CN116821642A - Building earthquake damage rapid assessment method and system based on data augmentation and deep learning - Google Patents
Building earthquake damage rapid assessment method and system based on data augmentation and deep learning Download PDFInfo
- Publication number
- CN116821642A CN116821642A CN202310701891.2A CN202310701891A CN116821642A CN 116821642 A CN116821642 A CN 116821642A CN 202310701891 A CN202310701891 A CN 202310701891A CN 116821642 A CN116821642 A CN 116821642A
- Authority
- CN
- China
- Prior art keywords
- strong vibration
- earthquake
- deep learning
- data
- damage
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
- 230000006378 damage Effects 0.000 title claims abstract description 149
- 238000000034 method Methods 0.000 title claims abstract description 67
- 238000013135 deep learning Methods 0.000 title claims abstract description 55
- 238000013434 data augmentation Methods 0.000 title claims abstract description 33
- 230000033001 locomotion Effects 0.000 claims abstract description 114
- 238000001228 spectrum Methods 0.000 claims abstract description 79
- 230000004044 response Effects 0.000 claims abstract description 49
- 238000010586 diagram Methods 0.000 claims abstract description 27
- 238000012549 training Methods 0.000 claims abstract description 25
- 238000011426 transformation method Methods 0.000 claims abstract description 22
- 238000005259 measurement Methods 0.000 claims abstract description 8
- 238000006243 chemical reaction Methods 0.000 claims description 27
- 238000013527 convolutional neural network Methods 0.000 claims description 20
- 230000009466 transformation Effects 0.000 claims description 19
- 238000013136 deep learning model Methods 0.000 claims description 18
- 238000004458 analytical method Methods 0.000 claims description 11
- 230000009471 action Effects 0.000 claims description 10
- 230000000737 periodic effect Effects 0.000 claims description 7
- 230000008569 process Effects 0.000 claims description 6
- 238000000638 solvent extraction Methods 0.000 claims description 3
- 238000006073 displacement reaction Methods 0.000 description 3
- 239000011229 interlayer Substances 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 230000001066 destructive effect Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 230000009528 severe injury Effects 0.000 description 2
- 230000003321 amplification Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000012916 structural analysis Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Tourism & Hospitality (AREA)
- Educational Administration (AREA)
- Development Economics (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention provides a method and a system for rapidly evaluating building earthquake damage based on data augmentation and deep learning, wherein the method comprises the following steps: acquiring actual measurement strong vibration record data; inputting a wavelet time-frequency diagram of the actually measured strong vibration record data into a pre-trained deep learning vibration damage prediction model to obtain a building vibration damage prediction result; the deep learning earthquake damage prediction model is obtained by training earthquake motion samples obtained by amplifying the strong vibration data, and the strong vibration data is widely used for constructing strong vibration data compatible with earthquake motion response spectrums based on a continuous wavelet transformation method. The embodiment of the invention adopts a continuous wavelet transformation method to construct strong vibration compatible with earthquake response spectrum for data enhancement, adopts a deep learning method to predict structural earthquake damages, can obtain more accurate results when being applied to rapid assessment of the structural earthquake damages, and improves the accuracy of rapid prediction of the structural earthquake damages.
Description
Technical Field
The invention relates to the technical field of rapid assessment of building earthquake damage, in particular to a rapid assessment method and a rapid assessment system of building earthquake damage based on data augmentation and deep learning.
Background
The method for quickly evaluating the building earthquake damage has important significance for post-earthquake emergency rescue and recovery, and the machine learning method provides an important means for quickly evaluating the building earthquake damage. The method generally adopts actually measured strong vibration data to calculate building earthquake response, utilizes various machine learning algorithms to train earthquake damage data, and develops building earthquake damage rapid assessment according to a trained model. The method needs to calculate and obtain the earthquake damage database by utilizing the strong vibration data, but even if the strong vibration observation technology is developed in recent years, the strong vibration data with larger destructive power recorded at present is still limited, which causes fewer severe strong vibration samples to be destroyed in the established earthquake damage database, such as severely destroyed and collapsed buildings, thereby influencing the accuracy of earthquake damage evaluation.
In the existing machine learning, a data augmentation method is generally adopted to increase the number and diversity of samples, and the traditional data augmentation method is to rotate, cut, scale, add noise and the like to data to generate new data, augment a data set and improve the accuracy and generalization capability of a model. However, the earthquake motion is a special data, has definite physical meaning in time domain and frequency domain, and is difficult to be applied to. For this reason, researchers typically use a seismic amplitude modulation method to increase the number of samples of strong vibration data with greater destructive power, but the number of severely damaged and collapsed samples is still small, resulting in insufficient accuracy in seismic damage assessment.
