A seismic motion simulation method based on broadband three-component hybrid method

Abstract

This invention discloses a seismic motion simulation method based on a broadband three-component hybrid method, belonging to the field of seismic motion simulation technology. This method establishes the relationship between the Fourier amplitude ratio of the mid-to-low frequency three components and source information and the Fourier amplitude ratio of the high frequency three components through neural network fitting. Then, the spectral element method is used to simulate mid-to-low frequency three-component seismic motions, thereby predicting the amplitude ratio of the high frequency three components. Next, an improved stochastic finite fault method is used to simulate high frequency three-component seismic motions with different amplitudes. Finally, matched filtering is used to superimpose the high frequency and mid-to-low frequency seismic motions, thereby obtaining a broadband three-component seismic motion that reflects the earthquake rupture process, three-dimensional inhomogeneous medium, and local site topographic conditions.

Description

Technical Field

[0001] This invention belongs to the field of seismic motion simulation technology, specifically relating to a seismic motion simulation method based on a broadband three-component hybrid method. Background Technology

[0002] In seismic design of structures, the rationality of seismic input is the premise of all structural dynamic calculations. For major projects such as dams and nuclear power plants, seismic time histories must be used directly for dynamic analysis, and artificial simulation of seismic motion is usually required to supplement the lack of real records.

[0003] Currently, commonly used seismic motion simulation methods in engineering include stochastic methods and deterministic methods. Stochastic methods (such as the stochastic finite fault method) have the advantage of simulating high-frequency ground motions relatively well; however, they cannot account for the differences between the three components of ground motion, struggle to incorporate the influence of local topographic relief on the simulation, and oversimplify the description of propagation paths and sites, failing to reflect their true characteristics. On the other hand, deterministic methods, such as the spectral element method, can address these issues; however, the required source and detailed crustal structural parameters are difficult to obtain. Furthermore, deterministic methods are often too costly to cover the frequency range usable in engineering, making them difficult to use directly in projects. Therefore, it is necessary to superimpose the high-frequency components of ground motions obtained using stochastic methods with the low-frequency components obtained using deterministic methods to obtain broadband ground motions usable in engineering.

[0004] However, the time histories of ground motions generated by the two methods are difficult to superimpose. In near-fault ground motion simulation, engineering projects require broadband (0.1Hz~25Hz) ground motion simulation results that can reflect the fault rupture process, three-dimensional inhomogeneous media, and local site topographic conditions. The stochastic finite fault method cannot reflect the differences in the three components of ground motion and the simulation effect at mid- and low frequencies is poor; the spectral element method is extremely costly and cannot meet the broadband requirement.

[0005] Therefore, existing methods are insufficient to reflect the differences in the three components of ground motion. For dams located in high mountain and canyon areas, the differences in ground motion intensity along the river, across the river, and vertically have a significant impact on the calculation results, making it difficult to achieve reasonable ground motion simulation using existing methods. Summary of the Invention

[0006] To address the aforementioned shortcomings in existing technologies, this invention provides a ground motion simulation method based on a broadband three-component hybrid method. This method solves the problem that existing stochastic finite fault methods and spectral element methods are unable to achieve broadband ground motion simulation effects that reflect the fault rupture process, three-dimensional inhomogeneous media, and local site topographic conditions in the process of simulating near-fault ground motion.

[0007] To achieve the aforementioned objectives, the present invention employs the following technical solution: a seismic motion simulation method based on a broadband three-component hybrid method, comprising the following steps:

[0008] S100. Collect regional strong earthquake records and then construct a dataset containing power spectral ratios for different frequency ranges.

[0009] S200. Using the dataset, construct and train an artificial neural network model to obtain a ground motion simulation model corresponding to different mixed frequencies;

[0010] S300, based on a broadband kinematic finite source model, uses the spectral element method to simulate low-frequency three-component ground motion;

[0011] S400. Input the power spectrum ratio obtained from the low-frequency three-component ground motion simulation into the ground motion simulation model of the corresponding mixed frequency to predict the power spectrum ratio of the high-frequency three components, and then use the stochastic finite fault method to perform high-frequency three-component ground motion simulation.

