Machine learning-based method and system for rapid assessment of building seismic resilience

Abstract

The present application relates to the field of machine learning evaluation, especially to a kind of building seismic toughness rapid evaluation method and system based on machine learning, by constructing the three-dimensional geometric model of target building group, the relative position relationship of several target buildings is clear, corresponding material properties are given to structural member, and the structure connection mode of structural member is determined, the building source information of target building group is integrated, the frequency information of earthquake wave is selected as the input layer of pre-earthquake simulation model, the number of hidden layers is determined, the standard pretreatment is carried out to building source information, the learning is carried out to building source pre-information, the propagation simulation result of earthquake wave in target building group is obtained, and the ground motion information of each target building is extracted, and dynamic response analysis is carried out, to obtain the collision possibility between adjacent target buildings, so as to accurately find the weak link of target building group under the action of earthquake, provide clear direction for the improvement of seismic toughness, reduce the collision risk between adjacent buildings when earthquake occurs.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of machine learning evaluation, in particular to a machine learning-based building seismic resilience rapid evaluation method and system. BACKGROUND

[0002] Traditional building seismic evaluation mainly relies on structural mechanics analysis, such as finite element analysis, etc. These methods require detailed modeling and mechanical calculation of each structural member of the building. Moreover, when the scale of the building group is large, the amount of calculation will increase exponentially, and it will take a long time.

[0003] With the rapid development of urbanization, the scale of building groups in cities is getting larger and larger, and the spacing between buildings is getting smaller and smaller. For example, in some high-density urban areas, high-rise buildings are densely distributed, and the mutual influence between buildings is more significant. In this case, the overall seismic performance evaluation of the building group becomes particularly important. The traditional single building seismic evaluation method has been unable to meet the seismic needs of such large-scale building groups.

[0004] In recent years, machine learning technology has developed rapidly, especially in data processing and pattern recognition. In the field of building seismic resistance, machine learning models can be used to learn a large amount of data such as building geometric information, material properties, and seismic wave information, so as to establish a prediction model of building seismic performance.

[0005] Chinese Patent No. CN113536195B discloses a community system seismic resilience calculation method and system at the moment of earthquake disaster, the method comprises: obtaining the community seismic resistance capacity when the earthquake has not occurred; determining the community seismic resistance capacity intact rate at the moment of earthquake occurrence based on the community seismic resistance capacity when the earthquake has not occurred; calculating the seismic resilience of the community at the moment of earthquake occurrence based on the community seismic resistance capacity intact rate at the moment of earthquake occurrence. The method in the invention can reflect the response ability of the community changing with time at the moment of crisis.

[0006] Chinese Patent No. CN114792020B discloses a machine learning-based building seismic resilience rapid evaluation method and system, belonging to the field of building seismic resilience evaluation. First, the building design model, structural design model, geographic location information, and geological condition information of the building to be evaluated are obtained. Then, according to the geographic location information and geological condition information, the seismic risk features of the building to be evaluated are extracted; according to the building design model, the building design features of the building to be evaluated are extracted; according to the structural design model, the structural design features of the building to be evaluated are extracted; and finally, the seismic risk features, building design features, and structural design features of the building to be evaluated are input into the trained building seismic resilience evaluation model to output the resilience evaluation result. This method can reduce the iterative adjustment work in the building design process and improve the efficiency of building design.

[0007] However, the above method has the following problems: the seismic performance analysis focuses on single buildings, and the mutual influence between buildings in the building group is not considered. SUMMARY

[0008] To this end, the present application provides a building seismic resilience rapid evaluation method and system based on machine learning to overcome the problem that the seismic performance analysis in the prior art focuses on single buildings and the mutual influence between buildings in the building group is not considered.