Disclosure of Invention
In order to solve the above problems, the embodiment of the invention provides a rapid assessment method for building earthquake damage based on data augmentation and deep learning, which comprises the following steps: acquiring actual measurement strong vibration record data; inputting the wavelet time-frequency diagram of the actually measured strong vibration recording data into a pre-trained deep learning earthquake damage prediction model to obtain a building earthquake damage prediction result; the deep learning earthquake damage prediction model is obtained by training earthquake motion samples obtained by amplifying strong vibration data, and the strong vibration data is obtained by constructing strong vibration data compatible with earthquake motion response spectrums based on a continuous wavelet transformation method.
Optionally, the training process of the deep learning earthquake damage prediction model is as follows: constructing strong vibration data compatible with earthquake motion response spectrums by using a continuous wavelet transformation method to amplify the strong vibration data; and generating a wavelet time-frequency diagram of the strong vibration data by wavelet transformation to serve as input data of a deep learning model, taking a building damage state label as an output result of the deep learning model, and obtaining the deep learning earthquake damage prediction model through training.
Optionally, the constructing the earthquake motion response spectrum compatible strong vibration data by using the continuous wavelet transformation method for strong vibration data augmentation comprises: calculating a destruction state label and a target reaction spectrum of a building corresponding to the target earthquake motion; the target earthquake motion is to randomly select M pieces of strong vibration data from a strong vibration database; randomly selecting N strong vibration actions with the field categories consistent with the target earthquake motions from the strong vibration database as seed earthquake motions; obtaining wavelet coefficients by performing continuous wavelet transform on the seed earthquake motion time course; calculating the ratio of the target reaction spectrum to the seed earthquake motion reaction spectrum at different periodic points, multiplying the ratio by the wavelet coefficient, and then performing inverse wavelet transformation to obtain an adjusted earthquake motion time course; comparing the error between the adjusted earthquake motion response spectrum and the target response spectrum; repeating the steps of calculating the wavelet coefficient and comparing the ratio with the error until the error is smaller than a given limit value or the maximum iteration number is reached, and obtaining strong vibration data based on response spectrum compatibility.
Optionally, the method further comprises: dividing the damage state of the existing strong vibration data based on a building earthquake damage analysis method to obtain the building damage state of the existing strong vibration data; the building destruction state includes: intact, slightly damaged, moderately damaged, severely damaged and collapsed.
Optionally, the deep learning model is a convolutional neural network.
The embodiment of the invention provides a rapid building earthquake damage assessment system based on data augmentation and deep learning, which comprises the following components: the actual measurement data acquisition module is used for acquiring actual measurement strong vibration record data; the prediction module is used for inputting the wavelet time-frequency diagram of the actually measured strong vibration record data into a pre-trained deep learning vibration damage prediction model to obtain a building vibration damage prediction result; the deep learning earthquake damage prediction model is obtained by training earthquake motion samples obtained by amplifying strong vibration data, and the strong vibration data is obtained by constructing strong vibration data compatible with earthquake motion response spectrums based on a continuous wavelet transformation method.
Optionally, the system further comprises a training module for: constructing strong vibration data compatible with earthquake motion response spectrums by using a continuous wavelet transformation method to amplify the strong vibration data; and generating a wavelet time-frequency diagram of the strong vibration data by wavelet transformation to serve as input data of a deep learning model, taking a building damage state label as an output result of the deep learning model, and obtaining the deep learning earthquake damage prediction model through training.
Optionally, the training module is specifically configured to: calculating a destruction state label and a target reaction spectrum of a building corresponding to the target earthquake motion; the target earthquake motion is to randomly select M pieces of strong vibration data from a strong vibration database; randomly selecting N strong vibration actions with the field categories consistent with the target earthquake motions from the strong vibration database as seed earthquake motions; obtaining wavelet coefficients by performing continuous wavelet transform on the seed earthquake motion time course; calculating the ratio of the target reaction spectrum to the seed earthquake motion reaction spectrum at different periodic points, multiplying the ratio by the wavelet coefficient, and then performing inverse wavelet transformation to obtain an adjusted earthquake motion time course; comparing the error between the adjusted earthquake motion response spectrum and the target response spectrum; repeating the steps of calculating the wavelet coefficient and comparing the ratio with the error until the error is smaller than a given limit value or the maximum iteration number is reached, and obtaining strong vibration data based on response spectrum compatibility.
