Earthquake vulnerability analysis method for earth-rockfill dam by fusing time-frequency space features

Abstract

The application discloses a kind of earth-rock dam seismic vulnerability analysis methods of fusion time-frequency space features, comprising the following steps: S1. constructing seismic time-frequency space feature fusion index;Wavelet transform is converted into wavelet time-frequency diagram by seismic acceleration time history data, deep features are obtained by using capsule neural network to wavelet time-frequency diagram feature extraction, and seismic time-frequency space feature fusion index is obtained by fusing deep features with traditional features;S2. constructing TPEMLP dam seismic demand model;The hyperparameters of MLP network model are optimized by TPE optimization algorithm to improve the precision and modeling efficiency of MLP network model, the relationship between seismic time-frequency space feature fusion index and dam seismic response is revealed, and then TPEMLP dam seismic demand model is established;S3. constructing dam seismic vulnerability analysis model of fusion time-frequency space features;By combining seismic time-frequency space feature fusion index with TPEMLP dam seismic demand model, the dam seismic vulnerability model of fusion time-frequency space features is constructed.

Description

Technical Field

[0001] This invention belongs to the field of water conservancy and hydropower engineering construction, and in particular relates to a method for analyzing the seismic vulnerability of earth-rock dams by integrating time-frequency spatial characteristics. Background Technology

[0002] Earth-rock dam seismic vulnerability analysis is of great significance for seismic design, seismic strengthening, and risk assessment. [1] Establishing seismic ground motion indices and dam seismic demand models are two key aspects of seismic vulnerability analysis for earth-rock dams. [2,3] However, existing seismic vulnerability analysis studies of earth-rock dams, using seismic ground motion indices such as peak ground acceleration and peak ground velocity, fail to fully reflect the temporal and spatial characteristics of seismic motions. Furthermore, traditional seismic demand models based on the assumption of a logarithmic spatial linear function struggle to reveal the complex nonlinear relationship between seismic ground motion indices and seismic response. Therefore, it is urgent to construct indices that fully reflect the complex temporal and spatial characteristics of seismic motions and to establish seismic demand models capable of revealing the complex nonlinear relationship between seismic ground motion indices and seismic response. Based on these models, seismic vulnerability analysis of earth-rock dams can be conducted, providing support for seismic risk assessment of earth-rock dams.

[0003] Regarding the establishment of seismic motion indices, existing studies commonly use two main categories: one uses a single index to characterize seismic motion, and the other uses two or more multivariate seismic motion indices. For single seismic motion indices, existing studies generally use spectral acceleration. [4,5] Peak acceleration [6,7] Peak speed [4,8] and peak displacement [8] Earthquake vulnerability studies are conducted using indicators such as [list of indicators]. In addition, some new ground motion indicators, such as those considering structural characteristics, are also being developed. [9] Seismic indices based on dominant modes

[10] and multi-period average spectral acceleration index

[11] Indicators such as [list of indices] have also been proposed to characterize ground motion. However, the information carried by a single ground motion index is limited, and it can only characterize some features of ground motion. In order to more fully reflect the complex information of ground motion, relevant scholars have proposed multivariate ground motion indices. For example, Alembagheri et al.

[12] Based on the proposed (S) a (T1),PGA), (S) a (T1),PGD), (PGA,PGD) and (S a (T1),S a (T1) / S a (T n Vulnerability studies were conducted using four multivariate ground motion indices (PGA); Baker et al. [13-15] It was proposed that (S) a(T1), Epsilon) and (S a (T1),S a (T1) / S a (T2) Two multivariate ground motion indices; Fotopoulou et al.

[16] Based on sufficiency and the correlation between ground motion indices, two multivariate ground motion indices, (PGA, Ia) and (PGV, Ky / PGV), were selected; Fan Shuli et al.

[17] Based on multi-dimensional ground motion index (S a (T1),S a (T2) Seismic vulnerability analysis was conducted on an arch dam. Although multi-element ground motion indices can reflect more ground motion information than single indices, due to the complex temporal and spatial characteristics of ground motion, multi-element ground motion indices composed of only a few single indices are still insufficient to fully characterize it.