[0009] To achieve the above-mentioned purpose, the present application provides a building seismic resilience rapid evaluation method based on machine learning, comprising:

[0010] A three-dimensional geometric model of the target building group is constructed using a three-dimensional modeling software, the relative position relationship of a plurality of target buildings in the target building group is determined, the corresponding material properties of the structural members of a single target building are given, and the structural connection mode of the structural members is determined. The building source information of the target building group is integrated, wherein,

[0011] The three-dimensional geometric model includes the floor height, wall thickness and structural member size of the target building;

[0012] The relative position relationship includes the floor spacing, relative height and floor-to-floor relative angle of a plurality of target buildings;

[0013] The material properties include the elastic modulus, Poisson's ratio and yield strength of the structural members;

[0014] The building source information includes the relative position relationship, the material properties and the structural connection mode;

[0015] The seismic wave frequency information of the target building group is selected as the input layer of the pre-seismic simulation model, and the number of hidden layers of the pre-seismic simulation model is determined;

[0016] The building source information is pre-processed to form corresponding building source pre-information, the pre-seismic simulation model is used to learn the building source pre-information, the propagation simulation result of the seismic wave in the target building group is obtained, and the seismic motion information of each target building is extracted from the propagation simulation result, wherein,

[0017] The seismic motion information includes the seismic motion acceleration time history, velocity time history and displacement time history;

[0018] The dynamic response analysis of the seismic motion information is performed using a collision simulation algorithm to obtain the collision possibility between adjacent target buildings, the collision possibility is compared with the possibility threshold, and a collision risk warning is issued when the collision possibility is greater than the possibility threshold.

[0019] Further, the step of determining the relative position relationship of the target buildings in the target building group comprises:

[0020] Importing the geographic coordinate data of the target buildings into the three-dimensional modeling software to generate a point map of the target buildings;

[0021] Drawing a polygon layer of the target buildings according to the point map to form a corresponding three-dimensional geometric model;

[0022] Adjusting and determining the relative position relationship of the target buildings according to the three-dimensional geometric model.

[0023] Further, the number of hidden layers is adjusted according to the data complexity of the historical ground motion record and the number of target buildings.

[0024] Further, the step of obtaining the seismic wave frequency information of the target building group comprises:

[0025] Collecting historical seismic waveform data of the area where the target building group is located;

[0026] Performing Fourier transform on the historical seismic waveform data to extract corresponding seismic wave frequency information, and taking the seismic wave frequency information as the input layer of the pre-earthquake simulation model.

[0027] Further, filtering the abnormal values in the building source information, standardizing the building source information, and forming corresponding building source pre-information, wherein,

[0028] The standard preprocessing is to divide the building source information according to a standard learning rate;

[0029] The standard learning rate is a learning rate that can be recognized by the pre-earthquake simulation model, and for a single sampling, the corresponding standard learning rate is a single learning rate.

[0030] Further, the step of extracting the ground motion information of the target building from the propagation simulation result comprises:

[0031] According to the propagation simulation result, extracting the acceleration time history of the target building;

[0032] Integrating the acceleration time history to obtain the corresponding velocity time history of the target building;

[0033] Integrating the velocity time history again to obtain the corresponding displacement time history of the target building.

[0034] Further, the seismic motion information is input as an initial condition into the collision simulation algorithm, and the collision simulation algorithm calculates relative motion between adjacent target buildings according to the initial condition to obtain a collision possibility between adjacent target buildings.

[0035] Further, the collision possibility is compared with the possibility threshold value, wherein,

[0036] The possibility threshold value is set according to a highest allowable range of the collision possibility, to determine whether the collision risk warning needs to be issued;

[0037] When the collision possibility is less than the possibility threshold value, it is determined that the seismic resilience of the target building group meets the normal standard.

[0038] Further, when the collision possibility is greater than the possibility threshold value, it is determined that the seismic resilience of the target building group does not meet the normal standard, and an alarm instruction is triggered to issue a corresponding collision risk warning.

[0039] In another aspect, the present application provides a building seismic resilience rapid evaluation system based on machine learning, comprising:

[0040] A three-dimensional modeling module is configured to construct a three-dimensional geometric model of a target building group, to determine relative position relationships of target buildings in the target building group, to assign corresponding material properties to structural members of a single target building, to determine structural connection modes of the structural members, and to integrate building source information of the target building group.

[0041] A model building module is connected to the three-dimensional modeling module, configured to select seismic wave frequency information of the target building group as an input layer of a pre-earthquake simulation model, and to determine a number of hidden layers of the pre-earthquake simulation model.

