A method, device and medium for cross-dimensional bayesian inversion of a rayleigh wave dispersion curve

Abstract

This invention discloses a method, apparatus, and medium for cross-dimensional Bayesian inversion of Rayleigh surface wave dispersion curves, relating to the field of surface wave exploration technology. This invention introduces a Dix-type physically guided soft prior, using a reference shear wave velocity profile as a probability anchor point to construct a Gaussian probability field around this profile. This guides the cross-dimensional Bayesian inversion to quickly enter a high-probability physically reasonable region in a probabilistic sense, thereby upgrading the Dix non-iterative inversion results from "fixed initial values / hard constraints" to "Gaussian probability anchor point soft priors." Simultaneously, based on the formation depth value corresponding to the subsurface shear wave velocity parameters to be inverted, a depth-related prior standard deviation of the Gaussian probability field is set to ensure the soft prior operates within the physical boundaries. This makes the final inversion results naturally biased towards the physically reasonable model region in a probabilistic sense, and physical boundary constraints are applied, eliminating systematic biases in the Dix inversion and ultimately obtaining inversion results consistent with actual geology.

Description

Technical Field

[0001] This invention relates to the field of surface wave exploration technology, and in particular to a method, apparatus, equipment and medium for transdimensional Bayesian inversion of Rayleigh surface wave dispersion curves. Background Technology

[0002] Rayleigh surface wave dispersion (phase velocity / group velocity varies with frequency) has a deterministic relationship with the mechanical parameters of the subsurface medium (Vs, Vp, density, layer thickness, etc.). Rayleigh surface wave dispersion inversion has become a commonly used technique for near-surface engineering exploration and shallow structural imaging. Rayleigh surface wave dispersion inversion utilizes the dispersion characteristics of surface waves propagating in layered media (i.e., phase velocity or group velocity varies with frequency) to invert the mechanical parameters of the subsurface medium, especially the shear wave velocity (Vs) structure. In Rayleigh surface wave dispersion curve inversion, the estimation of the subsurface shear wave velocity (Vs) structure is highly nonlinear and highly non-unique, and engineering data are often accompanied by mode mixing, picking errors, and noise disturbances.

[0003] Current inversion methods mainly include deterministic inversion methods, global optimization methods, and Bayesian MCMC and cross-dimensional rj-MCMC inversion methods. Deterministic inversion methods are linearized or locally iterative inversion methods. They are fast but heavily dependent on the initial model and are prone to getting trapped in local minima. Global optimization methods can alleviate local minima to some extent, but they usually output a single optimal solution and are not easy to quantify uncertainty. Bayesian MCMC and cross-dimensional rj-MCMC inversion methods can obtain uncertainty from posterior distribution sampling and allow adaptation of dimensions such as the number of layers. However, when there is a lack of effective prior information to guide them, their computational efficiency is extremely low and may even become unacceptable.

[0004] To address the problems existing in current inversion methods, researchers have proposed a non-perturbative, non-iterative Dix-type fast inversion technique. This technique can quickly obtain an approximate Vs profile from the dispersion curve, offering advantages such as low cost and strong interpretability. However, it simply introduces the Dix inversion results as a fixed initial model or hard constraint into the cross-dimensional Bayesian framework. While this can accelerate convergence, it forcibly transmits the systematic biases in the Dix inversion to the final inversion result. This suppresses the data-driven ability of the cross-dimensional Bayesian technique, which is originally based on the observation data itself for error correction and anomaly detection. Ultimately, this results in a significant discrepancy between the inversion results of the Rayleigh surface wave dispersion curve and the actual values, making it difficult to provide technical support for near-surface engineering exploration and shallow structure imaging. Summary of the Invention

[0005] This invention provides a method, apparatus, and medium for transdimensional Bayesian inversion of Rayleigh surface wave dispersion curves, which can solve the problems existing in the prior art.