Optionally, the system further comprises a destruction status partitioning module for: dividing the damage state of the existing strong vibration data based on a building earthquake damage analysis method to obtain the building damage state of the existing strong vibration data; the building destruction state includes: intact, slightly damaged, moderately damaged, severely damaged and collapsed.
Optionally, the deep learning model is a convolutional neural network.
According to the building earthquake damage rapid assessment method and system based on data augmentation and deep learning, the continuous wavelet transformation method is adopted to construct strong vibration compatible with earthquake response spectrums for data augmentation, the deep learning method is adopted to predict structural earthquake damage, more accurate results can be obtained when the method is applied to rapid assessment of building earthquake damage, and accuracy of rapid prediction of building earthquake damage is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are required to be used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only embodiments of the present invention, and that other drawings can be obtained according to the provided drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic flow chart of a method for rapidly evaluating building earthquake damage based on data augmentation and deep learning according to an embodiment of the invention;
FIG. 2 is a flow chart of another method for rapidly evaluating building earthquake damage based on data augmentation and deep learning according to an embodiment of the present invention;
FIG. 3 is a graph showing the contrast between the coefficient of variation of the maximum interlayer displacement angle and the 5% -95% of the critical duration of the vibration constructed by the spectrum compatible method and the existing amplitude modulation method according to the embodiment of the present invention;
FIG. 4 is a graph showing a comparison of the response spectra of seismic vibrations and true seismic vibrations constructed in accordance with an embodiment of the invention;
fig. 5 is a CNN network architecture according to an embodiment of the present invention;
FIG. 6 is a schematic diagram showing the comparison of the prediction results of different cases of the method according to the embodiment of the present invention and the conventional amplitude modulation method;
FIG. 7 is a graph showing average accuracy comparisons of damage, severe damage and collapse predictions in different cases according to an embodiment of the present invention;
fig. 8 is a schematic structural diagram of a rapid building earthquake damage assessment system based on data augmentation and deep learning according to an embodiment of the present invention.
Detailed Description
In order that the above objects, features and advantages of the invention will be readily understood, a more particular description of the invention will be rendered by reference to specific embodiments thereof which are illustrated in the appended drawings. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
The embodiment of the invention provides an intelligent rapid assessment method for building earthquake damage based on data augmentation and deep learning. The strong earthquake database constructed by the embodiment of the invention can obtain more accurate results when being applied to the rapid assessment of the building earthquake damage, and provides an important method for the amplification of the strong earthquake data and the rapid prediction of the building earthquake damage.
Referring to a flow chart of a method for rapidly evaluating building earthquake damage based on data augmentation and deep learning in the embodiment of the invention shown in fig. 1, the method comprises the following steps:
s102, acquiring actual measurement strong vibration record data.
S104, inputting the wavelet time-frequency diagram of the actually measured strong vibration record data into a pre-trained deep learning vibration damage prediction model to obtain a building vibration damage prediction result.
The deep learning earthquake damage prediction model is obtained by training earthquake motion samples obtained by amplifying strong vibration data, and the strong vibration data is widely obtained by constructing strong vibration data compatible with earthquake motion response spectrums based on a continuous wavelet transformation method.
In the embodiment, the earthquake motion method compatible with the earthquake motion reaction spectrum is constructed by utilizing continuous wavelet transformation, so that the earthquake motion similar to the target reaction spectrum can be generated, and the original characteristics of the earthquake motion can be reserved. The time-keeping characteristics of the earthquake motion constructed as described above and the degree of dispersion of the induced structural response are both smaller than those of the amplitude modulation method.
The wavelet time-frequency diagram obtained by the wavelet transformation gives consideration to the characteristics of the actual earthquake motion time domain and the actual earthquake motion frequency domain. The accuracy of deep learning prediction can be improved based on the spectrum compatible strong vibration data augmentation method.
According to the building earthquake damage rapid assessment method based on data augmentation and deep learning, the continuous wavelet transformation method is adopted to construct strong vibration compatible with earthquake response spectrums for data augmentation, the deep learning method is adopted to predict structural earthquake damage, more accurate results can be obtained when the method is applied to rapid assessment of building earthquake damage, and accuracy of rapid prediction of building earthquake damage is improved.