[18] .

[0004] Wavelet transform, with its powerful time-frequency analysis capabilities, can map one-dimensional ground motion signals onto a two-dimensional time-frequency plane to obtain wavelet time-frequency maps, thus fully reflecting the complex time-frequency characteristics of ground motion.

[19] However, the complex ground motion information represented by wavelet time-frequency maps is hidden within the time-frequency images, and the ground motion information and its features are inextricably linked in their spatial distribution on the time-frequency map. Therefore, how to consider the spatial distribution relationship of wavelet time-frequency map features and extract their features is an urgent problem to be solved. Capsule networks not only have strong feature extraction capabilities but also can learn the spatial relationships between features, and have already been applied in aerospace.

[20] Remote sensing images

[21] Fault detection

[22] This method has been widely applied in fields such as wavelet time-frequency mapping, providing an effective approach for extracting complex time-frequency spatial features from wavelet time-frequency maps. Therefore, this paper uses capsule networks to extract features from seismic ground motion wavelet time-frequency maps to obtain deep features that reflect the time-frequency spatial distribution of seismic ground motion. Furthermore, a feature concatenation method is used to fuse the deep time-frequency spatial features of seismic ground motion with traditional features to obtain a seismic ground motion fusion index, thereby fully revealing the complex time-frequency spatial features of seismic ground motion.

[0005] In establishing seismic demand models for dams, most existing studies assume that the relationship between seismic motion parameters and seismic response follows a logarithmic linear function, making it difficult to reveal the complex nonlinear relationship between these two parameters.

[23] MLP neural networks possess powerful data self-driving capabilities and the ability to handle high-dimensional nonlinear problems, and have been used in seismic assessment in recent years.

[24] Data denoising

[25] Trend Forecasting

[26] It has been widely used in fields such as earthquake motion index and earthquake response of earth-rock dam, providing a new way to reveal the complex nonlinear relationship between earthquake motion index and earthquake response of earth-rock dam. The selection of hyperparameters such as the number of neurons and learning rate in MLP neural network directly affects its performance

[27] . The traditional manual parameter tuning method has the problems of low accuracy and efficiency. The TPE optimization algorithm has the advantages of simple optimization mechanism, fast convergence speed and strong robustness, and can efficiently optimize the hyperparameters of MLP by using prior knowledge. Therefore, the TPE optimization algorithm is used to automatically optimize the hyperparameters of MLP, and then the TPE-MLP earth-rock dam earthquake demand model is established to improve the accuracy and construction efficiency of the model.

[0006] Current seismic vulnerability analyses using ground motion indices such as peak ground acceleration (PGA) and peak ground velocity (PGV) fail to fully reflect the temporal and spatial characteristics of ground motions. Furthermore, traditional seismic demand models based on the assumption of a logarithmic spatial linear function struggle to reveal the complex nonlinear relationship between ground motion indices and seismic response. Therefore, this paper proposes a method to improve the accuracy of seismic vulnerability analysis results for earth-rock dams. This involves using wavelet transform to obtain wavelet time-frequency maps of ground motions, followed by feature extraction from capsule neural networks to uncover deep features reflecting the temporal and spatial distribution of ground motions. Feature splicing is then used to fuse these deep temporal and spatial features with traditional features to obtain a fused ground motion index. Furthermore, an improved MLP based on TPE is used to establish a seismic demand model for dams, revealing the relationship between the fused ground motion index and the dam's seismic response. Finally, constructing a seismic vulnerability analysis model for earth-rock dams based on the fused ground motion temporal and spatial feature index and the dam seismic demand model is a crucial step in enhancing the accuracy of seismic vulnerability analysis results for earth-rock dams.

[0007] References:

[0008] [1] Zhang Chuhan, Jin Feng, Wang Jinting, et al. Key issues and research progress in seismic safety evaluation of high concrete dams [J]. Journal of Hydraulic Engineering, 2016, 47(03):253-264.