[0042] A pre-earthquake simulation module is connected to the model building module, configured to perform standard preprocessing on the building source information to form corresponding building source pre-information, to use the pre-earthquake simulation model to learn the building source pre-information, to obtain a propagation simulation result of seismic waves in the target building group, and to extract seismic motion information of each target building from the propagation simulation result.

[0043] A dynamic response analysis module is connected to the pre-earthquake simulation module, configured to perform dynamic response analysis on the seismic motion information to obtain a collision possibility between adjacent target buildings, and to issue a collision risk warning when the collision possibility is greater than a possibility threshold value.

[0044] Compared with the prior art, the application can find the weak link of the target building group under the action of the earthquake, provide a clear direction for the improvement of the seismic toughness, and reduce the collision risk between adjacent buildings when the earthquake occurs.

[0045] Further, by importing the geographic coordinate data of the target building into the three-dimensional modeling software, the specific position of each building in the geographic space can be determined with high precision. The geographic coordinate data has high accuracy and can provide a reliable basis for subsequent modeling and analysis, avoiding evaluation errors caused by inaccurate position information. After generating the point map, the polygon layer is drawn to form a three-dimensional geometric model, which can more carefully process the shape and boundary of the building, more intuitively display the outline and shape of the building, and more accurately reflect the actual layout and spatial relationship of the building, providing a more reliable model basis for subsequent analysis.

[0046] Further, by Fourier transforming the historical seismic waveform data, the corresponding seismic wave frequency information can be extracted to effectively obtain the seismic wave frequency information of the target building group and use it as the input of the pre-earthquake simulation model, providing a scientific basis for building seismic toughness evaluation.

[0047] Further, by setting a reasonable possibility threshold, false positives and false negatives can be effectively reduced to ensure the accuracy of the evaluation system. By collecting and analyzing the collision possibility data, data support can be provided for the research and development of seismic toughness evaluation technology, helping researchers better understand the dynamic response of target buildings under the action of the earthquake, and thus developing more effective seismic technology and methods. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 A flowchart of a building seismic toughness rapid evaluation method based on machine learning according to an embodiment of the application;

[0049] Figure 2 A flowchart for determining the relative position relationship of several target buildings in the target building group according to an embodiment of the application;

[0050] Figure 3 A flowchart for extracting the seismic motion information of the target building from the propagation simulation result according to an embodiment of the application;

[0051] Figure 4 Fig. 1 shows a structural schematic diagram of a machine learning-based building seismic resilience rapid evaluation system according to an embodiment of the present application. DETAILED DESCRIPTION

[0052] In order to make the objects and advantages of the present application clearer, the present application will be further described in conjunction with embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application.

[0053] The preferred embodiments of the present application will be described below with reference to the accompanying drawings. It should be understood by those skilled in the art that the embodiments are only used to explain the technical principles of the present application and are not used to limit the protection scope of the present application.

[0054] It should be noted that, in the description of the present application, the terms of direction or position relationship such as "upper", "lower", "left", "right", "inner", "outer" and the like are based on the direction or position relationship shown in the drawings, which is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present application.

[0055] In addition, it should also be noted that, in the description of the present application, unless otherwise explicitly specified and limited, the terms "mounting", "connecting", "connection" should be understood in a broad sense, for example, it can be fixed connection, or detachable connection, or integrally connected; it can be mechanical connection, or electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, or the internal communication of two elements. Those skilled in the art can understand the specific meaning of the above terms in the present application according to the specific circumstances.

[0056] Please refer to Figure 1 Fig. 1 shows a structural schematic diagram of a machine learning-based building seismic resilience rapid evaluation system according to an embodiment of the present application.

[0057] In step S1, a three-dimensional geometric model of the target building group is constructed using a three-dimensional modeling software, the relative position relationship of the target buildings in the target building group is determined, the corresponding material properties of the structural members of the single target building are given, the structural connection mode of the structural members is determined, and the building source information of the target building group is integrated, wherein,

[0058] The three-dimensional geometric model includes the floor height, wall thickness and structural member size of the target building;

[0059] The relative position relationship includes the floor spacing, relative height and floor-to-floor relative angle of the target buildings;

[0060] The material properties include the elastic modulus, Poisson's ratio and yield strength of the structural members;

[0061] The building source information includes relative position relationship, material properties and structural connection mode;

[0062] In step S2, the seismic wave frequency information of the target building group is selected as the input layer of the pre-earthquake simulation model, and the number of hidden layers of the pre-earthquake simulation model is determined.

