WO2022242435A1 - 一种基于人工智能的场地地震液化灾害快速评估方法 - Google Patents

一种基于人工智能的场地地震液化灾害快速评估方法 Download PDF

Info

Publication number
WO2022242435A1
WO2022242435A1 PCT/CN2022/089357 CN2022089357W WO2022242435A1 WO 2022242435 A1 WO2022242435 A1 WO 2022242435A1 CN 2022089357 W CN2022089357 W CN 2022089357W WO 2022242435 A1 WO2022242435 A1 WO 2022242435A1
Authority
WO
WIPO (PCT)
Prior art keywords
earthquake
site
frequency
information
liquefaction
Prior art date
Application number
PCT/CN2022/089357
Other languages
English (en)
French (fr)
Inventor
周燕国
陈仕海
刘凯
杨啸天
张东超
陈云敏
Original Assignee
浙江大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 浙江大学 filed Critical 浙江大学
Publication of WO2022242435A1 publication Critical patent/WO2022242435A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24—Querying
    • G06F16/245—Query processing
    • G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
    • G06F16/2462—Approximate or statistical queries
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/951—Indexing; Web crawling techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/26—Government or public services
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A10/00—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE at coastal zones; at river basins
    • Y02A10/40—Controlling or monitoring, e.g. of flood or hurricane; Forecasting, e.g. risk assessment or mapping

