A tunnel seismic three-dimensional full waveform inversion method

Abstract

The present application relates to the technical field of tunnel engineering, and discloses a tunnel seismic three-dimensional full waveform inversion method, comprising the following steps: S1, establishing a three-dimensional initial velocity model: a three-dimensional initial velocity model is established based on a common reflection point migration method of ellipsoid expansion; S2, forward simulation three-dimensional full space wave field characteristic analysis: forward simulation is carried out in the established typical adverse geological body model, a seismic source point is arranged on the tunnel wall or the tunnel face, a geophone point is arranged on the tunnel wall, and the characteristics of three-dimensional full space wave field signals received by the tunnel wall and the tunnel face are analyzed; S3, full waveform inversion numerical simulation: based on the established initial velocity model, a full waveform inversion numerical model experiment is realized, and simulation data are obtained; S5, three-dimensional full waveform inversion imaging; and S6, result verification. The present application can improve the tunnel seismic three-dimensional full waveform inversion precision, and realize a low-cost, safe, efficient and high-precision tunnel seismic three-dimensional advanced prediction method.

Description

Technical Field

[0001] This invention relates to the field of tunnel engineering technology, and specifically to a method for three-dimensional full-waveform inversion of tunnel seismic data. Background Technology

[0002] The construction of tunnels has demonstrated significant advantages in overcoming terrain constraints, improving road alignment, and shortening travel distances, becoming a key measure to enhance the quality of road transportation and services. Tunnel engineering, being deeply buried underground, faces complex and variable engineering geological and hydrogeological conditions. During construction, it frequently encounters adverse geological formations such as fault fracture zones, karst development zones, and water-rich zones. Due to the unclear geological conditions ahead of the tunnel face, failure to take timely and effective measures can easily lead to disasters such as large deformations, collapses, water inrushes, and mudslides, potentially causing significant loss of life and property. This situation is particularly pronounced in areas with complex geological conditions such as high mountains and canyons with intense tectonic activity and well-developed surface water systems.

[0003] To ensure the safety of the project and personnel, guarantee the smooth construction of the tunnel, and minimize economic losses, advanced geological forecasting is of paramount importance. Through advanced geological forecasting, it is possible to accurately understand the changes in the surrounding rock structure and geological hazards ahead of the tunnel face, thereby allowing for timely and appropriate modifications to the construction plan, the implementation of protective measures, and the avoidance or reduction of potential dangers. Therefore, the ability to accurately and promptly predict adverse geological factors ahead of the tunnel face is crucial to the success of tunnel construction.

[0004] Elastic wave detection is a crucial method for advanced geological detection in tunnels, accurately predicting the spatial distribution characteristics of adverse geological bodies within a certain range ahead of the tunnel face. Despite previous breakthroughs in tunnel accelerometers, seismic sources, and data processing, the accuracy of geological predictions remains limited by the low velocity accuracy of geological bodies in the advance direction. Three-dimensional elastic wave full-waveform inversion, currently the most accurate velocity analysis method, faces the following challenges: existing tunnel earthquake prediction methods lack data on the tunnel face and both sides of the tunnel walls; seismic sources often employ explosives and mechanical impact sources, which are costly but inefficient and produce small datasets; and prediction results are primarily two-dimensional or pseudo-three-dimensional, failing to achieve true three-dimensional detection and imaging.

[0005] Explosive-powered seismic sources, when used within boreholes, can avoid the effects of excavation damage zones, resulting in high signal-to-noise ratios (SNRs). Mechanically driven impact seismic sources, on the tunnel wall surface, are affected by excavation damage zones, leading to lower data quality. A small amount of high SNR data reduces the accuracy of the full waveform inversion due to its limited quantity, while a large amount of low SNR data also reduces the accuracy of the full waveform inversion due to its low SNR.

[0006] The accuracy of the initial model is a key parameter in full waveform inversion, directly affecting computational stability, iteration count, and accuracy. A high-accuracy initial model enables rapid convergence of the inversion calculation process, saving computation time. Existing initial velocity models established using direct wave fitting and tomographic inversion methods have low accuracy and deviate significantly from actual surrounding rock velocities, leading to unstable full waveform inversion results.