[0063] In step S3, the building source information is preprocessed to form corresponding building source pre-information, the pre-earthquake simulation model is used to learn the building source pre-information, the propagation simulation result of the seismic wave in the target building group is obtained, and the ground motion information of each target building is extracted from the propagation simulation result, wherein

[0064] The ground motion information includes ground motion acceleration time history, velocity time history and displacement time history.

[0065] In step S4, the dynamic response analysis is performed on the ground motion information using the collision simulation algorithm to obtain the collision possibility between adjacent target buildings, the collision possibility is compared with the possibility threshold, and the collision risk warning is issued when the collision possibility is greater than the possibility threshold.

[0066] By constructing the three-dimensional geometric model of the target building group, the relative position relationship of the target buildings is determined, the corresponding material properties are given to the structural members, and the structural connection mode of the structural members is determined, the building source information of the target building group is integrated, the seismic wave frequency information is selected as the input layer of the pre-earthquake simulation model, the number of hidden layers is determined, the building source information is preprocessed, the building source pre-information is learned, the propagation simulation result of the seismic wave in the target building group is obtained, the ground motion information of each target building is extracted, and the dynamic response analysis is performed to obtain the collision possibility between adjacent target buildings, so that the weak link of the target building group under the action of the earthquake can be accurately found, the direction for improving the seismic toughness is provided, and the collision risk between adjacent buildings when the earthquake occurs is reduced.

[0067] Please refer to Figure 2 As shown in the figure, it is a flow chart for determining the relative position relationship of the target buildings in the target building group according to the embodiment of the present application, which comprises the following steps:

[0068] In step S11, the geographic coordinate data of the target building is imported into the three-dimensional modeling software to generate a point map of the target building.

[0069] In step S12, a polygon layer of the target building is drawn according to the point map to form a corresponding three-dimensional geometric model.

[0070] In step S13, the relative position relationship of the target building is adjusted and determined according to the three-dimensional geometric model.

[0071] In practice, the geographic coordinates of each building can be obtained in various ways, such as by conducting on-site measurements using satellite positioning technology (such as GPS) or by extracting them from existing Geographic Information System (GIS) databases. If the building complex is newly constructed, coordinate information can also be obtained from architectural design drawings.

[0072] To ensure a consistent format for geographic coordinate data, latitude and longitude coordinates (WGS84 coordinate system) are typically used. Geographic coordinate data can be stored in tabular form, including fields such as building number, longitude, and latitude.

[0073] Choose suitable 3D modeling software, such as Revit, SketchUp, or Rhinoceros. These software programs have powerful 3D modeling capabilities and can import geographic coordinate data. Configure the 3D modeling software as needed to ensure it can correctly read and process geographic coordinate data. For example, set parameters such as coordinate system and units to ensure that the imported data can be accurately located in 3D space.

[0074] Import the collected geographic coordinate data into the 3D modeling software. Most 3D modeling software supports importing data from files in formats such as CSV and Excel. During the import process, the longitude and latitude fields need to be correctly mapped to the software's coordinate system. After importing the data, the software will generate a point map for each building based on the geographic coordinates. The point map displays the location of the building in two dimensions, with each point representing the center location or other key location points of a building. Based on the actual shape and size of the building, draw a polygon layer for each building on the point map. This can be done manually or using the software's automated tools. For example, in Revit, the "Draw Outline" function can be used to draw polygon boundaries based on the building's floor plan.

[0075] Combine polygon layers with building height information to generate a 3D geometric model. Height information can be manually entered or imported from architectural design files. For example, for a multi-story building, polygon layers can be drawn layer by layer and stacked to form a complete 3D model. After generating the 3D geometric model, carefully check the relative positions of the buildings in the model. If positional deviations or errors are found, the building positions in the model can be manually adjusted. For example, by moving and rotating, ensure that the spacing between buildings, relative heights, and relative angles conform to reality. During the adjustment process, actual architectural layout photos, satellite images, or other geospatial data can be referenced to ensure the accuracy of the model. For example, use tools such as Google Earth to view the actual layout of the building complex and compare it with the 3D model to make necessary adjustments.

