Rapid simulation method for horizontal layered medium seismic waves
By using the Shanks transformation and effective calculation time in the generalized inversion method, the problem of low computational efficiency of seismic wave simulation in horizontal layered medium is solved, and the seismic wave simulation efficiency is significantly improved. It is suitable for broadband earthquake simulation and seismic safety Evaluate.
Patent Information
- Application Number
- CN202410934155.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-12
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-07-12
AI Technical Summary
In the prior art, the generalized inverse transmission method is used for horizontal layered dielectric seismic wave simulation with low calculation efficiency, and it is difficult to meet the demand for regional seismic safety of broadband earthquake simulation.
Shanks transformation is used to calculate synthetic seismic waves by generalized inverse transmission method, and the effective calculation time is introduced, so as to improve the calculation efficiency by simplifying the stratigraphic structure model and optimizing the wave number domain integration process.
The calculation efficiency of horizontal layered dielectric seismic wave simulation has been improved, with a speed of nearly 5 times, significantly improving the simulation efficiency, and is suitable for broadband earthquake simulation and seismic safety assessment.
Smart Images

Figure CN118965690A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to a method for rapidly simulating seismic waves in a horizontal layered medium, and is applicable to the technical field of seismic services. Background Art
[0002] Seismic wave simulation is an important research direction in seismology. It has important theoretical and practical significance for understanding the excitation, propagation, development and changes of seismic waves and the causes of earthquake damage. Seismic wave simulation methods with a clear physical basis mainly include two categories: numerical methods and semi-analytical methods. Numerical methods mainly include finite difference method, spectral element method, pseudo-spectral method, etc., which are often used to simulate seismic wave propagation in complex media. Due to the extremely high computational complexity, numerical methods are mostly suitable for lower frequency seismic wave simulation. For areas where the lateral heterogeneity is much smaller than the vertical heterogeneity, the underground structure can be simplified as a horizontal layered medium. This simple medium model has a semi-analytical theoretical solution and can be used for broadband seismic motion simulation, source rupture process inversion, seismic surface wave characteristics research, and seismoelectric wave field simulation of layered porous media.
[0003] The generalized back-transmission method is currently an important method for synthesizing seismic waves in layered media. It has a wide range of applications, not only for simulating seismic wave fields but also for studying source rupture processes. However, this method is typically used for simulating low-frequency seismic waves, and its computational efficiency has not been a focus of attention. Broadband seismic motion is a major factor in causing earthquake disasters, and its simulation and calculation are even more crucial for regional seismic safety. The generalized back-transmission method has a clear physical meaning, and the simulated broadband Green function is more physically meaningful than the empirical Green function. However, its low computational efficiency has hindered its application.
[0004] Therefore, a method for rapidly simulating seismic waves in horizontal layered media with higher computational efficiency is needed in the prior art. Summary of the Invention
[0005] The present application relates to a method for rapid simulation of seismic waves in horizontal layered media, in which the Shanks transform is used to synthesize seismic waves using the generalized back-transmission method, and a method for effective calculation of duration is introduced, so that the computational efficiency of the simulation method of the present application can be greatly improved.
[0006] The present application relates to a method for rapidly simulating seismic waves in a horizontal layered medium, comprising the following steps:
[0007] (1) Based on the global velocity structure model of seismic waves, the underground stratum structure is simplified into a horizontal layered medium, and a horizontal layered medium model is established in the study area;
[0008] (2) Calculate the local reflection and transmission coefficient of the seismic wave at each interface based on the thickness and medium parameters of each layer in the horizontal layered medium model;
[0009] (3) Calculate the total reflection and transmission coefficient values of all layer interfaces based on the surface reflection coefficient, infinite boundary conditions and free surface boundary conditions, as well as the focal depth and radiation pattern;
[0010] (4) Calculate the kernel function containing the stratum structure information and the field point location information based on the calculated total reflection and transmission coefficient values of all layer interfaces;
[0011] (5) Calculate the wavenumber domain integral by Shanks transform based on the kernel function and Bessel function;
[0012] (6) Performing an inverse Fourier transform on the frequency domain integration result calculated in step (5) to obtain the seismic wave time history, i.e., the seismic wave of the layered medium.