[0012] S500: The low-frequency component of the ground motion obtained from the low-frequency three-component ground motion simulation and the high-frequency component of the ground motion obtained from the high-frequency three-component ground motion simulation are superimposed to obtain the broadband three-component ground motion simulation result.

[0013] Further, step S100 includes the following sub-steps:

[0014] S101. Collect regional strong earthquake records containing effective three-component acceleration time history records;

[0015] S102. Calculate the source, path and site parameters corresponding to each group of regional strong earthquake records;

[0016] S103. Based on the strong earthquake records of each group of regions, calculate the spectral ratio corresponding to different frequency ranges, and construct a dataset by combining the source, path and site parameters.

[0017] Further, in step S103, calculating the power spectral ratio corresponding to different frequency ranges of each group of regional strong earthquake records includes:

[0018] S103-1. Calculate the Fourier amplitude spectrum of each group of regional strong earthquake records, and calculate the power spectrum of each group in several frequency ranges divided by frequency range.

[0019] S103-2, Set the starting frequency to be less than the mixing frequency. The power spectrum is divided into 11 training ranges on average, and the spectral ratio independent variable is calculated for each training range; wherein, the frequency width of each training range is the same.

[0020] S103-3, Set the starting frequency to be greater than or equal to the mixing frequency. The power spectrum is divided into 11 prediction ranges on average, and the spectral ratio dependent variable in each prediction range is calculated; wherein, the frequency width of the prediction range increases linearly with the frequency.

[0021] Further, in step S103-1, the first The power spectrum of the d-th component within a frequency range Represented as:

[0022]

[0023] In the formula, and They represent the first The start and end frequencies within a frequency range, Indicates the first earthquake. Component frequency The corresponding Fourier amplitude spectrum, ;

[0024] In step S103-2, the first The spectral ratio independent variables for each training range include the horizontal power spectral ratio. and vertical / horizontal power spectrum ratio They are respectively:

[0025]

[0026]

[0027] In the formula, , They represent the first The starting and ending range values ​​of each training range, and satisfying , and They represent The starting and ending frequency values ​​for each training range. , and These represent the power spectra of the 1st, 2nd, and 3rd components, respectively.

[0028] In step S103-3, the first The spectral ratio dependent variable for each prediction range includes the horizontal power spectral ratio. and vertical / horizontal power spectrum ratio They are respectively:

[0029]

[0030]

[0031] In the formula, , They represent the first The starting and ending range values ​​of each prediction range, and satisfying the following conditions: , and They represent The starting and ending frequency values ​​for each prediction range.

[0032] Further, in step S200, for each mixed frequency, two artificial neural network models are independently constructed and trained: a horizontal ground motion simulation model that predicts the high-frequency horizontal power spectrum ratio dependent variable based on the independent variable and the low-frequency horizontal power spectrum ratio independent variable, and a vertical / horizontal ground motion simulation model that predicts the high-frequency vertical / horizontal power spectrum ratio dependent variable based on the independent variable and the low-frequency vertical / horizontal power spectrum ratio independent variable.

[0033] Each of the artificial neural network models includes a feedforward neural network and a linear output unit connected in sequence;

[0034] The inputs to the artificial neural network model include source, path, and site parameters, as well as the logarithm of the spectral ratio independent variable within different training ranges. The output of the artificial neural network model is the logarithm of the spectral ratio dependent variable within the corresponding prediction range.

[0035] Further, step S300 includes the following sub-steps:

[0036] S301. Construct a broadband kinematic finite-source model within the region that reflects the complex source rupture process;

[0037] S302. Collect the three-dimensional heterogeneous crustal medium and high-precision topography of the region, and construct a spectral unit model;

[0038] S303. Substitute the broadband kinematic finite source model into the spectral unit model to simulate the seismic wave propagation process and obtain the three-component low-frequency components of ground motion.

[0039] Further, step S400 includes the following sub-steps:

[0040] S401. Low-frequency three-component ground motion amplitude spectrum obtained from low-frequency three-component ground motion simulation. and Calculate the corresponding horizontal power spectral density ratio and vertical / horizontal power spectral density ratio;

[0041] S402. Calculate the source, path, and site parameters based on the focal mechanism, spatial location, platform location, and site conditions.