[0007] The three problems of insufficient data volume, low data signal-to-noise ratio, and low accuracy of initial velocity model severely restrict the accuracy of full waveform inversion of tunnel seismic data, thus limiting the prediction accuracy of adverse geological bodies. Summary of the Invention

[0008] To address the aforementioned shortcomings in existing technologies, this invention provides a three-dimensional full-waveform inversion method for tunnel seismic exploration, which solves the problems of high cost, low excitation efficiency of explosive sources, low signal-to-noise ratio and small data volume of mechanical impact source data.

[0009] To achieve the above-mentioned objectives, the technical solution adopted by this invention is: a three-dimensional full-waveform inversion method for tunnel seismic data, comprising the following steps:

[0010] S1. Establish a three-dimensional initial velocity model: Based on the common reflection point migration method of ellipsoidal expansion, perform three-dimensional migration imaging to obtain the velocity spectrum, select the velocity corresponding to the maximum amplitude as the optimal velocity, and establish a three-dimensional initial velocity model.

[0011] S2, Forward modeling wavefield characteristic analysis: Establish a typical adverse geological body model, which includes three-dimensional karst caves and fault fracture zones; perform forward modeling in the established typical adverse geological body model, set up source points at the tunnel wall or working face, set up detector points at the tunnel wall, and analyze the characteristics of seismic wave signals received at the tunnel wall and working face locations.

[0012] S3, Full waveform inversion numerical simulation: Using the seismic wave signal data received in S2, and based on the three-dimensional initial velocity model in S1, a full waveform inversion numerical model experiment is conducted to obtain simulation data;

[0013] S4, Deployment of observation system and data acquisition and processing: Deploy a symmetrical joint observation system in the actual tunnel, implement joint processing of dual source data from explosive source and mechanical impact source to obtain actual data;

[0014] S5, Three-dimensional full waveform inversion imaging: By comparing the simulation data obtained in S3 and the actual data obtained in S4, the residual is calculated, the gradient is calculated using the residual, and the gradient is used to update the parameters of the three-dimensional initial velocity model to obtain a reliable three-dimensional velocity model, thereby realizing the three-dimensional full waveform inversion of the tunnel and accurately calculating the physical quantities of the surrounding rock, including longitudinal wave velocity, transverse wave velocity, Poisson's ratio, Young's modulus, and volumetric modulus.

[0015] S6, Result Verification: Verify the accuracy of the calculation results in reverse, obtain the actual correction coefficients, and finally determine a reliable three-dimensional initial velocity model.

[0016] Furthermore, in the above-mentioned tunnel seismic three-dimensional full waveform inversion method, the forward modeling in S2 is the process of solving the wave equation. For three-dimensional multi-component wave equations, the wave equation is expressed as equation (1):

[0017]

[0018] Where x, y, z represent underground spatial coordinates, and m is the longitudinal wave velocity of the stratum. Let α represent the gradient of the objective function with respect to parameter m during the k-th iteration. k To update the step size.

[0019] Furthermore, in the above-mentioned three-dimensional full-waveform inversion method for tunnel seismic retrieval, the cross-correlation calculation formula for the forward extended wavefield from the shot point and the inverse time-delayed wavefield from the receiver point is Equation (2):

[0020]

[0021] Where S(x,y,z,t) and R(x,y,z,t) are the forward extended wave field and the reverse time extended wave field, respectively.

[0022] Furthermore, in the above-mentioned tunnel seismic three-dimensional full waveform inversion method, the deployment of the observation system and data acquisition and processing in S4 are specifically as follows: geophones and seismic source points are deployed on both sides of the tunnel wall, including explosive seismic sources and mechanical impact seismic sources; both types of seismic sources are excited, and elastic wave data at different locations in the tunnel are obtained through geophones; the longitudinal and transverse wave signals in the elastic wave data are extracted, and the surrounding rock velocity is calculated by fitting the direct wave velocity.

