Regional prediction model based on near-fault seismic oscillation input energy and processing equipment

Abstract

The invention relates to the field of seismic design based on energy, in particular to a regional prediction model and processing equipment based on near-fault seismic oscillation input energy, and the processing method of the model comprises the steps: obtaining acceleration time history data of a target region, calculating regional seismic oscillation input energy parameters based on the acceleration time history data, the regional seismic oscillation input energy parameters comprise elastic input energy and elastic-plastic input energy of seismic oscillation, establishing characteristic parameters, combining the characteristic parameters with the regional seismic oscillation input energy parameters to establish a characteristic database, and training a training set in the characteristic database. According to the method, a near-fault seismic oscillation input energy region prediction model which has universality for various scenes, is convenient to configure input data and is relatively high in processing efficiency can be obtained, and high-quality prediction requirements on region elasticity and elastic-plastic input energy under actual conditions can be met; and accurate and effective data support is provided for activity development of aseismic design engineering based on the behavior.

Description

Technical Field

[0001] The present invention relates to the field of energy-based seismic design, and in particular to a regional prediction model and processing module based on near-fault ground motion input energy, a computer-readable storage medium, and a processing device. Background Art

[0002] Earthquakes are caused by the movement of plates. The interaction of plates includes collision, separation and sliding, which leads to the accumulation of stress inside the earth's crust. When these stresses exceed the bearing capacity of rocks, rocks will break and shift. This may cause buildings to collapse, damage roads, bridges and other infrastructure, and may also trigger secondary disasters such as tsunamis and mudslides. Therefore, earthquakes are often accompanied by huge casualties and economic losses.

[0003] In the past, earthquake fortification mostly selected earthquake records based on earthquake amplitude. However, earthquake is a complex transient process, and a single amplitude index cannot fully represent all the information of earthquake motion. Therefore, it is very necessary to select a reasonable earthquake intensity index to represent earthquake motion information for structural earthquake resistance design. The response of a structure under earthquake motion is a continuous energy input and dissipation process. The earthquake input energy takes into account the characteristics of earthquake motion such as spectrum, duration, amplitude and energy dissipation, and can well predict the potential destructive capacity of earthquakes to structures. Therefore, in structural earthquake resistance design and earthquake risk assessment, input energy can be used as one of the most reasonable damage potential parameters. For example, the "General Principles for Seismic Design of Building Structures Based on Performance" also takes input energy as an important indicator, proving that the input energy parameter is widely used in seismic design based on performance.

[0004] However, the input energy prediction methods currently established mainly focus on regression analysis of empirical data. Such methods generally have the disadvantages of strong subjectivity, high cognitive uncertainty, and the entire research process is too complicated and has large errors, which is not conducive to the implementation of related research. Summary of the invention

[0005] The purpose of this application is to provide a regional prediction model and processing equipment based on near-fault ground motion input energy, aiming to solve the problems existing in the prior art.

[0006] The embodiment of the present application provides a regional prediction model based on near-fault ground motion input energy, and the processing method of the model includes the following steps:

[0007] Step S101, obtaining acceleration time history data of a target area;

[0008] Step S102, calculating regional earthquake input energy parameters based on acceleration time history data, where the regional earthquake input energy parameters include elastic input energy and elastoplastic input energy of earthquakes;

[0009] The calculation formula of elastic input energy is:

[0010]

[0011] Among them, E Ia is the absolute input energy, E Ir is the relative input energy, m is the mass of the single degree of freedom system, x t is the absolute displacement of the single degree of freedom system, x g is the displacement of ground motion, x is the relative displacement of the single-degree-of-freedom system;

[0012] In order to eliminate the influence of the mass of the single-degree-of-freedom system, the absolute and relative input energy are replaced by equivalent velocity, and the formula is:

[0013]

[0014] Among them, V Ea is the equivalent speed of absolute input energy, V Ir is the equivalent speed relative to the input energy;

[0015] The elastic-plastic input energy is calculated based on the acceleration time history data through the constitutive model, and the elastic-plastic input energy is replaced by the equivalent input;

[0016] Step S103, taking the moment magnitude, fault distance, fault type and site average shear wave velocity of the target area data as characteristic parameters;

[0017] Step S104, combining the characteristic parameters obtained in step S103 with the regional earthquake input energy parameters obtained in step S102 to establish a characteristic database;

[0018] Step S105, dividing the relevant feature combination data in the feature database obtained in step S104 into a training set and a test set;

[0019] Step S106, determining a machine learning algorithm, and using moment magnitude, fault distance, fault type, and site shear wave velocity as input parameters, and regional earthquake input energy parameters as output parameters, and using the training set data as a basis for model training to predict the corresponding regional earthquake input energy parameters;

[0020] Step S107, verifying the training model based on the test set;

[0021] Step S108, using moment magnitude, fault distance and site shear wave velocity as input parameters, estimating elastic and elastoplastic input energy parameters of a specific earthquake scenario.