[0042] S403. Input the horizontal power spectrum ratio and source, path and site parameters into the horizontal ground motion simulation model, and input the vertical / horizontal power spectrum ratio and source, path and site parameters into the vertical / horizontal ground motion simulation model to predict the high-frequency horizontal power spectrum ratio and the high-frequency vertical / horizontal power spectrum ratio respectively.

[0043] S404. Based on the high-frequency horizontal power spectrum ratio and the high-frequency vertical / horizontal power spectrum ratio, the random finite fault method is used to simulate the high-frequency three-component ground motion and obtain the high-frequency components of the three-component ground motion.

[0044] Furthermore, in step S404, during the high-frequency three-component ground motion simulation using the stochastic finite fault method based on the high-frequency horizontal power spectrum ratio and the high-frequency vertical / horizontal power spectrum ratio, the Fourier amplitude spectrum... The calculation formula is:

[0045]

[0046] In the formula, Represents the seismic moment of the sub-source. Indicates the source spectrum. This indicates the impact of propagation path attenuation. Describe the site effect. This represents the seismic motion type factor. When the output is displacement, velocity, and acceleration, the value of n is 0, 1, and 2, respectively. Indicates frequency; Indicates distance, Here are the Fourier amplitude spectra of the three components, where It can be H1, H2, or V. The adjustment factors for the three-component Fourier spectrum values ​​are as follows:

[0047]

[0048]

[0049]

[0050] in, Composed of 11 With 11 Interpolation to frequency It is confirmed above. Composed of 11 With 11 Interpolation to frequency It has been confirmed.

[0051] Further, step S500 includes the following sub-steps:

[0052] S501. For horizontal ground motion components in both low-frequency and high-frequency components, the attenuation relationship in BSSA2014 is used. The correlation coefficient is used to adjust the amplitude of ground motion intensity; the AS1997 attenuation relationship is used to adjust the amplitude of vertical ground motion intensity.

[0053] S502. The high-frequency components of the ground motion after the ground motion intensity amplitude is adjusted are low-pass filtered using a Butterworth filter; the low-frequency components of the ground motion after the ground motion intensity amplitude is adjusted are high-pass filtered using a Butterworth filter; wherein, the corner frequency of the high-pass / low-pass filtering is a mixed frequency.

[0054] S503. Mix the filtered high-frequency / low-frequency components of the ground motion to obtain broadband three-component ground motion simulation results.

[0055] The beneficial effects of this invention are as follows:

[0056] (1) The method of the present invention can obtain three-component ground motion in a wide frequency band (0.1Hz~25Hz) that can reflect the earthquake rupture process, three-dimensional inhomogeneous medium and local site topography.

[0057] (2) The method of the present invention only needs to simulate the spectral element method to 3~5Hz according to the specific engineering requirements, and can still obtain the wideband ground motion in the range of 0.1Hz~25Hz. If the traditional spectral element method is used completely, the calculation amount for simulating to 10Hz is more than 16 times that of the present method, and the calculation amount for simulating to 25Hz is more than 625 times that of the present method. Therefore, the present method greatly saves the cost required for simulation. Attached Figure Description

[0058] Figure 1 The flowchart of the ground motion simulation method based on the broadband three-component hybrid method provided by the present invention is shown.

[0059] Figure 2 A simplified schematic diagram of the training logic of the artificial neural network model provided by the present invention.

[0060] Figure 3 This is a schematic diagram illustrating the deviation between the simulated and measured values ​​of the seismic acceleration response spectrum provided by the present invention.

[0061] Figure 4 This is a schematic diagram comparing the acceleration, velocity, displacement, and acceleration response spectrum of a certain station with the measured records provided by the present invention.

[0062] Figure 5 This is a schematic diagram comparing the acceleration, velocity, displacement, and acceleration response spectrum of another station with the measured records provided by this invention. Detailed Implementation

[0063] 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.

[0064] This invention provides a seismic motion simulation method based on a broadband three-component hybrid method, such as... Figure 1 As shown, it includes the following steps:

[0065] S100. Collect regional strong earthquake records and then construct a dataset containing power spectral ratios for different frequency ranges.