[0009] [2] Chen Zhiyi, Yu Wei, Liu Zhiqian. Construction and performance verification of composite ground motion parameters based on overall damage characterization of underground structures [J]. Chinese Journal of Civil Engineering, 2023: 1-12.

[0010] [3] Song Zhiqiang, Zhang Jianfeng, Wang Fei, et al. Study on selection of seismic ground motion intensity index in seismic analysis of hydropower station powerhouse [J]. Vibration and Shock, 2022, 41(02):151-160.

[0011] [4]Kadkhodayan V,Aghajanzadeh SM,Mirzabozorg H.Seismic assessment of arch dams using fragility curves[J].Civil Engineering Journal,2015,1(2):14-20.

[0012] [5] Yu Xiaohui, Lü Dagang. Probabilistic earthquake demand analysis and earthquake vulnerability analysis based on cloud map-strip method [J]. Engineering Mechanics, 2016, 33(06):68-76.

[0013] [6] Jin Congcong, Chi Shichun, Nie Zhangbo. Seismic safety analysis of high earth-rock dams based on seismic deformation vulnerability [J]. Chinese Journal of Geotechnical Engineering, 2020, 42(02):334-343.

[0014] [7] Pang Rui, Kong Xianjing, Zou Degao, et al. Seismic Settlement Vulnerability Analysis of High Core Rockfill Dams Based on MSA Method [J]. Journal of Hydraulic Engineering, 2017, 48(07): 866-873.

[0015] [8]Pang R, Xu B, Kong

[0016] [9] Yao Xiaowen, Jiang Jianqun. Correlation between seismic intensity index and response of high arch dams [J]. Journal of Zhejiang University (Engineering Science), 2014, 48(02):254-261.

[0017]

[10] Xie Mingzhi, Yang Yongqing, Huang Shengqian, et al. Study on vulnerable ground motion intensity parameters of cable-stayed bridges with low towers for high-speed railways based on dominant modes [J]. China Railway Science, 2021, 42(04):41-50.

[0018]

[11] Zhang Yixin, Zheng Shansuo, Qin Qing, et al. Spectral acceleration index analysis applicable to high-rise RC structures [J]. Engineering Mechanics, 2017, 34(10): 149-157.

[0019]

[12] Alembagheri M.Investigating efficiency of vector-valued intensitymeasures in seismic demand assessment of concrete dams[J].Advances in CivilEngineering,2018,2018(2):1-12.

[0020]

[13] Baker J W.Probabilistic structural response assessment usingvector-valued intensity measures[J].Earthquake Engineering&StructuralDynamics,2007,36(13):1861-1883.

[0021]

[14] Baker J W,Cornell C A.Vector-valued intensity measuresincorporating spectral shape for prediction of structural response[J].Journalof Earthquake Engineering,2008,12(4):534-554.

[0022]

[15] Baker J W,Cornell C A.A vector-valued ground motion intensitymeasure consisting of spectral acceleration and epsilon[J].EarthquakeEngineering&Structural Dynamics,2005,34(10):1193-1217.

[0023]

[16] Fotopoulou SD, Pitilakis K D.Predictive relationships forseismically induced slope displacements using numerical analysis results[J].Bulletin of Earthquake Engineering,2015,13(11):3207-3238.

[0024]

[17] Fan Shuli, Tian Shuo, Chen Jianyun. Seismic vulnerability analysis of arch dams based on vector ground motion intensity index [J]. Journal of Hydraulic Engineering, 2019, 50(03):335-345.

[0025]

[18] Xu YJ, Lu XZ, Tian Y, et al. Real-time seismic damage prediction and comparison of various ground motion intensity measures based on machine learning [J]. Journal of Earthquake Engineering, 2022, 26(8): 4259-4279.

[0026]

[19] Zhang Shijie, Jin Faye, Yao Zhifeng, et al. Field test study on pressure pulsation and vibration characteristics of double-suction centrifugal pump station [J]. Journal of Hydraulic Engineering, 2021, 52(09):1047-1058.

[0027]

[20] Zhong, Xintong. Target detection of protrusions on aero-engine blades based on YOLO-CapsNet [D]. Qingdao University of Science and Technology, 2021.