[0022] Furthermore, the machine learning algorithm in step S106 is a particle swarm optimization support vector machine algorithm, and its quantization formula is:

[0023]

[0024] Among them, α i is the Lagrange multiplier vector value; α i * and b are the solutions to the dual optimization problem. K(X i ,X) represents the kernel function that must satisfy the Mercer condition.

[0025] Furthermore, the kernel function selects a radial basis function:

[0026] K(X i ,X)=exp(-γ||XX i || 2 );

[0027] Where γ is the kernel parameter; X is the input variable; X i is the i-th sample variable.

[0028] Furthermore, when optimizing the parameters in the model, the following formula is used:

[0029] Vi(k+1)=wVi(k)+c 1 ×rand(·)×[pbesti(k)-Xi(k)]+c 2 ×rand(·)×[gbesti(k)-X i (k)];

[0030] X i (k+1)=X i (k)+V i (k);

[0031] Where w is the inertia weight factor, k is the number of steps of particle movement, c 1 and c 2 is the learning factor, pbest i is the best individual particle, gbest i is the global best particle, X i is the current position of the particle, V i is the velocity of the particle.

[0032] A processing module loaded with the regional prediction model based on near-fault ground motion input energy, the device comprises:

[0033] A first acquisition unit, used to acquire acceleration time history data of a target area;

[0034] A second acquisition unit is used to acquire elastic and elastoplastic input energies corresponding to the acceleration time history parameters of the target area;

[0035] The third acquisition unit is used to obtain the moment magnitude, fault distance, fault type and site shear wave velocity of the target area;

[0036] A unit is established to determine the machine learning algorithm and establish a regional prediction model for elastic and elastoplastic input energy, with moment magnitude, fault distance, fault type, and site shear wave velocity as input parameters and elastic and elastoplastic input energy as output variables, to estimate the elastic and elastoplastic input energy of the corresponding earthquake scenario.

[0037] Furthermore, the module also includes an application unit, which is used to obtain the data to be processed in the target area, input the data to be processed into the establishment unit, and obtain the prediction results of the near-fault seismic input energy regional prediction model.

[0038] A computer-readable storage medium stores a computer program for implementing the regional prediction model based on near-fault seismic input energy; or a computer-readable storage medium stores a computer program for implementing the functions of each unit in the processing module.

[0039] A processing device comprises the computer-readable storage medium and a processor, wherein the processor executes the computer program in the computer-readable storage medium.

[0040] The beneficial effects of the present invention are as follows: the present invention provides a regional prediction model based on near-fault seismic input energy, which can obtain a regional prediction model for near-fault seismic input energy that is universal for various scenarios, convenient to configure input data and has high processing efficiency, and meets the high-quality prediction needs for regional elastic and elastoplastic input energy under actual conditions, and provides accurate and effective data support for the implementation of property-based seismic design engineering activities. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a flow chart of a processing method of a regional prediction model based on near-fault input energy of the present invention;

[0042] Figure 2 It is a schematic diagram of the principle of establishing a regression prediction model by a support vector machine in the present invention;

[0043] Figure 3 is a schematic diagram of a processing module of a regional prediction model based on near-fault input energy in Embodiment 2;

[0044] Figure 4 It is a schematic diagram of processing equipment of the regional prediction model based on near-fault input energy in Example 3.

[0045] Figure 5 It is a schematic diagram of the prediction residual result of the seismic input energy in the present invention. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] The regional prediction model based on near-fault input energy mentioned in the present application may be executed by a processing device of the regional prediction model of near-fault input energy, or a server integrating a processing device of the regional prediction model of near-fault input energy.

[0048] As an example, the processing device may include a first processing device that performs model training in the background and a second processing device that applies the model in the actual site. The model obtained by the first processing device through initial training or continuous updating after being put into actual application can be sent to the second processing device for model deployment, so as to carry out specific application of the model in the actual scene.

[0049] The division of modules appearing in this application is a logical division. There may be other division methods when implemented in actual applications. For example, multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection between modules may be electrical or other similar forms, which are not limited in this application.

[0050] Embodiment 1

[0051] A regional prediction model based on near-fault ground motion input energy, the model processing method includes the following Figure 1 The following steps are shown:

[0052] Step S101, obtaining acceleration time history data of a target area.