[0066] S200. Using the dataset, construct and train an artificial neural network model to obtain a ground motion simulation model corresponding to different mixed frequencies;

[0067] S300, based on a broadband kinematic finite source model, uses the spectral element method to simulate low-frequency three-component ground motion;

[0068] S400. Input the power spectrum ratio obtained from the low-frequency three-component ground motion simulation into the ground motion simulation model of the corresponding mixed frequency to predict the power spectrum ratio of the high-frequency three components, and then use the stochastic finite fault method to perform high-frequency three-component ground motion simulation.

[0069] S500: The low-frequency component of the ground motion obtained from the low-frequency three-component ground motion simulation and the high-frequency component of the ground motion obtained from the high-frequency three-component ground motion simulation are superimposed to obtain the broadband three-component ground motion simulation result.

[0070] Step S100 of this embodiment of the invention includes the following sub-steps:

[0071] S101. Collect regional strong earthquake records containing effective three-component acceleration time history records;

[0072] Specifically, in each group of regional strong earthquake records, the effective frequency should be no less than 10 Hz, and the total number of strong earthquake records should be N.

[0073] S102. Calculate the source, path and site parameters corresponding to each group of regional strong earthquake records;

[0074] Specifically, the focal parameters include magnitude M and focal depth. Fault dip angle Fault properties (slip angle) ); Path parameters include fault projection distance Cosine of the angle between the fault strike and the station azimuth Site parameters are ;

[0075] S103. Based on the strong earthquake records of each group of regions, calculate the spectral ratio corresponding to different frequency ranges, and construct a dataset by combining the source, path and site parameters.

[0076] In step S103 of this embodiment of the invention, the power spectral ratio corresponding to different frequency ranges of each group of regional strong earthquake records is calculated, including:

[0077] S103-1. Calculate the Fourier amplitude spectrum of each group of regional strong earthquake records, and calculate the power spectrum of each group in several frequency ranges divided by frequency range.

[0078] Specifically, all ground motion records were resampled to a time step of 0.02 s by downsampling (time step in multiples of 0.02) or piecewise cubic Hermite interpolation (time step not in multiples of 0.02); the time step of all ground motions was increased to 8192 steps by zero padding; the Fourier amplitude spectra of all three-component ground motions were calculated, resulting in 4096 Fourier amplitude spectra for each ground motion, corresponding to the frequencies. The frequency range is 0~25 Hz.

[0079] Furthermore, to reduce the volatility of the variables, the Fourier spectrum values ​​were divided into 256 smaller frequency ranges, and the power spectrum within each frequency range was calculated to obtain the... The power spectrum of the d-th component within a frequency range Represented as:

[0080]

[0081] In the formula, and They represent the first The start and end frequencies within a frequency range, Indicates the first earthquake. Component frequency The corresponding Fourier amplitude spectrum, ;

[0082] in:

[0083] Since the earthquake motion that needs to be simulated in the project may be in any direction, this embodiment does not specifically distinguish the directions of H1 and H2, and takes the first and second recorded directions respectively;

[0084] S103-2, Set the starting frequency to be less than the mixing frequency. The power spectrum was divided into 11 training ranges on average, and the spectral ratio independent variable was calculated for each training range; wherein, the frequency width of each training range was the same.

[0085] Specifically, the initial frequency is less than The power spectrum was further divided into 11 training ranges, each with the same frequency width, resulting in 11 independent variables each in the horizontal and vertical / horizontal directions. If a certain training range... The start and end range is ~ , No. The spectral ratio independent variables for each training range include the horizontal power spectral ratio. and vertical / horizontal power spectrum ratio They are respectively:

[0086]

[0087]

[0088] In the formula, , They represent the first The starting and ending range values ​​of each training range, and satisfying , and They represent The starting and ending frequency values ​​for each training range. , and These represent the power spectra of the 1st, 2nd, and 3rd components, respectively.

[0089] S103-3, Set the starting frequency to be greater than or equal to the mixing frequency. The power spectrum is divided into 11 prediction ranges on average, and the spectral ratio dependent variable is calculated for each prediction range; the frequency width of the prediction range increases linearly with increasing frequency.

