is the configuration that minimizes the cross entropy loss value. The application effect of the dispersion curve mode number prediction model is introduced below.

[0060] First, numerical experiments are carried out based on the MATLAB platform.

[0061] Step 0: Using the seismological calculation program (CPS330) developed by Herrmann, generate simulated Rayleigh surface wave data according to the theoretical earth model parameters in Table 1. Figure 4 In order to simulate the actual situation, the dispersion curve of the second higher-order mode is intentionally omitted when generating the theoretical Rayleigh surface wave data. Based on these data, the frequency-Bessel transform method is used to generate a Figure 4 The Rayleigh surface wave dispersion energy diagram is shown in Figure 1. First, determine the fundamental mode dispersion curve (e.g. Figure 5 The black solid dotted line in the figure) and select the first high-order modal dispersion curve to extract the frequency points corresponding to the two modal dispersion curves. The frequency points on the x-axis are and .

[0062] Set the parameter search space of the earth model according to Table 1 and randomly generate N = Theoretical multi-order modal dispersion curves of 600 earth models (the maximum number of modes is set to K =3). Select one of the theoretical multi-order modes corresponding to the earth model, extract the reference multi-order modal dispersion curve according to the determined frequency point range, and form three sample vectors with the fundamental modal dispersion curve in the extracted multi-order modal dispersion curve and each of the extracted high-order modal dispersion curves. For each theoretical multi-order modal dispersion curve generated for the earth model, three sample vectors are generated in the same way as above, for a total of sample vectors, where the input sample vector of the sample vector is:

[0063] in, , .

[0064] The output sample of the sample vector is:

[0065] in, . Table 1

[0066] Step 1: Randomly select from 1800 sample vectors = 1500 sample vectors are used as training sample vectors.

[0067] The training sample vector includes input training samples and output training samples, which are expressed as: as well as .

[0068] The output training samples are generated by one-hot encoding, so if k indivual is equal to 1, then this k is the modal number. The remaining The sample vectors are used as validation sample vectors.

[0069] Step 2: Use the data obtained in step 1 The width learning network is trained using training sample vectors.

[0070] The complexity of the width learning network is determined by the number of neurons in each mapping feature layer of the width learning network. , the number of mapping feature layers And the number of enhanced nodes Decision, which can form a hyperparameter space .

[0071] During network training, the grid search method Calculation is performed in the parameter space. Cross entropy loss value , which is used to evaluate candidate network structure configurations. The optimal network complexity is the configuration that minimizes the cross entropy loss value. For the validation dataset i Earth model k The first sample m The one-hot encoded true value of the entries, is the corresponding softmax probability output by a trained wideband learning network of specified complexity.

[0072] Step 3: Use the 300 validation sample vectors obtained in step 1 to select the optimal complexity combination for the width learning network. Specifically, use the 300 validation sample vectors to calculate the cross entropy loss value under each model complexity combination, and select the complexity combination with the smallest cross entropy loss value as the optimal complexity combination. Figure 5 The figure shows the cross-entropy loss over time for the first and third high-order modes identified, corresponding to the optimal complexity combination. This time span includes both sample generation time and network training time. As can be seen from this figure, the model converges within approximately 10 seconds, demonstrating its efficiency and the significance of the complexity combination chosen.

[0073] Step 4: Set the node combination of the width learning network to the optimal complexity combination to obtain the dispersion curve mode number prediction model. The observed fundamental mode dispersion curve and the selected target high-order mode dispersion curve are input into this model to predict the mode number of the target high-order mode dispersion curve.

[0074] Table 2 shows the Softmax output probabilities in the numerical examples of two target high-order modal dispersion curves. The probability at the first high-order mode corresponding to the target high-order modal dispersion curve A is the largest, indicating that the mode number of the target high-order modal dispersion curve A to be identified is 1; similarly, the probability at the third high-order mode corresponding to the target high-order modal dispersion curve B is the largest, so the mode number of the target high-order modal dispersion curve B is 3.

[0075] Table 2

[0076] Figure 5 The green solid circle in the middle is the first high-order mode dispersion curve identified, and the blue solid circle is the third high-order mode dispersion curve identified. Figure 5 This indicates that the dispersion curve modal number prediction model provided by the embodiment of the present invention can accurately identify the high-order modal number of the Rayleigh surface wave dispersion curve.

[0077] After completing the numerical experiments, experiments were conducted using actual field observation data.