[0028]

[21] Xu Zhengsen, Guan Haiyan, Peng Daifeng, et al. Attention capsule network algorithm for building extraction from high-resolution remote sensing images [J]. Journal of Remote Sensing, 2022, 26(08):1636-1649.

[0029]

[22] Zhang Hui, Ge Baojun, Han Bin, et al. Fault diagnosis method for motor bearings based on GAF-CapsNet [J]. Journal of Electrical Engineering, 2023, 38(10):2675-2685.

[0030]

[23] Chen Zhiyi, Huang Pengfei. Vulnerability analysis of subway stations based on probabilistic neural network [J]. Journal of Tongji University (Natural Science Edition), 2021, 49(06):791-798.

[0031]

[24] Yang Shun, Chang Chaoyu, Yu Sihan, et al. A comprehensive discrimination method for rapid assessment of seismic stability of loess slopes - a case study of Xiji County, Ningxia [J]. Journal of Earthquake Engineering, 2022, 44(01):54-61.

[0032]

[25] Wang Qiqi, Tang Jingtian, Zhang Liang, et al. Denoising of seismic data using multilayer perceptron [J]. Petroleum Geophysical Exploration, 2020, 55(02):272-281.

[0033]

[26] Chen Peng, Wu Yifan, Cai Shuang, et al. Assessment and prediction of deterioration trend of pumped storage units based on autoencoder compression and multi-scale feature extraction [J]. Journal of Hydraulic Engineering, 2022, 53(06):747-756.

[0034]

[27] Zhang Jun, Yu Jia, Ren Bingyu, et al. Research on simulation model of construction of high core wall rockfill dam with cold and low temperature consideration [J]. Journal of Hydraulic Engineering, 2022, 53(02):200-211. Summary of the Invention

[0035] The purpose of this invention is to overcome the shortcomings of existing technologies. Currently used seismic vulnerability analysis methods, such as peak ground acceleration (PGA) and peak ground velocity (PGV), fail to fully reflect the temporal and spatial characteristics of ground motion. Furthermore, traditional seismic demand models based on the assumption of a logarithmic spatial linear function struggle to reveal the complex nonlinear relationship between seismic motion indicators and seismic response. This invention provides a seismic vulnerability analysis method for earth-rock dams that integrates temporal and spatial characteristics.

[0036] The objective of this invention is achieved through the following technical solution:

[0037] A method for seismic vulnerability analysis of earth-rock dams that integrates time-frequency spatial characteristics includes the following steps:

[0038] S1. Construct a fusion index for the temporal and frequency spatial characteristics of ground motion;

[0039] The seismic ground motion acceleration time history data is converted into a wavelet time-frequency map by wavelet transform. Deep features are obtained by extracting features from the wavelet time-frequency map using a capsule neural network. The deep features are then fused with traditional features to obtain the seismic ground motion time-frequency spatial feature fusion index.

[0040] S2. Construct the TPEMMLP dam seismic demand model;

[0041] The hyperparameters of the MLP network model are optimized by the TPE optimization algorithm to improve the accuracy and modeling efficiency of the MLP network model, reveal the relationship between the fusion index of seismic motion time-frequency spatial characteristics and the seismic response of the dam, and then establish the TPE MLP dam seismic demand model.

[0042] S3. Construct a seismic vulnerability analysis model for dams that integrates time-frequency spatial characteristics;

[0043] By combining the temporal and spatial characteristics of ground motion with the TPEMLP dam seismic demand model, a dam seismic vulnerability model integrating temporal and spatial characteristics is constructed.

[0044] Furthermore, in step S1, the conventional features include peak acceleration, peak velocity, peak displacement, cumulative absolute velocity, duration of strong earthquake, Arias intensity, squared average acceleration, Housner spectral intensity, peak spectral acceleration, peak spectral velocity, peak spectral displacement, average spectral acceleration, average spectral velocity, average spectral displacement, acceleration spectral intensity, velocity spectral intensity, displacement spectral intensity, effective peak spectral acceleration, effective peak spectral velocity, and effective peak spectral displacement.