[0076] By importing the geographic coordinate data of the target buildings into three-dimensional modeling software, the specific location of each building in geographic space can be determined with high precision. The geographic coordinate data has high accuracy, which can provide a reliable basis for subsequent modeling and analysis, avoiding evaluation errors caused by inaccurate location information. After generating the point map, the polygon layer is drawn to form a three-dimensional geometric model, which can more meticulously process the shape and boundary of the building, more intuitively display the outline and shape of the building, and more accurately reflect the actual layout and spatial relationship of the building, providing a more reliable model basis for subsequent analysis.

[0077] Specifically, the number of hidden layers is adjusted according to the data complexity of historical seismic records and the number of target buildings.

[0078] Embodiment 1:

[0079] Assuming that the target building group has 8 buildings, the historical seismic records contain acceleration, velocity and displacement characteristics, and the data distribution is relatively complex and has certain nonlinear relationship. The specific steps are as follows:

[0080] Preliminary selection of the number of hidden layers: according to the data complexity and the number of buildings, 3 hidden layers are preliminarily selected.

[0081] Model construction and training: use 3 hidden layers to construct a neural network model, train the model and record the performance indicators.

[0082] Verification and adjustment: use 5-fold cross-validation to evaluate the model performance. If the model performance is found to be poor, try to increase to 4 hidden layers; if the model is overfitting, try to reduce to 2 hidden layers.

[0083] Final selection: through multiple verifications, the number of hidden layers with the best performance on the validation set is finally selected (assuming 3 layers).

[0084] Through the above method, the number of hidden layers can be reasonably adjusted according to the data complexity of historical seismic records and the number of target buildings, so as to improve the evaluation accuracy and reliability of the model.

[0085] Specifically, the steps of obtaining the seismic wave frequency information of the target building group include:

[0086] Collect historical seismic waveform data of the area where the target building group is located;

[0087] Perform Fourier transform on the historical seismic waveform data to extract corresponding seismic wave frequency information, and use the seismic wave frequency information as the input layer of the pre-earthquake simulation model.

[0088] In specific implementation, ensure that the collected seismic waveform data format is uniform, usually time series data, including time stamp and corresponding acceleration, velocity or displacement value. Common data file formats include SEED, SAC, etc.

[0089] Fourier transform can convert time domain seismic waveform data into frequency domain data, so as to extract frequency information of seismic wave. Frequency information reflects energy distribution and propagation characteristics of seismic wave, which is crucial for evaluating seismic performance of building.

[0090] Arrange the extracted seismic wave frequency information into a format suitable for inputting into pre-earthquake simulation model. Usually, frequency information can be represented as a vector or matrix, where each row or each column corresponds to a frequency component and its amplitude. Arrange the frequency information as the input layer of the pre-earthquake simulation model. The pre-earthquake simulation model is a neural network or other machine learning model, which is used to learn the response characteristics of building group under different seismic wave frequencies. In the model training process, use frequency information as input and seismic response of building (such as acceleration, velocity, displacement) as output, train the model to predict the dynamic response of building under different seismic wave frequencies.

[0091] By Fourier transform on historical seismic waveform data, extract corresponding seismic wave frequency information, which can effectively obtain seismic wave frequency information of target building group and use it as input of pre-earthquake simulation model, providing scientific basis for building seismic resilience evaluation.

[0092] Specifically, filter the abnormal values in the building source information, standardize the building source information, and form the corresponding building source pre-information, wherein,

[0093] Standard preprocessing is to divide the building source information according to the standard learning rate;

[0094] The standard learning rate is the learning rate that can be recognized by the pre-earthquake simulation model, and for single sampling, the corresponding standard learning rate is a single learning rate.