[0023] Furthermore, in the above-mentioned three-dimensional full-waveform inversion method for tunnel seismic sources, the joint processing of explosive sources and mechanical impact sources adopts the empirical Green's function calculation method, the expression of which is Equation (3):

[0024]

[0025] Here, x is a variable belonging to the spatial range A, x∈A, r0 is the distance in space A, u(x,t) is the observation record at the position of x, and v(x+r0,t) is the observation record at the position of x+r0.

[0026] Furthermore, in the above-mentioned tunnel seismic three-dimensional full waveform inversion method, the simulation data in S3 includes rock physical quantities, including P-wave velocity, S-wave velocity, Poisson's ratio, Young's modulus, and volumetric modulus.

[0027] The beneficial effects of this invention are as follows: The dual-source joint symmetrical observation system of this invention considers the impact of tunnel excavation damage zone on data signal-to-noise ratio and resolution, selects the optimal excitation point and receiver point, effectively increases the amount of data on both sides of the tunnel wall, and provides richer basic data for subsequent data processing and interpretation. The dual-source data joint processing method improves the signal-to-noise ratio of a large amount of low signal-to-noise ratio mechanical impact source data by constraining a small amount of high signal-to-noise ratio source data, thereby improving the overall data quality and providing more accurate data support for subsequent three-dimensional full waveform inversion. The three-dimensional initial velocity modeling method based on ellipsoidal expansion and common reflection point migration sets the velocity scanning range, and selects the velocity corresponding to the maximum amplitude value in the three-dimensional migration scan map as the optimal velocity of that spatial point. Combined with massive data calculations, the relationship between the velocity corresponding to the maximum amplitude energy in the three-dimensional migration data and the actual velocity of the surrounding rock is established, and empirical correction coefficients for different lithologies are summarized. Based on these correction coefficients, a reliable three-dimensional initial velocity model is established, thereby improving the accuracy of three-dimensional full waveform inversion and thus improving prediction accuracy. Attached Figure Description

[0028] Figure 1 This is a flowchart of the method;

[0029] Figure 2 This is a schematic diagram based on the offset of the common reflection point of the ellipsoid.

[0030] Figure 3 This is a diagram of a symmetrical joint observation system with two seismic sources. Detailed Implementation

[0031] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0032] like Figure 1 As shown, this embodiment provides a method for three-dimensional full waveform inversion of tunnel earthquakes. This embodiment studies a dual-source symmetrical joint observation system, joint processing of dual-source earthquake data, and an initial velocity modeling method based on the common reflection point migration of ellipsoidal expansion, so as to improve the accuracy of three-dimensional full waveform inversion of tunnel earthquakes and realize a low-cost, safe, efficient and high-precision method for three-dimensional advance prediction of tunnel earthquakes.

[0033] The common reflection point migration method based on ellipsoidal expansion can be used to achieve migration imaging of a point in space. It assumes that given the correct velocity, the optimal amplitude value at that point can be obtained; conversely, the velocity corresponding to the strongest amplitude value at that point is the optimal velocity for that point. Based on this fundamental theory, a certain velocity range is defined for all points in space. Three-dimensional migration imaging is performed with a fixed velocity step size to obtain the velocity spectrum. The velocity corresponding to the maximum amplitude value at a point in space is selected as the optimal velocity, resulting in a three-dimensional initial velocity model. Based on this, three-dimensional full-waveform inversion imaging can be performed.

[0034] like Figure 2 As shown, P0 to P2 are the common reflection points of curves C1 and C2 on the ellipsoid. G1 and G2 are the coordinates of the receiver point, and S is the coordinate of the shot point. a and b are the major and minor semi-axises of curve C2 on the ellipsoid. Given a velocity v, the amplitude value of point P0(x,y,z) in the tunnel advance direction is obtained; conversely, the velocity corresponding to the maximum amplitude value at point P0 is the optimal velocity.