Definitions

  • the invention belongs to the field of earthquake engineering geological disaster prevention and computer technology application field, a rapid assessment method of site earthquake liquefaction disaster, specifically a rapid assessment method of site earthquake liquefaction disaster based on artificial intelligence.
  • the current assessment of site earthquake liquefaction disasters mostly uses on-site surveys, or builds empirical models based on historical earthquake surveys in many places to assess earthquake damage.
  • the shortcoming of on-site surveys is to carry out investigations after the fact.
  • the data points in include a wide variety of soil properties, and the results evaluated are often the most unfavorable and generally conservative.
  • the use of empirical methods to assess earthquake damage will not take into account the changes in site properties after strong earthquakes, and will only use the existing on-site survey results as a basis for on-site earthquake damage assessment.
  • Strong earthquake stations can monitor the situation of the site in real time. This information not only reflects the nature of the earthquake, but also reflects the nature of the site. Using the information collected by strong earthquake stations as a supplement to earthquake damage assessment can better accurately reflect the real-time changes in the nature of the site.
  • the present invention proposes a A rapid assessment method for site earthquake liquefaction (softening) disasters based on artificial intelligence algorithms, which can quickly assess the performance parameters of various aspects of the site after the earthquake or within a period of time after the earthquake, and can quickly assess the degree of site earthquake damage and its comparison under given earthquake conditions.
  • Comprehensive field service performance can quickly assess the performance parameters of various aspects of the site after the earthquake or within a period of time after the earthquake, and can quickly assess the degree of site earthquake damage and its comparison under given earthquake conditions.
  • the establishment of historical earthquake and site information database includes sequentially connected demand input module, web crawler module, data processing module and database module;
  • the demand input module receives the input strong earthquake station information and earthquake signal collection time range and then sends it to the web crawler module;
  • the strong earthquake station information is a specific strong earthquake station number or a given latitude and longitude range and then a strong earthquake station within the given range.
  • At least 4 or more major earthquakes greater than magnitude 6.0 occurred within the time range of earthquake signal collection, and the time interval between the two major earthquakes should not be less than 1 month. If the input time range does not meet the requirements, it will be automatically filled to The required time frame is met.
  • the web crawler module receives the strong earthquake station information and the collection time range of the earthquake signal from the demand input module, and uses the web crawler to crawl on the strong earthquake database website according to the strong earthquake station number and the collection time range of the earthquake signal.
  • the strong earthquake station corresponding to the strong earthquake station number and the earthquake signal of the strong earthquake station collect all public earthquake signals, basic earthquake information and site information within the time range, and make a summary table of basic earthquake information; , basic earthquake information and site information and send them to the data processing module, and send the summary table of basic earthquake information to the database module;
  • the ground motion signal refers to the acceleration time history signal in three orthogonal directions including two horizontal directions and one vertical direction.
  • Each earthquake corresponds to a strong earthquake station and may have zero or one ground motion signal .
  • the basic earthquake information refers to the earthquake number, earthquake time, magnitude, epicentral location, epicentral distance, focal depth, azimuth of the earthquake source relative to the strong earthquake station, etc. corresponding to all earthquake signals within a period of time.
  • the number is composed of the station number of the strong earthquake and the time when the earthquake occurred.
  • the site information includes soil layer distribution, soil and topographical parameters where the strong earthquake station is located.
  • the data processing module receives the ground motion signal from the web crawler module for analysis and processing, and obtains the main frequency of the acceleration time history signal in the two horizontal directions, three kinds of ground motion intensity parameters and the excellent frequency of the site where the strong motion station is located. , organize into a summary table of earthquake supplementary information and send it to the database module;
  • the database module receives the summary table of basic earthquake information from the web crawler module and the summary table of supplementary information of earthquakes from the data processing module and merges them according to the ground motion signals to obtain the general table of earthquake information, and then the earthquake information of all strong earthquake stations
  • the general table is divided into multiple sub-tables according to different strong earthquake stations, each strong earthquake station corresponds to a sub-table, and the information in each sub-table is arranged in chronological order to obtain the earthquake information of a single strong earthquake station
  • the summary table corresponds the site information of the site where the strong earthquake station is located to the summary table of earthquake information for each single strong earthquake station.
  • the excellent frequency of the site where the strong earthquake station is located is processed and obtained:
  • the median curve excellent frequency obtained in the last cycle iteration is used as the site excellent frequency of the ground motion signal.
  • the S2 includes:
  • the excellent frequency prediction model is trained, tested and saved, and the ground motion information, ground motion intensity, main frequency of acceleration in the horizontal direction and site information in the seismic information summary table of a single strong earthquake station are used as the input of the model, and the site excellent frequency is used as the input of the model.
  • the output of the model is input into the excellent frequency prediction model for training, and the test is saved after training;
  • the ground motion information to be predicted the earthquake intensity index, the main frequency of the acceleration in the horizontal direction, and the site information are input into the excellent frequency prediction model after the training test is saved, and the excellent frequency of the post-earthquake site is obtained.
  • the S3 includes:
  • the site earthquake damage degree and seismic performance parameters are obtained based on the average shear wave velocity of the site.
  • the site earthquake damage degree refers to the settlement, lateral displacement and pore pressure rise degree of the foundation
  • the site seismic performance parameter refers to the anti-liquefaction strength and response spectrum of the site. curve, and the reduction of the bearing capacity of the pile foundation.
  • the average shear wave velocity of the site is obtained by calculating the post-earthquake site preeminent frequency of the site where the strong earthquake station is located;
  • V savg 4Hf
  • V savg the average shear wave velocity of the site
  • H the thickness of the soil layer of the site
  • f the pre-earthquake site preeminent frequency obtained by S2 prediction.
  • r u (t) excess pore pressure ratio
  • r u u/ ⁇ ' v0
  • u excess pore water pressure
  • ⁇ ' v0 effective overlying stress
  • f 0 (t) strong Pre-earthquake site preeminent frequency
  • the site category is determined by the thickness of the soil layer and the average shear wave velocity of the site, and then the site response spectrum curve is given.
  • V s after the earthquake the distribution of shear wave velocity after the earthquake
  • V s before the earthquake the distribution of shear wave velocity before the earthquake
  • V savg after the earthquake the average shear wave velocity of the site after the earthquake
  • V savg Before the earthquake the average shear wave velocity of the site before the earthquake; in the present invention, it is considered that the shear wave velocity distribution of the site before and after the earthquake is similar.
  • the anti-liquefaction strength of each soil layer on the site is obtained by the following formula:
  • CRR the liquefaction resistance strength of the soil layer
  • ⁇ c the reduction coefficient of multi-directional seismic vibration
  • P a the normalized reference pressure, which is taken as 100kpa
  • k N the ratio between the strength and stiffness of the soil mass Correlation coefficient between , determined by experiment
  • ⁇ soil density
  • F(e min ) a function only related to the minimum void ratio
  • V s1 shear wave velocity corrected by overlying effective stress
  • K 0 lateral pressure coefficient of soil layer, 0.5 for normally consolidated soil
  • P a 100kpa , ⁇ ’ v0 — overlying effective stress of soil layer
  • Shear wave speed the normalized reference pressure
  • the anti-liquefaction safety factor FS is calculated by the following formula:
  • CSR the cyclic shear stress ratio caused by the earthquake
  • a max the peak horizontal acceleration of the surface
  • g the acceleration of gravity
  • ⁇ v0 the total overlying stress of the soil layer
  • ⁇ ' v0 the overlying stress of the soil layer overlying effective stress
  • ⁇ d resistance coefficient of shear stress along depth.
  • the reduction amount of the bearing capacity of the pile foundation is determined according to the anti-liquefaction safety factor.
  • F a represents the reference value of liquefaction safety factor
  • the seismic lateral displacement is obtained according to the maximum shear strain processing; if it is a horizontal site, the seismic settlement is obtained according to the maximum shear strain processing:
  • the seismic lateral displacement is determined by the following formula:
  • LD semiconductor lateral displacement
  • C h parameters related to local soil properties, obtained from the local ground subsidence data observed in past earthquakes:
  • z max represents the thickness of the inclined soil layer;
  • the seismic settlement is determined by the following formula:
  • ⁇ e the change in the void ratio of the foundation soil
  • e 0 the initial void ratio of the soil layer
  • e max the maximum void ratio of the soil layer
  • e min the minimum void ratio of the soil layer
  • S the total settlement
  • H i thickness of the i-th soil layer
  • R 0 * represents the reference value of residual volume strain, which is a constant
  • m represents the power correlation coefficient between the maximum shear strain and residual volume change, which is a constant.
  • the process of the method of the present invention is automatically implemented by python programming to input the number corresponding to the strong earthquake station and the excellent frequency of the site, as shown in Figure 3, each sub-step can be automatically performed, and the degree of earthquake damage and the seismic performance of the site are given.
  • the invention includes establishing a historical earthquake and site information database, establishing a neural network model to predict the post-earthquake site excellence frequency, and obtaining site earthquake damage degree and site seismic performance parameters based on the site average shear wave velocity.
  • the invention solves the problem of quickly evaluating site earthquake damage and site seismic performance parameters after an earthquake, and can quickly evaluate site liquefaction or softening damage degree and site seismic performance parameters under given earthquake conditions.
  • the invention adopts the earthquake signal received by the strong earthquake station as the basis, and based on this, the property of the site is reflected in real time.
  • Fig. 1 is a technical flow chart of the inventive method
  • Figure 2 is a schematic diagram of the decline and recovery of the site's preeminent frequency after the earthquake
  • Fig. 3 is the technical flowchart of establishing historical earthquake and site information database
  • Figure 4 is a technical flow chart for obtaining the site earthquake damage degree and seismic performance parameters based on the site average shear wave velocity
  • Fig. 5 is the window selection schematic diagram of HVSR method (horizontal vertical spectral ratio method).
  • Fig. 6 is a schematic diagram of horizontal and vertical spectrum ratio spectrum and determining the site's prominent frequency according to it;
  • Fig. 7 is a schematic diagram of a data table compiled by a strong earthquake station
  • Figure 8 is a section view of the soil layer of a certain site, including the variation of shear wave velocity and cone tip resistance with depth;
  • Figure 9 is an effect diagram of the superior frequency prediction model.
  • Fig. 10 is a flow chart of the present invention to obtain the preeminent frequency of the site.