[0090] Specifically, the starting frequency is greater than or equal to the mixing frequency. The power spectrum is divided into 11 prediction ranges, resulting in 11 dependent variables each in the horizontal and vertical / horizontal directions. However, since the prediction effect worsens with higher prediction frequencies, and the horizontal difference in ground motion due to directional effects decreases with higher frequencies, the width of these 11 prediction ranges increases linearly with increasing frequency.

[0091] Among them, the closest The prediction range width is the same as the frequency width of the training range. If a certain prediction range... The start and end range is ~ Then the first The spectral ratio dependent variable for each prediction range includes the horizontal power spectral ratio. and vertical / horizontal power spectrum ratio They are respectively:

[0092]

[0093]

[0094] In the formula, , They represent the first The starting and ending range values ​​of each prediction range, and satisfying the following conditions: , and They represent The starting and ending frequency values ​​for each prediction range.

[0095] In this embodiment, based on the above-described division of the training range and prediction range, the center frequencies corresponding to the 11 training ranges and 11 prediction ranges are calculated. and They are respectively:

[0096]

[0097] .

[0098] In step S200 of this embodiment of the invention, in order to cover the frequency range usable in engineering, and considering the maximum simulation frequency of the spectral element method, this method selects 19 mixed frequencies in the range of 1~10Hz; for each mixed frequency Two artificial neural network models were independently constructed and trained, one based on the independent variable and the other on the low-frequency horizontal power spectrum ratio. Predicting the high-frequency horizontal power spectrum ratio dependent variable A horizontal ground motion simulation model, and independent variables based on the vertical / horizontal power spectrum ratio at low frequencies. Predicting the dependent variable of the high-frequency vertical / horizontal power spectrum ratio Vertical / horizontal ground motion simulation model.

[0099] In this embodiment, each artificial neural network model includes a feedforward neural network and a linear output unit connected in sequence, wherein the feedforward neural network includes 30 sigmoid hidden neurons (activation functions).

[0100] In this embodiment, the input to the artificial neural network model includes source, path, and site parameters, as well as the logarithm of the spectral ratio independent variable within different training ranges. The output of the artificial neural network model is the logarithm of the spectral ratio dependent variable within the corresponding prediction range. Specifically, for each... Each artificial neural network model is a multiple-input multiple-output model, with the number of input layer nodes equal to the number of input variables, including 7 independent variables S and 11 independent variables. or Each of the two models has 18 input layer nodes; the number of output nodes equals the number of dependent variables. or Each of the two models has 11 output layer nodes.

[0101] In this embodiment, the aforementioned artificial neural network model is trained using an error backpropagation algorithm, that is, in single-step teaching, the calculated error signal is backpropagated to all connected neurons; backpropagation requires supervised learning (the correct output corresponding to the input is known), and the weights are adjusted based on the error gradient descent method. Figure 2 The diagram shown is a simplified illustration of the ANN training logic.

[0102] Specifically, during the training process, for each mixing frequency During the training of the two artificial neural network models, the entire dataset (N sets of inputs and outputs) was randomly divided into three subsets: (1) a training set, used to calibrate the adjustable weights of the ANN; (2) a validation set, containing different patterns from the training set, used to monitor the accuracy of the model during training; and (3) a test set, which did not participate in training and validation, but was only used to evaluate the network's generalization ability to new data. An early stopping mechanism was introduced during training, and training was terminated when the error in the validation set began to rise. Specifically, the ratio of training / validation / test sets was set to 75% / 15% / 10%. Before selecting the final network for prediction, 41 different artificial neural network models with 10 to 50 hidden nodes were constructed. Each artificial neural network model was trained based on a subset randomly selected from 90% of the records, and the artificial neural network with the smallest mean square error on the remaining 10% of the dataset was finally selected as the optimal model.

[0103] Step S300 in this embodiment of the invention includes the following sub-steps:

[0104] S301. Construct a broadband kinematic finite-source model within the region that reflects the complex source rupture process;

[0105] S302. Collect the three-dimensional heterogeneous crustal medium and high-precision topography of the region, and construct a spectral unit model;

[0106] Specifically, the minimum inter-node element size in the simulation is determined by the mixing frequency. Determined according to the following empirical formula:

[0107]

[0108] In the formula, This represents the number of Gauss-Lobatto-Legendre integration points in the spectral unit, which is typically 5th order. For the reference dimensions of an 8-node hexahedral element with 5 Gauss-Lobatto-Legendre integration points, a maximum frequency of 1 Hz, a surface shear wave velocity of 1000 m / s, we can take... ; The number of segments for nodes on the element edge, for an 8-node hexahedral element. 27-node hexahedral surface element ;

[0109] S303. Substitute the broadband kinematic finite source model into the spectral unit model to simulate the seismic wave propagation process and obtain the three-component low-frequency components of ground motion.