[0045] Furthermore, in step S2, the TPE optimization algorithm is used to automatically optimize the three hyperparameters of the MLP network model: learning rate, number of hidden layer neurons, and activation function, to obtain the optimal hyperparameter setting scheme of the MLP network model. Then, a dam seismic demand model based on TPE MLP is established to reveal the relationship between the high-dimensional fusion index of ground motion and the seismic response of the dam.

[0046] Furthermore, in step S3,

[0047] The seismic vulnerability model of dams that integrates time-frequency spatial characteristics is as follows:

[0048] F(x)=P[D≥C|IMs] (1)

[0049] In the formula: IMs represents the fusion index of temporal and frequency spatial characteristics of ground motion; F(x) represents the seismic vulnerability model of the dam; C represents the seismic resistance capacity of the dam; D represents the seismic demand of the dam; P represents the probability;

[0050] The expression for the dam's earthquake demand D is as follows:

[0051] D = d TPEMLP +ε (2)

[0052] In the formula: d TPEMLP The predicted value from the TPEMLP dam seismic demand model is represented by dam crest settlement; ε represents random error, assumed to follow a standard deviation of β. D|IMs The normal distribution;

[0053] Based on equations (1) and (2), the expression for the seismic vulnerability model F(x) is derived:

[0054]

[0055] In the formula: c i(i=1,2,3) represents the ultimate failure state. The dam crest settlement rates of 0.3%, 0.6%, and 1% are used as the criteria for classifying the failure level, corresponding to three failure levels: slight failure (c1), moderate failure (c2), and severe failure (c3), respectively. Φ represents a standard normal distribution.

[0056] Standard deviation β D|IMs It can be obtained from the following formula:

[0057]

[0058] In the formula: D i (i = 1, 2, ..., n) represents the seismic dynamic analysis sample of the dam; n is the number of samples.

[0059] Furthermore, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method for analyzing the seismic vulnerability of earth-rock dams by fusing time-frequency spatial features.

[0060] Furthermore, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for analyzing the seismic vulnerability of earth-rock dams by fusing time-frequency spatial features.

[0061] Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

[0062] 1. This invention utilizes the powerful feature extraction capabilities of deep learning networks and their ability to learn the spatial relationships between features to propose a deep spatial feature extraction method for seismic motion based on capsule networks. Furthermore, it constructs a fusion index that integrates deep features and traditional features to reveal the complex time-frequency spatial characteristics of seismic motion, providing a solid foundation for accurate prediction of dam seismic response and reliable analysis of seismic vulnerability.

[0063] 2. This invention utilizes the powerful high-dimensional nonlinear problem processing capabilities of neural network models to establish an improved MLP model based on the TPE algorithm to reveal the deep nonlinear relationship between seismic motion fusion index and dam seismic response. It can achieve accurate and rapid prediction of dam response and failure under different seismic actions, providing strong support for seismic optimization design, dam safety and reliability analysis, and post-earthquake dam structure repair.

[0064] 3. Based on the established fusion index of seismic time-frequency spatial characteristics and the TPEMLP dam seismic demand model, a dam seismic vulnerability analysis model integrating time-frequency spatial characteristics was constructed, which improved the reliability of dam seismic vulnerability analysis and accurately obtained the probability of different levels of damage to the dam body under different seismic actions. This is of great significance for the seismic performance design, seismic reinforcement, earthquake relief and loss assessment of earth-rock dams. Attached Figure Description

[0065] Figure 1 Flowchart for constructing fusion indicators of temporal and frequency spatial characteristics of ground motion.

[0066] Figure 2 Optimize the MLP neural network process for the TPE algorithm. Detailed Implementation

[0067] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only for explaining the present invention and are not intended to limit the present invention.