  • Fig. 11 is an effect diagram of automatically selecting the coda wave window of the earthquake signal in the present invention.
  • Fig. 12 is an effect diagram of the excellent frequency of the site obtained by automatically performing horizontal and vertical spectral ratio analysis in the present invention.
  • the establishment of the historical earthquake and site information database includes a demand input module, a web crawler module, a data processing module, and a database module that are connected in sequence; in the specific implementation, this process is realized by python programming, and the demand is input in the demand input module Once the data is collected, all subsequent steps can be automatically performed and organized into a format that is easy to read and processed by artificial intelligence algorithms.
  • the demand input module receives the input strong earthquake station information and the collection time range of the earthquake signal and then sends it to the web crawler module;
  • the web crawler module receives the strong earthquake station information and the time range of earthquake signal collection from the demand input module, and uses the web crawler on the public strong earthquake database websites around the world according to the strong earthquake station number and the time range of earthquake signal collection Crawl the strong earthquake station corresponding to the strong earthquake station number and all the public earthquake motion signals, basic earthquake information and site information within the collection time range of the strong earthquake station and the strong earthquake signal under the strong earthquake station, and mainly make the basic earthquake information from the basic earthquake information Summary table; send the earthquake signal, basic earthquake information and site information to the data processing module, and send the summary table of basic earthquake information to the database module;
  • the main frequency of the acceleration signal in the two horizontal directions of the site is obtained by fast Fourier transform to obtain the Fourier spectrum in the two horizontal directions, and the frequency corresponding to the point with the largest amplitude is the main frequency of the acceleration.
  • the main frequency signal of the site where the strong earthquake station is located refers to the horizontal vertical spectral ratio (HVSR) analysis of the ground motion signal, and the coda part of the ground motion signal (the coda wave signal refers to the acceleration amplitude that is significantly smaller than the main shock part)
  • HVSR horizontal vertical spectral ratio
  • the coda wave signal refers to the acceleration amplitude that is significantly smaller than the main shock part
  • the data processing module receives the ground motion signals from the web crawler module for analysis and processing, and obtains the main frequency of the acceleration time-history signals in the two horizontal directions, three kinds of ground motion intensity parameters and the site excellence frequency of the strong motion station, and organizes them into Seismic supplementary information summary table and sent to the database module;
  • the three earthquake intensity parameters include Arias intensity Ia, two absolute cumulative accelerations CAV and CAV5.
  • HVSR horizontal vertical spectral ratio
  • the coda wave signal refers to the acceleration amplitude significantly smaller than the main shock.
  • Part take two to three consecutive windows without overlapping parts, and perform the following operations on each window: perform Fourier transform on the two components in the horizontal direction and compare the sum of the squares of the two spectral amplitudes Add the root sign again as the horizontal acceleration spectrum H, perform Fourier transform on the acceleration signal in the vertical direction as the vertical acceleration spectrum V, and then divide the horizontal acceleration spectrum and the vertical acceleration spectrum to obtain the horizontal and vertical spectrum ratio spectrum H/V.
  • the abscissa of the horizontal acceleration spectrum H, the vertical acceleration spectrum V, and the horizontal-vertical ratio spectrum H/V is frequency, and the ordinate is the amplitude corresponding to the frequency.
  • the current operation of the horizontal and vertical spectral ratio method often relies on manually selecting the window, using the visual method to check the results obtained by the HVSR method, and re-analyzing the unqualified ones.
  • this method has been widely used in decades of practice , but the window of each signal needs to be manually selected and checked, which is time-consuming and labor-intensive.
  • the method for obtaining the excellent frequency of the site proposed by the present invention can automatically select the uncontaminated time window, delete the polluted time window, and input the earthquake signal without human intervention to directly obtain the excellent frequency of the site.
  • the flow chart is shown in the figure 10 shown.
  • the excellent frequency of the site where the strong earthquake station is located is obtained by processing:
  • HVSR method horizontal and vertical spectral ratio method
  • the ground motion signal is divided into n parts according to the window length, and the window length of the specific implementation is selected as 4 seconds, and the horizontal and vertical spectral ratio (HVSR) analysis is carried out to n windows to obtain n horizontal and vertical spectral ratios of the ground motion signal Spectrum (H/V spectrum), a window is analyzed by HVSR to obtain a horizontal and vertical spectrum ratio spectrum.
  • the frequency corresponding to the maximum value of the horizontal and vertical spectrum ratio is taken as the excellent frequency of the window;
  • the specific implementation selects 50 times of the sampling period as the minimum value of the window length.
  • the median curve excellent frequency obtained in the last cycle iteration is used as the site excellent frequency of the ground motion signal.
  • Figure 11 is a schematic diagram of the acceleration time-history signal in three directions of a ground motion signal. The final result of window selection is shown in the light-colored box in the figure. A total of 16 windows were selected, 12 windows were deleted, and finally left 4 windows.
  • Figure 12 is the analysis result of the ground motion signal.
  • the left figure is the initial HVSR analysis result without window deletion, and the right figure is the HVSR analysis result after the window affected by noise is deleted. It can be seen that the initial HVSR analysis results
  • the frequency of the signal is scattered and the standard deviation (the range of the gray shaded part in the figure is also relatively large). After the program automatically deletes some windows, the excellent frequency obtained is relatively concentrated, which can meet the needs of actual use.
  • the signal is not obtained by window deletion.
  • the prominent frequency of the median curve is 1.83 Hz. After deleting the window, the prominent frequency of the median curve is 1.74 Hz.
  • the final analysis result of 1.74 Hz is taken as the site’s prominent frequency.
  • the database module receives the summary table of basic earthquake information from the web crawler module and the summary table of supplementary information of earthquakes from the data processing module and merges them according to the ground motion signals to obtain a general table of earthquake information, and then compiles the general table of earthquake information of all strong earthquake stations into Different strong earthquake stations are divided into multiple sub-tables, each strong earthquake station corresponds to a sub-table, and the information in each sub-table is arranged in chronological order to form time series information, and the single strong earthquake station earthquake
  • the information summary table corresponds the site information of the site where the strong earthquake station is located to the earthquake information summary table of each single strong earthquake station, as shown in Figure 7.
  • GRU gated recurrent unit
  • the ground motion information to be predicted the earthquake intensity index, the main frequency of the acceleration in the horizontal direction, and the site information parameters are input into the excellent frequency prediction model output after the training test is saved to obtain the excellent frequency of the site after the earthquake.
  • the site excellent frequency in the seismic information summary table of a single strong earthquake station is used as the output of the model, and other parts are used as the input of the model, which is divided into a training set and a test set.
  • the test set needs to contain the excellent frequency after a major earthquake For descent and recovery, the number of data in the test set is 1/3 to 1/4 of the number of data in the training set.
  • the average absolute error of the model is used as the loss function, and the minimum training loss function of the test set is used as the target of the model. Bayesian hyperparameter adjustment is performed on the model, and finally the model with the smallest training loss is saved for use.
  • Hyperparameters refer to parameters such as the number of hidden layers in the GRU neural network, the number of hidden layer units, the learning rate, and the dropout ratio.
  • the process of hyperparameter optimization adopts empirically determined parameters hyperparameter selection methods, such as random search, grid search and Bayesian parameter tuning.
  • the Bayesian optimization method assumes the hyperparameters and the probability distribution form of the corresponding objective function, and then adjusts the relevant parameters of the hyperparameters and the probability distribution of the objective function according to the pros and cons of the test results, and then adjusts the hyperparameters.
  • This process is realized by python programming. After receiving the seismic information summary table of a single strong earthquake station and inputting the earthquake information to be predicted into the model, the next steps can be automatically carried out and the excellent frequency of the site will be output.
  • the average shear wave velocity of the site is obtained by calculating the post-earthquake site preeminent frequency of the site where the strong earthquake station is located;
  • the average shear wave velocity of the site is obtained by the following formula:
  • V savg 4Hf
  • V savg the average shear wave velocity of the site
  • H the thickness of the soil layer of the site
  • f the pre-earthquake site preeminent frequency obtained by S2 prediction.
  • the excess static pore pressure ratio of the site is obtained as the degree of pore pressure rise:
  • the average shear wave velocity distribution along depth is obtained according to the following formula:
  • the anti-liquefaction strength of each soil layer on the site is obtained by the following formula:
  • the anti-liquefaction safety factor FS is calculated by the following formula:
  • the seismic lateral displacement is obtained according to the maximum shear strain processing; if it is a horizontal site, the seismic settlement is obtained according to the maximum shear strain processing:
  • the seismic lateral displacement is determined by the following formula:
  • the seismic settlement is determined by the following formula:
  • the ground motion signals of a strong earthquake station were collected from 2010 to 2020.
  • the soil profile information of the area where the strong earthquake station is located is shown in Figure 8, and the thickness of the overlying soft soil layer is 20m.
  • each ground motion signal obtains the main frequency of the horizontal component corresponding to each ground motion signal, and ground motion intensity parameters, use the HVSR method to analyze the excellent frequency of the site, and then collect the earthquake corresponding to each ground motion signal
  • the motion information including the earthquake time, magnitude, focal depth, focal distance, and azimuth of the focal point relative to the strong earthquake station, is sorted and collected into a table, as shown in Figure 7.
  • the dots are the actual site excellence frequency data
  • the square dots are the predicted values of the model on the training set
  • the triangle dots are the model on the test set predicted value of .
  • the earthquake occurrence time is set to 1 day after the last earthquake signal received, the magnitude is set to 6.5, the focal depth and distance are both set to 10, and the level Set peak acceleration 1 to 2000, horizontal peak acceleration 2 to 2000, vertical peak acceleration to 1000, azimuth angle to 250, absolute cumulative acceleration to 25, arias strength to 5, cav5 to 20, horizontal
  • the main frequency of the directional acceleration signal 1 is set to 2
  • the main frequency of the horizontal acceleration signal 2 is set to 2
  • input into the saved model, the excellent frequency of the site is 1.36914163hz.
  • the actual measured site preeminent frequency of the last earthquake in 2020 is 2.317hz.
  • the preeminent frequency of the site obtained by the GRU neural network model is 1.36914163hz.
  • the present invention can quickly evaluate the change of site performance under given earthquake conditions, and solves the problem that only the static seismic performance of soil is considered in previous earthquake damage assessments.