[0006] This invention provides a method for transdimensional Bayesian inversion of Rayleigh surface wave dispersion curves, comprising the following steps: Rayleigh surface wave data of the geological area to be tested were collected, and multimode dispersion curve data were extracted from the Rayleigh surface wave data; Based on multimodal dispersion curve data, the Dix-type inversion method is used to invert the underground medium model constructed for the geological area under test, obtain the reference shear wave velocity profile, and interpolate and map the reference shear wave velocity profile onto the parameterized grid of the underground medium model so that each underground shear wave velocity parameter to be inverted in the underground medium model has a corresponding reference shear wave velocity value. For each subsurface shear wave velocity parameter to be inverted, a Gaussian probability density function with the corresponding reference shear wave velocity value as the mean is constructed, and the Gaussian probability density function is used as the soft prior probability distribution field of the subsurface shear wave velocity parameter to be inverted. At the same time, the formation depth value to which the subsurface shear wave velocity parameter to be inverted belongs is obtained from the subsurface medium model, a scaling factor is set based on the depth value, the scaling factor is multiplied by the reference shear wave velocity value corresponding to the subsurface shear wave velocity parameter to be inverted, and the product is used as the depth-related prior standard deviation of the Gaussian probability density function. Based on the soft prior probability distribution field of each subsurface shear wave velocity parameter to be inverted, the Bayesian inversion method is used to perform velocity inversion of the subsurface medium model and obtain the inversion results of the geological area to be measured.

[0007] Preferably, the step of interpolating and mapping the reference shear wave velocity profile onto the parameterized mesh of the subsurface medium model includes: Multimodal dispersion curve data were extracted from the Rayleigh surface wave data. d , represented as: ; in: f j Indicates the first j One frequency sampling point; q Indicates the modal number; c obs ( f j , q ) represents the corresponding observed phase velocity, and simultaneously obtains the relative observation error coefficient corresponding to the Rayleigh surface wave data; Based on multimodal dispersion curve data Based on the relative observation error coefficient, the Dix-type inversion method was used to invert the subsurface medium model constructed for the geological area under test to obtain the reference shear wave velocity profile. The reference shear wave velocity profile is interpolated and mapped onto the parameterized grid of the subsurface medium model; The parameterized mesh is either a depth mesh based on the Voronoi kernel or a fixed layered mesh.

[0008] Preferably, obtaining the Gaussian probability density function includes: For each subsurface shear wave velocity parameter to be inverted Construct a system based on the corresponding reference shear wave velocity value. The Gaussian probability density function with mean is expressed as: ; in: Indicates the prior standard deviation of deep correlation; K Indicates the dimension of the Voronoi kernel; m This represents the model to be inverted.

[0009] Preferably, obtaining the depth-related prior standard deviation of the Gaussian probability density function includes: The depth-related prior standard deviation of the Gaussian probability density function is , represented as: ; in: α ∈(0,1) represents the proportionality coefficient; v min ( z )and v min ( z ) represent depth z Lower and upper bounds of the physically feasible region for transverse wave velocity; The scaling factor is determined based on the formation depth of the subsurface shear wave velocity parameter to be inverted; when α When the value is less than the preset threshold, the depth-related prior standard deviation Small, narrow soft prior probability distribution field, strong constraint on the parameters of the underground medium model; when α When the value is greater than the preset threshold, the depth-related prior standard deviation The large, soft prior probability distribution field provides weak constraints on the parameters of the underground medium model.

[0010] Preferably, the velocity inversion of the subsurface medium model using the Bayesian inversion method includes: Based on the soft prior probability distribution field of each subsurface shear wave velocity parameter to be inverted, the target posterior distribution characterized by the Bayesian inversion method is constructed, expressed as: ; in: The soft prior probability distribution field representing the subsurface shear wave velocity parameters to be inverted; Indicates data likelihood; And set the posterior energy of the target posterior distribution as: ; in: Represents the mismatch function; Indicates basic priors; The soft prior probability distribution field representing the subsurface shear wave velocity parameters to be inverted; Based on the target posterior distribution and the set posterior energy, the underground shear wave velocity parameters to be inverted in the underground medium model are inverted.