[0078] The field data were processed to obtain four dispersion energy images, such as Figure 6 Shown are dispersion energy diagrams for passive source surveys and active source surveys using different linear array configurations. In the passive source dispersion energy diagrams, only the fundamental mode energy is present. In the three active source dispersion energy diagrams, the fundamental mode energy is represented by the lowest energy band in the lower left corner of each diagram, while the remaining two energy bands correspond to higher-order mode energies. Note that the three active source dispersion energy diagrams do not include the first higher-order mode dispersion curve.

[0079] The earth model is used to generate 1,800 sample vectors, of which 1,500 sample vectors are selected for model training and 300 sample vectors are used for model validation.

[0080] Figure 7 The results of high-order modal number identification based on measured data are presented. Figure 7 The black solid dots in the figure represent the identified fundamental mode dispersion curve, and the red implementation dots represent the target high-order mode dispersion curves that need to be identified.

[0081] Table 3 shows the Softmax output probabilities for two target high-order modal dispersion curves based on measured data. Target high-order modal dispersion curve A has the highest output probability for the second high-order mode, indicating that the mode number of target high-order modal dispersion curve A is 2. Similarly, target high-order modal dispersion curve B has the highest output probability for the third high-order mode, indicating that the mode number of target high-order modal dispersion curve B is 3. Based on the results given in Table 3, the mode numbers corresponding to each high-order modal dispersion curve state are determined.

[0082] Table 3

[0083] like Figure 7 As shown, the light blue solid dots are the identified second-order mode dispersion curves, and the blue solid dots are the identified third-order mode dispersion curves. Figure 7 The cross-entropy loss curves corresponding to the two modal number identification processes are shown. The total computation time included in each loss function curve includes two stages: training sample generation and network training, demonstrating the efficiency of the dispersion curve modal number prediction model provided by the embodiment of the present invention in achieving fast and accurate high-order modal number identification. The above results demonstrate the effectiveness and efficiency of the dispersion curve modal number prediction model proposed in the embodiment of the present invention in accurately identifying high-order modal numbers in field measurements.

[0084] In order to execute the corresponding steps in the above embodiments and various possible methods, a method for implementing the shear wave velocity profile reconstruction device 100 is given below. Figure 8 , Figure 8 This is a functional block diagram of a shear wave velocity profile reconstruction device 100 provided in an embodiment of the present invention. It should be noted that the basic principles and technical effects of the shear wave velocity profile reconstruction device 100 provided in this embodiment are the same as those of the above-mentioned embodiments. For the sake of brevity, any parts not mentioned in this embodiment may be referred to the corresponding contents of the above-mentioned embodiments. The shear wave velocity profile reconstruction device 100 includes: The acquisition module 101 is used to acquire a fundamental modal dispersion curve and a target high-order modal dispersion curve from a Rayleigh surface wave dispersion energy map corresponding to the observation area.

[0085] The identification module 102 is used to identify the modal number of the target high-order modal dispersion curve using a pre-trained dispersion curve modal number prediction model, wherein the training samples of the dispersion curve modal number prediction model are generated based on the frequency point range of the sample fundamental modal dispersion curve and the sample high-order modal dispersion curve.

[0086] The inversion module 103 is used to jointly invert the fundamental modal dispersion curve and the target higher-order modal dispersion curve based on the mode number of the target higher-order modal dispersion curve, and reconstruct the shear wave velocity profile corresponding to the observation area.

[0087] The training module 104 is used to extract a first frequency point range of the sample fundamental modal dispersion curve and a second frequency point range of the sample high-order modal dispersion curve; for each preset earth model, a reference multi-order modal dispersion curve corresponding to the preset earth model is extracted from the theoretical multi-order modal dispersion curve corresponding to the preset earth model according to the first frequency point range and the second frequency point range, so as to obtain a reference multi-order modal dispersion curve corresponding to each preset earth model; a training sample set and a verification sample set are generated using the reference multi-order modal dispersion curve corresponding to each preset earth model; a plurality of pre-constructed width learning networks corresponding to different network complexities are trained using the training sample set to obtain a plurality of trained width learning networks; a plurality of trained width learning networks are evaluated using the verification sample set, and the dispersion curve modal number prediction model is determined from the plurality of trained width learning networks.