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Tourism & Hospitality (AREA)
  • Databases & Information Systems (AREA)
  • Economics (AREA)
  • Human Resources & Organizations (AREA)
  • Strategic Management (AREA)
  • Data Mining & Analysis (AREA)
  • Probability & Statistics with Applications (AREA)
  • General Engineering & Computer Science (AREA)
  • General Business, Economics & Management (AREA)
  • Marketing (AREA)
  • Development Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Computational Linguistics (AREA)
  • Game Theory and Decision Science (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Educational Administration (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Geophysics And Detection Of Objects (AREA)

Abstract

本发明公开了一种基于人工智能的场地地震液化灾害快速评估方法。建立历史地震与场地信息数据库，数据库包括依次连接的需求输入模块、网络爬虫模块、数据处理模块和数据库模块；神经网络模型预测获得震后场地卓越频率；基于震后场地卓越频率获得场地震害程度与抗震性能参数。本发明解决了快速评估震后场地震害、场地抗震性能参数的问题，可快速评估在给定地震条件下，场地液化或软化震害程度和场地抗震性能参数。

Description

[根据细则37.2由ISA制定的发明名称]　一种基于人工智能的场地地震液化灾害快速评估方法 技术领域
本发明属于地震工程地质灾害防治和计算机技术应用领域的一种场地地震液化灾害快速评估方法，具体为一种基于人工智能的场地地震液化灾害快速评估方法。
背景技术
随着人口和财富越来越向城市集中，在地震灾害作用下的城市建筑，不仅需要考虑地震灾害下建筑物的强度是否满足要求，以保证人民的生命安全，还需要考虑地震灾害造成的建筑物使用功能的中断所造成的影响，抗震韧性设计的需求应运而生，即对于一些重要建筑物如医院、能源、通信、交通系统的控制中心，这些设施功能的损坏会导致社会功能的大规模瘫痪，需要估测其在灾害后能够恢复其原有功能所需要的时间和耗费的资源。
强震会使场地的抗震性能迅速下降，并随着时间逐渐恢复到先前的水平，如图2所示，地震发生的时间和大小难以预测，由于客观需求贸然使用建筑物，有可能在小于最初设计标准地震的作用下，地基产生较大的破坏，危及建筑物中的生命财产安全。
当前对于场地地震液化灾害的评估多是采用现场调查，或是根据许多地方的历史地震的调查来建立经验模型以评估震害，现场调查的缺点是事后进行调查，经验方法的缺点是会使得数据库中的数据点包含多种多样的土层性质，所评估的结果往往是最不利的情况，通常来说偏于保守。而且采用经验方法来评估震害不会考虑强震以后场地性质的改变，只会以已有的现场调查结果作为依据来进行现场震害的评估。
强震台站可以实时监测所在场地的情况，这些信息不仅反映了地震的性质，而且还反映了场地的性质，以强震台站的搜集到的信息来作为震害评估的补充，能够更好地反映场地性质的实时变化情况。
发明内容
为解决传统方法评估场地性能时，未考虑地震对场地造成性能的下降以及场地性能随着时间逐渐恢复，以震前现场测试参数作为判断场地各方面性能是否满足要求的依据，本发明提出了一种基于人工智能算法的场地地震液化(软 化)灾害快速评估方法，可以快速评估震后或震后一段时间内，场地各方面的性能参数，可以快速评估在给定地震条件下场地震害程度及较为全面的场地服役性能。
如图1所示，本发明所采用的技术方案是：
S1：建立历史地震与场地信息数据库，其中包含有场地卓越频率；
S2：根据S1的结果经神经网络模型预测获得震后场地卓越频率；
S3：基于震后场地卓越频率获得场地震害程度与抗震性能参数。
所述S1中，建立历史地震与场地信息数据库包括依次连接的需求输入模块、网络爬虫模块、数据处理模块和数据库模块；
所述需求输入模块，接收输入的强震台站信息和地震动信号搜集时间范围进而发送到网络爬虫模块；
所述的强震台站信息是具体的强震台站编号或者给定经纬度范围进而给出范围区域内的强震台站。
地震动信号搜集时间范围内至少发生4次或4次以上大于6.0级的大地震，且两次大震时间需要间隔不小于1个月，若输入时间范围没有达到要求，则自动进行补齐到满足要求的时间范围。
所述网络爬虫模块，接收来自需求输入模块的强震台站信息和地震动信号搜集时间范围，根据强震台站编号和地震动信号搜集时间范围，在强震数据库网站上使用网络爬虫爬取强震台站编号对应的强震台站及其强震台站下地震动信号搜集时间范围内的所有公开的地震动信号、地震基本信息和场地信息，制作地震基本信息汇总表；将地震动信号、地震基本信息和场地信息并发送到数据处理模块，将地震基本信息汇总表发送到数据库模块；
所述地震动信号是指包含两个水平方向和一个竖直方向的三个正交方向上的加速度时程信号，每一次地震对应于一个强震台站可能会有零个或一个地震动信号。
所述地震基本信息是指一段时间内所有地震动信号对应的地震动编号、地震时间、震级、震源位置、震中距、震源深度、震源相对于强震台站的方位角等，所述地震动编号是由强震台站编号和地震发生时间组成。
所述场地信息包括强震台站所在地的土层分布、土性和地形参数。
所述的数据处理模块，接收来自网络爬虫模块的地震动信号进行分析处理，得到两个水平方向上加速度时程信号的主频、三种地震动强度参数和强震台站所在地的场地卓越频率，整理成地震补充信息汇总表并发送到数据库模块；
所述的数据库模块，接收来自网络爬虫模块的地震基本信息汇总表以及来 自数据处理模块的地震补充信息汇总表根据地震动信号进行合并，得到地震信息总表，再将所有强震台站的地震信息总表按照不同的强震台站进行拆分为多个子表，每个强震台站对应于一个子表，每个子表中的信息按时间先后顺序进行排列，得到单强震台站地震信息汇总表，将强震台站所在场地的场地信息对应每个单强震台站地震信息汇总表。
所述的数据处理模块中，根据地震动信号按照以下方式处理获得强震台站所在地的场地卓越频率：
计算过程按照以下步骤：
S1.1、将地震动信号按窗口长度分为n份，对n个窗口进行横向纵向谱比分析，得到地震动信号的n个横向纵向谱比谱，横向纵向谱比谱的横坐标是频率，纵坐标是横向纵向谱比值，每个窗口的横向纵向谱比谱中以横向纵向谱比最大值对应的频率作为该窗口的卓越频率；
S1.2、每个窗口的横向纵向谱比谱中，将横向纵向谱比谱的横坐标的频率不变，纵坐标的横向纵向谱比值取平均值，进行曲线平滑得到横向纵向谱比谱所对应窗口的中值曲线，取中值曲线的幅值最大对应的频率作为窗口的参考频率f t,m,0；
S1.3、将所有窗口的卓越频率构建第一卓越频率集合X 0＝[f t,1,f t,2,···，f t,n]，f t,n表示t时刻下第n个窗口的卓越频率，计算第一卓越频率集合X 0的均值
Figure PCTCN2022089357-appb-000001
和标准差σ ft,0，进而计算偏移度
Figure PCTCN2022089357-appb-000002
S1.4、第一卓越频率集合X 0中，将卓越频率落在
Figure PCTCN2022089357-appb-000003
以外的窗口删掉，将剩余的m个窗口的横向纵向谱比谱均取平均值并进行曲线平滑，得到m个窗口各自的中值曲线卓越频率f t,m,1；
S1.5、将剩余的m个窗口的卓越频率构建第二卓越频率集合X 1＝[f t,1,f t,2,···，f t,m]中，计算第二卓越频率集合X 1的均值
Figure PCTCN2022089357-appb-000004
和标准差σ ft,1，进而计算偏移度
Figure PCTCN2022089357-appb-000005
S1.6、以删除窗口前后的第一卓越频率集合X 0和第二卓越频率集合X 1的标准差之差的绝对值为第一偏移参数ε 1，即ε 1＝|σ ft,1-σ ft,0|；以删除窗口前后的第一卓越频率集合X 0和第二卓越频率集合X 1的偏移度之差的绝对值为第二偏移参数ε 2，即ε 2＝|y 1-y 0|/y 0；
S1.7、根据第一偏移参数ε 1和第二偏移参数ε 2进行判断：
当第一偏移参数ε 1和第二偏移参数ε 2均大于等于0.01时，以剩余的m个窗口作为所有窗口回到S1.3进行处理，进行下一次循环迭代；
当第一偏移参数ε 1和第二偏移参数ε 2均小于0.01时，以最后一次循环迭 代获得的中值曲线卓越频率作为地震动信号的场地卓越频率。
所述S2中，包括：
建立神经网络的卓越频率预测模型，建立门控循环单元模型，用来处理单强震台站地震信息汇总表；
对卓越频率预测模型进行训练、测试以及保存，将单强震台站地震信息汇总表中的地震动信息、地震动强度、水平方向上加速度主频以及场地信息作为模型的输入，场地卓越频率作为模型的输出，输入卓越频率预测模型中进行训练，训练后测试保存；
卓越频率预测模型的预测，将需要预测的地震动信息、地震动强度指标、水平方向上加速度主频以及场地信息输入到训练测试保存后的卓越频率预测模型输出得到震后场地卓越频率。
所述S3中，包括：