[0068] With the implementation of national strategies such as the West-to-East Power Transmission Project and the Tibet-to-Other Power Transmission Project, a number of large-scale earth-rock dam projects are under construction or about to be constructed in Southwest my country. These regions are prone to earthquakes of high intensity, and damage to these dams from strong earthquakes would result in severe socio-economic losses. Earth-rock dam seismic vulnerability analysis can effectively reflect the conditional probability of the dam structure reaching or exceeding different damage levels under varying seismic motion intensities, and is of great significance for seismic performance design, seismic reinforcement, earthquake relief, and loss assessment. Constructing temporal and frequency spatial characteristic indicators of seismic motion to characterize seismic ground motion signals is one of the key steps in seismic vulnerability analysis. However, current traditional indicators such as peak ground acceleration (PGA), peak velocity (PV), and spectral acceleration, which are mainly used in seismic vulnerability analysis, cannot fully reflect the complex temporal and frequency spatial characteristics of seismic ground motion. Furthermore, traditional seismic demand models based on the assumption of a logarithmic spatial linear function relationship are insufficient to reveal the complex nonlinear relationship between seismic ground motion indicators and seismic response.

[0069] First, addressing the issue that peak ground acceleration (PGA), peak velocity (PFR), and spectral acceleration (SFR) indicators primarily used in pre-earthquake vulnerability analysis cannot fully reflect the complex time-frequency spatial characteristics of ground motion, this paper utilizes the advantage of Capsule Neural Networks (CapsNet) in capturing and representing the spatial distribution of target features. Feature extraction is performed on the time-frequency wavelet map of ground motion to obtain deep features reflecting the time-frequency spatial distribution of ground motion. Then, a feature concatenation method is used to fuse these deep time-frequency spatial features with traditional features to obtain a fused time-frequency spatial feature index. This can fully reveal the complex time-frequency spatial characteristics of ground motion, improve the accuracy of dam seismic response prediction, and thus ensure the accuracy of vulnerability analysis results. Second, addressing the problem that traditional seismic demand models based on the assumption of a logarithmic spatial linear function relationship struggle to reveal the complex nonlinear relationship between ground motion indicators and seismic response, this paper employs the TPE optimization algorithm to optimize the hyperparameters of the MLP network model to improve model accuracy and modeling efficiency. Then, based on the TPE-improved MLP, a dam seismic demand model is established that reveals the relationship between the fused ground motion index and the dam seismic response. This can accurately reveal the deep nonlinear relationship between the fused time-frequency spatial feature index of ground motion and the dam seismic response. Finally, based on the fusion index of seismic motion time-frequency spatial characteristics and the TPEMLP dam seismic demand model, the expression for the seismic vulnerability function was derived, and a seismic vulnerability analysis model for dams was constructed. This model can predict the probability of different damage levels to the dam under different seismic motions.

[0070] Specifically, this embodiment proposes an improved MLP model for the seismic vulnerability of earth-rock dams that integrates time-frequency spatial features. The process for establishing the fusion index of seismic motion time-frequency spatial features is as follows: Figure 1 As shown, the TPE algorithm optimizes the MLP neural network process as follows: Figure 2 As shown, the specific implementation method is as follows:

[0071] 1. Construct a fusion index of temporal and frequency spatial characteristics of ground motion.

[0072] a) Filter the seismic ground acceleration time history data from the PEER database and format it into a uniform format, namely two columns of txt text data separated by commas. The first column is the time series with the time interval adjusted to 0.02 seconds, and the second column is the acceleration series with the unit uniformly in g.

[0073] b) Based on the structural dimension parameters of the earth-rock dam, a three-dimensional physical model of the earth-rock dam was established using Rhino software, and a hexahedral mesh was generated for it, thereby constructing a three-dimensional finite element model of the earth-rock dam.

[0074] c) Using the seismic ground acceleration time history data and the three-dimensional finite element model of the earth-rock dam obtained above, three-dimensional dynamic time history analysis was performed using Abaqus software to obtain a sample dataset of the dam's seismic response.