Optionally, the training process of the deep learning earthquake damage prediction model is as follows:
firstly, constructing strong vibration data compatible with earthquake motion response spectrums by using a continuous wavelet transformation method to amplify the strong vibration data; and secondly, a wavelet time-frequency diagram of strong vibration data generated by wavelet transformation is used as input data of a deep learning model, a building damage state label is used as an output result of the deep learning model, and the deep learning earthquake damage prediction model is obtained through training. Illustratively, the deep learning model is a convolutional neural network (Convolutional Neural Networks, CNN).
Wherein data augmentation may be performed as follows:
(1) Calculating a destruction state label and a target reaction spectrum of a building corresponding to the target earthquake motion; the target earthquake motion is to randomly select M pieces of strong vibration data from a strong vibration database;
(2) Randomly selecting N strong vibration actions with the field categories consistent with the target earthquake motions from the strong vibration database as seed earthquake motions;
(3) Obtaining wavelet coefficients by carrying out continuous wavelet transformation on the seed earthquake motion time course, calculating the ratio of target reaction spectrums of different periodic points to the seed earthquake motion reaction spectrums, multiplying the ratio by the wavelet coefficients, and then carrying out inverse wavelet transformation to obtain an adjusted earthquake motion time course;
(4) Comparing the error between the adjusted earthquake motion response spectrum and the target response spectrum;
(5) And (3) repeating the step (3) of calculating the ratio to compare errors until the errors are smaller than a given limit value or the maximum iteration number is reached, and obtaining strong vibration data based on response spectrum compatibility.
In the embodiment, the structural vibration state and the target reaction spectrum of the target earthquake motion are calculated through randomly selected seed earthquake motion and target earthquake motion, and the seed earthquake motion can construct earthquake motion with compatible earthquake motion reaction spectrum by using a continuous wavelet transformation method to amplify strong vibration data.
Table 1 shows the data amounts for the different cases in the embodiments of the present invention.
TABLE 1
Further, the damage state of the existing strong vibration data can be divided based on the building earthquake damage analysis method, and the building damage state of the existing strong vibration data can be obtained. The building destruction status may include: intact, slightly damaged, moderately damaged, severely damaged and collapsed. The structural damage state is analyzed by inputting existing strong vibration data and utilizing a building vibration damage analysis method, and the earthquake is divided according to the structural damage state to form a strong vibration database.
Referring to fig. 2, a flow chart of a method for quickly evaluating building earthquake damage based on data augmentation and deep learning according to an embodiment of the invention includes the following steps:
s201, inputting existing strong vibration data by using a building earthquake damage analysis method, and dividing earthquake motions in five defined damage states of good damage, slight damage, medium damage, serious damage and collapse.
The earthquake motion analysis method is used for inputting existing strong vibration data, and dividing earthquake motion into five predefined damage states of good damage, slight damage, medium damage, serious damage and collapse.
S202, constructing earthquake motion response spectrum compatible earthquake motion by utilizing a continuous wavelet transformation method according to the divided earthquake motion to amplify the strong vibration data.
Randomly selecting M pieces of strong vibration data from a strong vibration database as target earthquake motions, and then calculating the destruction state (namely earthquake damage labels) of the building and the response spectrum of the earthquake motions; randomly selecting N strong vibration actions with the field categories consistent with the target earthquake motions from the strong vibration database as seed earthquake motions; the seismic vibration compatible with the seismic vibration response spectrum can be obtained by utilizing a continuous wavelet transformation method through the seed seismic vibration and the target response spectrum to amplify the strong vibration data.
Specifically, a wavelet coefficient is obtained by carrying out continuous wavelet transformation on the seed earthquake motion time course, the ratio of the target reaction spectrum to the earthquake motion reaction spectrum at different periodic points is calculated, the ratio is multiplied by the wavelet coefficient, the adjusted earthquake motion time course is obtained by inverse wavelet transformation, the error between the adjusted earthquake motion reaction spectrum and the target reaction spectrum is compared again, the steps are repeated until the error is smaller than a given limit value or the maximum iteration number is reached, and then the strong vibration record based on the compatibility of the reaction spectrum can be obtained.
Illustratively, 20 pieces of earthquake motion are randomly selected for the serious damage and collapse of the label by the earthquake motion constructed by the method of the embodiment of the invention and the earthquake motion constructed by the existing amplitude modulation method, and the coefficient of variation (COV) of the important holding time and the maximum interlayer displacement angle of 5% -95% is calculated. FIG. 3 is a graph showing the contrast between the coefficient of variation (COV) of the maximum interlayer displacement angle and the 5% -95% of the critical duration of the vibration constructed by the spectrum compatible method and the existing amplitude modulation method according to the embodiment of the present invention.