[0053] Determine the target area, collect the acceleration seismic data of the target area, and preliminarily screen the data according to data quality, delete the data with abnormal waveforms, and preprocess the near-field seismic data. Preprocessing includes baseline correction and filtering. At present, a relatively mature preprocessing method for seismic data has been formed, which will not be repeated here.

[0054] Step S102: Calculate regional earthquake input energy parameters including elastic and elastoplastic input energies of earthquake motion based on acceleration time history data.

[0055] Based on the acceleration time history data, the elastic input energy is calculated, and the quantitative formula involved is:

[0056]

[0057] Among them, E Ia is the absolute input energy, E Ir is the relative input energy, m is the mass of the single degree of freedom system, x t is the absolute displacement of the single degree of freedom system, x g is the displacement of ground motion, and x is the relative displacement of the single-degree-of-freedom system.

[0058] In order to eliminate the influence of the mass of the single-degree-of-freedom system, the absolute and relative input energy are replaced by equivalent velocity, and the quantitative formula involved is:

[0059]

[0060]

[0061] Among them, V Ea is the equivalent speed of absolute input energy, V Ir is the equivalent speed relative to the input energy.

[0062] Select an appropriate constitutive model, such as the ideal elastic-plastic model, calculate the elastic-plastic input energy based on the acceleration time history data, and replace the elastic-plastic input energy with an equivalent input.

[0063] The essential difference between elastic input energy and elastoplastic input energy lies in whether the deformation of a single-degree-of-freedom system is elastic or not. Therefore, when calculating elastoplastic input energy, it is necessary to first determine the constitutive model, such as the ideal elastoplastic model. Since the research on constitutive models has a fairly mature theoretical system, it will not be elaborated here.

[0064] Step S103: The moment magnitude, fault distance, fault type and site average shear wave velocity V of the target area data are S30 as a characteristic parameter.

[0065] Step S104: Combine the characteristic parameters with the regional earthquake input energy parameters to establish a characteristic database. The processed acceleration time history data can be screened by setting the thresholds of the earthquake peak acceleration and fault distance, such as the peak acceleration greater than 5 cm / s 2 ,The fault distance is less than 30km, which is used to establish the characteristic database of the target area.

[0066] Step S105 , the relevant feature combination data in the feature database are divided into a ratio of 8:2 into a training set and a test set.

[0067] Step S106, determining a machine learning algorithm, in this embodiment, a particle swarm optimization support vector machine algorithm is used, such as Figure 2 As shown; and the moment magnitude, fault distance, fault type, site shear wave velocity V S30 The input parameter is the regional earthquake input energy parameter, and the output parameter is the training set data as the basis for model training to predict the corresponding regional earthquake input energy parameter.

[0068] After completing the moment magnitude, fault distance, fault type, and site shear wave velocity V used in the training model, S30 After inputting energy parameters (these data can be used as training samples for training models), specific model training work can be carried out to train a regional prediction model specifically used to predict near-fault seismic input energy.

[0069] It can be understood that for the near-fault seismic input energy regional prediction model to be configured in this application, the model type involved, the model training framework or loss function used in the training process, can adopt existing schemes, or improve on the basis of existing schemes, or adopt self-developed novel schemes, all of which are possible in actual situations.

[0070] Specifically, in this embodiment, the particle swarm optimization support vector machine algorithm is used to perform regression algorithm for model training, and the quantization formula involved is specifically as follows:

[0071]

[0072] Among them, α i is the Lagrange multiplier vector value; α i * and b are the solutions to the dual optimization problem. K(X i , X) represents a kernel function that must satisfy the Mercer condition. In this embodiment, a radial basis function (RBF) is selected as the kernel function to implement kernel mapping, and the quantization formula involved is specifically:

[0073] K(X i ,X)=exp(-γ||XX i || 2 );

[0074] Where γ is the kernel parameter; X is the input variable; X i is the i-th sample variable.

[0075] Furthermore, when optimizing the parameters in the model, the particle swarm optimization algorithm is used for optimization, and the quantitative formula involved is specifically as follows:

[0076] V i (k+1)=wV i (k)+c 1 ×rand(·)×[pbest i (k)-X i (k)]+c 2 ×rand(·)×[gbest i (k)-

[0077] X i (k)];

[0078] X i (k+1)=X i (k)+V i (k);

[0079] Where w is the inertia weight factor. When its specific value is larger, the optimization ability is stronger. k is the number of steps of particle movement. 1 and c 2 is the learning factor, which is set to 2 in this embodiment. i is the best individual particle, gbest i is the global best particle; X i is the current position of the particle; V i is the velocity of the particle.