[0110] Step S400 of this embodiment of the invention includes the following sub-steps:

[0111] S401. Low-frequency three-component ground motion amplitude spectrum obtained from low-frequency three-component ground motion simulation. and Calculate the corresponding horizontal power spectral density ratio and vertical / horizontal power spectral density ratio;

[0112] S402. Calculate the source, path, and site parameters based on the focal mechanism, spatial location, platform location, and site conditions.

[0113] S403. Input the horizontal power spectrum ratio and source, path and site parameters into the horizontal ground motion simulation model, and input the vertical / horizontal power spectrum ratio and source, path and site parameters into the vertical / horizontal ground motion simulation model to predict the high-frequency horizontal power spectrum ratio and the high-frequency vertical / horizontal power spectrum ratio respectively.

[0114] S404. Based on the high-frequency horizontal power spectrum ratio and the high-frequency vertical / horizontal power spectrum ratio, the random finite fault method is used to simulate the high-frequency three-component ground motion and obtain the high-frequency components of the three-component ground motion.

[0115] In this embodiment, the calculation parameters required for high-frequency three-component ground motion simulation using the stochastic finite fault method include stress drop, path duration, geometric spread function, quality factor, and high-frequency attenuation coefficient. Based on this, during the high-frequency three-component ground motion simulation using the stochastic finite fault method based on the high-frequency horizontal power spectrum ratio and the high-frequency vertical / horizontal power spectrum ratio, the Fourier amplitude spectrum... The calculation formula is:

[0116]

[0117] In the formula, Represents the seismic moment of the sub-source. Indicates the source spectrum. This indicates the impact of propagation path attenuation. Describe the site effect. This represents the seismic motion type factor. When the output is displacement, velocity, and acceleration, the value of n is 0, 1, and 2, respectively. Indicates frequency; Indicates distance, Here are the Fourier amplitude spectra of the three components, where It can be H1, H2, or V. The adjustment factors for the three-component Fourier spectrum values ​​are as follows:

[0118]

[0119]

[0120]

[0121] in, Composed of 11 With 11 Interpolation to frequency It is confirmed above. Composed of 11 With 11 Interpolation to frequency It has been confirmed.

[0122] In this embodiment, it should be noted that when using the random finite fault method for simulation, the same sub-source division scheme and slip model as the spectral element method are used. For the Fourier spectrum of each sub-source, the difference in Fourier spectrum values ​​caused by the three-component Fourier spectrum value adjustment coefficient formula is considered, and the ground motion time histories of each sub-source are superimposed to obtain the three-component high-frequency components of the ground motion.

[0123] Step S500 of this embodiment of the invention includes the following sub-steps:

[0124] S501. For horizontal ground motion components in both low-frequency and high-frequency components, the attenuation relationship in BSSA2014 is used. The correlation coefficient is used to adjust the amplitude of ground motion intensity; the AS1997 attenuation relationship is used to adjust the amplitude of vertical ground motion intensity.

[0125] S502. The high-frequency components of the ground motion after the ground motion intensity amplitude is adjusted are low-pass filtered using a Butterworth filter; the low-frequency components of the ground motion after the ground motion intensity amplitude is adjusted are high-pass filtered using a Butterworth filter; wherein, the corner frequency of the high-pass / low-pass filtering is a mixed frequency.

[0126] S503. Mix the filtered high-frequency / low-frequency components of the ground motion to obtain broadband three-component ground motion simulation results.

[0127] In an embodiment of the present invention, Figure 3 The paper presents the deviations between the simulated and measured values ​​of the ground motion acceleration response spectrum recorded at 34 stations when simulating the Northridge earthquake using the method of this invention. It can be seen that, compared with the traditional stochastic finite fault method, the broadband hybrid method proposed in this invention exhibits only small systematic errors between the simulated and observed values ​​of the three-component ground motion of Northridge across the entire ground motion broadband range (0.1Hz~25Hz).