[0013] In step (5), the generalized inverse transmission algorithm is used to synthesize the seismic waves in the layered medium, and the following oscillation function is used for wavenumber domain integration:
[0014] I n (w)=∫F(k,w)·J n (kr)dk (1)
[0015] Where r is the epicentral distance, k is the spatial wave number, w is the angular frequency, and J n (kr) is the nth-order Bessel function; F(k,w) is the kernel function calculated in step (4);
[0016] Discretize the integral of the above formula, that is
[0017]
[0018] The sequence I obtained after discretization integration n (w) Perform Shanks transformation to construct I n (w); when the Shanks transform sequence reaches convergence, the converged value is the integral result of the wavenumber domain integral. Preferably, at least two orders of Shanks transform can be performed, and the interval dx of the Shanks transform can be taken as:
[0019]
[0020] In step (6), the actual ground motion time series of the site is recorded as L1(t'),
[0021]
[0022] Where t' is the actual ground motion time, w is the angular frequency, r is the epicenter distance, S(w) is the source time function spectrum, I n(w) is the propagation path spectrum containing stratum structure information and field location information, T s is the propagation time of seismic waves in the medium. Its discrete form is used in actual calculations:
[0023]
[0024] Preferably, the effective computing time can be calculated using the following formula:
[0025]
[0026] Among them, r0 represents the straight-line distance from the earthquake source to the field point, T represents the total earthquake wave time, and v max is the maximum wave velocity of the model.
[0027] To improve the simulation efficiency of seismic waves in horizontally layered media, this application adopts two methods:
[0028] (1) Implementing Shanks transform: The generalized inverse transmission method synthesizes seismic waves. It is necessary to calculate the wavenumber domain integral of an oscillation function consisting of a series of kernel functions multiplied by Bessel functions in the frequency domain. In order to accelerate the convergence of these integrals, this application implements the Shanks transform and provides a basis for selecting the Shanks transform step size. This allows the integral to converge before the function oscillation ends, without having to wait until the oscillation stops.
[0029] (2) The effective calculation time is introduced: the generalized inverse transmission method needs to calculate the time domain seismic waves by the spectrum inverse Fourier transform. For a certain bandwidth, the number of discrete spectra that need to be calculated is determined by the spectrum interval, which is equal to the inverse of the calculation time. Therefore, the shorter the calculation time, the fewer the number of spectra calculated, and the higher the calculation efficiency. This application defines the effective calculation time as mainly composed of the ground motion time of the field point, excluding the quiet period when the ground does not move, and provides a formula for estimating the effective calculation time according to the medium parameters;
[0030] (3) The present application simultaneously implements Shanks transformation and effective duration synthetic seismic waves, which can increase the speed by nearly 5 times, thereby improving the simulation efficiency of seismic waves in horizontal layered media. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] Figure 1 Schematic diagram of the horizontal layered medium model of this application.
[0032] Figure 2 Show this application's A n Schematic diagram of the convergence process of a slow sum approaching one million terms.
[0033] Figure 3 2 is a time comparison chart of the ZC algorithm in the embodiment and the algorithm of the present application.
[0034] Figure 4 Schematic diagram of the time ratio between the ZC algorithm in the embodiment and the algorithm of the present application. DETAILED DESCRIPTION
[0035] To make the purpose, technical solutions and advantages of this application more clear, the embodiments of this application will be described in detail below with reference to the accompanying drawings. It should be noted that, unless there is a conflict, the embodiments and features in the embodiments of this application can be combined with each other in any way.
[0036] Studying the velocity structure of seismic wave propagation requires the use of seismic wave velocity models. These models describe the variations in seismic wave propagation velocity in different media within the Earth. Using these models, we can simulate the propagation of seismic waves within the Earth, thereby studying their propagation paths and velocity variations. The velocity structure within the Earth primarily includes crustal, mantle, and core velocities. Studying the propagation paths and velocity structure of seismic waves is crucial for predicting earthquake activity and preventing and controlling earthquake disasters. By studying the propagation paths of seismic waves, we can determine the path taken by an earthquake's epicenter to reach the surface, thereby predicting the propagation of seismic waves in different regions. This helps facilitate early earthquake warning and emergency preparedness.