[0088] Optionally, the theoretical multi-order modal dispersion curve includes a theoretical fundamental modal dispersion curve and multiple different theoretical higher-order modal dispersion curves, and the reference multi-order modal dispersion curve includes a reference fundamental modal dispersion curve and multiple different higher-order modal dispersion curves; when the training module 104 is used to extract the reference multi-order modal dispersion curve corresponding to the preset earth model from the theoretical multi-order modal dispersion curve corresponding to the preset earth model according to the first frequency point range and the second frequency point range, it is specifically used to use the part of the theoretical fundamental modal dispersion curve within the first frequency point range as the reference fundamental modal dispersion curve; and use the part of each theoretical high-order modal dispersion curve within the second frequency point range as a reference high-order modal dispersion curve to obtain multiple reference high-order modal dispersion curves.

[0089] Optionally, when the training module 104 is used to generate a training sample set and a verification sample set using the reference multi-order modal dispersion curve corresponding to each preset earth model, it is specifically used to generate, for each preset earth model, a plurality of sample vectors corresponding to the preset earth model using the reference fundamental modal dispersion curve corresponding to the preset earth model and each reference higher-order modal dispersion curve corresponding to the preset earth model; and divide all sample vectors into a plurality of training sample vectors and a plurality of verification sample vectors, the training sample set including a plurality of training sample vectors, and the verification sample set including a plurality of verification sample vectors.

[0090] Optionally, when the training module 104 is used to evaluate multiple trained width learning networks using a validation sample set and determine a dispersion curve modal number prediction model from multiple trained width learning networks, it is specifically used to input the validation sample set into each trained width learning network respectively to obtain the cross-entropy loss value of each trained width learning network; and the trained width learning network with the smallest cross-entropy loss value is used as the dispersion curve modal number prediction model.

[0091] Furthermore, the embodiment of the present invention also provides an electronic device 200, please refer to Figure 9 , the electronic device 200 includes a memory 210 and a processor 220.

[0092] The processor 220 may be a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), a network processor (NP), a digital signal processor (DSP), a field-programmable gate array (FPGA), or other programmable logic device, or a transistor logic device, or a discrete hardware component. Each step of the above-described shear wave velocity profile reconstruction method may be accomplished by hardware integrated logic circuits in the processor 220 or by software instructions.

[0093] The memory 210 may be a ROM or other type of static storage device capable of storing static information and instructions, a RAM or other type of dynamic storage device capable of storing information and instructions, an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, an optical disc storage (including a compact disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 210 may exist independently and be connected to the processor 220 via a communication bus. The memory 210 may also be integrated with the processor 220. The memory 210 is used to store machine executable instructions for executing the scheme of the present invention. The processor 220 is used to execute the machine executable instructions stored in the memory 210 to implement the embodiment of the above-mentioned shear wave velocity profile reconstruction method.

[0094] An embodiment of the present invention further provides a computer-readable storage medium storing a computer program, which, when executed by the processor 220, implements the shear wave velocity profile reconstruction method disclosed in the above embodiments.

[0095] In the embodiments provided by the present invention, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions, and operations of the devices, methods, and computer program products according to multiple embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or part of a code, which contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, as well as the combination of boxes in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or can be implemented using a combination of dedicated hardware and computer instructions.

[0096] In addition, the functional modules in the various embodiments of the present invention may be integrated together to form an independent part, or each module may exist independently, or two or more modules may be integrated to form an independent part.

[0097] If the functions are implemented as software modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or the portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage media include various media capable of storing program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0098] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A shear wave velocity profile reconstruction method, characterized in that: The method comprises: Obtain the fundamental mode dispersion curve and target high-order mode dispersion curve from the Rayleigh surface wave dispersion energy diagram corresponding to the observation area; Identifying the modal number of the target high-order modal dispersion curve using a pre-trained dispersion curve modal number prediction model, wherein training samples of the dispersion curve modal number prediction model are generated based on the frequency point range of the sample fundamental modal dispersion curve and the sample high-order modal dispersion curve; Based on the mode number of the target high-order modal dispersion curve, the fundamental modal dispersion curve and the target high-order modal dispersion curve are jointly inverted to reconstruct the shear wave velocity profile corresponding to the observation area.

2. The method according to claim 1, wherein The method further includes a step of training a dispersion curve modal number prediction model, wherein the training step of the dispersion curve modal number prediction model includes: Extracting a first frequency point range of the sample fundamental modal dispersion curve and a second frequency point range of the sample high-order modal dispersion curve; For each preset earth model, extracting a reference multi-order modal dispersion curve corresponding to the preset earth model from the theoretical multi-order modal dispersion curve corresponding to the preset earth model according to the first frequency point range and the second frequency point range, to obtain a reference multi-order modal dispersion curve corresponding to each preset earth model; Generate a training sample set and a validation sample set using the reference multi-order modal dispersion curve corresponding to each of the preset earth models; Using the training sample set to train a plurality of pre-constructed width learning networks corresponding to different network complexities to obtain a plurality of trained width learning networks; The plurality of trained width learning networks are evaluated using the validation sample set, and the dispersion curve mode number prediction model is determined from the plurality of trained width learning networks.