[0128] Figure 4-5 The simulation of acceleration, velocity, displacement, and acceleration response spectrum (0.1Hz~25Hz) at two stations during the Northridge earthquake using this method is compared with the measured records. It can be seen that the broadband three-component hybrid method not only fits the acceleration response spectrum of the original records well, but also demonstrates good comparability between the original records and the simulation of acceleration, velocity, displacement amplitudes, and time-frequency non-stationary characteristics, illustrating the rationality of the ground motion simulation method proposed in this study.

[0129] Specific embodiments have been used to illustrate the principles and implementation methods of this invention. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

[0130] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. A seismic motion simulation method based on a broadband three-component hybrid method, characterized in that, Includes the following steps: S100. Collect regional strong earthquake records and then construct a dataset containing power spectral ratios for different frequency ranges. S200. Using the dataset, construct and train an artificial neural network model to obtain a ground motion simulation model corresponding to different mixed frequencies; S300, based on a broadband kinematic finite source model, uses the spectral element method to simulate low-frequency three-component ground motion; S400. Input the power spectrum ratio obtained from the low-frequency three-component ground motion simulation into the ground motion simulation model of the corresponding mixed frequency to predict the power spectrum ratio of the high-frequency three components, and then use the stochastic finite fault method to perform high-frequency three-component ground motion simulation. S500: The low-frequency component of the ground motion obtained from the low-frequency three-component ground motion simulation and the high-frequency component of the ground motion obtained from the high-frequency three-component ground motion simulation are superimposed to obtain the broadband three-component ground motion simulation result.

2. The seismic motion simulation method based on the broadband three-component hybrid method according to claim 1, characterized in that, Step S100 includes the following sub-steps: S101. Collect regional strong earthquake records containing effective three-component acceleration time history records; S102. Calculate the source, path and site parameters corresponding to each group of regional strong earthquake records; S103. Based on the strong earthquake records of each group of regions, calculate the spectral ratio corresponding to different frequency ranges, and construct a dataset by combining the source, path and site parameters.

3. The seismic motion simulation method based on the broadband three-component hybrid method according to claim 2, characterized in that, In step S103, the power spectral ratio corresponding to different frequency ranges of each group of regional strong earthquake records is calculated, including: S103-1. Calculate the Fourier amplitude spectrum of each group of regional strong earthquake records, and calculate the power spectrum of each group in several frequency ranges divided by frequency range. S103-2, Set the starting frequency to be less than the mixing frequency. The power spectrum is divided into 11 training ranges on average, and the spectral ratio independent variable is calculated for each training range; wherein, the frequency width of each training range is the same. S103-3, Set the starting frequency to be greater than or equal to the mixing frequency. The power spectrum is divided into 11 prediction ranges on average, and the spectral ratio dependent variable in each prediction range is calculated; wherein, the frequency width of the prediction range increases linearly with the frequency.

4. The seismic motion simulation method based on the broadband three-component hybrid method according to claim 3, characterized in that, In step S103-1, the first The power spectrum of the d-th component within a frequency range Represented as: In the formula, and They represent the first The start and end frequencies within a frequency range, Indicates the earthquake's first Each component frequency The corresponding Fourier amplitude spectrum, ; In step S103-2, the first The spectral ratio independent variables for each training range include the horizontal power spectral ratio. and vertical / horizontal power spectrum ratio They are respectively: In the formula, , They represent the first The starting and ending range values ​​of each training range, and satisfying the following conditions: , and They represent The starting and ending frequency values ​​for each training range. , and These represent the power spectra of the 1st, 2nd, and 3rd components, respectively. In step S103-3, the first The spectral ratio dependent variable for each prediction range includes the horizontal power spectral ratio. and vertical / horizontal power spectrum ratio They are respectively: In the formula, , They represent the first The starting and ending range values ​​of each prediction range, and satisfying the following conditions: , and They represent The starting and ending frequency values ​​for each prediction range.