[0037] Seismic waves propagate primarily from the earthquake's source to the surface, then through various media within the Earth. The velocity structure of seismic waves describes the variations in their propagation velocity within different media within the Earth. Seismic records from seismic monitoring equipment, observations from seismic networks, and velocity models allow for the study of seismic wave propagation paths and velocity structure. These studies can reveal structural information about the Earth's interior, explain the mechanisms of seismic activity, and provide a scientific basis for earthquake disaster prevention and control.
[0038] The present application relates to a method for rapidly simulating seismic waves in a horizontal layered medium, which comprises the following steps:
[0039] (1) Based on the global velocity structure model of seismic waves, the underground stratum structure is simplified into a horizontal layered medium, and a horizontal layered medium model is established in the study area;
[0040] (2) According to the thickness of each layer and the medium parameters (λ j , μ j , ρ j ), calculate the local reflection and transmission coefficient of the seismic wave at each interface (R j ,T j );
[0041] (3) According to the free surface reflection coefficient and the infinite boundary R N+1= 0, and the focal depth and radiation pattern are used to calculate the total reflection and transmission coefficient values of all layer interfaces.
[0042] (4) According to the calculated total reflection and transmission coefficient value of all layer interfaces Calculate the kernel function F(k,w) that includes stratum structure information and field point location information;
[0043] (5) According to the kernel function and Bessel function, the wave number domain integral I is calculated by Shanks transformation. n (w)=∫F(k,w)·J n (kr)dk; In this step, the Shanks transform is implemented, which speeds up the generalized inverse transmission algorithm by more than 2 times;
[0044] (6) The frequency domain integration result I calculated in step (5) n Perform inverse Fourier transform to obtain the seismic wave time history Wave(t). In this step, the effective calculation time is introduced to speed up the generalized inverse transmission algorithm by about 2.5 times. The calculation formula of inverse Fourier transform is:
[0045] The above steps (1) to (4) can all be implemented using existing technologies and will not be described in detail here.
[0046] About Shanks Transform
[0047] The biggest advantage of Shanks sequence transformation is that it can make a sequence that converges slowly converge quickly. For a sequence S in the form of equation (3) n
[0048] S n =A+αq n (3)
[0049] If q < 0 and |q| < 1, the sequence is an instantaneous convergence sequence, and the convergence result is A, that is,
[0050] S n =A,n→∞ (4)
[0051] And A can pass the sequence S n The n-1, n and n+1 items of
[0052]
[0053] Combining the above three expressions, we can solve the sequence S n The limit converges to A when n→∞, that is,
[0054]
[0055] Therefore, for any sequence {a i}, if the sum of the sequence To be determined. Add a driving model to this sequence to make it an instantaneous convergence sequence S n =A+αq n , can achieve the goal of improving the sequence convergence speed. Definition {a i The sum of the parts of} is:
[0056]
[0057] Construct a new sequence {A n}, if the sequence converges, when n→∞A n will converge to the limit A ∞ . Sequence A n Shanks transform {S(A n )} is the new sequence
[0058]
[0059] Usually the convergence speed of this sequence is faster than that of the original sequence. Repeated use of Shanks transform can further speed up the calculation, that is,
[0060] S 2 (A n )=S(S(A n )) (9)
[0061] S 3 (A n )=S(S(S(A n ))) (10)
[0062] Here is a simple example of the Shanks transform: Consider the series Its harmony It is a convergent sequence, but it converges very slowly, requiring millions of sums to converge to six-digit precision. Figure 2 Display A n Convergence of a slow summation approaching one million terms, where it is shown that after one million summations, A n Still fluctuating around 0.6931472. Table 1 shows the A n The convergence process of the 3rd order Shanks transform. As can be seen from the table, only 12 {a n}, the third-order Shanks change reaches a convergence value of 0.6931472.