[0011] Preferably, obtaining the inversion results of the geological area to be tested includes: When inverting each subsurface shear wave velocity parameter to be inverted within the subsurface medium model, based on the target posterior distribution and the set posterior energy, and using reversible skip Markov chain Monte Carlo sampling, the current subsurface medium model is... m Proposal to a new subsurface medium model m′ ; In the current underground medium model m Proposal to a new subsurface medium model m′ During the process, the acceptance rate is set as follows: ; in: This indicates a proposal ratio or Jacobian term for a cross-dimensional proposal; T Indicates temperature parameter; Indicates the increase in energy; T Indicates temperature; and These represent the forward and reverse proposal densities, respectively. Based on the set acceptance rate, the reversible jumping Markov chain is guided to Monte Carlo sampling to explore physical regions with high posterior probabilities, and the posterior probability density distribution, interface depth histogram, maximum posterior probability model, mean profile model and layer posterior distribution of the underground medium model are obtained.

[0012] Preferably, in the process of Monte Carlo sampling using reversible jumping Markov chains, a parallel tempering method is adopted, in which multiple Markov chains at different temperatures are set up for parallel sampling and periodically exchanged to accelerate the exploration of the global model space, and finally, cold chain samples are used for posterior statistics.

[0013] This invention also provides a trans-Viges Bayesian inversion device for Rayleigh surface wave dispersion curves, comprising: The data processing module is used to collect Rayleigh surface wave data of the geological area to be measured and extract multimode dispersion curve data from the Rayleigh surface wave data; The Dix inversion module is used to invert the subsurface medium model constructed for the geological area under test based on multimodal dispersion curve data and the Dix inversion method to obtain a reference shear wave velocity profile. The reference shear wave velocity profile is then interpolated and mapped onto the parameterized grid of the subsurface medium model so that each subsurface shear wave velocity parameter to be inverted in the subsurface medium model has a corresponding reference shear wave velocity value. The soft prior construction module is used to construct a Gaussian probability density function with the corresponding reference shear wave velocity value as the mean for each subsurface shear wave velocity parameter to be inverted, and use the Gaussian probability density function as the soft prior probability distribution field of the subsurface shear wave velocity parameter to be inverted; at the same time, it obtains the formation depth value of the subsurface shear wave velocity parameter to be inverted from the subsurface medium model, sets a scaling factor based on the depth value, multiplies the scaling factor with the reference shear wave velocity value corresponding to the subsurface shear wave velocity parameter to be inverted, and uses the product as the depth-related prior standard deviation of the Gaussian probability density function; The probability inversion module is used to perform velocity inversion of the subsurface medium model based on the soft prior probability distribution field of each subsurface shear wave velocity parameter to be inverted, and to obtain the inversion results of the geological area to be measured.

[0014] This invention also provides an electronic device, including a memory and a processor; The memory is used to store computer programs; When the processor executes the computer program stored in the memory, it implements the steps of the Rayleigh surface wave dispersion curve transdimensional Bayesian inversion method as described above.

[0015] This invention also provides a computer-readable storage medium for storing a computer program, which, when executed by a processor, implements the steps of a Rayleigh surface wave dispersion curve transdimensional Bayesian inversion method as described above.

[0016] This invention provides a method, apparatus, and medium for transdimensional Bayesian inversion of Rayleigh surface wave dispersion curves. Compared with existing technologies, its advantages are as follows: This invention introduces a Dix-type physical-guided soft prior, using the reference shear wave velocity profile obtained from Dix ​​rapid inversion as a probabilistic anchor point. A Gaussian probability field is constructed around this reference shear wave velocity profile to guide the cross-dimensional Bayesian inversion rapidly into a high-probability physically reasonable region in a probabilistic sense. This upgrades the Dix non-iterative inversion results from "fixed initial values / hard constraints" to "Gaussian probability anchor point soft prior." Simultaneously, based on the formation depth value corresponding to each subsurface shear wave velocity parameter to be inverted within the subsurface medium model, the depth-related prior standard deviation of the Gaussian probability field is set to ensure that the soft prior works within the physical boundary. In this process, the physical-guided soft prior exists in a probabilistic form rather than a mandatory command, forming a mechanism of probabilistic guidance and physical boundary constraints with the physical boundary constraints. This makes the final inversion results naturally biased towards the physically reasonable model region in a probabilistic sense, while simultaneously implementing physical boundary constraints, eliminating systematic biases in Dix inversion, and ultimately obtaining inversion results consistent with actual geology. This provides technical support for near-surface engineering exploration and shallow structure imaging. Attached Figure Description