The duration of the earthquake motion constructed by the embodiment of the invention shown in fig. 3 and the discrete degree of the structural response are smaller, and the requirements of deep learning strong vibration database construction can be met.
FIG. 4 shows a graph of the contrast of the response spectrum of the earthquake motion constructed in accordance with the embodiment of the present invention and the actual earthquake motion. As shown in FIG. 4, it can be seen that the earthquake constructed in accordance with the embodiments of the present invention can obtain a earthquake motion time course closer to the true earthquake motion response spectrum. The earthquake motion constructed by the embodiment of the invention can obtain more accurate analysis results when being applied to structural analysis, which further illustrates the advantages of the embodiment of the invention.
S203, adopting CNN as a deep learning algorithm, converting strong vibration data into a wavelet time-frequency diagram, taking the wavelet time-frequency diagram as input of a deep learning network, taking a damage state label as network output, and obtaining a CNN earthquake damage prediction model through training.
Optionally, according to the data-augmented strong vibration database, a wavelet time-frequency diagram for generating strong vibration data through wavelet transformation is used as input data of the input CNN network, a damage state label is used as network output, and a CNN vibration damage prediction model can be obtained through training. Illustratively, fig. 5 shows a CNN network architecture established by an embodiment of the present invention. The accuracy of deep learning prediction can be improved based on the spectrum compatible strong vibration data augmentation method.
S204, converting the actually measured earthquake motion into an earthquake motion wavelet time-frequency diagram serving as input data of a trained CNN prediction model.
S205, inputting the input data into a CNN earthquake damage prediction model to obtain a building earthquake damage prediction result.
Specifically, the actually measured earthquake motion is converted into an earthquake motion wavelet time-frequency diagram to be used as input data of a trained CNN prediction model. And converting the actually measured earthquake motion into an earthquake motion wavelet time-frequency diagram, and inputting the earthquake motion wavelet time-frequency diagram into input data of a CNN prediction model to obtain a building earthquake damage prediction result.
FIG. 6 is a schematic diagram showing the comparison of the prediction results of the different cases of the method according to the embodiment of the present invention and the existing amplitude modulation method. The prediction results of cases a (original database), B (original database + amplitude modulation method), C (original database + spectrum compatible method) are shown.
Fig. 7 shows a comparison of average accuracy of damage, severe damage, and collapse predictions in different cases in an embodiment of the present invention. The average accuracy of the predictions for cases a (original database), B (original database + amplitude modulation method), C (original database + spectrum compatible method) are shown.
According to the building earthquake damage intelligent rapid assessment method based on data augmentation and deep learning, existing strong vibration data are input, a building earthquake damage analysis method is utilized to analyze the structural damage state, and earthquake is divided according to the structural damage state to form a strong vibration database.
According to the intelligent rapid evaluation method for building earthquake damages based on data augmentation and deep learning, the structural earthquake damage state and the target response spectrum of the target earthquake damages are calculated through randomly selected seed earthquake damages and target earthquake damages, and the seed earthquake damages can be used for constructing earthquake damages with compatible earthquake response spectrums by using a continuous wavelet transformation method to carry out strong earthquake data augmentation.
According to the building earthquake damage intelligent rapid assessment method based on data augmentation and deep learning, strong vibration data are converted into a wavelet time-frequency diagram through wavelet transformation, the wavelet time-frequency diagram is used as input of a deep learning network, and a CNN earthquake damage prediction model can be obtained through training.
According to the intelligent rapid assessment method for building earthquake damage based on data augmentation and deep learning, the building earthquake damage prediction result can be obtained by inputting the actually measured earthquake motion wavelet time-frequency diagram into the trained CNN earthquake damage prediction model.
Fig. 8 shows a schematic structural diagram of a rapid building earthquake damage assessment system based on data augmentation and deep learning, which includes:
the measured data acquisition module 801 is configured to acquire measured strong vibration record data;
the prediction module 802 is configured to input the wavelet time-frequency chart of the actually measured strong vibration record data into a pre-trained deep learning vibration prediction model, so as to obtain a building vibration prediction result; the deep learning earthquake damage prediction model is obtained by training earthquake motion samples obtained by amplifying strong vibration data, and the strong vibration data is obtained by constructing strong vibration data compatible with earthquake motion response spectrums based on a continuous wavelet transformation method.
