Machine learning-based record-free site seismic oscillation simulation method and equipment

Abstract

The invention discloses a machine learning-based no-record site seismic oscillation simulation method and equipment, and the method comprises the steps: predicting a pseudo acceleration response spectrum of a target site through the actual measurement of seismic oscillation records of a plurality of stations near the target site under a seismic event; the pseudo acceleration response spectrum serves as a target spectrum, fitting is conducted on the actually-measured seismic oscillation acceleration time history of a station near a target site, and the seismic oscillation acceleration time history of the target site is obtained; transforming the seismic oscillation acceleration time history obtained by fitting to obtain an amplitude spectrum grey-scale map, inputting random noise into a trained generator of the generative adversarial network, and generating a pseudo amplitude spectrum image having similar characteristics with the amplitude spectrum grey-scale map; reconstructing a phase spectrum image corresponding to the pseudo amplitude spectrum image; and obtaining a one-dimensional seismic oscillation acceleration time history of the target site based on the pseudo amplitude spectrum image and the reconstructed phase spectrum image. According to the method, the seismic oscillation simulation data of the target site without any record can be generated.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of seismic motion simulation, in particular to a method and device for simulating seismic motion of unrecorded sites based on machine learning. BACKGROUND

[0002] Seismic monitoring instruments, as the core tools for recording seismic wave characteristics and analyzing seismic causes, are important technical supports for modern earthquake prevention and disaster reduction systems. According to the latest statistics, the total number of global seismic monitoring stations has exceeded ten thousand, mainly distributed in the Pacific Ring of Fire, the Eurasian Seismic Belt and other active seismic areas. However, the existing monitoring network has a significant uneven distribution problem: the station density is higher in the active seismic areas of developed countries, while there are still a large number of monitoring gaps in the ocean areas, remote mountainous areas and developing countries, which account for 71% of the earth's surface.

[0003] In the field of earthquake engineering, the time history of seismic acceleration is a key input parameter for evaluating the seismic response of structures. Although the global seismic monitoring capacity continues to improve, the existing station network still cannot effectively cover all seismic risk areas. For building structures lacking direct seismic observation records of the site, the post-earthquake performance evaluation and nonlinear seismic response analysis face the dilemma of missing basic data. Therefore, how to generate reliable seismic acceleration time history at sites without direct seismic observation records has become a difficult problem to be solved.

[0004] The current mainstream seismic time history simulation methods can be divided into two categories: the first category is the numerical simulation method based on physical mechanism: by establishing a refined model (such as fault model, velocity structure model, etc.) containing source characteristics, propagation path and site effects to simulate the seismic wave propagation process. Although this method can accurately reflect the wave characteristics, it has strict requirements for input data and high computational cost; the second category is the coherent function method based on statistical characteristics: using mutual power spectral density function and self-power spectral density function to construct the spatial relationship model of seismic motion. Although this method simplifies the calculation process, it still needs complete site parameters and propagation characteristic data, which has great limitations in practical application.

[0005] With the development of artificial intelligence technology, machine learning has shown unique advantages in ground motion simulation. Compared with traditional methods, data-driven machine learning technology can break away from the theoretical constraints of specific coherence functions or power spectra, and establish a prediction model by independently mining data characteristics. In addition, machine learning methods can efficiently capture the spatial variability of ground motion, maintain the physical rationality of simulation results, and significantly improve computational efficiency, supporting the rapid generation of large-scale ground motion time histories. However, existing methods still face challenges under complex geological conditions: when the distance between stations is large, the difference in soil layers is significant, which causes significant changes in seismic wave propagation characteristics, limiting the applicability of traditional methods. There are related patents for near-fault ground motion fitting, which can simulate artificial ground motion that fits the target response spectrum. However, under the premise of known measured ground motion records at some sites, there is still no reliable method to simulate ground motion records at any location under the same earthquake event.