[0017] Figure 1 A schematic diagram of the overall process of a transdimensional Bayesian inversion method for Rayleigh surface wave dispersion curves provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the Dix-type physical guidance soft prior mechanism provided in an embodiment of the present invention. Detailed Implementation

[0018] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of the present invention. However, the present invention can be practiced in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0019] Rayleigh surface wave dispersion (phase velocity / group velocity varies with frequency) has a deterministic relationship with the mechanical parameters of the subsurface medium (Vs, Vp, density, layer thickness, etc.). Surface wave dispersion inversion has become a commonly used technique for near-surface engineering exploration and shallow structural imaging. In the inversion of Rayleigh surface wave dispersion curves, the estimation of the subsurface shear wave velocity (Vs) structure is highly nonlinear and highly non-unique, and engineering data are often accompanied by mode mixing, picking errors and noise disturbances.

[0020] Current inversion methods mainly include: deterministic inversion, global optimization methods, Bayesian MCMC, and cross-dimensional rj-MCMC methods. Deterministic inversion is linearization / local iteration, which is fast but heavily dependent on the initial model and prone to getting trapped in local minima. Global optimization methods alleviate local minima to some extent, but usually output a single optimal solution and it is not easy to quantify uncertainty. Bayesian MCMC and cross-dimensional rj-MCMC can obtain uncertainty from posterior distribution sampling and allow adaptation of dimensions such as the number of layers, but in engineering, the computation is often unbearable due to the excessively broad prior.

[0021] In addition to deterministic inversion, global optimization methods, Bayesian MCMC, and cross-dimensional rj-MCMC methods, researchers have proposed the idea of ​​introducing prior information to accelerate sampling. However, this idea relies heavily on external data such as boreholes, interfaces, or region boundaries. When external data is missing, it is still difficult to avoid blind search and long warm-up problems. Meanwhile, in the surface wave domain, there is a non-perturbative, non-iterative Dix-type fast inversion that can quickly obtain an approximate Vs profile from the dispersion curve. It has the advantages of low cost and strong interpretability, but its results contain approximate errors. If it is directly used as a fixed initial model or hard constraint, the deterministic bias will be transmitted to the inversion results.

[0022] To address the problems existing in current inversion methods, such as Figure 1 As shown, this invention proposes a Dix-based Soft Prior mechanism. This mechanism constructs a Gaussian probability anchor point from the reference Vs profile obtained by Dix fast inversion, forming a "probability corridor" that guides cross-dimensional rj-MCMC to quickly enter a high-probability physically plausible region in a probabilistic sense; through the prior strength parameter ( σ prior This achieves "adjustable tightness," ensuring that even when observed data contradicts the Dix reference, the posterior can still adaptively deviate and correct under likelihood-driven conditions; specifically: The cross-dimensional Bayesian inversion method provided by this invention has a target posterior distribution composed of data likelihood, basic prior, and Dix-type soft prior, expressed as: .

[0023] in: A Gaussian soft prior probability field is constructed around the Dix reference profile to form a "probability corridor".

[0024] The specific implementation process of this invention includes: Step S1: Input data and preprocessing.

[0025] Acquire surface wave records and extract multimodal dispersion curve data d , represented as: .

[0026] in, f j For the first j Each frequency sampling point q Modal numbering, c obs ( f j , q The corresponding observed phase velocity is obtained, and the relative observation error coefficient corresponding to the Rayleigh surface wave data is obtained at the same time.

[0027] Step S2: Fast Dix-type non-perturbative inversion.

[0028] The reference shear wave velocity profile can be directly calculated from the dispersion curve. And interpolate the mapping to the inverted depth grid or Voronoi kernel depth location.

[0029] Step S3: Construct a Dix-type physical boot soft prior.

[0030] For each core / layer parameter , construct The Gaussian prior with mean is expressed as: .

[0031] in: K This represents the dimension of the Voronoi kernel. m The model to be inverted; The deep-related prior standard deviation, serving as a "knob" for soft constraint strength, is constructed as follows: .

[0032] in, α ∈(0,1) represents the proportionality coefficient. v min ( z )and v min ( z ) represent depth z The lower and upper bounds of the physically feasible region for transverse wave velocity; and through The feasible region is pruned to ensure that the soft prior operates within the physical boundaries. Here, we take... n =1 represents a confidence band multiple, used to indicate that... The soft prior corridor half-width is of scale, corresponding to approximately 68% Gaussian confidence band.