The rapid building earthquake damage assessment system based on data augmentation and deep learning provided by the embodiment of the invention adopts the continuous wavelet transformation method to construct strong earthquake with compatible earthquake response spectrum for data augmentation, adopts the deep learning method to predict structural earthquake damage, can obtain more accurate results when being applied to rapid building earthquake damage assessment, and improves the accuracy of rapid building earthquake damage prediction.
Optionally, the system further comprises a training module for: constructing strong vibration data compatible with earthquake motion response spectrums by using a continuous wavelet transformation method to amplify the strong vibration data; and generating a wavelet time-frequency diagram of the strong vibration data by wavelet transformation to serve as input data of a deep learning model, taking a building damage state label as an output result of the deep learning model, and obtaining the deep learning earthquake damage prediction model through training.
Optionally, the training module is specifically configured to: calculating a destruction state label and a target reaction spectrum of a building corresponding to the target earthquake motion; the target earthquake motion is to randomly select M pieces of strong vibration data from a strong vibration database; randomly selecting N strong vibration actions with the field categories consistent with the target earthquake motions from the strong vibration database as seed earthquake motions; obtaining wavelet coefficients by performing continuous wavelet transform on the seed earthquake motion time course; calculating the ratio of the target reaction spectrum to the seed earthquake motion reaction spectrum at different periodic points, multiplying the ratio by the wavelet coefficient, and then performing inverse wavelet transformation to obtain an adjusted earthquake motion time course; comparing the error between the adjusted earthquake motion response spectrum and the target response spectrum; repeating the steps of calculating the wavelet coefficient and comparing the ratio with the error until the error is smaller than a given limit value or the maximum iteration number is reached, and obtaining strong vibration data based on response spectrum compatibility.
Optionally, the system further comprises a destruction status partitioning module for: dividing the damage state of the existing strong vibration data based on a building earthquake damage analysis method to obtain the building damage state of the existing strong vibration data; the building destruction state includes: intact, slightly damaged, moderately damaged, severely damaged and collapsed.
Optionally, the deep learning model is a convolutional neural network.
It will be appreciated by those skilled in the art that implementing all or part of the above-described methods in the embodiments may be implemented by a computer level to instruct a control device, where the program may be stored in a computer readable storage medium, where the program may include the above-described methods in the embodiments when executed, where the storage medium may be a memory, a magnetic disk, an optical disk, or the like.
In this document, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
In the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, and identical and similar parts between the embodiments are all enough to refer to each other.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. The method for quickly evaluating the earthquake damage of the building based on data augmentation and deep learning is characterized by comprising the following steps of:
acquiring actual measurement strong vibration record data;
inputting the wavelet time-frequency diagram of the actually measured strong vibration recording data into a pre-trained deep learning earthquake damage prediction model to obtain a building earthquake damage prediction result; the deep learning earthquake damage prediction model is obtained by training earthquake motion samples obtained by amplifying strong vibration data, and the strong vibration data is obtained by constructing strong vibration data compatible with earthquake motion response spectrums based on a continuous wavelet transformation method.
2. The method of claim 1, wherein the training process of the deep learning earthquake damage prediction model is as follows:
constructing strong vibration data compatible with earthquake motion response spectrums by using a continuous wavelet transformation method to amplify the strong vibration data;
and generating a wavelet time-frequency diagram of the strong vibration data by wavelet transformation to serve as input data of a deep learning model, taking a building damage state label as an output result of the deep learning model, and obtaining the deep learning earthquake damage prediction model through training.
3. The method of claim 2, wherein constructing seismic response spectrum compatible strong vibration data for strong vibration data augmentation using a continuous wavelet transform method comprises:
calculating a destruction state label and a target reaction spectrum of a building corresponding to the target earthquake motion; the target earthquake motion is to randomly select M pieces of strong vibration data from a strong vibration database;
randomly selecting N strong vibration actions with the field categories consistent with the target earthquake motions from the strong vibration database as seed earthquake motions;
obtaining wavelet coefficients by performing continuous wavelet transform on the seed earthquake motion time course;
calculating the ratio of the target reaction spectrum to the seed earthquake motion reaction spectrum at different periodic points, multiplying the ratio by the wavelet coefficient, and then performing inverse wavelet transformation to obtain an adjusted earthquake motion time course;
comparing the error between the adjusted earthquake motion response spectrum and the target response spectrum;
repeating the steps of calculating the wavelet coefficient and comparing the ratio with the error until the error is smaller than a given limit value or the maximum iteration number is reached, and obtaining strong vibration data based on response spectrum compatibility.