所述基于场地平均剪切波速获得场地震害程度和抗震性能参数，场地震害程度是指地基的沉降、侧移和孔压上升程度，场地抗震性能参数是指场地的抗液化强度和反应谱曲线，以及桩基承载力的折减。
S3.1、由强震台站所在地的震后场地卓越频率进行计算获得场地平均剪切波速；
S3.2、计算强震台站所在地的场地的孔压上升程度；
S3.3、根据场地平均剪切波速和场地的土层厚度处理得到场地类别和场地反应谱曲线；
S3.4、计算震后场地的剪切波速沿深度分布，根据剪切波速沿深度分布处理获得各土层的抗液化强度和抗液化安全系数；
S3.5、根据抗液化安全系数处理获得桩基承载力折减量；
S3.6、根据抗液化安全系数计算最大剪应变，根据最大剪应变再进行计算获得对于水平场地的地震沉降量或者对于倾斜场地的地震侧移量。
所述S3.1中，场地平均剪切波速由以下公式获得：
V savg＝4Hf
式中，V savg——场地平均剪切波速，H——场地土层厚度，f——S2预测获得的震后场地卓越频率。
所述S3.2中，按照以下公式获得场地的超静孔压比作为孔压上升程度：
Figure PCTCN2022089357-appb-000006
式中，r u(t)——超静孔压比，r u＝u/σ’ v0，u为超静孔隙水压力，σ’ v0为有效上覆应力，f 0(t)——强震后场地卓越频率，f 0,t＝0——震前最初的场地卓越频率。
所述S3.3中，通过土层厚度和场地平均剪切波速确定场地类别，再给出场地反应谱曲线。
所述S3.4中，根据下式获得平均剪切波速沿深度分布：
Figure PCTCN2022089357-appb-000007
式中，V s震后——震后剪切波速的分布情况，V s震前——震前剪切波速的分布情况，V savg震后——震后场地的平均剪切波速，V savg震前——震前场地平均剪切波速；本发明中认为震前震后场地剪切波速分布相似。
然后根据平均剪切波速沿深度分布分别计算不同土层的抗液化强度CRR和抗液化安全系数：
由下式来获得场地各土层的抗液化强度：
Figure PCTCN2022089357-appb-000008
Figure PCTCN2022089357-appb-000009
式中，CRR——土层的抗液化强度，γc——地震多向振动作用折减系数，P a——归一化的参考压力，取100kpa，k N——土体的强度与刚度之间的相关系数，由试验确定，ρ——土体的密度，F(e min)——只与最小孔隙比相关的函数，V s1——经过上覆有效应力修正后的剪切波速，K 0——土层的侧压力系数，对于正常固结土取0.5，P a＝100kpa，σ’ v0——土层的上覆有效应力；V s1表示考虑上覆有效应力修正的归一化剪切波速；
然后根据抗液化强度由下式来计算抗液化安全系数FS：
Figure PCTCN2022089357-appb-000010
Figure PCTCN2022089357-appb-000011
式中，CSR——地震引起的循环剪应力比，a max——地表峰值水平加速度，g——重力加速度，σ v0——土层的上覆总应力，σ’ v0——土层的上覆有效应力，γ d——剪应力沿深度的折减系数。
所述S3.5中，根据抗液化安全系数，确定桩基承载力的折减量。
所述S3.6中，由以下式处理根据抗液化安全系数获得最大剪应变γ max：
Figure PCTCN2022089357-appb-000012
Figure PCTCN2022089357-appb-000013
其中，F a表示液化安全系数参考值；
然后判断场地地形特征：如果是倾斜场地，根据最大剪应变处理获得地震侧移量；如果是水平场地，根据最大剪应变处理获得地震沉降量：
由下式确定地震侧移量：
Figure PCTCN2022089357-appb-000014
式中，LD——地震侧移量，C h——与当地土性相关的参数，由过去地震中观察到当地的地面沉降数据获得：z max表示倾斜土层的厚度；
由下式确定地震沉降量：
Figure PCTCN2022089357-appb-000015
Figure PCTCN2022089357-appb-000016
式中，Δe——地基土孔隙比改变量；e 0——土层初始孔隙比；e max——土层最大孔隙比；e min——土层最小孔隙比；S——总沉降量；H i——第i土层的厚度；R 0*表示残余体应变参考值，是一个常数，m表示最大剪应变与残余体变的幂次相关系数，是一个常数。
由此根据抗液化安全系数沿土层深度的分布，结合公式获得强震下的最大剪应变γ max。
本发明方法过程由python编程自动实现向其中输入对应强震台站对应的编号和场地卓越频率，如图3所示，可自动进行各个子步骤，给出场地震害程度与抗震性能。
本发明包括建立历史地震与场地信息数据库，建立神经网络模型来预测震后场地卓越频率，基于场地平均剪切波速获得场地震害程度、场地抗震性能参数。
本发明的有益效果是：
本发明解决了快速评估震后场地震害、场地抗震性能参数的问题，可快速 评估在给定地震条件下，场地液化或软化震害程度和场地抗震性能参数。
本发明采用强震台站接收到的地震动信号为基础，并以此为基础，实时反映场地的性质。
并且提出了一种可以没有人为干涉，自动进行横向竖向谱比分析，由三个方向上的地震动信号获得场地卓越频率的方法。传统横向竖向谱比分析中需要人为选取尾波窗口，并且目测选取结果是否合理，导致不同工程人员进行横向竖向谱比分析可能出现的结果不一致，本发明的方案能有效地解决这一问题。
附图说明
图1是本发明方法的技术流程图；
图2是震后场地卓越频率的下降以及恢复示意图；
图3是建立历史地震与场地信息数据库的技术流程图；
图4是基于场地平均剪切波速获得场地震害程度和抗震性能参数的技术流程图；
图5是HVSR方法(横向竖向谱比法)的窗口选取示意图；
图6是横向竖向谱比谱和据其确定场地卓越频率的示意图；
图7是某强震台站整理得到的数据表格示意图；
图8是某场地土层的剖面图，同时包括剪切波速和锥尖阻力随深度的变化情况；
图9是卓越频率预测模型的效果图。
图10是本发明获得场地卓越频率的处理流程图。
图11是本发明自动选择地震动信号尾波窗口的效果图。
图12是本发明自动进行横向竖向谱比分析得到场地卓越频率的效果图。
具体实施方式
下面结合附图和具体实施对本发明作进一步说明。
本发明的实施例的实施过程如下：
S1：建立历史地震与场地信息数据库；
如图3所示，建立历史地震与场地信息数据库包括依次连接的需求输入模块、网络爬虫模块、数据处理模块和数据库模块；具体实施中，这个过程采用python编程实现，在需求输入模块中输入需求数据后，即可自动进行之后的所有步骤，并整理成便于阅读和人工智能算法处理的格式。
需求输入模块，接收输入的强震台站信息和地震动信号搜集时间范围进而发送到网络爬虫模块；
网络爬虫模块，接收来自需求输入模块的强震台站信息和地震动信号搜集时间范围，根据强震台站编号和地震动信号搜集时间范围，在世界各地公开的强震数据库网站上使用网络爬虫爬取强震台站编号对应的强震台站及其强震台站下地震动信号搜集时间范围内的所有公开的地震动信号、地震基本信息和场地信息，主要由地震基本信息制作地震基本信息汇总表；将地震动信号、地震基本信息和场地信息并发送到数据处理模块，将地震基本信息汇总表发送到数据库模块；
场地两个水平方向上加速度信号的主频由快速傅里叶变换得到两个水平方向上的傅氏谱，幅值最大的点对应的频率即为加速度的主频。
获取强震台站所在场地的主频信号是指，对地震动信号进行横向竖向谱比(HVSR)分析，对地震动信号的尾波部分(尾波信号是指加速度幅值明显小于主震的部分)取两至三个没有重合部分的窗口，对每个窗口进行以下操作，在水平方向上的两个分量进行傅里叶变换并将这两个谱幅值的平方的和进行相加再开根号作为水平加速度谱H，将竖直方向上的加速度信号进行傅里叶变换作为竖直加速度谱V，再将水平加速度谱与竖向加速度谱相除得到横向竖向谱比谱H/V，H/V最大值对应的横坐标作为场地的卓越频率，水平加速度谱H、竖直加速度谱V和横向竖向谱比谱H/V的横坐标均为频率，纵坐标为该频率对应的幅值。对加速度时程信号进行尾波选取的示意图如图5所示。
数据处理模块，接收来自网络爬虫模块的地震动信号进行分析处理，得到两个水平方向上加速度时程信号的主频、三种地震动强度参数和强震台站所在地的场地卓越频率，整理成地震补充信息汇总表并发送到数据库模块；
三种地震动强度参数包括了阿里亚斯强度Ia、两种绝对累计加速度CAV和CAV5。
获取强震台站所在场地的卓越频率是指，对地震动信号进行横向竖向谱比(HVSR)分析，对地震动信号的尾波部分(尾波信号是指加速度幅值明显小于主震的部分)取两至三个连续且没有重合部分的窗口，对每个窗口进行以下操作：在水平方向上的两个分量进行傅里叶变换并将这两个谱幅值的平方的和进行相加再开根号作为水平加速度谱H，将竖直方向上的加速度信号进行傅里叶变换作为竖直加速度谱V，再将水平加速度谱与竖向加速度谱相除得到横向竖向谱比谱H/V。
水平加速度谱H、竖直加速度谱V和横向竖向谱比谱H/V的横坐标均为频率，纵坐标为该频率对应的幅值。
对于这两三个窗口幅值最高点对应频率差距较大的结果需要进行重新分析 和计算，对于合格的结果，由这两三个窗口确定横向纵向谱比谱的上界下界以及均值，得到的幅值最大点对应的频率作为场地的卓越频率。
当前对于横向纵向谱比法的操作往往是依赖人工选取窗口，使用目测法对HVSR方法得到的结果进行校验，不合格的进行重新分析，这种方法虽在几十年的实践中被广泛采用，但每个信号的窗口都需要人工进行选取和校核，费时费力。
本发明提出的场地卓越频率获得方法可以自动选取未被污染时间窗口，删除被污染的时间窗口，不需人为干涉，输入地震动信号，即可直接获得场地卓越频率的办法，其流程图如图10所示。
数据处理模块中，根据地震动信号按照以下方式处理获得强震台站所在地的场地卓越频率：
采用基于横向纵向谱比法(HVSR方法)的方法进行场地卓越频率自动计算，计算过程按照以下步骤：