[0006] The disclosed patent CN 114442153 A discloses a near-fault ground motion fitting method, which artificially simulates near-fault ground motion that fits the response spectrum, but requires prior knowledge of the response spectrum to fit. For sites without any records, ground motion simulation is still not possible. SUMMARY

[0007] To solve the problems in the prior art, the purpose of the present application is to provide a machine learning-based record-free site ground motion simulation method and device that simulates target site ground motion based on nearby station seismic information.

[0008] To achieve the above-mentioned purpose of the application, the present application provides the following technical solutions:

[0009] A machine learning-based record-free site ground motion simulation method, comprising the following steps:

[0010] (1) Predict the pseudo-acceleration response spectrum of the target site by measuring the ground motion records of several stations near the target site under the earthquake event;

[0011] (2) Fit the measured ground motion acceleration time history of the stations near the target site with the pseudo-acceleration response spectrum as the target spectrum to obtain the ground motion acceleration time history of the target site;

[0012] (3) transforming the obtained seismic acceleration time history to obtain an amplitude spectrum gray image, and inputting the amplitude spectrum gray image to a generator of a trained generative adversarial network as random noise to generate a pseudo amplitude spectrum image having similar characteristics with the amplitude spectrum gray image; wherein the generative adversarial network comprises the generator and a discriminator, in the training, the generator is used to generate a new pseudo amplitude spectrum image based on random noise, the discriminator is used to judge the similarity between the pseudo amplitude spectrum image generated by the generator and the amplitude spectrum gray image, and according to the similarity, the pseudo amplitude spectrum image generated by the generator is made close to the amplitude spectrum gray image through back propagation;

[0013] (4) processing the generated pseudo amplitude spectrum image to reconstruct a phase spectrum image corresponding to the pseudo amplitude spectrum image;

[0014] (5) combining the generated pseudo amplitude spectrum image and the reconstructed phase spectrum image to perform inverse short-time Fourier transform to obtain a one-dimensional seismic acceleration time history of the target site.

[0015] Further, the step (1) specifically comprises:

[0016] (1.1) obtaining station measured seismic information from a plurality of station measured seismic records near the target site, including normalized station position coordinates, the logarithm of the average shear wave velocity of the top 30m soil layer of the station site and the corresponding pseudo acceleration response spectrum;

[0017] (1.2) establishing a Gaussian process regression model as:

[0018]

[0019] In the formula, denotes the pseudo acceleration response spectrum, denotes the Gaussian process function, denotes the covariance function of the Gaussian process, which is obtained by training, denotes the station measured seismic information, denotes the normalized station position coordinates x, y, z axis coordinates, denotes the logarithm of the average shear wave velocity of the top 30m soil layer of the station site ; denotes the seismic information of the target site, denotes the normalized target site position coordinates x, y, z axis coordinates, denotes the logarithm of the average shear wave velocity of the top 30m soil layer of the target site ;

[0020] (1.3) inputting the station measured seismic information into the Gaussian process regression model for training to obtain the trained Gaussian process regression model;

[0021] (1.4) inputting the ground motion information of the target site into the Gaussian process regression model to obtain the pseudo-acceleration response spectrum of the target site.

[0022] Further, the step (2) specifically comprises:

[0023] (2.1) obtaining a station measured ground motion time history from a plurality of station measured ground motion records near the target site;

[0024] (2.2) superimposing a wavelet function on the station measured ground motion time history to obtain a new ground motion time history, so that the pseudo-acceleration response spectrum corresponding to the new ground motion time history matches the pseudo-acceleration response spectrum of the target site, and the new ground motion time history is the ground motion acceleration time history of the target site.

[0025] Further, the calculation process of the amplitude spectrum gray image is:

[0026] Performing short-time Fourier transform on the fitted ground motion acceleration time history, adjusting the window length and overlap length to obtain an amplitude spectrum image of a preset size, and performing normalization processing to obtain an amplitude spectrum gray image.

[0027] Further, the generator comprises an input layer, a full connection layer, two up-sampling layers, four convolution layers connected in sequence, a batch normalization operation and a LeakyReLU activation function are arranged between the third convolution layer and the fourth convolution layer, and a Tanh activation function is arranged after the fourth convolution layer.