[0063] Table 1 Sequence A n The third-order Shanks transform of
[0064]
[0065] Regarding the generalized inverse transmission algorithm in step (5)
[0066] The generalized inverse transmission algorithm synthesizes seismic waves in layered media, which requires completing the wavenumber domain integration of multiple sets of oscillation functions in the form of equation (11):
[0067] I n (w)=∫F(k,w)·J n (kr)dk (11)
[0068] Where r is the epicentral distance, k is the spatial wave number, w is the angular frequency, and J n (kr) is the nth-order Bessel function; F(k,w) is the complex kernel function, which contains information about layered media and focal depth. n (w) The integral convergence speed is slow, which is the most time-consuming calculation part of the synthetic seismic wave function. Therefore, increasing I n The convergence speed of (w) is the key to improving the efficiency of the generalized inverse transmission algorithm.
[0069] This application implements Shanks transformation on equation (11) to improve I n Since Equation (11) does not have a direct integral value, it must be discretized, that is,
[0070]
[0071] At this time, the partial integrals of a single frequency I1(w), I2(w), I3(w)…, I n (w),… form a new sequence, and sequence I n (w) Perform Shanks transformation to construct I n (w) The Shanks transform sequence. When the Shanks transform sequence reaches convergence, its value is the integral result of equation (11).
[0072] After comparative calculations, this application found that the second-order Shanks transform is more suitable for synthesizing layered seismic waves using the generalized back-transmission method, eliminating the need for a higher-order Shanks transform. Since only a lower-order Shanks transform is involved, equations (9-10) can be used to meet the computational requirements for equation (11).
[0073] Define dx = k i -k i-1 is the interval of Shanks transform. In this application, the interval dx of Shanks transform is taken as:
[0074]
[0075] After multiple model experiments, this value is valid for multilayer media and other layered media. The key reason is that before the kernel function reaches the minimum period oscillation, the Shanks transform has already made I n The sequence reaches convergence, so the value of dx in equation (13) is quite conservative and can meet the computational requirements of many models.
[0076] Regarding the effective calculation time in step (6)
[0077] The generalized inverse transmission method synthesizes the seismic waves of the field point. First, it is necessary to complete the calculation of the discrete frequency spectrum C(f) of the seismic wave of the field point, and then transform it into the time domain through the inverse Fourier transform to obtain the time series L(t) of the seismic wave of the field point. Here, the frequency spectrum of each field point is equal to the product of the source spectrum and the propagation path spectrum, that is,
[0078] C(w)=S(w)I n (w) (14)
[0079] Where w is the angular frequency, r is the epicenter distance, S(w) is the source time function spectrum, I n (w) is the propagation path spectrum, which is determined by parameters such as medium parameters, source depth, and epicenter distance. C(w) already includes the propagation time of the seismic wave and the ground motion time after arriving at the site. Using equation (14) and inverse Fourier transform, the time domain series L(t) can be obtained:
[0080]
[0081] Where w is the angular frequency, t is the time, and r is the epicenter distance. Using discrete calculation for equation (15), if the total duration of L(t) is T and the maximum frequency is f max , the corresponding discrete frequency spectrum interval is df = 1 / T, and the total number of frequencies that need to be discretely calculated is f max ×T.
[0082] The total length T of L(t) includes the propagation time of the seismic wave in the medium and the ground motion time of the field point. No ground fluctuation occurs before the ground motion of the field point. The actual ground motion time series of the field point is recorded as L1(t'), and its length is T1. This period of time is mainly the field point movement time, and less includes the propagation time of the wave from the source to the field point. The waveform vibration part of L(t) and L1(t') is exactly the same, the only difference is that L(t) is T longer than the total length of L1(t'). s Considering L(t) and L1(t') as simple independent time series, their continuous spectra are:
[0083]
[0084] Since T = T l +Ts , t=t'+T s , so from the above two formulas we get
[0085]
[0086] Taking T1 as the calculation length, the spectrum sequence C(w j ),Right now
[0087] C(w j )=S(w j )I n (w k ) (17)
[0088] in At this time C(w k ) contains the propagation time.