S1.1、将地震动信号按窗口长度分为n份，具体实施的窗口长度选取4秒，对n个窗口进行横向纵向谱比(HVSR)分析，得到地震动信号的n个横向纵向谱比谱(H/V谱)，一个窗口进行HVSR分析得到一个横向纵向谱比谱，横向纵向谱比谱的横坐标是频率，纵坐标是横向纵向谱比值，每个窗口的横向纵向谱比谱中以横向纵向谱比最大值对应的频率作为该窗口的卓越频率；
具体实施选取采样周期的50倍作为窗口长度的最小值。
S1.2、每个窗口的横向纵向谱比谱中，将横向纵向谱比谱的横坐标的频率不变，纵坐标的横向纵向谱比值取平均值，进行曲线平滑得到横向纵向谱比谱所对应窗口的中值曲线，即中值HVSR曲线，取中值曲线的幅值最大对应的频率作为窗口的参考频率f t,m,0；
S1.3、将所有窗口的卓越频率构建第一卓越频率集合X 0＝[f t,1,f t,2,···，f t,n]，f t,n表示t时刻下第n个窗口的卓越频率，计算第一卓越频率集合X 0的均值
Figure PCTCN2022089357-appb-000017
和标准差σ ft,0，进而计算偏移度
Figure PCTCN2022089357-appb-000018
S1.4、第一卓越频率集合X 0中，将卓越频率落在
Figure PCTCN2022089357-appb-000019
以外的窗口删掉，在地震动信号中将受噪声污染的窗口删掉，将剩余的m个窗口的横向纵向谱比谱均取平均值并进行曲线平滑，得到m个窗口各自的中值曲线卓越频率f t,m,1；
S1.5、将剩余的m个窗口的卓越频率构建第二卓越频率集合X 1＝[f t,1,f t,2,···，f t,m]中，计算第二卓越频率集合X 1的均值
Figure PCTCN2022089357-appb-000020
和标准差σ ft,1，进而计算偏移度
Figure PCTCN2022089357-appb-000021
S1.6、以删除窗口前后的第一卓越频率集合X 0和第二卓越频率集合X 1的标准差之差的绝对值为第一偏移参数ε 1，即ε 1＝|σ ft,1-σ ft,0|；以删除窗口前后的第一卓越频率集合X 0和第二卓越频率集合X 1的偏移度之差的绝对值为第二偏移参数ε 2，即ε 2＝|y 1-y 0|/y 0；
S1.7、根据第一偏移参数ε 1和第二偏移参数ε 2进行判断：
当第一偏移参数ε 1和第二偏移参数ε 2均大于等于0.01时，以剩余的m个窗口作为所有窗口回到S1.3进行处理，进行下一次循环迭代；
当第一偏移参数ε 1和第二偏移参数ε 2均小于0.01时，以最后一次循环迭代获得的中值曲线卓越频率作为地震动信号的场地卓越频率。
图11是某地震动信号三个方向上加速度时程信号的示意图，窗口选取的最终结果是图中浅色框所示，一共选取了16个窗口，删掉了12个窗口，最终留下了4个窗口。
图12是该地震动信号的分析结果，左图是未进行窗口删除时的最初HVSR分析结果，右图是删除了受噪声干扰的窗口后的HVSR分析结果，可以看到，最初HVSR分析所得到的频率较为分散，标准差(图中的灰色阴影部分范围也比较大)，在程序自动删除一些窗口后，得到的卓越频率较为集中，能够满足实际使用的需要，该信号未执行窗口删除得到的中值曲线卓越频率为1.83赫兹，删除窗口后得到中值曲线的卓越频率为1.74赫兹，取最后分析结果1.74赫兹作为场地的卓越频率。
数据库模块，接收来自网络爬虫模块的地震基本信息汇总表以及来自数据处理模块的地震补充信息汇总表根据地震动信号进行合并，得到地震信息总表，再将所有强震台站的地震信息总表按照不同的强震台站进行拆分为多个子表，每个强震台站对应于一个子表，每个子表中的信息按时间先后顺序进行排列形成时间序列信息，得到单强震台站地震信息汇总表，将强震台站所在场地的场地信息对应每个单强震台站地震信息汇总表，如图7所示。
S2：神经网络模型预测获得震后场地卓越频率；
建立神经网络的卓越频率预测模型，基于神经网络算法建立门控循环单元(GRU)模型，用来处理单强震台站地震信息汇总表的时间序列信息；
对卓越频率预测模型进行训练、测试以及保存，将单强震台站地震信息汇总表中的需要预测的地震动信息、地震动强度、水平方向上加速度主频以及场地信息输入卓越频率预测模型中进行训练，训练后测试保存；
卓越频率预测模型的预测，将需要预测的地震动信息、地震动强度指标、水平方向上加速度主频以及场地信息参数输入到训练测试保存后的卓越频率预 测模型输出得到震后场地卓越频率。
具体实施中，将单强震台站地震信息汇总表中场地卓越频率作为模型的输出，其他部分作为模型的输入，并划分为训练集和测试集，测试集需要包含一次大震以后卓越频率的下降和恢复，测试集数据个数是训练集数据个数的1/3～1/4。训练中，将模型的平均绝对误差作为损失函数，以测试集训练损失函数最小作为模型的目标，对模型进行贝叶斯超参数调节，最后将训练损失最小的模型进行保存，以便使用。
平均绝对误差越小说明模型预测值与真实值之间越接近，模型的表现越好。超参数是指例如GRU神经网络中的隐藏层层数，隐藏层单元个数，学习率，dropout比率等参数。超参数优化的过程采用参数超参数选择方式有经验确定，例如随机搜索，网格搜索和贝叶斯参数调整。贝叶斯优化方法是通过假设超参数和对应的目标函数的概率分布形式，再根据一次次试验的结果的优劣来调整超参数和目标函数的概率分布的相关参数，进而调整超参数。
这个过程采用python编程实现，在接收到单强震台站地震信息汇总表，并向模型中输入需要预测的地震信息之后，即可自动进行接下来的步骤，并输出场地的卓越频率。
S3：基于震后场地卓越频率获得场地震害程度与抗震性能参数，如图4所示，完成场地地震液化灾害的快速评估。
S3.1、由强震台站所在地的震后场地卓越频率进行计算获得场地平均剪切波速；
场地平均剪切波速由以下公式获得：
V savg＝4Hf
式中，V savg——场地平均剪切波速，H——场地土层厚度，f——S2预测获得的震后场地卓越频率。
S3.2、计算强震台站所在地的场地的孔压上升程度；
按照以下公式获得场地的超静孔压比作为孔压上升程度：
Figure PCTCN2022089357-appb-000022
S3.3、根据场地平均剪切波速和场地的土层厚度处理得到场地类别和场地反应谱曲线；
S3.4、计算震后场地的剪切波速沿深度分布，根据剪切波速沿深度分布处理获得各土层的抗液化强度CRR和抗液化安全系数；
根据下式获得平均剪切波速沿深度分布：
Figure PCTCN2022089357-appb-000023
然后根据平均剪切波速沿深度分布分别计算不同土层的抗液化强度CRR和抗液化安全系数：
由下式来获得场地各土层的抗液化强度：
Figure PCTCN2022089357-appb-000024
Figure PCTCN2022089357-appb-000025
然后根据抗液化强度由下式来计算抗液化安全系数FS：
Figure PCTCN2022089357-appb-000026
Figure PCTCN2022089357-appb-000027
S3.5、根据抗液化安全系数处理获得桩基承载力折减量；
S3.6、根据抗液化安全系数计算最大剪应变，根据最大剪应变再进行不同公式分别计算获得对于水平场地的地震沉降量或者对于倾斜场地的地震侧移量。
由以下式处理根据抗液化安全系数获得最大剪应变γ max：
Figure PCTCN2022089357-appb-000028
Figure PCTCN2022089357-appb-000029
然后判断场地地形特征：如果是倾斜场地，根据最大剪应变处理获得地震侧移量；如果是水平场地，根据最大剪应变处理获得地震沉降量：
由下式确定地震侧移量：
Figure PCTCN2022089357-appb-000030
由下式确定地震沉降量：
Figure PCTCN2022089357-appb-000031
Figure PCTCN2022089357-appb-000032
实例情况：
搜集某强震台站进行从2010年到2020年的地震动信号，该强震台站所在区域的土层剖面信息如图8所示，其上覆软弱土层厚度为20m。
对每一个地震动信号进行分析，获得每个地震动信号对应的水平分量的分量的主频，地震动强度参数，用HVSR方法分析得到场地的卓越频率，再搜集每个地震动信号对应的地震动信息，包括地震时间、震级、震源深度、震源距离，震源相对于强震台站的方位角，并将其整理搜集到表格之中，如图7所示。
将数据分为训练集和测试集，并将场地卓越频率作为模型的输出，将其他数据作为模型的输入。
建立GRU神经网络模型，以在训练集上损失最低为目标，进行贝叶斯超参数优化。最后得到模型在训练集和测试集上的表现如图9所示，圆点是实际的场地卓越频率数据，方点是在模型在训练集上的预测值，三角形的点是模型在测试集上的预测值。
输入一个在模型最后一次收到地震动信号之后较大的地震动信号地震发生时间设为最后一次接收到地震动信号后的1天，震级设为6.5，震源深度和距离都设为10，水平峰值加速度1设为2000，水平峰值加速度2设为2000，竖直峰值加速度设为1000，方位角设为250，绝对累计加速度设为25，阿里亚斯强度设为5，cav5设为20，水平方向加速度信号1的主频设为2，水平方向加速度信号2的主频设为2，输入保存好的模型之中，得到场地的卓越频率为1.36914163hz。
实际测得2020年最后一次地震的场地卓越频率为2.317hz，在经受上述地震作用下，由GRU神经网络模型得到场地的卓越频率为1.36914163hz。
计算场地的平均剪切波速。
强震前：
V savg1＝4Hf＝4×20×2.317＝185.36m/s
强震后：
V savg2＝4Hf＝4×20×1.369＝109.52m/s
结合建筑抗震设计规范可知场地由Ⅱ类场地变为了Ⅲ类场地。
由此实施可见，本发明能够快速评估在给定地震条件下场地性能的变化情况，解决了以往震害评估中，只考虑土体静态抗震性能的问题。