[0075] d) Perform time-frequency analysis on the ground motion acceleration time history data used in finite element calculations, and convert the ground motion acceleration time history data into wavelet time-frequency graphs using wavelet time-frequency graph conversion;

[0076] e) Use capsule neural networks to extract features from wavelet time-frequency graphs to obtain 16-dimensional deep features;

[0077] f) By referring to relevant studies and selecting based on the principles of comprehensiveness, computational efficiency and availability, 20 traditional seismic motion characteristics were obtained;

[0078] g) The 16-dimensional deep features are fused with traditional features using feature splicing to obtain the seismic motion time-frequency spatial feature fusion index;

[0079] 2. Construct a seismic demand model for dams.

[0080] h) Based on the dam seismic response sample dataset, the TPE optimization algorithm was used to optimize the three hyperparameters of the MLP network, and the TPE MLP dam seismic demand model was trained; the specific optimization process is detailed in [link to optimization process]. Figure 2 .

[0081] 3. Construct a seismic vulnerability analysis model for dams that integrates time-frequency spatial characteristics.

[0082] k) Define the seismic demand expression for the dam;

[0083] D = d TPEMLP +ε (1)

[0084] In the formula: d TPEMLP The predicted value from the TPEMLP dam seismic demand model is represented by dam crest settlement; ε represents random error, assumed to follow a standard deviation of β. D|IMs It follows a normal distribution.

[0085] l) Define the expression for the seismic vulnerability function of the dam;

[0086] F(x)=P[D≥C|IMs] (2)

[0087] In the formula: IMs represents the fusion index of time-frequency spatial characteristics of ground motion; F(x) represents the seismic vulnerability model of the dam; C represents the seismic resistance capacity of the dam; D represents the seismic demand of the dam; P represents the probability.

[0088] m) Based on equations (1) and (2), the expression for the seismic vulnerability model of the dam is derived;

[0089]

[0090] In the formula: c i(i = 1, 2, 3) represents the ultimate failure state. The dam crest settlement rates of 0.3%, 0.6%, and 1% are used as the criteria for classifying the failure level, corresponding to three failure levels: minor (c1), moderate (c2), and severe (c3), respectively. Φ represents a standard normal distribution.

[0091] n) Define the standard deviation β D|IMs expression;

[0092]

[0093] In the formula: D i (i = 1, 2, ..., n) represents the seismic dynamic analysis sample of the dam; n is the number of samples.

[0094] o) The TPEMLP dam seismic demand model is used to predict the dam seismic response and obtain the predicted value d. TPEMLP ;

[0095] p) Calculate the random error ε between the predicted and simulated values, and its standard deviation β. D|IMs ;

[0096] q) The destruction level threshold and standard deviation β D|IMs and predicted value d TPEMLP Substituting these values ​​into the vulnerability function expression yields the probability of different damage levels to the dam under different seismic motions, i.e., seismic vulnerability.

[0097] Preferably, embodiments of this application also provide a specific implementation of an electronic device capable of implementing all steps in the earthquake vulnerability analysis method for earth-rock dams that integrates time-frequency spatial features as described in the above embodiments. The electronic device specifically includes the following:

[0098] Processor, memory, communications interface, and bus;

[0099] The processor, memory, and communication interface communicate with each other via a bus; the communication interface is used to realize information transmission between server-side devices, metering devices, and user-side devices.

[0100] The processor is used to call the computer program in the memory. When the processor executes the computer program, it implements all the steps in the earthquake vulnerability analysis method for earth-rock dams that integrates time-frequency spatial features in the above embodiments.

[0101] Embodiments of this application also provide a computer-readable storage medium capable of implementing all steps of the earthquake vulnerability analysis method for earth-rock dams that integrates time-frequency spatial features as described in the above embodiments. The computer-readable storage medium stores a computer program that, when executed by a processor, implements all steps of the earthquake vulnerability analysis method for earth-rock dams that integrates time-frequency spatial features as described in the above embodiments.

[0102] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0103] While this application provides method operation steps as shown in the embodiments or flowcharts, more or fewer operation steps may be included based on conventional or non-inventive labor. The order of steps listed in the embodiments is merely one possible execution order among many and does not represent the only execution order. In actual device or client product execution, the method can be executed sequentially as shown in the embodiments or drawings, or in parallel (e.g., in a parallel processor or multi-threaded processing environment).