[0089] According to equation (16), equation (17) is multiplied by the time shift factor Then perform the inverse Fourier transform to obtain the L1(t') sequence that does not include the propagation time.
[0090]
[0091] Actual calculations require discrete form
[0092] For discrete frequencies w j , the discrete number of total computation time is w max T, using the effective time discrete number w max T1, obviously w max T1<w max T, so using T1 to calculate seismic waves saves more time than using T. T1 is defined as the effective time, which does not include or less includes the ground motion time.
[0093] The maximum and minimum wave velocities of the computational model can be used to roughly estimate the effective computation time. If r0 represents the straight-line distance from the earthquake source to the field point, and T represents the total seismic wave time, then the time it takes for the wave to propagate from the earthquake source to the field point at the fastest wave velocity is approximately equal to the quiet time before the field point moves. Therefore, the effective computation time estimate is Equation (19):
[0094]
[0095] Among them, v max is the maximum wave velocity of the model.
[0096] The calculation results show that the effective duration can speed up the process by nearly 2 times compared to the total duration. This simple transformation brings very obvious results.
[0097] Example
[0098] Table 2 Parameters of layered medium model
[0099]
[0100] This embodiment applies the Shanks transform and effective duration described above to improve the computational efficiency of synthesizing seismic waves using the generalized inverse transmission algorithm, that is, simultaneously applies these two methods to the four-layer model in Table 2.
[0101] For the four-layer model in Table 2, different field points within 0-300 km are selected as calculation points. The calculation bandwidth of each field point is 0-30 Hz. The comparison time is the total calculation time for completing the 0-30 Hz frequency spectrum for each field point and converting it into a time series through inverse Fourier transformation. Figure 3 The red line represents the calculation time of each field point by the ZC algorithm, and the black line represents the calculation time of each field point using both the effective calculation time and the Shanks change. Five total calculation times are set for field points at different epicenter distances: when the epicenter distance is less than 20 km, the total calculation time is 30 seconds; when the epicenter distance is greater than 20 km and less than 50 km, the total calculation time is 50 seconds; when the epicenter distance is greater than 50 km and less than 100 km, the total calculation time is 100 seconds; when the epicenter distance is greater than 100 km and less than 150 km, the total calculation time is 120 seconds; and when the epicenter distance is greater than 150 km, the total calculation time is 150 seconds. For these five total calculation times, the ZC calculation time represented by the red line shows a step-by-step mutation increase, and the increase is very obvious. The SH time (algorithm of this application) calculation time represented by the black line is the effective time calculated according to equation (19), and the calculation time changes much less drastically with the epicenter distance than the ZC. Moreover, for sites farther than 150 km, the absolute value of the calculation time of the "Shanks transform + effective duration" is much smaller than that of the ZC algorithm. For example, at a site 300 km away, the ZC algorithm takes about 3000 seconds, and the "Shanks transform + effective duration" takes about 500 seconds. The absolute value difference between the two is about 2500 seconds. Figure 4 The figure shows the relative comparison between the calculation time of the ZC algorithm and the calculation time of the "Shanks transform + effective duration". The horizontal axis represents the distance of the field point, and the vertical axis represents the ratio of the calculation time of the two. As shown in the figure, for field points 50 km away, the "Shanks transform + effective duration" is about 5 times faster than the ZC algorithm, and it is even faster within 50 km. Therefore, the method proposed in this application to improve the efficiency of generalized inverse transmission calculation has a very significant speed-up effect, and its introduction can promote the further application of the generalized inverse transmission method in seismology. The above-mentioned ZC algorithm is the abbreviation of an algorithm proposed by Zhang, Chen and Chang in 2003, which belongs to the prior art.