[0028] Further, the discriminator comprises an input layer, four convolution layers, a flattening layer and a full connection layer connected in sequence, and a batch normalization, a LeakyReLU activation function and a random inactivation processing are arranged after each convolution layer.

[0029] Further, the step (4) specifically comprises:

[0030] (4.1) performing inverse normalization processing on the generated pseudo-amplitude spectrum image to obtain an amplitude spectrum color map;

[0031] (4.2) using the Griffin-Lim algorithm to reconstruct a phase spectrum image corresponding to the amplitude spectrum color map.

[0032] A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the above method.

[0033] A computer readable storage medium having a computer program / instruction stored thereon, wherein the computer program / instruction, when executed by a processor, implements the above method.

[0034] A computer program product includes a computer program / instructions that, when executed by a processor, implement the above-described method.

[0035] Compared with the prior art, the beneficial effects of this invention are:

[0036] (1) Based on the hybrid machine learning framework of Gaussian process regression and deep convolutional generative adversarial network, this application conducts innovative research on the problem of ground motion simulation in unrecorded sites and proposes a two-stage simulation strategy of "response spectrum guidance-time history generation". While ensuring the physical rationality of ground motion, it significantly improves the computational efficiency and is applicable to ground motion prediction under any geological conditions. While improving the simulation accuracy, it reduces the dependence on prior data and enhances the adaptability to complex site conditions, providing a brand-new technical path for data-driven simulation in the field of earthquake engineering.

[0037] (2) This application constructs a seismic motion generation system based on the synergistic optimization of spectral features and temporal features, breaking through the technical bottlenecks of traditional methods in terms of physical model dependence and statistical constraints. Single methods are limited by their theoretical basis and application scope, making it difficult to take into account the physical characteristics and spatial variability of seismic motions, and they also suffer from problems such as power spectrum distortion and coherence function deviation.

[0038] (3) This application can not only efficiently generate seismic motion time histories that conform to the characteristics of specified earthquake events, but also maintain the spatial correlation and physical consistency of seismic motion, providing reliable data support for seismic analysis of building structures, seismic risk assessment and optimization of seismic design of engineering. Attached Figure Description

[0039] Figure 1 A flowchart illustrating the machine learning-based ground motion simulation method for unrecorded sites provided by this invention.

[0040] Figure 2 The geographical distribution of the stations;

[0041] Figure 3 The pseudo-acceleration response spectrum predicted by the Gaussian process regression model in the east-west direction for station S1;

[0042] Figure 4 To compare the fitted ground motion acceleration time history with the fitted ground motion response spectrum and the predicted response spectrum, (a) shows the ground motion acceleration time history fitted with the superimposed wavelet function, and (b) shows the ground motion response spectrum fitted with the superimposed wavelet function and the predicted response spectrum.

[0043] Figure 5 The images show the amplitude spectrum grayscale images before and after normalization. (a) is the amplitude spectrum image before normalization, and (b) is the amplitude spectrum grayscale image after normalization.

[0044] Figure 6 The generator architecture provided by the present application is shown in the figure;

[0045] Figure 7 The discriminator provided by the present application will be composed of the figure;

[0046] Figure 8 For the iterative error of the Griffin-Lim algorithm and the reconstructed ground motion time history, (a) is the iterative error of the Griffin-Lim algorithm, (b) is the reconstructed ground motion time history under the optimal number of iterations of the Griffin-Lim algorithm;

[0047] Figure 9 For the S1 station result comparison, (a) is the comparison of the pseudo-acceleration response spectrum of the measured ground motion in the east-west direction of the S1 station and 150 artificial simulated ground motions, (b) is the comparison of the peak acceleration and duration of the measured ground motion in the east-west direction of the S1 station and 150 artificial simulated ground motions. DETAILED DESCRIPTION

[0048] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application.