[0104] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0105] This invention is not limited to the embodiments described above. The above description of specific embodiments is intended to illustrate and explain the technical solutions of this invention. The specific embodiments described above are merely illustrative and not restrictive. Without departing from the spirit and scope of the claims, those skilled in the art can make many specific modifications based on the teachings of this invention, and these modifications all fall within the scope of protection of this invention.

Claims

1. A method for analyzing the seismic vulnerability of earth-rock dams by integrating time-frequency spatial characteristics, characterized in that, Includes the following steps: S1. Construct a fusion index for the temporal and frequency spatial characteristics of ground motion; The seismic ground motion acceleration time history data is converted into a wavelet time-frequency map by wavelet transform. Deep features are obtained by extracting features from the wavelet time-frequency map using a capsule neural network. The deep features are then fused with traditional features to obtain the seismic ground motion time-frequency spatial feature fusion index. S2. Construct the TPEMMLP dam seismic demand model; The hyperparameters of the MLP network model are optimized by the TPE optimization algorithm to improve the accuracy and modeling efficiency of the MLP network model, reveal the relationship between the fusion index of seismic motion time-frequency spatial characteristics and the seismic response of the dam, and then establish the TPE MLP dam seismic demand model. S3. Construct a seismic vulnerability analysis model for dams that integrates time-frequency spatial characteristics; By combining the time-frequency spatial characteristics of ground motion with the TPEMLP dam seismic demand model, a dam seismic vulnerability model integrating time-frequency spatial characteristics is constructed. The seismic vulnerability model of dams that integrates time-frequency spatial characteristics is as follows: （1） In the formula: Indicators representing the fusion of temporal and frequency spatial characteristics of ground motion; This represents a model of the dam's seismic vulnerability. Indicates the dam's earthquake resistance; Indicates the earthquake demand of the dam; P Represents probability; Dam earthquake demand The expression is as follows: （2） In the formula: The predicted values ​​from the TPEMLP dam seismic demand model are represented by dam crest settlement. Let the random error be represented, and assume it follows a standard deviation of . The normal distribution; The seismic vulnerability model is derived based on equations (1) and (2). expression: （3） In the formula: ( The dam crest settlement rate of 0.3%, 0.6%, and 1% is used as the standard for classifying the damage level, corresponding to minor damage, respectively, representing the ultimate failure state. Moderate damage Severe damage Three levels of damage; It follows a standard normal distribution; Standard deviation It can be obtained from the following formula: （4） In the formula: This represents a sample of seismic dynamic analysis of a dam. This represents the number of samples.

2. The method for analyzing the seismic vulnerability of earth-rock dams by integrating time-frequency spatial characteristics according to claim 1, characterized in that, In step S1, the conventional features include peak acceleration, peak velocity, peak displacement, cumulative absolute velocity, duration of strong earthquake, Arias intensity, squared average acceleration, Housner spectral intensity, peak spectral acceleration, peak spectral velocity, peak spectral displacement, average spectral acceleration, average spectral velocity, average spectral displacement, acceleration spectral intensity, velocity spectral intensity, displacement spectral intensity, effective peak spectral acceleration, effective peak spectral velocity, and effective peak spectral displacement.

3. The method for analyzing the seismic vulnerability of earth-rock dams by integrating time-frequency spatial characteristics according to claim 1, characterized in that, In step S2, the TPE optimization algorithm is used to automatically optimize the three hyperparameters of the MLP network model: learning rate, number of hidden layer neurons, and activation function, to obtain the optimal hyperparameter setting scheme of the MLP network model. Then, a dam seismic demand model based on TPE MLP is established to reveal the relationship between the high-dimensional fusion index of ground motion and the seismic response of the dam.

4. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method for analyzing the seismic vulnerability of earth-rock dams by fusing time-frequency spatial features as described in any one of claims 1 to 3.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the method for analyzing the seismic vulnerability of earth-rock dams by fusing time-frequency spatial features as described in any one of claims 1 to 3.

Images