[0080] Step S107, verify the training model based on the test set, calculate the residual based on the model trained in step S106, and perform residual analysis. If the distribution of the residual and the response explanatory variable is approximately unbiased, then the obtained model can be considered reasonable and can be put into practical use. Figure 5 shown.

[0081] Step S108, using moment magnitude, fault distance, and site shear wave velocity V S30 For the input parameters, the elastic and elastoplastic input energy parameters of the specific earthquake scenario are estimated, including relative input energy and absolute input energy, and elastoplastic input energy.

[0082] Embodiment 2

[0083] A processing module based on a regional prediction model of near-fault ground motion input energy, such as Figure 3 As shown, module 200 includes:

[0084] A first acquisition unit 201 is used to acquire acceleration time history data of a target area;

[0085] A second acquisition unit 202 is used to acquire elastic and elastoplastic input energies corresponding to acceleration time history parameters of the target area;

[0086] The third acquisition unit 203 is used to acquire the moment magnitude, fault distance, fault type and site shear wave velocity of the target area.

[0087] Establishing unit 204, for determining a machine learning algorithm, establishing a regional prediction model of elastic and elastoplastic input energy, based on moment magnitude, fault distance, fault type, site shear wave velocity V S30 The elastic and elastoplastic input energy equivalent velocities are used as input parameters and the elastic and elastoplastic input energy equivalent velocities corresponding to the earthquake scenario are estimated.

[0088] In this embodiment, the second acquisition unit 202 is used to:

[0089] Based on the acceleration time history data, the elastic input energy is calculated, and the quantitative formula involved is:

[0090]

[0091] Among them, E Ia is the absolute input energy, E Ir is the relative input energy, m is the mass of the single degree of freedom system, x t is the absolute displacement of the single degree of freedom system, x g is the displacement of ground motion, and x is the relative displacement of the single-degree-of-freedom system.

[0092] In order to eliminate the influence of the mass of the single-degree-of-freedom system, the absolute and relative input energy are replaced by equivalent velocity, and the quantitative formula involved is:

[0093]

[0094] Among them, V Ea is the equivalent speed of absolute input energy, V Ir is the equivalent speed relative to the input energy.

[0095] Select an appropriate constitutive model, such as the ideal elastic-plastic model, calculate the elastic-plastic input energy based on the acceleration time history data, and replace the elastic-plastic input energy with an equivalent input.

[0096] In this embodiment, the establishing unit 204 is used to:

[0097] The particle swarm optimization support vector machine algorithm is used to perform regression algorithm model training. The quantitative formula involved is as follows:

[0098]

[0099] Among them, α iis the Lagrange multiplier vector value; α i * and b are the solutions to the dual optimization problem. K(X i , X) represents a kernel function that must satisfy the Mercer condition. In this embodiment, a radial basis function (RBF) is selected as the kernel function to implement kernel mapping, and the quantization formula involved is specifically:

[0100] K(X i ,X)=exp(-γ||XX i || 2 );

[0101] Where γ is the kernel parameter; X is the input variable; X i is the i-th sample variable.

[0102] Furthermore, when optimizing the parameters in the model, the particle swarm optimization algorithm is used for optimization, and the quantitative formula involved is specifically as follows:

[0103] V i (k+1)=wV i (k)+c 1 ×rand(·)×[pbest i (k)-Xi(k)]+c 2 ×rand(·)×[gbest i (k)-X i (k)];

[0104] X i (k+1)=X i (k)+V i (k);

[0105] Where w is the inertia weight factor. When its specific value is larger, the optimization ability is stronger. k is the number of steps of particle movement. 1 and c 2 is the learning factor, which is set to 2 in this embodiment. i is the best individual particle, gbest i is the global best particle; X i is the current position of the particle; V i is the velocity of the particle.

[0106] In this embodiment, the module also includes an application unit 205, which is used to: obtain the data to be processed in the target area, the data to be processed include the target moment magnitude, fault distance, target fault type and target site shear wave velocity; input the data to be processed into the establishment unit 204 to obtain the prediction results of the near-fault seismic input energy regional prediction model.

[0107] Embodiment 3

[0108] A processing device, which may have a structure such as Figure 4 As shown, the processing device includes a processor 301, a memory 302 and an input / output device 303. The processor 301 is used to implement the steps of the near-fault ground motion input energy regional prediction model in the first embodiment when executing the computer program stored in the memory 302.

[0109] Alternatively, the processor 301 is used to implement the functions of each unit in the second embodiment when executing the computer program stored in the memory 302, and the memory 302 is used to store the computer program required for the processor 301 to execute the regional prediction model based on near-fault seismic input energy in the first embodiment.