[0095] In specific implementation, abnormal values refer to data points that deviate significantly from normal range, which may be caused by measurement error, data entry error or special circumstances. In building seismic resilience evaluation, the existence of abnormal values may interfere with the learning process of the model, leading to inaccurate evaluation results. Calculate the mean, standard deviation, median, quartile and other statistics of the data, and identify abnormal values through these statistics. According to the above method, the abnormal values identified are removed from the data set or replaced with reasonable values (such as median or mean). For example, for elastic modulus data, if a value is found to deviate significantly from the normal range, it can be replaced with the median of the data set.

[0096] The building source information is divided into multiple subsets according to a standard learning rate. The standard learning rate refers to a learning rate that can be recognized by the pre-earthquake simulation model. For a single sampling, the corresponding standard learning rate is a single learning rate. Preferably, the standard learning rate is 0.01, and the building source information can be divided into multiple subsets according to this learning rate, and each subset contains a certain number of samples.

[0097] Referring to Figure 3 shown, which is a flowchart of extracting the ground motion information of the target building from the propagation simulation results according to an embodiment of the present application, comprising:

[0098] Step S21, according to the propagation simulation results, extracting the acceleration time history of the target building;

[0099] Step S22, integrating the acceleration time history to obtain the corresponding velocity time history of the target building;

[0100] Step S23, integrating the velocity time history again to obtain the corresponding displacement time history of the target building.

[0101] In specific implementation, the propagation simulation results are the propagation of seismic waves in the building group, including the acceleration response of each target building at different time points. From the propagation simulation results, according to the identification (such as building number or location) of the target building, the acceleration time history under the action of the earthquake is extracted. The acceleration time history is a time series that records the acceleration value at each time point after the earthquake occurs. The acceleration time history can be converted into a velocity time history by integration, and the velocity time history reflects the velocity change of the building under the action of the earthquake.

[0102] Integration method: using numerical integration method (such as trapezoidal integration method) to integrate the acceleration time history.

[0103] The velocity time history can be converted into a displacement time history by integration, and the displacement time history reflects the displacement change of the building under the action of the earthquake. Similarly, the numerical integration method (such as trapezoidal integration method) is used to integrate the velocity time history.

[0104] Through the above steps, the acceleration time history, velocity time history and displacement time history of the target building can be obtained, which are crucial for evaluating the seismic performance and collision risk of the building.

[0105] Specifically, the ground motion information is input as an initial condition into the collision simulation algorithm, and the collision simulation algorithm calculates the relative motion between adjacent target buildings according to the initial condition to obtain the collision possibility between adjacent target buildings.

[0106] In a specific implementation, the ground motion information of the target buildings extracted from the propagation simulation results includes acceleration time histories, velocity time histories, and displacement time histories. The ground motion information is organized into a format suitable for input into the collision simulation algorithm. Typically, this includes the displacement, velocity, and acceleration of each building at different time points.

[0107] A suitable collision simulation algorithm is selected, such as the discrete element method (DEM), the finite element method (FEM), or a rule-based collision detection algorithm. The collision simulation algorithm calculates the relative motion between adjacent target buildings based on the input initial conditions. This includes the relative displacement, relative velocity, and relative acceleration. The collision simulation algorithm detects whether the distance between adjacent buildings is less than a certain threshold value (such as the size of the building components) to determine whether a collision has occurred. Based on the results of the relative motion calculation and collision detection, the algorithm gives the collision probability between adjacent target buildings.

[0108] Specifically, the collision probability is compared with a probability threshold value, wherein,

[0109] The probability threshold value is set according to the highest allowable range of the collision probability to determine whether a collision risk warning needs to be issued;

[0110] When the collision probability is less than the probability threshold value, it is determined that the seismic resilience of the target building group meets the normal standard.

[0111] Specifically, when the collision probability is greater than the probability threshold value, it is determined that the seismic resilience of the target building group does not meet the normal standard, and an alarm instruction is triggered to issue the corresponding collision risk warning.

[0112] Example 2:

[0113] The collision probability evaluation results of the target building group are as follows:

[0114] Collision probability: 0.4;

[0115] Probability threshold value: 0.5;

[0116] The collision probability does not exceed 0.5, and it is determined that the seismic resilience of the target building group meets the normal standard.