Claims (9)

  1. 一种基于人工智能的场地地震液化灾害快速评估方法，其特征在于包括以下步骤：
    S1：建立历史地震与场地信息数据库；
    S2：根据S1的结果经神经网络模型预测获得震后场地卓越频率；
    S3：基于震后场地卓越频率获得场地震害程度与抗震性能参数。
  2. 根据权利要求1所述一种基于人工智能的场地地震液化灾害快速评估估测方法，其特征在于：所述S1中，建立历史地震与场地信息数据库包括依次连接的需求输入模块、网络爬虫模块、数据处理模块和数据库模块；
    所述需求输入模块，接收输入的强震台站信息和地震动信号搜集时间范围进而发送到网络爬虫模块；
    所述网络爬虫模块，接收来自需求输入模块的强震台站信息和地震动信号搜集时间范围，根据强震台站编号和地震动信号搜集时间范围，在强震数据库网站上使用网络爬虫爬取强震台站编号对应的强震台站及其强震台站下地震动信号搜集时间范围内的所有公开的地震动信号、地震基本信息和场地信息，制作地震基本信息汇总表；将地震动信号、地震基本信息和场地信息并发送到数据处理模块，将地震基本信息汇总表发送到数据库模块；
    所述的数据处理模块，接收来自网络爬虫模块的地震动信号进行分析处理，得到两个水平方向上加速度时程信号的主频、三种地震动强度参数和强震台站所在地的场地卓越频率，整理成地震补充信息汇总表并发送到数据库模块；
    所述的数据库模块，接收来自网络爬虫模块的地震基本信息汇总表以及来自数据处理模块的地震补充信息汇总表根据地震动信号进行合并，得到地震信息总表，再将所有强震台站的地震信息总表按照不同的强震台站进行拆分为多个子表，每个强震台站对应于一个子表，每个子表中的信息按时间先后顺序进行排列，得到单强震台站地震信息汇总表，将强震台站所在场地的场地信息对应每个单强震台站地震信息汇总表。
  3. 根据权利要求2所述一种基于人工智能的场地地震液化灾害快速评估估测方法，其特征在于：
    所述的数据处理模块中，根据地震动信号按照以下方式处理获得强震台站所在地的场地卓越频率：
    计算过程按照以下步骤：
    S1.1、将地震动信号按窗口长度分为n份，对n个窗口进行横向纵向谱比分 析，得到地震动信号的n个横向纵向谱比谱，横向纵向谱比谱的横坐标是频率，纵坐标是横向纵向谱比值，每个窗口的横向纵向谱比谱中以横向纵向谱比最大值对应的频率作为该窗口的卓越频率；
    S1.2、每个窗口的横向纵向谱比谱中，将横向纵向谱比谱的横坐标的频率不变，纵坐标的横向纵向谱比值取平均值，进行曲线平滑得到横向纵向谱比谱所对应窗口的中值曲线，取中值曲线的幅值最大对应的频率作为窗口的参考频率f t,m,0；
    S1.3、将所有窗口的卓越频率构建第一卓越频率集合X 0＝[f t,1,f t,2,···，f t,n]，f t,n表示t时刻下第n个窗口的卓越频率，计算第一卓越频率集合X 0的均值
    Figure PCTCN2022089357-appb-100001
    和标准差σ ft,0，进而计算偏移度
    Figure PCTCN2022089357-appb-100002
    S1.4、第一卓越频率集合X 0中，将卓越频率落在
    Figure PCTCN2022089357-appb-100003
    以外的窗口删掉，将剩余的m个窗口的横向纵向谱比谱均取平均值并进行曲线平滑，得到m个窗口各自的中值曲线卓越频率f t,m,1；
    S1.5、将剩余的m个窗口的卓越频率构建第二卓越频率集合X 1＝[f t,1,f t,2,···，f t,m]中，计算第二卓越频率集合X 1的均值
    Figure PCTCN2022089357-appb-100004
    和标准差σ ft,1，进而计算偏移度
    Figure PCTCN2022089357-appb-100005
    S1.6、以删除窗口前后的第一卓越频率集合X 0和第二卓越频率集合X 1的标准差之差的绝对值为第一偏移参数ε 1，即ε 1＝|σ ft,1-σ ft,0|；以删除窗口前后的第一卓越频率集合X 0和第二卓越频率集合X 1的偏移度之差的绝对值为第二偏移参数ε 2，即ε 2＝|y 1-y 0|/y 0；
    S1.7、根据第一偏移参数ε 1和第二偏移参数ε 2进行判断：
    当第一偏移参数ε 1和第二偏移参数ε 2均大于等于0.01时，以剩余的m个窗口作为所有窗口回到S1.3进行处理，进行下一次循环迭代；
    当第一偏移参数ε 1和第二偏移参数ε 2均小于0.01时，以最后一次循环迭代获得的中值曲线卓越频率作为地震动信号的场地卓越频率。
  4. 根据权利要求1所述一种基于人工智能的场地地震液化灾害快速评估方法，其特征在于：所述S2中，包括：
    建立神经网络的卓越频率预测模型，建立门控循环单元模型，用来处理单强震台站地震信息汇总表；
    对卓越频率预测模型进行训练、测试以及保存，将单强震台站地震信息汇总表中的地震动信息、地震动强度、水平方向上加速度主频以及场地信息作为模型的输入，场地卓越频率作为模型的输出，输入卓越频率预测模型中进行训练，训练后测试保存；
    卓越频率预测模型的预测，将需要预测的地震动信息、地震动强度指标、水平方向上加速度主频以及场地信息输入到训练测试保存后的卓越频率预测模型输出得到震后场地卓越频率。
  5. 根据权利要求1所述一种基于人工智能的场地地震液化灾害快速评估方法，其特征在于：所述S3中，包括：
    S3.1、由强震台站所在地的震后场地卓越频率进行计算获得场地平均剪切波速；
    S3.2、计算强震台站所在地的场地的孔压上升程度；
    S3.3、根据场地平均剪切波速和场地的土层厚度处理得到场地类别和场地反应谱曲线；
    S3.4、计算震后场地的剪切波速沿深度分布，根据剪切波速沿深度分布处理获得各土层的抗液化强度和抗液化安全系数；
    S3.5、根据抗液化安全系数处理获得桩基承载力折减量；
    S3.6、根据抗液化安全系数计算最大剪应变，根据最大剪应变再进行计算获得对于水平场地的地震沉降量或者对于倾斜场地的地震侧移量。
  6. 根据权利要求4所述一种基于人工智能的场地地震液化灾害快速评估方法，其特征在于：所述S3.1中，场地平均剪切波速由以下公式获得：
    V savg＝4Hf
    式中，V savg——场地平均剪切波速，H——场地土层厚度，f——S2预测获得的震后场地卓越频率。
  7. 根据权利要求4所述一种基于人工智能的场地地震液化灾害快速评估方法，其特征在于：所述S3.2中，按照以下公式获得场地的超静孔压比作为孔压上升程度：
    Figure PCTCN2022089357-appb-100006
    式中，r u(t)——超静孔压比，r u＝u/σ’ v0，u为超静孔隙水压力，σ’ v0为有效上覆应力，f 0(t)——强震后场地卓越频率，f 0,t＝0——震前最初的场地卓越频率。
  8. 根据权利要求5所述一种基于人工智能的场地地震液化灾害快速评估方法，其特征在于：所述S3.4中，根据下式获得平均剪切波速沿深度分布：
    Figure PCTCN2022089357-appb-100007
    式中，V s震后——震后剪切波速的分布情况，V s震前——震前剪切波速的分布 情况，V savg震后——震后场地的平均剪切波速，V savg震前——震前场地平均剪切波速；
    然后根据平均剪切波速沿深度分布分别计算不同土层的抗液化强度CRR和抗液化安全系数：
    由下式来获得场地各土层的抗液化强度：
    Figure PCTCN2022089357-appb-100008
    Figure PCTCN2022089357-appb-100009
    式中，CRR——土层的抗液化强度，γc——地震多向振动作用折减系数，P a——归一化的参考压力，k N——土体的强度与刚度之间的相关系数，由试验确定，ρ——土体的密度，F(e min)——只与最小孔隙比相关的函数，V s1——经过上覆有效应力修正后的剪切波速，K 0——土层的侧压力系数，σ’ v0——土层的上覆有效应力；V s1表示考虑上覆有效应力修正的归一化剪切波速；
    然后根据抗液化强度由下式来计算抗液化安全系数FS：
    Figure PCTCN2022089357-appb-100010
    Figure PCTCN2022089357-appb-100011
    式中，CSR——地震引起的循环剪应力比，a max——地表峰值水平加速度，g——重力加速度，σ v0——土层的上覆总应力，σ’ v0——土层的上覆有效应力，γ d——剪应力沿深度的折减系数。
  9. 根据权利要求5所述一种基于人工智能的场地地震液化灾害快速评估方法，其特征在于：所述S3.6中，由以下式处理根据抗液化安全系数获得最大剪应变γ max：
    Figure PCTCN2022089357-appb-100012
    Figure PCTCN2022089357-appb-100013
    其中，F a表示液化安全系数参考值，V s1表示经过上覆有效应力修正后的剪切波速，FS表示抗液化安全系数；
    然后判断场地地形特征：如果是倾斜场地，根据最大剪应变处理获得地震侧移量；如果是水平场地，根据最大剪应变处理获得地震沉降量：
    由下式确定地震侧移量：
    Figure PCTCN2022089357-appb-100014
    式中，LD——地震侧移量，C h——与当地土性相关的参数：z max表示倾斜土层的厚度；
    由下式确定地震沉降量：
    Figure PCTCN2022089357-appb-100015
    Figure PCTCN2022089357-appb-100016
    式中，Δe——地基土孔隙比改变量；e 0——土层初始孔隙比；e max——土层最大孔隙比；e min——土层最小孔隙比；S——总沉降量；H i——第i土层的厚度；R 0*表示残余体应变参考值，m表示最大剪应变与残余体变的幂次相关系数。
PCT/CN2022/089357 2021-05-21 2022-04-26 一种基于人工智能的场地地震液化灾害快速评估方法 WO2022242435A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110558271.9A CN113379105B (zh) 2021-05-21 2021-05-21 一种基于人工智能的场地地震液化灾害快速评估方法
CN202110558271.9 2021-05-21

Publications (1)

Publication Number Publication Date
WO2022242435A1 true WO2022242435A1 (zh) 2022-11-24

Family

ID=77571546

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/089357 WO2022242435A1 (zh) 2021-05-21 2022-04-26 一种基于人工智能的场地地震液化灾害快速评估方法

Country Status (2)

Country Link
CN (1) CN113379105B (zh)
WO (1) WO2022242435A1 (zh)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115965246A (zh) * 2023-03-16 2023-04-14 山东省地质矿产勘查开发局八〇一水文地质工程地质大队(山东省地矿工程勘察院) 一种岩溶塌陷灾害的预警分析方法
CN116046079A (zh) * 2023-04-03 2023-05-02 山东省地质调查院(山东省自然资源厅矿产勘查技术指导中心) 一种基于位置的地质环境专题数据集成管理系统
CN116186825A (zh) * 2022-11-29 2023-05-30 清华大学 基于图节点分类图神经网络的剪力墙设计方法和装置
CN116433032A (zh) * 2023-04-26 2023-07-14 中国农业科学院农业环境与可持续发展研究所 基于网络爬虫方式的智能评估方法
CN116720352A (zh) * 2023-06-08 2023-09-08 大连理工大学 一种面向长大结构平动-转动六分量多维多点地震动场的人工模拟方法
CN117542153A (zh) * 2024-01-03 2024-02-09 深圳市纳泽科技有限公司 一种基于九轴传感器的入侵检测方法、系统、围栏和设备
CN117574705A (zh) * 2023-11-07 2024-02-20 哈尔滨工业大学 一种考虑反应谱约束的rc框架建筑地震时程响应预测方法
CN117828481A (zh) * 2024-03-04 2024-04-05 烟台哈尔滨工程大学研究院 基于动态集成框架共轨船用燃油系统故障诊断方法及介质