[0033] Step S4: Model parameterization and forward modeling.

[0034] An equivalent layered model is obtained by using a one-dimensional Voronoi kernel or layered parameterization; Vp and density can be constructed using empirical relationships; multimodal dispersion forward modeling is performed using algorithms such as Thomson-Haskell. .

[0035] Step S5: Construct the likelihood function and posterior energy.

[0036] An independent Gaussian error model is used, and scaling is achieved using relative error to obtain the mismatch function. Likelihood The posterior energy is defined as follows: .

[0037] in, Represents the mismatch function; The initial soft prior probability distribution field for the Dix model; Based on prior knowledge.

[0038] Step S6: Cross-dimensional rj-MCMC sampling (Change / Move / Birth / Death).

[0039] Propose a new model m′ from the current model m, including: Change: Perturbation of nuclear velocity .

[0040] Move: Perturbation kernel depth .

[0041] Birth / Death: Add / remove kernels to make dimension K variable.

[0042] Step S7: Standardize the acceptance rate.

[0043] The general acceptance rate at temperature T is expressed as: .

[0044] in: For energy increment; T For temperature; and These represent the positive and negative proposal densities, respectively. The proposal ratio / Jacobi term for cross-dimensional proposals (taken as 1 under common constructions).

[0045] The key point is: As a priori energy term, it explicitly enters the acceptance rate, achieving "probability-guided rather than hard boundary".

[0046] Step S8: Parallel tempering PT accelerates global exploration.

[0047] Multiple temperature chains are set up and periodically exchanged. The exchange acceptance rate is determined by the energy difference and temperature difference, and ultimately only... T =1 chain sample statistical posterior.

[0048] Step S9: Output and Statistics.

[0049] Outputs include: posterior vs. probability density, interface depth histogram, layer posterior distribution, MAP / ML / mean / median profiles, burn-in and ESS efficiency metrics, etc.

[0050] like Figure 2 As shown, the red solid line represents the reference Vs profile obtained from Dix-type deterministic inversion and interpolated to the inversion mesh; the gray shading represents the search space defined by the hard boundary; dark blue and light blue respectively represent the search space defined by the hard boundary. v dix ( z The Gaussian soft priors ±1σ (approximately 68%) and ±2σ (approximately 95%) confidence bands were constructed using the mean.

[0051] This invention upgrades the Dix non-iterative inversion results from "fixed initial value / hard constraint" to "Gaussian probability anchor soft prior" on the one hand, and on the other hand, it improves the soft prior strength through deep correlation. The construction of the data (especially determined by the feasible speed-bandwidth ratio and pruned to hard boundaries) allows soft priors to enter the cross-dimensional rj-MCMC unified acceptance rate, rewarding closeness to the Dix reference and penalizing deviations in a probabilistic sense, but allowing deviations and corrections under likelihood dominance, while adapting to Voronoi kernel parameterization or layered parameterization, and can be used for multimodal dispersion data.

[0052] The advantages of this invention are: ① Significantly shortened burn-in period: sampling enters the high-probability physical reasonable region more quickly; ② Improved sampling efficiency and stability: smaller error dispersion in repeated experiments and more consistent convergence process; ③ Reduced false thin layers and fragmented stratification: more concentrated posterior layer number, closer to the complexity of the real model; ④ No need for external borehole / interface data: "Physically guided prior" is generated endogenously using the same data, with strong engineering generalization; ⑤ Avoids system bias caused by hard constraints: soft prior allows the posterior to deviate from the reference profile with data support, retaining the ability of data-driven discovery of abnormal structures; ⑥ Compatible with parallel tempering and cross-dimensional mechanisms: still robust under multi-peak posterior.

[0053] This invention utilizes the same set of dispersion data used for inversion to generate a physical reference profile through rapid Dix-type inversion, and transforms it into probabilistic guidance information, achieving a closed loop of "endogenous prior knowledge." This means that efficient guidance can be obtained without any external boreholes or prior geological knowledge, and can be widely applied to scenarios such as new area exploration, rapid site assessment, and geological hazard investigation without prior data, filling the gap in the application of existing technologies in areas without data. The soft prior introduced by this invention ensures that even in cross-dimensional proposals, the newly generated layer parameters are attracted by the Dix reference probability field, thereby significantly improving the proposal acceptance rate.