[0035] Based on the above, three-dimensional elastic wave numerical simulations were conducted to establish three-dimensional velocity models for adverse geological bodies such as karst caves and weak interlayers. Numerical simulations yielded three-part seismic records. Ellipsoidal common reflection point migration imaging was performed using the model data to obtain three-dimensional migration images, which were then compared with the three-dimensional velocity model to verify the accuracy of the migration method. Given a velocity range and velocity step size, constant-velocity three-dimensional scanning was used to obtain an initial three-dimensional velocity model, which was then compared with a correct three-dimensional velocity model to statistically determine empirical correction coefficients.

[0036] Based on the above, multiple tunnel experiments were conducted, and methods such as automatic gain control, bandpass filtering, P-wave and S-wave separation, FK wavefield separation, and velocity spectrum scanning were applied to process the measured seismic data of the tunnel. Following the simulation data analysis process, empirical correction coefficients for the measured data were obtained, and finally, a reliable three-dimensional initial velocity model was established.

[0037] Forward modeling is the foundation and prerequisite for full-waveform inversion. This involves deriving the three-dimensional multi-wave, multi-component forward modeling equations for radiometric waves, the perfectly matched layer boundary conditions, and the source loading stress equations. The partial derivative equations are transformed into second-order time and twelfth-order spatial difference equations, and the forward modeling is implemented through programming. The adjoint equations, objective functional, and stability factors for full-waveform inversion are derived and implemented using Fortran programming. The formula for model updates during inversion calculations is as follows: Where x, y, z represent underground spatial coordinates, and m is the longitudinal wave velocity of the stratum. This represents the gradient of the objective function with respect to parameter m during the k-th iteration. α kThe update step size is typically a positive constant. Full waveform inversion uses the adjoint state method to calculate the gradient, avoiding the enormous computational cost of calculating the Frechet derivative. The model gradient can be obtained by cross-correlation between the forward extended wavefield from the shot point and the inverse time-extended wavefield from the receiver point, calculated using the following formula: S(x,y,z,t) and R(x,y,z,t) represent the forward and reverse time-extended wavefields, respectively. The excitation function for the forward extension is a wavelet estimated using the reciprocal bispectral method. The excitation function for the reverse time-extended wavefield is transformed into the residuals between the simulated and observed data.

[0038] During implementation, a symmetrical joint observation system was deployed inside the actual tunnel, and joint processing of dual-source data was carried out. For example... Figure 3 As shown, the deployment of the observation system and data acquisition and processing are as follows: One geophone and 24 seismic source points are placed on each side of the tunnel wall, totaling 2 geophones and 48 seismic source points. This requires drilling 2 geophone holes and 2 boreholes. The 2 boreholes are 1.5m deep, and the 2 geophone holes are 2m deep. During data acquisition, two shots are sequentially fired using explosive sources at the borehole locations, and 48 shots are fired using mechanical impact sources at the other source points. Elastic wave data from different locations in the tunnel are acquired through the geophones. The P-wave and S-wave signals are extracted from the elastic wave data, and the surrounding rock velocity is calculated using direct wave velocity fitting to obtain high signal-to-noise ratio seismic data from both sides of the tunnel wall. The joint processing of the explosive source and mechanical impact source uses the empirical Green's function calculation method, the expression of which is: In this model, x is a variable belonging to the spatial range A, x∈A, r0 is the distance in space A, u(x,t) is the observation record at point x, and v(x+r0,t) is the observation record at point x+r0. Combining the symmetrical observation system, u(x,t) can be considered the mechanically driven impact source records of shots 1 to 24, and v(x+r0,t) is the explosive source record of shot 24. If the explosive source record is missing, v(x+r0,t) is the mechanically driven impact source record of shot 24. Based on this, a reliable three-dimensional initial velocity model is established using the common reflection point migration method based on ellipsoidal expansion. Finally, true three-dimensional full-waveform inversion imaging of the tunnel seismic system is implemented to accurately calculate the P-wave and S-wave velocities and Poisson's ratio parameters of the surrounding rock. Subsequent excavation results are used to verify the accuracy of the calculation results, obtain the actual correction coefficients, and finally determine the reliable three-dimensional initial velocity model.