[0049] Embodiment one

[0050] The present application provides a machine learning-based record-free site ground motion simulation method, as shown in the figure, which includes the following steps: Figure 1

[0051] (1) By the measured ground motion records of several stations near the target site under a seismic event, the pseudo-acceleration response spectrum of the target site is predicted.

[0052] This step specifically includes:

[0053] (1.1) Obtain the station measured ground motion information from the measured ground motion records of several stations near the target site, including the normalized station position coordinates, the average shear wave velocity V s30 of the top 30m soil layer of the station site and the corresponding pseudo-acceleration response spectrum.

[0054] In the experiment, the Mw 7.1 Ridgecrest earthquake in California, USA in 2019 is taken as the target seismic event, and the measured ground motion data is obtained from the 149 community seismic network records of the California Institute of Technology. Among them, the data of 148 stations is used for the training set, and the data of the remaining 1 station is used for the test set. The geographical position distribution of the 149 stations is shown in the figure Figure 2 ​The input information includes the position coordinates of the stations and the average shear wave velocity of the top 30m soil layer of the station site . Considering the significant differences in soil characteristics of different station sites, to reduce the dispersion degree in the training process, logarithmic transformation is performed. Thus, the measured ground motion information , represents the normalized station position coordinates x, y, z axis coordinates, = To avoid the adverse effects of different orders of magnitude data on the training process of the Gaussian process regression model, the data in the four dimensions are normalized before training to ensure that the data distribution in each dimension has a mean of zero and a standard deviation of one. The pseudo-acceleration response spectrum is the pseudo-acceleration response spectrum with a damping ratio of 5%, containing 85 specific periods between 0.1 seconds and 20 seconds. The pseudo-spectrum acceleration values differ significantly at long and short periods, and such different orders of magnitude data may cause large bias and instability in the training process of the machine learning model. Therefore, the pseudo-spectrum acceleration values of the measured ground motion of different stations are logarithmically transformed to reduce the numerical difference between different periods. Thus, the training data set is formed as wherein represents the input four-dimensional site information of the station and the output pseudo-acceleration response spectrum observation value of the measured ground motion of the station at 85 specified periods. The test data set consists of the site information and pseudo-spectrum acceleration data of stations S1 and S2, i.e. wherein .

[0055] (1.2) The Gaussian process regression model is established as:

[0056]

[0057] wherein, y= represents the pseudo-acceleration response spectrum, represents the Gaussian process function, represents the covariance function of the Gaussian process, which is obtained by training, represents the measured ground motion information of the station, represents the normalized station position coordinates x, y, z axis coordinates, represents the logarithm of the measured ground motion information of the station . represents the ground motion information of the target site, represents the normalized target site position coordinates x, y, z axis coordinates, a pseudo-acceleration response spectrum of a target site a logarithm of a target site; wherein,

[0058]

[0059] a pseudo-acceleration response spectrum of a target site, represents , represents a prior mean vector of a target site, represents a prior mean vector of a target site, represents a covariance matrix of a target site, and represents and a covariance matrix between and, represents a covariance matrix of a target site.

[0060] (1.3) inputting a number of station measured seismic information into a Gaussian process regression model to train, obtaining a trained Gaussian process regression model;

[0061] inputting a training data set into a Gaussian process regression model to train, can obtain:

[0062]

[0063]

[0064] represents a variance of a Gaussian distribution of a predicted value .

[0065] (1.4) inputting seismic information of a target site into a Gaussian process regression model, obtaining a pseudo-acceleration response spectrum of the target site;

[0066] Specifically, the pseudo-acceleration response spectrum of the target site can be represented as:

[0067]

[0068] wherein, represents an input of a training set; represents seismic information of a target site; under the condition that an input and an output of a training data set are known, can be inputted, to obtain a pseudo-acceleration response spectrum of the target site.

[0069] In this example, the predicted response spectrum of the S1 station east-west direction Gaussian process regression model and the predicted response spectrum of the measured ground motion of the station are shown in Figure 3 .

[0070] (2) Using the pseudo-acceleration response spectrum as the target spectrum, the measured ground motion acceleration time history of the stations near the target site is fitted to obtain the ground motion acceleration time history of the target site.