[0054] This invention cleverly utilizes the physical information in the source data by transforming the Dix fast inversion results into a "probability corridor" type soft prior. While maintaining the data-driven advantages of Bayesian inversion, it significantly accelerates sampling convergence, improves inversion stability, suppresses spurious stratification, avoids hard constraint bias, and requires no external borehole data.

[0055] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A method for trans-Vigeal Bayesian inversion of Rayleigh surface wave dispersion curves, characterized in that, Includes the following steps: Rayleigh surface wave data of the geological area to be tested were collected, and multimode dispersion curve data were extracted from the Rayleigh surface wave data; Based on multimodal dispersion curve data, the Dix-type inversion method is used to invert the underground medium model constructed for the geological area under test, obtain the reference shear wave velocity profile, and interpolate and map the reference shear wave velocity profile onto the parameterized grid of the underground medium model so that each underground shear wave velocity parameter to be inverted in the underground medium model has a corresponding reference shear wave velocity value. For each subsurface shear wave velocity parameter to be inverted, a Gaussian probability density function with the corresponding reference shear wave velocity value as the mean is constructed, and the Gaussian probability density function is used as the soft prior probability distribution field of the subsurface shear wave velocity parameter to be inverted. At the same time, the formation depth value to which the subsurface shear wave velocity parameter to be inverted belongs is obtained from the subsurface medium model, a scaling factor is set based on the depth value, the scaling factor is multiplied by the reference shear wave velocity value corresponding to the subsurface shear wave velocity parameter to be inverted, and the product is used as the depth-related prior standard deviation of the Gaussian probability density function. Based on the soft prior probability distribution field of each subsurface shear wave velocity parameter to be inverted, the Bayesian inversion method is used to perform velocity inversion of the subsurface medium model and obtain the inversion results of the geological area to be measured.

2. The method for transdimensional Bayesian inversion of Rayleigh surface wave dispersion curves according to claim 1, characterized in that, The process of interpolating and mapping the reference shear wave velocity profile onto the parameterized mesh of the subsurface medium model includes: Extracting multimodal dispersion curve data from Rayleigh surface wave data d , is represented as: ； in: f j Indicates the first j One frequency sampling point; q Indicates the modal number; c obs ( f j , q ) represents the corresponding observed phase velocity, and simultaneously obtains the relative observation error coefficient corresponding to the Rayleigh surface wave data; Based on multimodal dispersion curve data Based on the relative observation error coefficient, the Dix-type inversion method was used to invert the subsurface medium model constructed for the geological area under test to obtain the reference shear wave velocity profile. The reference shear wave velocity profile is interpolated and mapped onto the parameterized grid of the subsurface medium model; The parameterized mesh is either a depth mesh based on the Voronoi kernel or a fixed layered mesh.

3. The method for transdimensional Bayesian inversion of Rayleigh surface wave dispersion curves according to claim 2, characterized in that, The acquisition of the Gaussian probability density function includes: For each subsurface shear wave velocity parameter to be inverted Construct a system based on the corresponding reference shear wave velocity value. The Gaussian probability density function with mean is expressed as: ； in: Indicates the prior standard deviation of deep correlation; K Indicates the dimension of the Voronoi kernel; m This represents the model to be inverted.

4. The method for transdimensional Bayesian inversion of Rayleigh surface wave dispersion curves according to claim 3, characterized in that, The acquisition of the depth-related prior standard deviation of the Gaussian probability density function includes: The depth-related prior standard deviation of the Gaussian probability density function is , is represented as: ； in: α ∈(0,1) represents the proportionality coefficient; v min ( z )and v min ( z ) represent depth z Lower and upper bounds of the physically feasible region for transverse wave velocity; The scaling factor is determined based on the formation depth of the subsurface shear wave velocity parameter to be inverted; when α When the value is less than the preset threshold, the depth-related prior standard deviation Small, narrow soft prior probability distribution field, strong constraint on the parameters of the underground medium model; when α When the value is greater than the preset threshold, the depth-related prior standard deviation The large, soft prior probability distribution field provides weak constraints on the parameters of the underground medium model.