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113379105B (zh) * 2021-05-21 2022-05-31 浙江大学 一种基于人工智能的场地地震液化灾害快速评估方法
CN114741758B (zh) * 2022-04-12 2024-04-05 大连理工大学 一种基于机器学习的建筑抗震韧性初步设计方法及系统

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104361219A (zh) * 2014-10-31 2015-02-18 华北水利水电大学 一种评价建筑物抗震性能的方法
JP2017026569A (ja) * 2015-07-28 2017-02-02 清水建設株式会社 免震部材応答推定装置及び免震部材応答推定方法
CN111458748A (zh) * 2020-03-30 2020-07-28 青岛理工大学 基于三层数据集神经网络的性能地震动危险性分析方法
CN112001565A (zh) * 2020-09-08 2020-11-27 清华大学合肥公共安全研究院 基于Softmax回归模型的地震灾害损失预测与评估的方法及系统
CN112630827A (zh) * 2020-12-07 2021-04-09 中国地震局工程力学研究所 地表加速度峰值参数预测方法及系统
CN113379105A (zh) * 2021-05-21 2021-09-10 浙江大学 一种基于人工智能的场地地震液化灾害快速估测方法

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103164420A (zh) * 2011-12-12 2013-06-19 国家电网公司 地质信息的处理方法及装置
ES2914062T3 (es) * 2014-11-10 2022-06-07 Schreder Método para detectar terremotos y localizar epicentros mediante una red de luces
CN105044776B (zh) * 2015-07-31 2017-08-11 中国能源建设集团江苏省电力设计院有限公司 基于abaqus的土体地基液化研究方法
TWI661214B (zh) * 2016-11-29 2019-06-01 National Applied Research Laboratories 自動化校正地盤特性之現地型地震預警系統及相關方法
CN107067656A (zh) * 2017-01-23 2017-08-18 重庆三峡学院 一种地质灾害监测系统
CN110134682A (zh) * 2019-04-16 2019-08-16 同济大学 基于随机地震动数据库构建的数据交互方法及数据库装置
CN111553107B (zh) * 2020-05-12 2024-05-24 江南大学 可液化场地桩基随机地震反应分析与安全评价方法
CN112508124A (zh) * 2020-12-22 2021-03-16 三峡大学 一种基于贝叶斯网络的砂砾土地震液化判别方法

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104361219A (zh) * 2014-10-31 2015-02-18 华北水利水电大学 一种评价建筑物抗震性能的方法
JP2017026569A (ja) * 2015-07-28 2017-02-02 清水建設株式会社 免震部材応答推定装置及び免震部材応答推定方法
CN111458748A (zh) * 2020-03-30 2020-07-28 青岛理工大学 基于三层数据集神经网络的性能地震动危险性分析方法
CN112001565A (zh) * 2020-09-08 2020-11-27 清华大学合肥公共安全研究院 基于Softmax回归模型的地震灾害损失预测与评估的方法及系统
CN112630827A (zh) * 2020-12-07 2021-04-09 中国地震局工程力学研究所 地表加速度峰值参数预测方法及系统
CN113379105A (zh) * 2021-05-21 2021-09-10 浙江大学 一种基于人工智能的场地地震液化灾害快速估测方法

Cited By (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116186825A (zh) * 2022-11-29 2023-05-30 清华大学 基于图节点分类图神经网络的剪力墙设计方法和装置
CN116186825B (zh) * 2022-11-29 2023-10-31 清华大学 基于图节点分类图神经网络的剪力墙设计方法和装置
CN115965246B (zh) * 2023-03-16 2023-05-19 山东省地质矿产勘查开发局八〇一水文地质工程地质大队(山东省地矿工程勘察院) 一种岩溶塌陷灾害的预警分析方法
CN115965246A (zh) * 2023-03-16 2023-04-14 山东省地质矿产勘查开发局八〇一水文地质工程地质大队(山东省地矿工程勘察院) 一种岩溶塌陷灾害的预警分析方法
CN116046079A (zh) * 2023-04-03 2023-05-02 山东省地质调查院(山东省自然资源厅矿产勘查技术指导中心) 一种基于位置的地质环境专题数据集成管理系统
CN116046079B (zh) * 2023-04-03 2023-06-30 山东省地质调查院(山东省自然资源厅矿产勘查技术指导中心) 一种基于位置的地质环境专题数据集成管理系统
CN116433032B (zh) * 2023-04-26 2024-04-09 中国农业科学院农业环境与可持续发展研究所 基于网络爬虫方式的智能评估方法
CN116433032A (zh) * 2023-04-26 2023-07-14 中国农业科学院农业环境与可持续发展研究所 基于网络爬虫方式的智能评估方法
CN116720352A (zh) * 2023-06-08 2023-09-08 大连理工大学 一种面向长大结构平动-转动六分量多维多点地震动场的人工模拟方法
CN116720352B (zh) * 2023-06-08 2024-01-30 大连理工大学 一种面向长大结构地震动场的人工模拟方法
CN117574705A (zh) * 2023-11-07 2024-02-20 哈尔滨工业大学 一种考虑反应谱约束的rc框架建筑地震时程响应预测方法
CN117542153A (zh) * 2024-01-03 2024-02-09 深圳市纳泽科技有限公司 一种基于九轴传感器的入侵检测方法、系统、围栏和设备
CN117542153B (zh) * 2024-01-03 2024-03-15 深圳市纳泽科技有限公司 一种基于九轴传感器的入侵检测方法、系统、围栏和设备
CN117828481A (zh) * 2024-03-04 2024-04-05 烟台哈尔滨工程大学研究院 基于动态集成框架共轨船用燃油系统故障诊断方法及介质

Also Published As

Publication number Publication date
CN113379105A (zh) 2021-09-10
CN113379105B (zh) 2022-05-31

Similar Documents

Publication Publication Date Title
WO2022242435A1 (zh) 一种基于人工智能的场地地震液化灾害快速评估方法
KR101642951B1 (ko) Ｇｉｓ 기반 실시간 지진피해 예측 방법
van Ballegooy et al. Assessment of various CPT based liquefaction severity index frameworks relative to the Ishihara (1985) H1–H2 boundary curves
Erdik et al. Rapid earthquake hazard and loss assessment for Euro-Mediterranean region
KR100982447B1 (ko) 지공간 상관관계 통합기법을 이용한 산사태 발생 예측시스템 및 이를 이용한 산사태 발생 예측방법
CN112541666B (zh) 考虑地震易损性模型不确定性的盾构隧道风险评估方法
Chang et al. An empirical approach of accounting for the amplification effects induced by near-fault directivity
Palanci et al. A statistical assessment on global drift ratio demands of mid-rise RC buildings using code-compatible real ground motion records
Grimaz et al. The ASSESS project: assessment for seismic risk reduction of school buildings in the Friuli Venezia Giulia region (NE Italy)
CN105427189A (zh) 一种滑坡灾害下的电网易损性评测方法
Yakhchalian et al. Reliable seismic collapse assessment of short-period structures using new proxies for ground motion record selection
CN113536433A (zh) 基于bim平台的建筑灾后疏散动态逃生路线优化系统
Chieffo et al. A Simplified Approach to Estimate Seismic Vulnerability and Damage Scenarios Including Site Effects. Application to the Historical Centre of Horta, Azores, Portugal
Sonmezer et al. Effects of the use of the surface spectrum of a specific region on seismic performances of R/C structures
Pan Damage prediction of low-rise buildings under hurricane winds
CN113887019A (zh) 一种基于灰色关联的基坑整体的预警状态评价方法
Du et al. Methodology for estimating human perception to tremors in high-rise buildings
Du et al. Urban seismic loss estimation by incorporating ground‐motion simulation, site effect, and kriging techniques: An application in Singapore
Sonmezer et al. Seismic risk estimation of the Kirikkale province through street survey based rapid assessment method (SSRA)
CN112946740A (zh) 一种智能化地震震源搜寻定位系统
Kafali System performance under multihazard environment
Ruggieri et al. Using transfer learning technique to define seismic vulnerability of existing buildings through mechanical models
Kongar et al. Evaluating desktop methods for assessing liquefaction-induced damage to infrastructure for the insurance sector
D’Amato et al. Risk analysis of existing building heritage through damage assessment after L’Aquila earthquake 2009
Nikellis Risk-informed decision making for civil infrastructure subjected to single and multiple hazards

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22803759

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 22803759

Country of ref document: EP

Kind code of ref document: A1