[0071] Step (2) specifically includes:

[0072] (2.1) Obtain the station measured ground motion time history from the measured ground motion records of several stations near the target site;

[0073] (2.2) Superimpose a wavelet function on the station measured ground motion time history to obtain a new ground motion time history, so that the pseudo-acceleration response spectrum corresponding to the new ground motion time history matches the pseudo-acceleration response spectrum of the target site. The new ground motion time history is the ground motion acceleration time history of the target site.

[0074] The form of the wavelet function used is as follows:

[0075]

[0076] wherein is the wavelet function at frequency and time t; is the time when the maximum response of the single-degree-of-freedom system at the jth frequency occurs in the initial ground motion time history; is the time difference between the time when the maximum value of the wavelet function occurs and the time when the maximum value of the response function occurs; n is an integer, representing the number of cycles of the sine function, used to adjust the duration of the wavelet function; is a parameter related to the frequency and the number of cycles, which needs to be determined to ensure that the velocity and displacement time histories of the wavelet function do not have zero-line drift.

[0077] In this example, the measured ground motion acceleration time histories of the 10 surrounding stations are adjusted so that the acceleration response spectrum of the adjusted ground motion time history matches the pseudo-acceleration response spectrum of the target site. The fitted ground motion acceleration time history is shown in Figure 4 (a), where the duration of the ground motion time history is 200s, and the time interval is 0.02 seconds, i.e., each ground motion time history consists of 10001 data points. The comparison of the response spectrum of the fitted ground motion acceleration time history and the target response spectrum is shown in Figure 4 (b).

[0078] ​The step can fit the same number of seismic motion acceleration time histories as the seismic motion records in step (1), for example, N seismic motions are used in step (1), and then N seismic motion time histories can be fitted in step (2). Due to the small number of measured seismic motions, only a small number of target site seismic motions can be obtained through step (2). These seismic motion data are used as a training set, and a large number of new artificial seismic motion data are generated through the subsequent generative adversarial network model.

[0079] (3) The fitted seismic motion acceleration time history is transformed to obtain an amplitude spectrum gray image, and the image is input into the trained generator of the generative adversarial network as random noise to generate a pseudo-amplitude spectrum image with similar characteristics to the amplitude spectrum gray image.

[0080] The calculation process of the amplitude spectrum gray image is as follows: the fitted seismic motion acceleration time history is subjected to short-time Fourier transform, and the time and frequency of the original seismic motion time history are not truncated in order to completely retain the time and frequency domain information of the seismic motion. Considering the convolution operation of the generator and the discriminator, the window length is set to 198 data points, and the overlap length is set to 96 data points. The amplitude spectrum gray image obtained by short-time Fourier transform is a two-dimensional square image with a size of 100x100 pixels, as shown in Figure 5 (a). Due to the large difference in peak acceleration of different seismic motion time histories, the amplitude spectrum after short-time Fourier transform also presents a large difference in numerical value. In order to reduce the instability that may be caused by the discrete nature of the input data during the model training process, the two-dimensional time-frequency domain amplitude spectrum corresponding to 10 seismic motions in the training set is normalized, and the normalized amplitude spectrum is used as the input of the model in the form of a single-channel gray image. Figure 5 (b) shows the seismic motion amplitude spectrum gray image used for training the generative adversarial network model.

[0081] The generative adversarial network includes a generator and a discriminator. During training, the generator is used to generate new pseudo-amplitude spectrum images based on random noise, the discriminator is used to determine the similarity between the pseudo-amplitude spectrum images generated by the generator and the amplitude spectrum gray image, and the pseudo-amplitude spectrum images generated by the generator are made close to the amplitude spectrum gray image according to the similarity.