5. The method for transdimensional Bayesian inversion of Rayleigh surface wave dispersion curves according to claim 1, characterized in that, The velocity inversion of the subsurface medium model using the Bayesian inversion method includes: Based on the soft prior probability distribution field of each subsurface shear wave velocity parameter to be inverted, the target posterior distribution characterized by the Bayesian inversion method is constructed, expressed as: ； in: The soft prior probability distribution field representing the subsurface shear wave velocity parameters to be inverted; Indicates data likelihood; And set the posterior energy of the target posterior distribution as: ； in: Represents the mismatch function; Indicates basic priors; The soft prior probability distribution field representing the subsurface shear wave velocity parameters to be inverted; Based on the target posterior distribution and the set posterior energy, the underground shear wave velocity parameters to be inverted in the underground medium model are inverted.

6. The method for transdimensional Bayesian inversion of Rayleigh surface wave dispersion curves according to claim 5, characterized in that, The inversion results obtained for the geological area to be tested include: When inverting each subsurface shear wave velocity parameter to be inverted within the subsurface medium model, based on the target posterior distribution and the set posterior energy, and using reversible skip Markov chain Monte Carlo sampling, the current subsurface medium model is... m Proposal to a new subsurface medium model m′ ; In the current underground medium model m Proposal to a new subsurface medium model m′ During the process, the acceptance rate is set as follows: ； in: This indicates a proposal ratio or Jacobian term for a cross-dimensional proposal; T Indicates temperature parameter; Indicates the increase in energy; T Indicates temperature; and These represent the forward and reverse proposal densities, respectively. Based on the set acceptance rate, the reversible jumping Markov chain is guided to Monte Carlo sampling to explore physical regions with high posterior probabilities, and the posterior probability density distribution, interface depth histogram, maximum posterior probability model, mean profile model and layer posterior distribution of the underground medium model are obtained.

7. The method for transdimensional Bayesian inversion of Rayleigh surface wave dispersion curves according to claim 6, characterized in that, In the process of Monte Carlo sampling using reversible jumping Markov chains, a parallel tempering method is adopted, in which multiple Markov chains at different temperatures are set up for parallel sampling and periodically exchanged to accelerate the exploration of the global model space. Finally, cold chain samples are used for posterior statistics.

8. A trans-Viges Bayesian inversion device for Rayleigh surface wave dispersion curves, characterized in that, include: The data processing module is used to collect Rayleigh surface wave data of the geological area to be measured and extract multimode dispersion curve data from the Rayleigh surface wave data; The Dix inversion module is used to invert the subsurface medium model constructed for the geological area under test based on multimodal dispersion curve data and the Dix inversion method to obtain a reference shear wave velocity profile. The reference shear wave velocity profile is then interpolated and mapped onto the parameterized grid of the subsurface medium model so that each subsurface shear wave velocity parameter to be inverted in the subsurface medium model has a corresponding reference shear wave velocity value. The soft prior construction module is used to construct a Gaussian probability density function with the corresponding reference shear wave velocity value as the mean for each subsurface shear wave velocity parameter to be inverted, and use the Gaussian probability density function as the soft prior probability distribution field of the subsurface shear wave velocity parameter to be inverted; at the same time, it obtains the formation depth value of the subsurface shear wave velocity parameter to be inverted from the subsurface medium model, sets a scaling factor based on the depth value, multiplies the scaling factor with the reference shear wave velocity value corresponding to the subsurface shear wave velocity parameter to be inverted, and uses the product as the depth-related prior standard deviation of the Gaussian probability density function; The probability inversion module is used to perform velocity inversion of the subsurface medium model based on the soft prior probability distribution field of each subsurface shear wave velocity parameter to be inverted, and to obtain the inversion results of the geological area to be measured.

9. An electronic device, characterized in that, include: Memory and processor; The memory is used to store computer programs; When the processor executes the computer program stored in the memory, it implements the steps of the Rayleigh surface wave dispersion curve transdimensional Bayesian inversion method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Used to store a computer program, which, when executed by a processor, implements the steps of a transdimensional Bayesian inversion method for Rayleigh surface wave dispersion curves as described in any one of claims 1 to 7.

Images