4. The method according to claim 1, wherein the method further comprises:
dividing the damage state of the existing strong vibration data based on a building earthquake damage analysis method to obtain the building damage state of the existing strong vibration data; the building destruction state includes: intact, slightly damaged, moderately damaged, severely damaged and collapsed.
5. The method of any one of claims 1-4, wherein the deep learning model is a convolutional neural network.
6. A rapid assessment system for building earthquake damage based on data augmentation and deep learning, the system comprising:
the actual measurement data acquisition module is used for acquiring actual measurement strong vibration record data;
the prediction module is used for inputting the wavelet time-frequency diagram of the actually measured strong vibration record data into a pre-trained deep learning vibration damage prediction model to obtain a building vibration damage prediction result; the deep learning earthquake damage prediction model is obtained by training earthquake motion samples obtained by amplifying strong vibration data, and the strong vibration data is obtained by constructing strong vibration data compatible with earthquake motion response spectrums based on a continuous wavelet transformation method.
7. The system of claim 6, further comprising a training module for:
constructing strong vibration data compatible with earthquake motion response spectrums by using a continuous wavelet transformation method to amplify the strong vibration data;
and generating a wavelet time-frequency diagram of the strong vibration data by wavelet transformation to serve as input data of a deep learning model, taking a building damage state label as an output result of the deep learning model, and obtaining the deep learning earthquake damage prediction model through training.
8. The system according to claim 7, wherein the training module is specifically configured to:
calculating a destruction state label and a target reaction spectrum of a building corresponding to the target earthquake motion; the target earthquake motion is to randomly select M pieces of strong vibration data from a strong vibration database;
randomly selecting N strong vibration actions with the field categories consistent with the target earthquake motions from the strong vibration database as seed earthquake motions;
obtaining wavelet coefficients by performing continuous wavelet transform on the seed earthquake motion time course;
calculating the ratio of the target reaction spectrum to the seed earthquake motion reaction spectrum at different periodic points, multiplying the ratio by the wavelet coefficient, and then performing inverse wavelet transformation to obtain an adjusted earthquake motion time course;
comparing the error between the adjusted earthquake motion response spectrum and the target response spectrum;
repeating the steps of calculating the wavelet coefficient and comparing the ratio with the error until the error is smaller than a given limit value or the maximum iteration number is reached, and obtaining strong vibration data based on response spectrum compatibility.
9. The system of claim 6, further comprising a destroy state partitioning module for:
dividing the damage state of the existing strong vibration data based on a building earthquake damage analysis method to obtain the building damage state of the existing strong vibration data; the building destruction state includes: intact, slightly damaged, moderately damaged, severely damaged and collapsed.
10. The system of any of claims 6-9, wherein the deep learning model is a convolutional neural network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310701891.2A CN116821642A (en) | 2023-06-13 | 2023-06-13 | Building earthquake damage rapid assessment method and system based on data augmentation and deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310701891.2A CN116821642A (en) | 2023-06-13 | 2023-06-13 | Building earthquake damage rapid assessment method and system based on data augmentation and deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116821642A true CN116821642A (en) | 2023-09-29 |
Family
ID=88112026
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310701891.2A Pending CN116821642A (en) | 2023-06-13 | 2023-06-13 | Building earthquake damage rapid assessment method and system based on data augmentation and deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116821642A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200183035A1 (en) * | 2018-12-11 | 2020-06-11 | Exxonmobil Upstream Research Company | Data Augmentation for Seismic Interpretation Systems and Methods |
| CN114842280A (en) * | 2021-01-14 | 2022-08-02 | 西南科技大学 | Automatic micro-seismic signal identification algorithm based on convolutional neural network |