[0082] The deep convolutional network model structure of the generator is as follows: Figure 6The input of the generator is a one-dimensional random noise variable with a length of 100. After the input, the variable is first expanded to an 80000x1 matrix through a fully connected layer, and then the matrix is adjusted to a 25x25x128 three-dimensional tensor. Then, the size of the tensor is expanded to 50x50x128 through the first upsampling operation. Subsequently, the size of the tensor is maintained at 50x50x128 through two convolution operations. Then, the size of the tensor is expanded to 100x100x128 through the second upsampling operation. Next, the size of the tensor is adjusted to 100x100x64 through the third convolution. Finally, the output of the fourth convolution is a 100x100 matrix, and the data of the matrix is plotted as a grayscale image, which is the final output of the generator, i.e., a pseudo-amplitude spectrum image. Batch normalization is performed on the output of the first three convolution operations to speed up the training of the neural network and improve stability, and the momentum parameter of the batch normalization is set to 0.8. Then, the batch-normalized result is activated, and the activation function is LeakyReLU function. After the fourth convolution operation, the output is directly activated, and the activation function is Tanh function.

[0083] The structure of the deep convolutional neural network model of the discriminator is shown in FIG. 3. Figure 7 The input of the discriminator is a pseudo-amplitude spectrum image with a size of 100x100. After the first convolution layer, a three-dimensional tensor with a size of 50x50x32 is generated. Then, the size of the tensor is reduced to 25x25x64 through the second convolution operation. Subsequently, the third convolution converts the tensor to a three-dimensional tensor with a size of 13x13x128, and the fourth convolution obtains a tensor with a size of 13x13x256. Next, the three-dimensional tensor is flattened into a vector with a length of 42364, and finally a scalar is output through a fully connected layer as the score of the pseudo-amplitude spectrum image. Batch normalization, activation and random inactivation are performed in turn after each convolution operation. In the batch normalization process, the momentum parameter is set the same as that of the generator. The activation function of the non-linear transformation is LeakyReLU function. The parameter of the random inactivation is set to 0.25, i.e., a part of the neuron output is randomly discarded with a probability of 0.25 to reduce the risk of model overfitting. When the flattened vector is processed through the fully connected layer to obtain the final output result, the activation function is Sigmoid function to ensure that the output of the discriminator is between 0 and 1. The numerical size of the output result is used to determine whether the amplitude spectrum image input into the discriminator is real or generated by the generator. The closer the score is to 1, the closer the discriminator thinks the image is to the real image. In addition, the discriminator needs to learn the features of the amplitude spectrum grayscale image first to ensure that the score of the output amplitude spectrum grayscale image is 1. Then, the pseudo-spectrum image generated by the generator is input into the discriminator to give the score of the pseudo-spectrum image. According to the score, the generator and the discriminator continuously optimize the parameters of the convolutional neural network to finally reach a dynamic balance.

[0084] This step can generate a large number of pseudo-amplitude spectrum images by the generator, providing a data basis for the subsequent.

[0085] (4) The generated pseudo-amplitude spectrum image is processed to reconstruct the phase spectrum image corresponding to the pseudo-amplitude spectrum image.

[0086] The present application adopts Griffin-Lim algorithm for reconstruction. Before reconstruction, the generated amplitude spectrum needs to be inversely normalized first to obtain the amplitude spectrum color image. Since the random generation of the initial phase spectrum may lead to slow convergence speed of the Griffin-Lim algorithm. Considering that the phase spectrum corresponding to the trained pseudo-amplitude spectrum image is known, it can be used as the initial phase spectrum, thereby improving the efficiency of the Griffin-Lim algorithm for reconstructing the seismic time history. The number of iterations is an important parameter of the Griffin-Lim algorithm. Taking the first seismic acceleration time history in the training set as an example, the optimal number of iterations is selected by calculating the average absolute error between the reconstructed signal and the original signal of the amplitude spectrum under different numbers of iterations of the Griffin-Lim algorithm. The iteration error of the Griffin-Lim algorithm is shown in (a) of Figure 8 When the iteration is performed to 990 times, the error of the reconstructed signal is the smallest, and at this time the reconstruction effect of the signal is the best. The comparison between the reconstructed acceleration time history and the original time history is shown in (b) of Figure 8 .