[0102] This application proposes two improvements to improve the computational efficiency of the rapid simulation method for seismic waves in horizontal layered media:
[0103] (1) The Shanks transform is applied to the wavenumber domain integral of the product of the kernel function and the Bessel function in the generalized reverse transmission (GRT) irregular high-frequency oscillation function, and the basis for selecting the Shanks transform interval is given. It is verified that the use of the second-order Shanks transform can greatly improve the convergence speed of the wavenumber domain integral of the irregular high-frequency oscillation function, thereby greatly reducing the amount of calculation. This method improves the simulation efficiency by 2 to 5 times.
[0104] (2) Effective calculation time: from the propagation time and movement time of each field point, the propagation time is subtracted as much as possible, and the ground movement time of the field point is retained to obtain a shorter effective calculation time. The shorter effective calculation time reduces the number of spectral points that need to be calculated for the field point, that is, reduces the number of oscillation integral calculations, thereby improving the calculation efficiency by about 1 times. For the four-layer medium model of this application, the above two methods are implemented simultaneously, which makes the calculation efficiency of large epicentral distances increase by about 5 times, and the small epicentral distances increase even more, up to more than ten times.
Claims
1. A method for rapid simulation of seismic waves in horizontal layered media, characterized in that: The following steps are involved: (1) Based on the global velocity structure model of seismic waves, the underground strata structure is simplified into a horizontal layered medium, and a horizontal layered medium model in the study area is established; (2) Calculate the local reflection and transmission coefficient of the seismic wave at each interface according to the thickness and medium parameters of each layer in the horizontal layered medium model; (3) Calculate the total reflection and transmission coefficient values of all layer interfaces based on the surface reflection coefficient, infinite boundary conditions and free surface boundary conditions, as well as the focal depth and radiation pattern; (4) Calculating a kernel function including stratum structure information and field point location information based on the calculated local reflection and transmission coefficients of all layer interfaces and the total reflection and transmission coefficient values; (5) Calculate the wave number domain integral by Shanks transformation based on the kernel function and Bessel function; (6) Performing an inverse Fourier transform on the frequency domain integration result calculated in step (5) to obtain the seismic wave time history, i.e., the layered medium seismic wave.
2. The rapid simulation method of seismic waves in horizontal layered media according to claim 1, characterized in that: In step (5), the layered medium seismic waves are synthesized using the generalized inverse transmission algorithm, and the following oscillation function is used for wavenumber domain integration: AND n (w)=∫F(k,w) J n (kr)dk Where r is the epicentral distance, k is the spatial wave number, w is the angular frequency, and J n (kr) is the nth-order Bessel function; F(k,w) is the kernel function calculated in step (4); Discretize the above equation and integrate it, that is, The sequence I obtained after discretization integration n (w) Perform Shanks transformation to construct each frequency I at each field point n (w); when the Shanks transform sequence reaches convergence, the converged value is the integral result of the wavenumber domain integration.
3. The rapid simulation method of seismic waves in horizontal layered media according to claim 2, characterized in that: In step (5), at least two-order Shanks transformation is performed.
4. The rapid simulation method of seismic waves in horizontal layered media according to claim 2 or 3, characterized in that: In step (5), the interval dx of the Shanks transform is taken as:
5. The method for rapid simulation of seismic waves in horizontal layered media according to any one of claims 1 to 3, characterized in that: In step (6), the actual ground motion time series of the site is recorded as L1(t'), Where t' is the actual ground motion time, w is the angular frequency, r is the epicentral distance, S(w) is the frequency spectrum of the source time function, I n (w) is the propagation path spectrum containing stratum structure information and field point location information, T s is the propagation time of seismic waves in the medium.
6. The rapid simulation method of seismic waves in horizontal layered media according to claim 5, characterized in that: In step (6), the effective calculation time T1 is calculated using the following formula: Among them, r0 represents the straight-line distance from the earthquake source to the field point, T represents the total seismic wave time, and v max is the maximum wave speed of the model.
Citation Information
Patent Citations
Wave travel time obtaining method and device, imaging method and device
CN108254787A
Point source excitation seismic wave field forward modeling method in horizontal layered double-pore medium
CN116256797A
System and method for estimating one-way propagation operators
US20220317325A1