5. The seismic motion simulation method based on the broadband three-component hybrid method according to claim 3, characterized in that, In step S200, for each mixed frequency, two artificial neural network models are independently constructed and trained: a horizontal ground motion simulation model that predicts the high-frequency horizontal power spectrum ratio dependent variable based on the independent variable and the low-frequency horizontal power spectrum ratio independent variable, and a vertical / horizontal ground motion simulation model that predicts the high-frequency vertical / horizontal power spectrum ratio dependent variable based on the independent variable and the low-frequency vertical / horizontal power spectrum ratio independent variable. Each of the artificial neural network models includes a feedforward neural network and a linear output unit connected in sequence; The inputs to the artificial neural network model include source, path, and site parameters, as well as the logarithm of the spectral ratio independent variable within different training ranges. The output of the artificial neural network model is the logarithm of the spectral ratio dependent variable within the corresponding prediction range.

6. The seismic motion simulation method based on the broadband three-component hybrid method according to claim 1, characterized in that, Step S300 includes the following sub-steps: S301. Construct a broadband kinematic finite-source model within the region that reflects the complex source rupture process; S302. Collect the three-dimensional heterogeneous crustal medium and high-precision topography of the region, and construct a spectral unit model; S303. Substitute the broadband kinematic finite source model into the spectral unit model to simulate the seismic wave propagation process and obtain the three-component low-frequency components of ground motion.

7. The seismic motion simulation method based on the broadband three-component hybrid method according to claim 5, characterized in that, Step S400 includes the following sub-steps: S401. Low-frequency three-component ground motion amplitude spectrum obtained from low-frequency three-component ground motion simulation. and Calculate the corresponding horizontal power spectral density ratio and vertical / horizontal power spectral density ratio; S402. Calculate the source, path, and site parameters based on the focal mechanism, spatial location, platform location, and site conditions. S403. Input the horizontal power spectrum ratio and source, path and site parameters into the horizontal ground motion simulation model, and input the vertical / horizontal power spectrum ratio and source, path and site parameters into the vertical / horizontal ground motion simulation model to predict the high-frequency horizontal power spectrum ratio and the high-frequency vertical / horizontal power spectrum ratio respectively. S404. Based on the high-frequency horizontal power spectrum ratio and the high-frequency vertical / horizontal power spectrum ratio, the random finite fault method is used to simulate the high-frequency three-component ground motion and obtain the high-frequency components of the three-component ground motion.

8. The seismic motion simulation method based on the broadband three-component hybrid method according to claim 7, characterized in that, In step S404, during the high-frequency three-component ground motion simulation using the stochastic finite fault method based on the high-frequency horizontal power spectrum ratio and the high-frequency vertical / horizontal power spectrum ratio, the Fourier amplitude spectrum... The calculation formula is: In the formula, Represents the seismic moment of the sub-source. Indicates the source spectrum. This indicates the impact of propagation path attenuation. Describe the site effect. This represents the seismic motion type factor. When the output is displacement, velocity, and acceleration, the value of n is 0, 1, and 2, respectively. Indicates frequency; Indicates distance, Here are the Fourier amplitude spectra of the three components, where It can be H1, H2, or V. The adjustment factors for the three-component Fourier spectrum values ​​are as follows: in, Composed of 11 With 11 Interpolation to frequency It is confirmed above. Composed of 11 With 11 Interpolation to frequency It has been confirmed.

9. The seismic motion simulation method based on the broadband three-component hybrid method according to claim 1, characterized in that, Step S500 includes the following sub-steps: S501. For horizontal ground motion components in both low-frequency and high-frequency components, the attenuation relationship in BSSA2014 is used. The correlation coefficient is used to adjust the amplitude of ground motion intensity; the AS1997 attenuation relationship is used to adjust the amplitude of vertical ground motion intensity. S502. The high-frequency components of the ground motion after the ground motion intensity amplitude is adjusted are low-pass filtered using a Butterworth filter; the low-frequency components of the ground motion after the ground motion intensity amplitude is adjusted are high-pass filtered using a Butterworth filter; wherein, the corner frequency of the high-pass / low-pass filtering is a mixed frequency. S503. Mix the filtered high-frequency / low-frequency components of the ground motion to obtain broadband three-component ground motion simulation results.

Images