[0087] (5) The inverse short-time Fourier transform is performed on the generated pseudo-amplitude spectrum image and the reconstructed phase spectrum image to obtain the one-dimensional seismic acceleration time history of the target site.

[0088] In this example, 150 amplitude spectrum images generated by the generator and 150 phase spectrum images reconstructed by the Griffin-Lim algorithm are combined for the S1 station, and 150 artificial seismic acceleration time histories are generated by inverse short-time Fourier transform. The pseudo-acceleration response spectrum of the artificial simulated seismic motion is shown in (a) of Figure 9 , and the duration and peak acceleration distribution of the simulated seismic motion are shown in (b) of Figure 9 . It can be seen from (a) of Figure 9 and (b) of Figure 9 that the method proposed in the present application realizes the artificial simulation of seismic motion for sites without seismic instrument recording. The characteristics of the simulated seismic motion in the frequency domain are close to the measured values, and the peak acceleration and duration of the simulated seismic motion fluctuate around the corresponding values of the measured seismic motion.

[0089] Example Two

[0090] The embodiment of the present application provides a computer device, and the embodiment of the present application provides services for the implementation of the method of the above embodiment one of the present application. The device can comprise: a memory storing a computer executable program; a processor coupled with the memory; the processor invokes the computer executable program stored in the memory, and is used for executing the steps in the method described in the embodiment one.

[0091] The memory can comprise a computer system readable medium in the form of volatile memory, such as a random access memory (RAM) and / or cache memory. The device can further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the memory can be used for reading from and writing to a non-removable, non-volatile magnetic media, commonly called a "hard drive". Program / utility, having a set of program modules, can be stored in, for example, memory, such as operating system, one or more application programs, other program modules, and program data, each or some combination thereof, can include implementation of a network environment. The computer executable program of the program modules typically performs the functions and / or methodologies of the embodiments described in the present application.

[0092] The processor performs various function applications and data processing by running the program stored in the memory, for example, the method provided in the embodiment one of the present application.

[0093] The code of the computer executable program can be written in one or more programming languages or combinations thereof, and the programming languages include object-oriented programming languages such as Java, Smalltalk, C++, and conventional procedural programming languages such as "C" language or similar programming languages.

[0094] Embodiment three

[0095] The embodiment of the present application provides a storage medium containing a computer executable program, and the computer executable program is used for executing the method of the embodiment one when executed by a computer processor.

[0096] The storage medium of the embodiments of the present application can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may, for example, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or apparatus, or any combination thereof. More specific examples (non-exhaustive list) of the computer-readable storage medium include an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, the computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus or device.

[0097] Of course, the storage medium provided by the embodiments of the present application includes computer executable programs, and the computer executable programs are not limited to the method operations described above, but can also perform related operations in the method provided by any embodiment of the present application.

[0098] Embodiment four

[0099] The embodiments of the present application also provide a computer program product, such as an app on a mobile phone, a tablet, an installation program on a computer, etc. The product includes computer programs / instructions, which, when executed by a processor, implement the method described in embodiment one. The code of the computer executable program for performing the operations of the present application can be written in one or more programming languages or combinations thereof, including object-oriented programming languages such as Java, Smalltalk, C++, and conventional procedural programming languages such as "C" language or similar programming languages. The program code can be executed entirely on a user computer, partially on a user computer, as an independent software package, partially on a user computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, through the Internet using an Internet service provider).

[0100] It should be understood that the above embodiments and descriptions described in the specification are only the principles, main features and advantages of the present application, and various changes and improvements can be made to the present application without departing from the spirit and scope of the present application, and these changes and improvements all fall within the scope of protection of the present application.