[0117] Example 3:

[0118] The collision probability evaluation results of the target building group are as follows:

[0119] Collision probability: 0.6;

[0120] Probability threshold value: 0.5;

[0121] The collision probability exceeds 0.5, and it is determined that the seismic resilience of the target building group does not meet the normal standard, triggering an alarm instruction to issue a collision risk warning.

[0122] In specific implementation, when the collision risk warning is issued, relevant departments and personnel can take emergency measures in advance, such as evacuating personnel, starting emergency plans, etc. This helps to reduce personnel casualties and property losses when an earthquake occurs. By issuing the warning in a timely manner, the public's awareness of earthquake risk and safety consciousness can be improved, and the public can be encouraged to actively participate in earthquake resistance and disaster reduction work.

[0123] By setting a reasonable possibility threshold, false positives and false negatives can be effectively reduced, ensuring the accuracy of the evaluation system. By collecting and analyzing collision possibility data, data support can be provided for the research and development of seismic resilience evaluation technology, helping researchers better understand the dynamic response of target buildings under earthquake action, and thus developing more effective seismic technology and methods.

[0124] Please refer to Figure 4 The structural diagram of a building seismic resilience rapid evaluation system based on machine learning is shown in the figure, which includes:

[0125] The three-dimensional modeling module is used to construct a three-dimensional geometric model of the target building group, to clearly define the relative position relationship of the target buildings in the target building group, to assign appropriate material properties to the structural members of individual target buildings, and to determine the structural connection mode of the structural members, and to integrate the building source information of the target building group.

[0126] The model building module is connected to the three-dimensional modeling module, and is used to select the seismic wave frequency information of the target building group as the input layer of the pre-earthquake simulation model, and to determine the number of hidden layers of the pre-earthquake simulation model.

[0127] The pre-earthquake simulation module is connected to the model building module, and is used to standardize the preprocessing of the building source information to form corresponding building source pre-information, to use the pre-earthquake simulation model to learn the building source pre-information, to obtain the propagation simulation results of the seismic wave in the target building group, and to extract the seismic motion information of each target building from the propagation simulation results.

[0128] The dynamic response analysis module is connected to the pre-earthquake simulation module, and is used to analyze the dynamic response of the seismic motion information to obtain the collision possibility between adjacent target buildings, and to issue a collision risk warning when the collision possibility is greater than the possibility threshold.

[0129] Thus, the technical solutions of the present application have been described in conjunction with the preferred embodiments shown in the accompanying drawings, but those skilled in the art will readily understand that the protection scope of the present application is obviously not limited to these specific embodiments. Those skilled in the art can make equivalent changes or replacements to the related technical features without departing from the principles of the present application, and the technical solutions after such changes or replacements will all fall within the protection scope of the present application.

[0130] The above merely provides the preferred embodiments of the present application, but not for limiting the present application; for those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A machine learning based method for rapid assessment of seismic resilience of buildings, characterized in that, The application relates to a method for earthquake risk warning of a target building group. The method comprises the following steps: a three-dimensional geometric model of the target building group is constructed by using three-dimensional modeling software, the relative position relationship of a plurality of target buildings in the target building group is determined, material properties of structural members of a single target building are given, the structural connection mode of the structural members is determined, and building source information of the target building group is integrated, wherein, the three-dimensional geometric model comprises floor height, wall thickness and structural member size of the target building; the relative position relationship comprises inter-floor spacing, relative height and relative angle between floors of the target buildings; the material properties comprise elastic modulus, Poisson's ratio and yield strength of the structural members; the building source information comprises the relative position relationship, the material properties and the structural connection mode; seismic wave frequency information of the target building group is selected as an input layer of a pre-earthquake simulation model, and the number of hidden layers of the pre-earthquake simulation model is determined; standard preprocessing is performed on the building source information to form corresponding building source pre-information, the pre-earthquake simulation model is used to learn the building source pre-information, propagation simulation results of seismic waves in the target building group are obtained, and seismic motion information of each target building is extracted from the propagation simulation results, wherein, the seismic motion information comprises acceleration time history, velocity time history and displacement time history of seismic motion; 2.The machine learning based rapid building seismic resilience assessment method according to claim 1, wherein, dynamic response analysis is performed on the seismic motion information by using a collision simulation algorithm to obtain a collision possibility between adjacent target buildings, the collision possibility is compared with a possibility threshold value, and a collision risk warning is issued when the collision possibility is greater than the possibility threshold value. The step of determining the relative position relationship of a plurality of target buildings in the target building group comprises: geographic coordinate data of the target buildings are imported into the three-dimensional modeling software to generate a point map of the target buildings; a polygon layer of the target buildings is drawn according to the point map to form a corresponding three-dimensional geometric model; 3.The machine learning based rapid building seismic resilience assessment method of claim 2, wherein, the relative position relationship of the target buildings is adjusted and determined according to the three-dimensional geometric model.