Claims

1. A method for three-dimensional full-waveform inversion of tunnel seismic data, characterized in that, Includes the following steps: S1. Establish a three-dimensional initial velocity model: Based on the common reflection point migration method of ellipsoidal expansion, perform three-dimensional migration imaging to obtain the velocity spectrum, select the velocity corresponding to the maximum amplitude as the optimal velocity, and establish a three-dimensional initial velocity model. S2, Forward modeling wavefield characteristic analysis: Establish a typical adverse geological body model, which includes three-dimensional karst caves and fault fracture zones; perform forward modeling in the established typical adverse geological body model, set up source points at the tunnel wall or working face, set up detector points at the tunnel wall, and analyze the characteristics of seismic wave signals received at the tunnel wall and working face locations. S3, Full waveform inversion numerical simulation: Using the seismic wave signal data received in S2, and based on the three-dimensional initial velocity model in S1, a full waveform inversion numerical model experiment is conducted to obtain simulation data; S4, Deployment of observation system and data acquisition and processing: Deploy a symmetrical joint observation system in the actual tunnel, implement joint processing of dual source data from explosive source and mechanical impact source to obtain actual data; S5, Three-dimensional full waveform inversion imaging: By comparing the simulation data obtained in S3 and the actual data obtained in S4, the residual is calculated, the gradient is calculated using the residual, and the gradient is used to update the parameters of the three-dimensional initial velocity model to obtain a reliable three-dimensional velocity model, realizing the three-dimensional full waveform inversion of the tunnel, and accurately calculating the physical quantities of the surrounding rock, including the longitudinal wave velocity, transverse wave velocity, Poisson's ratio, Young's modulus and volumetric modulus; S6, Result Verification: Verify the accuracy of the calculation results in reverse, obtain the actual correction coefficients, and finally determine a reliable three-dimensional initial velocity model.

2. The tunnel seismic three-dimensional full waveform inversion method according to claim 1, characterized in that, The forward modeling in S2 is the process of solving the wave equation. For three-dimensional multi-component wave equations, the wave equation is expressed as equation (1): Where x, y, z represent underground spatial coordinates, and m is the longitudinal wave velocity of the stratum. Let α represent the gradient of the objective function with respect to parameter m during the k-th iteration. k To update the step size.

3. The tunnel seismic three-dimensional full waveform inversion method according to claim 2, characterized in that, The cross-correlation formula for the forward extended wavefield from the shot point and the reverse time-delayed extended wavefield from the receiver point is given by equation (2): Where S(x,y,z,t) and R(x,y,z,t) are the forward extended wave field and the reverse time extended wave field, respectively.

4. The tunnel seismic three-dimensional full waveform inversion method according to claim 1, characterized in that, The deployment of the observation system and data acquisition and processing in S4 specifically involves: deploying geophones and seismic source points on both sides of the tunnel wall, including explosive seismic sources and mechanical impact seismic sources; exciting both types of seismic sources and acquiring elastic wave data at different locations in the tunnel through the geophones; extracting the longitudinal and transverse wave signals from the elastic wave data and using the direct wave velocity fitting to calculate the surrounding rock velocity.

5. The tunnel seismic three-dimensional full waveform inversion method according to claim 4, characterized in that, The combined processing of the explosive source and the mechanical impact source adopts the empirical Green's function calculation method, and its expression is Equation (3): Here, x is a variable belonging to the spatial range A, x∈A, r0 is the distance in space A, u(x,t) is the observation record at the position of x, and v(x+r0,t) is the observation record at the position of x+r0.

6. The tunnel seismic three-dimensional full waveform inversion method according to claim 1, characterized in that, The simulation data in S3 includes rock physical quantities, including P-wave velocity, S-wave velocity, Poisson's ratio, Young's modulus, and volumetric modulus.

Images