Claims

1. A method for simulating ground motion in a recordless site based on machine learning, characterized in that, Includes the following steps: (1) The pseudo-acceleration response spectrum of the target site is predicted by using the actual ground motion records of several stations near the target site under the earthquake event; (2) Using the pseudo-acceleration response spectrum as the target spectrum, the measured ground motion acceleration time history of the stations near the target site is fitted to obtain the ground motion acceleration time history of the target site. (3) Transform the fitted earthquake acceleration time history to obtain an amplitude spectrum grayscale image, and input the random noise into the generator of the trained generative adversarial network to generate a pseudo amplitude spectrum image with similar features to the amplitude spectrum grayscale image; wherein, the generative adversarial network includes a generator and a discriminator. During training, the generator is used to generate a new pseudo amplitude spectrum image based on random noise, and the discriminator is used to determine the similarity between the pseudo amplitude spectrum image generated by the generator and the amplitude spectrum grayscale image, and backpropagate according to the similarity to make the pseudo amplitude spectrum image generated by the generator close to the amplitude spectrum grayscale image; (4) Process the generated pseudo amplitude spectrum image to reconstruct the phase spectrum image corresponding to the pseudo amplitude spectrum image; (5) Combine the generated pseudo amplitude spectrum image and the reconstructed phase spectrum image to perform inverse short-time Fourier transform to obtain the one-dimensional ground motion acceleration time history of the target site.

2. The method according to claim 1, characterized in that: Step (1) specifically includes: (1.1) Obtain the measured ground motion information from several stations near the target site, including the normalized station location coordinates and the average shear wave velocity of the top 30m soil layer of the site where the station is located. The logarithm and the corresponding pseudo-acceleration response spectrum; (1.2) The Gaussian process regression model is established as follows: ， In the formula, This represents the pseudo-acceleration response spectrum. Represents the Gaussian process function. The covariance function representing the Gaussian process is obtained through training. This indicates the actual ground motion information measured by the station. This represents the normalized x, y, and z coordinates of the station location. Indicates station The logarithm of ; This indicates the seismic motion information of the target site. This represents the normalized x, y, and z coordinates of the target site location. Indicates the target site The logarithm of ; (1.3) Input the measured ground motion information from several stations into the Gaussian process regression model for training to obtain the trained Gaussian process regression model; (1.4) Input the ground motion information of the target site into the Gaussian process regression model to obtain the pseudo-acceleration response spectrum of the target site.

3. The method according to claim 1, characterized in that: Step (2) specifically includes: (2.1) Obtain the time history of measured ground motion from several stations near the target site; (2.2) A wavelet function is superimposed on the measured ground motion time history at the station to obtain a new ground motion time history, so that the pseudo-acceleration response spectrum corresponding to the new ground motion time history matches the pseudo-acceleration response spectrum of the target site. The new ground motion time history is the ground motion acceleration time history of the target site.

4. The method according to claim 1, characterized in that: In step (3), the calculation process of the amplitude spectrum grayscale image is as follows: A short-time Fourier transform is performed on the fitted seismic acceleration time history. By adjusting the window length and overlap length, an amplitude spectrum image of a preset size is obtained, and then normalized to obtain a grayscale image of the amplitude spectrum.

5. The method according to claim 1, characterized in that: The generator includes an input layer, a fully connected layer, two upsampling layers, and four convolutional layers connected in sequence. A batch normalization operation and a LeakyReLU activation function are provided between the third and fourth convolutional layers, and a Tanh activation function is provided after the fourth convolutional layer.

6. The method according to claim 1, characterized in that, The discrimination process includes one input layer, four convolutional layers, one flattening layer, and one fully connected layer connected in sequence. Each convolutional layer is followed by batch normalization, LeakyReLU activation function, and random deactivation.

7. The method according to claim 1, characterized in that, Step (4) specifically includes: (4.1) Perform inverse normalization on the generated pseudo amplitude spectrum image to obtain the amplitude spectrum color image; (4.2) The phase spectrum image corresponding to the amplitude spectrum color image is reconstructed using the Griffin-Lim algorithm.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: The processor executes the computer program to implement the method as described in any one of claims 1-7.

9. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that, The computer program / instructions, when executed by a processor, implement the method of any one of claims 1-7.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the method of any one of claims 1-7.

Images