4. The machine learning based rapid assessment method of building seismic resilience according to claim 3, wherein, The number of hidden layers is adjusted according to the data complexity of historical seismic motion records and the number of target buildings. The step of obtaining seismic wave frequency information of the target building group comprises: historical seismic waveform data of an area where the target building group is located are collected; 5. The machine learning based rapid building seismic resilience assessment method according to claim 4, wherein, Fourier transform is performed on the historical seismic waveform data to extract corresponding seismic wave frequency information, and the seismic wave frequency information is used as an input layer of the pre-earthquake simulation model. Abnormal values in the building source information are filtered, the building source information is subjected to standard preprocessing to form corresponding building source pre-information, wherein, the standard preprocessing is to segment the building source information according to a standard learning rate; 6.The method of claim 5, wherein, the standard learning rate is a learning rate that can be recognized by the pre-earthquake simulation model, and for single sampling, the corresponding standard learning rate is a single learning rate. The step of extracting seismic motion information of a target building from propagation simulation results comprises: acceleration time history of the target building is extracted according to the propagation simulation results; integration is performed on the acceleration time history to obtain corresponding velocity time history of the target building; The velocity-time history is integrated again to obtain a displacement-time history corresponding to the target building.

7. The machine learning based rapid assessment method of building seismic resilience according to claim 6, wherein, The seismic motion information is input as an initial condition into the collision simulation algorithm, and the collision simulation algorithm calculates relative motion between adjacent target buildings according to the initial condition to obtain a collision possibility between adjacent target buildings.

8. The machine learning based rapid assessment method of building seismic resilience according to claim 7, wherein, The collision possibility is compared with the possibility threshold value, wherein, The possibility threshold value is set according to a highest allowable range of the collision possibility, and is used to determine whether the collision risk warning needs to be issued; When the collision possibility is less than the possibility threshold value, it is determined that the seismic resilience of the target building group meets the normal standard. 9.The machine learning based rapid building seismic resilience assessment method of claim 8, wherein, When the collision possibility is greater than the possibility threshold value, it is determined that the seismic resilience of the target building group does not meet the normal standard, and an alarm instruction is triggered to issue a corresponding collision risk warning.

10. A system for rapid assessment of seismic resilience of buildings based on a machine learning based method according to any one of claims 1-9, characterized in that, It comprises: a three-dimensional modeling module, which is used to construct a three-dimensional geometric model of a target building group, to clearly define the relative position relationship of a plurality of target buildings in the target building group, to assign corresponding material properties to the structural members of a single target building, to determine the structural connection mode of the structural members, and to integrate the building source information of the target building group; a model building module connected with the three-dimensional modeling module, which is used to select seismic wave frequency information of the target building group as an input layer of a pre-earthquake simulation model, and to determine the number of hidden layers of the pre-earthquake simulation model; a pre-earthquake simulation module connected with the model building module, which is used to standardize the building source information for preprocessing to form corresponding building source pre-information, to use the pre-earthquake simulation model to learn the building source pre-information, to obtain a propagation simulation result of seismic waves in the target building group, and to extract seismic motion information of each target building from the propagation simulation result; a dynamic response analysis module connected with the pre-earthquake simulation module, which is used to perform dynamic response analysis on the seismic motion information to obtain a collision possibility between adjacent target buildings, and to issue a collision risk warning when the collision possibility is greater than a possibility threshold value.

Images