3. The method according to claim 2, wherein The theoretical multi-order modal dispersion curve includes a theoretical fundamental modal dispersion curve and a plurality of different theoretical higher-order modal dispersion curves, and the reference multi-order modal dispersion curve includes a reference fundamental modal dispersion curve and a plurality of different reference higher-order modal dispersion curves; The step of extracting the reference multi-order modal dispersion curve corresponding to the preset earth model from the theoretical multi-order modal dispersion curve corresponding to the preset earth model according to the first frequency point range and the second frequency point range comprises: Using the portion of the theoretical fundamental-order modal dispersion curve within the first frequency point range as the reference fundamental-order modal dispersion curve; The portion of each theoretical high-order modal dispersion curve that is within the second frequency point range is used as a reference high-order modal dispersion curve to obtain a plurality of reference high-order modal dispersion curves.

4. The method according to claim 3, wherein The step of generating a training sample set and a verification sample set by using the reference multi-order modal dispersion curve corresponding to each preset earth model includes: For each of the preset earth models, using a reference fundamental modal dispersion curve corresponding to the preset earth model and each of the reference higher-order modal dispersion curves corresponding to the preset earth model, a plurality of sample vectors corresponding to the preset earth model are generated; All the sample vectors are divided into a plurality of training sample vectors and a plurality of validation sample vectors, the training sample set includes the plurality of training sample vectors, and the validation sample set includes the plurality of validation sample vectors.

5. The method according to claim 2, wherein The step of evaluating the plurality of trained width learning networks using the validation sample set and determining the dispersion curve mode number prediction model from the plurality of trained width learning networks comprises: Inputting the verification sample set into each of the trained width learning networks respectively to obtain a cross entropy loss value of each of the trained width learning networks; The trained width learning network with the smallest cross entropy loss value is used as the dispersion curve mode number prediction model.

6. A shear wave velocity profile reconstruction device, characterized in that: The device comprises: An acquisition module is used to obtain the fundamental mode dispersion curve and the target high-order mode dispersion curve from the Rayleigh surface wave dispersion energy map corresponding to the observation area; an identification module, configured to identify the modal number of the target high-order modal dispersion curve using a pre-trained dispersion curve modal number prediction model, wherein the training samples of the dispersion curve modal number prediction model are generated based on the frequency point range of the sample fundamental modal dispersion curve and the sample high-order modal dispersion curve; An inversion module is used to jointly invert the fundamental modal dispersion curve and the target higher-order modal dispersion curve based on the mode number of the target higher-order modal dispersion curve, and reconstruct the shear wave velocity profile corresponding to the observation area.

7. The device according to claim 6, characterized in that The device also includes a training module; The training module is configured to extract a first frequency point range of the sample fundamental modal dispersion curve and a second frequency point range of the sample higher-order modal dispersion curve; for each preset earth model, extract a reference multi-order modal dispersion curve corresponding to the preset earth model from the theoretical multi-order modal dispersion curve corresponding to the preset earth model according to the first frequency point range and the second frequency point range, and obtain a reference multi-order modal dispersion curve corresponding to each preset earth model; Generate a training sample set and a validation sample set using the reference multi-order modal dispersion curve corresponding to each of the preset earth models; Using the training sample set to train a plurality of pre-constructed width learning networks corresponding to different network complexities to obtain a plurality of trained width learning networks; The plurality of trained width learning networks are evaluated using the validation sample set, and the dispersion curve mode number prediction model is determined from the plurality of trained width learning networks.

8. The device according to claim 7, wherein When the training module is used to evaluate the multiple trained width learning networks using the verification sample set and determine the dispersion curve modal number prediction model from the multiple trained width learning networks, it is specifically used to input the verification sample set into each of the trained width learning networks respectively to obtain the cross-entropy loss value of each of the trained width learning networks; and use the trained width learning network with the smallest cross-entropy loss value as the dispersion curve modal number prediction model.

9. An electronic device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the method for reconstructing a shear wave velocity profile according to any one of claims 1 to 5 is implemented.

10. A computer-readable storage medium, characterized in that The computer program is stored therein, and when the computer program is executed by a processor, the shear wave velocity profile reconstruction method according to any one of claims 1 to 5 is implemented.

Images