[0110] It is obvious to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, the embodiments should be considered as exemplary and non-restrictive in all respects.

Claims

1. A regional prediction model based on near-fault ground motion input energy, characterized in that: The model processing method includes the following steps: Step S101, obtaining acceleration time history data of a target area; Step S102, calculating regional earthquake input energy parameters based on acceleration time history data, where the regional earthquake input energy parameters include elastic input energy and elastoplastic input energy of earthquakes; The calculation formula of elastic input energy is: Among them, E Ia is the absolute input energy, E Ir is the relative input energy, m is the mass of the single degree of freedom system, x t is the absolute displacement of the single degree of freedom system, x g is the displacement of ground motion, x is the relative displacement of the single-degree-of-freedom system; In order to eliminate the influence of the mass of the single-degree-of-freedom system, the absolute and relative input energy are replaced by equivalent velocity, and the formula is: Among them, V Ea is the equivalent speed of absolute input energy, V Ir is the equivalent speed relative to the input energy; The elastic-plastic input energy is calculated based on the acceleration time history data through the constitutive model, and the elastic-plastic input energy is replaced by the equivalent input; Step S103, taking the moment magnitude, fault distance, fault type and site average shear wave velocity of the target area data as characteristic parameters; Step S104, combining the characteristic parameters obtained in step S103 with the regional earthquake input energy parameters obtained in step S102 to establish a characteristic database; Step S105, dividing the relevant feature combination data in the feature database obtained in step S104 into a training set and a test set; Step S106, determining a machine learning algorithm, and using moment magnitude, fault distance, fault type, and site shear wave velocity as input parameters, and regional earthquake input energy parameters as output parameters, and using the training set data as a basis for model training to predict the corresponding regional earthquake input energy parameters; Step S107, verifying the training model based on the test set; Step S108, using moment magnitude, fault distance and site shear wave velocity as input parameters, estimating elastic and elastoplastic input energy parameters of a specific earthquake scenario.

2. The regional prediction model based on near-fault ground motion input energy according to claim 1 is characterized in that: The machine learning algorithm in step S106 is a particle swarm optimization support vector machine algorithm, and its quantization formula is: Among them, α i is the Lagrange multiplier vector value; α i * and b are the solutions to the dual optimization problem, K(X i ,X) represents the kernel function that must satisfy the Mercer condition.

3. The regional prediction model based on near-fault ground motion input energy according to claim 2 is characterized in that: The kernel function selects the radial basis function: K(X i ,X)=exp(-γ||X-X i || 2 ); Where γ is the kernel parameter; X is the input variable; X i is the i-th sample variable.

4. The regional prediction model based on near-fault ground motion input energy according to claim 3 is characterized in that: When optimizing the parameters in the model, the following formula is used: V i (k+1)=wV i (k)+c1×rand(·)×[pbest i (k)-X i (k)]+c2×rand(·)×[gbest i (k)-X i (k)]; X i (k+1)=X i (k)+V i (k); Where w is the inertia weight factor, k is the number of steps of particle movement, c1 and c2 are learning factors, and pbest i is the best individual particle, gbest i is the global best particle, X i is the current position of the particle, V i is the velocity of the particle.

5. A processing module loaded with the regional prediction model based on near-fault ground motion input energy according to claim 1, characterized in that: The module (200) comprises: A first acquisition unit (201), used for acquiring acceleration time history data of a target area; A second acquisition unit (202) is used to acquire elastic and elastoplastic input energies corresponding to the acceleration time history parameters of the target area; A third acquisition unit (203) is used to acquire moment magnitude, fault distance, fault type and site shear wave velocity of the target area; A unit (204) is established to determine a machine learning algorithm, establish a regional prediction model for elastic and elastoplastic input energy, use moment magnitude, fault distance, fault type, and site shear wave velocity as input parameters, and elastic and elastoplastic input energy as output variables to estimate the elastic and elastoplastic input energy of the corresponding earthquake scenario.

6. The processing module according to claim 5, characterized in that The module (200) also includes an application unit (205), which is used to obtain the data to be processed in the target area, input the data to be processed into the establishment unit (204), and obtain the prediction result of the near-fault seismic input energy regional prediction model.

7. A computer-readable storage medium, characterized in that: A computer program for implementing the regional prediction model for near-fault seismic input energy as described in claim 1 is stored in a computer-readable storage medium; or a computer program for implementing the functions of each unit in the processing module as described in claim 5 is stored in a computer-readable storage medium.

8. A processing device, characterized in that: The computer-readable storage medium according to claim 7 further comprises a processor (301) which executes a computer program in the computer-readable storage medium.

Images