| CN114912798A (en) * | 2022-05-16 | 2022-08-16 | 中震科建(广东)防灾减灾研究院有限公司 | Earthquake loss evaluation system based on random forest and earthquake damage big data |
| CN115017591A (en) * | 2022-06-20 | 2022-09-06 | 长沙理工大学 | Building structure group earthquake-resistant performance evaluation method |
| CN115201902A (en) * | 2022-07-04 | 2022-10-18 | 中海石油(中国)有限公司 | Fault intelligent identification method and system based on deep learning |
-
2023
- 2023-06-13 CN CN202310701891.2A patent/CN116821642A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200183035A1 (en) * | 2018-12-11 | 2020-06-11 | Exxonmobil Upstream Research Company | Data Augmentation for Seismic Interpretation Systems and Methods |
| CN114842280A (en) * | 2021-01-14 | 2022-08-02 | 西南科技大学 | Automatic micro-seismic signal identification algorithm based on convolutional neural network |
| CN114912798A (en) * | 2022-05-16 | 2022-08-16 | 中震科建(广东)防灾减灾研究院有限公司 | Earthquake loss evaluation system based on random forest and earthquake damage big data |
| CN115017591A (en) * | 2022-06-20 | 2022-09-06 | 长沙理工大学 | Building structure group earthquake-resistant performance evaluation method |
| CN115201902A (en) * | 2022-07-04 | 2022-10-18 | 中海石油(中国)有限公司 | Fault intelligent identification method and system based on deep learning |
Non-Patent Citations (5)
| Title |
|---|
| LUIS A MONTEJO 等: "An improved CWT-based algorithm for the generation of spectrum-compatible records", 《INTERNATIONAL JOURNAL OF ADVANCED STRUCTURAL ENGINEERING》, pages 1 - 7 * |
| SIDDHARTH S. PARIDA 等: "SVD enabled data augmentation for machine learning based surrogate modeling of non-linear structures", 《ENGINEERING STRUCTURES》, pages 1 - 15 * |
| XINZHENG LU 等: "A deep learning approach to rapid regional post-event seismic damage assessment using time-frequency distributions of ground motions", 《EARTHQUAKE ENGINEERING STRUCTURAL DYNAMICS》, vol. 50, no. 6, pages 1612 - 1627 * |
| 程庆乐 等: "基于数据增强和深度学习的建筑震害智能化快速评估方法", 《工程力学》, pages 1 - 10 * |
| 程庆乐: "基于实测地震动的近实时震损评价方法研究", 《中国博士学位论文全文数据库基础科学辑》, pages 011 - 6 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Huang et al. | Robust Bayesian compressive sensing for signals in structural health monitoring | |
| US11215719B2 (en) | Performance-level seismic motion hazard analysis method based on three-layer dataset neural network | |
| Hammer et al. | A seismic‐event spotting system for volcano fast‐response systems | |
| Ching et al. | Reliability estimation for dynamical systems subject to stochastic excitation using subset simulation with splitting | |
| Wang et al. | Damage identification scheme based on compressive sensing | |
| Kim et al. | Automated peak picking using region‐based convolutional neural network for operational modal analysis | |
| Beuzen et al. | Ensemble models from machine learning: an example of wave runup and coastal dune erosion | |
| Yao et al. | Blind modal identification using limited sensors through modified sparse component analysis by time‐frequency method | |
| Baker | Identification of near-fault velocity pulses and prediction of resulting response spectra | |
| CN113488058A (en) | Voiceprint recognition method based on short voice | |
| Tomar et al. | Active learning method for risk assessment of distributed infrastructure systems | |
| JP2005049869A (en) | Method for detecting component of non-stationary signal | |
| Cartwright et al. | Tricycle: Audio representation learning from sensor network data using self-supervision | |
| CN112086100B (en) | Quantization error entropy based urban noise identification method of multilayer random neural network | |
| Kim et al. | Near-real-time identification of seismic damage using unsupervised deep neural network | |
| Zhang et al. | Damage detection of nonlinear structures using probability density ratio estimation | |
| Antoniadis et al. | Model selection using wavelet decomposition and applications | |
| Verma et al. | Speaker-independent source cell-phone identification for re-compressed and noisy audio recordings | |
| Assessment | Pacific earthquake engineering research center | |
| Ning et al. | Convolutional variational autoencoder for ground motion classification and generation toward efficient seismic fragility assessment | |
| CN116821642A (en) | Building earthquake damage rapid assessment method and system based on data augmentation and deep learning | |
| JP6216809B2 (en) | Parameter adjustment system, parameter adjustment method, program | |
| CN116312639A (en) | Speech emotion recognition method and device, storage medium and computer equipment | |
| Ete et al. | Forecasting the Number of Tourist Arrivals to Batam by applying the Singular Spectrum Analysis and the Arima Method | |
| Chan et al. | Equalization of speech and audio signals using a nonlinear dynamical approach |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |