Method for calculating influence index of flood discharge vibration multi-vibration source on site

Abstract

The invention discloses a method for calculating the influence index of multiple vibration sources of flood discharge vibration on a site, and the method comprises the steps: generating an artificial vibration time travel curve according to flood discharge vibration in-situ observation or the dominant frequency and amplitude of flood discharge vibration waves at different vibration sources; establishing a flood discharge vibration field numerical model, inputting an artificial vibration time travel curve at a vibration source, and performing field flood discharge vibration numerical simulation by adopting a static boundary and a free field boundary; and outputting a simulated vibration source and slope site monitoring point acceleration time travel curve, and establishing a proxy model based on a machine learning algorithm to perform Sobol global sensitivity analysis to obtain an influence index of the vibration source on the site monitoring point. Compared with an existing method for calculating the vibration source influence index through the root mean square of the accelerated speed of a monitoring point obtained through simulation, flood discharge vibration numerical simulation is carried out through a free field boundary, a proxy model is established through a machine learning algorithm to carry out Sobol global sensitivity analysis, and the influence index of different flood discharge vibration multiple vibration sources on site vibration is obtained; the problem of flood discharge vibration multi-vibration-source identification is solved.

Description

Technical Field

[0001] The present invention relates to the field of numerical analysis of hydraulic structures and rock mass engineering, and in particular to a method for calculating an index of influence of multiple vibration sources of flood discharge vibration on a site. Background Art

[0002] Flood discharge and energy dissipation from high dams are directly related to the safety of hydropower projects, surrounding buildings, sites, and slopes. While the amplitude of site vibration induced by high dam discharge is much smaller than that of earthquakes, its duration, spectral characteristics, and location and mode of action are much more complex than those studied in earthquake dynamics.

[0003] Due to the large discharge height, large discharge flow rate, and high water velocity, the pulsating load of the water flow will induce vibration in the discharge structure, such as the bottom plate and side walls of the water cushion pond or stilling basin, and other hydraulic structures such as gates and guide walls. On the one hand, this vibration causes the hydraulic structures in the hub area to vibrate. On the other hand, it is transmitted through the dam foundation and rock mass to the slopes and surrounding areas of the hub area, which has an adverse impact on the safety of the dam structure, the stability of the slopes, and the lives of urban residents. At the same time, the pulsating load of the water flow acts on hydraulic structures such as gates, guide walls, the bottom plate and side walls of the water cushion pond or stilling basin, generating multiple vibration sources. Therefore, the vibration of the slopes and surrounding areas of the hub area caused by the high dam discharge is a multi-source combined excitation induced vibration. The resulting vibration source identification problem is a major challenge to clarify the mechanism of flood discharge vibration and then carry out targeted prevention and control.

[0004] Existing methods for flood discharge vibration source influence index include correlation analysis of the root mean square acceleration of flood discharge vibration monitoring points; frequency domain correlation analysis of vibration signals based on Fourier transform; correlation analysis of vibration signals in time domain and frequency domain based on cross wavelet transform. Most of these methods process and generalize vibration signals twice or multiple times, which may lead to the loss of vibration information during signal processing, resulting in reduced reliability of vibration source influence index identification.

[0005] Therefore, how to obtain the vibration time history curves of each vibration source and the site, and then analyze and obtain the impact index of multiple vibration sources of flood discharge vibration on the site, is a technical problem that needs to be solved urgently. Summary of the Invention

[0006] Purpose of the invention: In view of the problems in the prior art that the mechanism of the action of multiple vibration sources of flood discharge vibration on the vibration of the site slope is complex and the influence index of each vibration source on the stability of the site slope is difficult to confirm, the present invention proposes a method for calculating the influence index of multiple vibration sources of flood discharge vibration on the site. By inputting artificial vibration waves and using static boundaries and free field boundaries to perform numerical simulation of flood discharge vibration, the vibration time history curves of each vibration source and the site are obtained. A proxy model is established through a machine learning algorithm to perform Sobol global sensitivity analysis, and the influence index of different multiple vibration sources of flood discharge vibration on the site is obtained. Compared with the existing method of calculating the vibration source influence index by the root mean square acceleration of the monitoring point obtained by experiments or numerical simulations, the present invention solves the problem of identifying multiple vibration sources of flood discharge vibration through numerical simulation and machine learning methods, and provides technical support for flood discharge safety during the operation period of high dams.

[0007] Technical Solution: The present invention provides a method for calculating the impact index of multiple vibration sources of flood discharge vibration on a site, including generating an artificial flood discharge vibration time history curve, numerical simulation of flood discharge vibration, and establishing a proxy model based on a machine learning algorithm to perform Sobol global sensitivity analysis of the vibration source impact index. The process is as follows:

[0008] Step (1) is to manually modify the seismic wave based on the amplitude and main frequency characteristics of the displacement, velocity, or acceleration time history curve of the monitoring point in the recorded in-situ observation of flood discharge vibration or the hydroelastic model test of flood discharge vibration, so as to generate a displacement, velocity, or acceleration time history curve that conforms to the amplitude and frequency characteristics of the flood discharge vibration. The process of manually modifying the seismic wave includes two steps: filtering the high-frequency components of the seismic motion and correcting the baseline of the vibration wave.

[0009] Step (1.1) filters the high-frequency components of the earthquake motion. Referring to the recorded flood discharge vibration wave, the amplitude and main frequency characteristics of the displacement, velocity or acceleration time history curve, based on the earthquake wave time history curve, the high-frequency components of the vibration time history curve are filtered through a Butterworth filter to obtain a time history curve in which the main frequency and amplitude of the displacement, velocity or acceleration time history curve are consistent with the flood discharge vibration process; the frequency of the corrected vibration time history curve is made consistent with the time and frequency characteristics of the flood discharge vibration wave, wherein the Butterworth filter is used to filter the high-frequency components of the vibration time history curve:

[0010]

[0011] Where: f low is the normalized low-frequency cutoff frequency; f high is the normalized high frequency cutoff frequency; f s is the sampling frequency lowcut, highcut is the actual filtering cutoff frequency (Hz).

[0012] In step (1.2), if the beginning and end of the flood discharge vibration wave time history curve are zero, a linear method is used to perform baseline correction on the vibration time history curve after Butterworth filtering the high-frequency component in step (1.1), and an artificially corrected flood discharge vibration displacement, velocity or acceleration time history curve is obtained; the purpose of artificially correcting the baseline of the vibration wave is to remove drift, low-frequency trend or bias, so that the mean value of the signal is stable at zero or a certain expected value, and the beginning and end are guaranteed to be 0:

[0013] x corr (t) = x(t) - (at + b) # (3)

[0014] Where: x(t) is the original signal; a and b are trend terms fitted by the least squares method.

[0015] Step (2) is to perform numerical simulation on the slope of the engineering site based on the artificial correction of the displacement, velocity or acceleration time history curve of the flood discharge vibration, and obtain the displacement, velocity or acceleration time history curve of the site slope monitoring point during the flood discharge vibration process. It is divided into three parts: establishing a numerical model of the flood discharge vibration site, setting the static boundary and free field boundary, and numerical simulation of the flood discharge vibration. The process is as follows:

[0016] Step (2.1) is to establish a numerical model of the flood discharge vibration site. The modeling focuses on the basic topography of the site, stratum distribution, and fault structure surface distribution. The numerical model of the flood discharge vibration site is established based on the topographic contour lines, geological profiles, stratum and fault distribution of the engineering geological data.

[0017] In step (2.2), the source of flood discharge vibration is determined based on the terrain and the location of hydraulic structures. A static boundary is added at the source of flood discharge vibration to absorb the vibration wave. The static boundary added for the numerical simulation of flood discharge vibration is:

[0018] t n =-ρC p ν n #(4)

[0019] t s =-ρC s ν s #(5)

[0020] Among them, ν n and ν s are the normal and shear components of the velocity at the static boundary; ρ is the mass density; C p and C s are the speeds of p-wave and s-wave.

[0021] Input the displacement, velocity or acceleration artificial flood discharge vibration time history curve generated in step (1) at the vibration source in the form of surface load, add free field boundaries as the four boundaries of the flood discharge vibration site numerical model to reduce the influence of the boundaries on vibration propagation; the free field boundaries added for flood discharge vibration numerical simulation are:

[0022]

[0023] Where A is the influence area of the free-field grid; and is the velocity of the grid points in the main grid of the side boundary in the x, y and z directions; and is the velocity of the grid point in the side free field; and is the free-field grid point force under the stress in the free-field region surrounding the grid point.

[0024] In step (2.3), the finite difference method is used to perform numerical simulation of the site slope flood discharge vibration, and monitoring points are set in the flood discharge vibration site numerical model to detect the displacement, velocity or acceleration historical time-history response curves at the vibration source and the monitoring points.

[0025] Step (3) is to perform Fourier transform on the displacement, velocity or acceleration history response curves of the vibration source and the monitoring points on the slope during the output numerical simulation of flood discharge vibration, and obtain the frequency domain characteristics of the vibration, namely the amplitude-frequency curve A(f), and extract the main frequency of vibration; and use machine learning algorithm and Sobol global sensitivity analysis method to calculate the influence index of each flood discharge vibration source on the vibration of the site slope monitoring point from both time domain and frequency domain. It includes three steps: Fourier transform of the displacement, velocity and acceleration history curves of the site slope monitoring point; based on the machine learning method, with the time domain and frequency domain characteristic curves at the flood discharge vibration source as input and the time domain and frequency domain characteristic curves at the monitoring point as output, to establish a flood discharge vibration linear regression proxy model:

[0026]

[0027] Where, is the predicted value, i.e., the time domain or frequency domain characteristic curve at the monitoring point; β0 is the bias term; β1~β n is the regression coefficient; x1~x n is the input variable;

[0028] The Sobol global sensitivity analysis method is used to calculate the vibration impact index of the flood discharge vibration source on the site slope monitoring point in the time domain and frequency domain:

[0029]

[0030] Among them, ST iThe x axis component of the horizontal and vertical axes of the curve at the input variable vibration source i The global sensitivity index, is the conditional expected value of the model output, ~x i is x i A vector of all input variables except .

[0031] In step (1), the site vibration caused by flood discharge is a narrow-band low-frequency random vibration. Therefore, based on the amplitude and vibration main frequency characteristics of the displacement, velocity or acceleration time history curve of the monitoring point in the in-situ observation of flood discharge vibration or the flood discharge vibration hydroelastic model test, the seismic wave is artificially corrected to generate an artificial flood discharge vibration time history curve of displacement, velocity or acceleration that is consistent with the amplitude and frequency characteristics of the flood discharge vibration.

[0032] In step (1.1), the transfer function of the Butterworth filter is:

[0033]

[0034] Where: s is a complex frequency variable; ω c is the filter cutoff angular frequency; n is the filter order, which is 3 to 4 in the process of artificially correcting ground motion.

[0035] In step (2), the artificial flood discharge vibration time history curve generated in step (1) is input as a surface load at the corresponding vibration source position of the flood discharge vibration site numerical model according to the test observation results, and the flood discharge vibration simulation is performed by the finite difference method to obtain a time history curve including displacement, acceleration and velocity.

[0036] In step (2.2), to ensure that the vibration waves propagate without distortion within the numerical model of the flood discharge vibration site, a static boundary is used locally at the vibration source input position so that the numerical model of the flood discharge vibration site absorbs the vibration waves. At the same time, a free-field boundary is used to simulate the infinite domain outside the model so that the vibration waves are completely transmitted at the model boundary, reducing the influence of the boundary on the numerical calculation results.

[0037] In step (2.3), based on the determination of the vibration source location and boundary conditions, i.e., the input flood discharge vibration wave, the Mohr-Coulomb constitutive model is used for the rock and soil material, and the finite difference method is used to perform numerical simulation of the site slope flood discharge vibration.

[0038] In step (3), manual sampling is performed through the linear regression proxy model of the site slope flood discharge vibration to obtain sample data, and the Sobol global sensitivity analysis method is used to obtain the influence index of the flood discharge vibration source on the site slope measuring point.

[0039] In step (3), spectrum analysis is performed on the vibration source and the time history curve of the slope site monitoring point output by the numerical simulation of the flood discharge vibration. The Fourier transform of the monitoring point flood discharge vibration time history curve and the extraction process of the vibration main frequency include determining the sampling frequency, generating a Hanning window in the Fourier transform process, discrete Fourier transform and vibration main frequency extraction.

[0040] The sampling frequency of the flood discharge vibration time history curve is calculated using Fourier transform:

[0041]

[0042] Among them, f s is the sampling frequency; Δw is the sampling time interval (s).

[0043] In order to accurately perform Fourier transform on the flood discharge vibration time history curve, a Hanning window is generated to reduce the spectrum leakage phenomenon in FFT:

[0044]

[0045] Where: w(n) is the window function weight; N is the number of signal data points, and n is the index of the current data point.

[0046] The vibration frequency components are obtained by performing discrete Fourier transform on the flood discharge vibration time history curve:

[0047]

[0048] Where: X(k) is the jth frequency component, x(n) is the nth time domain data point; N is the signal length (number of data points); j is the imaginary unit, j 2 =-1.

[0049] The frequency axis of the Fourier transform is:

[0050]

[0051] Where f(k) is the corresponding frequency; N is the number of data points; Δt is the sampling time interval (s).

[0052] Extract the main vibration frequency of the flood discharge vibration frequency domain characteristic curve. The flood discharge vibration frequency-amplitude curve A(f) consists of the frequency f(k) and the corresponding vibration frequency component X(k). Use the maximum amplitude sorting to extract the vibration frequency-amplitude curve A(f) obtained by processing the vibration time history curve using formulas (7) to (10) to obtain the main vibration frequency:

[0053] f main =argmaxA(f)#(11)

[0054] Among them, f mainis the main frequency (Hz), and A(f) is the frequency-amplitude curve obtained by numerical simulation and Fourier transform of flood discharge vibration.

[0055] In step (3), the spectrum analysis is performed on the time history curves of the vibration source and the monitoring points of the slope site output by the numerical simulation of the flood discharge site vibration. The horizontal and vertical time and frequency curves at different vibration sources are used as variables, and the time and frequency curves at the monitoring points are used as dependent variables. The linear regression method is established based on the machine learning method to establish the proxy model of the site slope flood discharge vibration. The linear regression process of establishing the proxy model of the site slope flood discharge vibration is as follows:

[0056] For the time domain or frequency domain characteristic curve of flood discharge vibration displacement, velocity or acceleration, the horizontal and vertical axis components of the curve at the vibration source are taken as the independent variables x. i , the horizontal and vertical components of the monitoring point curve are the dependent variable y, and a linear regression proxy model is established for the dependent variable in the time domain and frequency domain:

[0057]

[0058] Where, is the predicted value, i.e., the time domain or frequency domain characteristic curve at the monitoring point; β0 is the bias term; β1~β n is the regression coefficient; x1~x n is the input variable.

[0059] Use the least squares method to solve the linear regression proxy model coefficients:

[0060]

[0061] Where m is the number of samples and n is the number of features.

[0062] In step (3), a proxy model is used to obtain a flood discharge vibration sample data set, and a Sobol global sensitivity analysis based on variance contribution is performed to obtain the influence index of the vibration source on the vibration of the site monitoring point based on time domain analysis and the influence index of the vibration source on the vibration of the site monitoring point based on frequency domain analysis.

[0063] Working principle: The present invention obtains the main vibration frequency range of each vibration source based on the Fourier spectrum of the in-situ test observation results of flood discharge vibration or the model test results. According to the measured vibration main frequency and amplitude, a Butterworth filter is used to filter the high-frequency components of the vibration time history curve, perform baseline correction, and generate an artificial vibration time history curve.

[0064] Based on engineering geological data, a numerical model of the slope flood discharge vibration site is established. Artificial vibration time history curves with different main vibration frequencies and amplitudes are input at the multiple vibration source locations of the flood discharge vibration. Static boundaries and free field boundaries are used for zoning to carry out numerical simulation of the site flood discharge vibration. Monitoring points are set in the numerical model of the flood discharge vibration site to monitor the displacement, velocity and acceleration responses at the vibration sources and measuring points.

[0065] During the numerical simulation of flood discharge vibration, the acceleration time history curves of the vibration source and the monitoring points on the slope site were output, and spectral analysis was performed. A proxy model was established based on the machine learning regression algorithm. The flood discharge vibration sample data set was obtained through sampling the proxy model, and a Sobol global sensitivity analysis was performed to identify the intrinsic correlation between the vibration source and the vibration response of the monitoring points on the slope. The influence index of the horizontal and vertical time and frequency characteristics of each vibration source on the horizontal and vertical vibration of the site monitoring points was calculated in the time domain and frequency domain respectively.

[0066] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0067] (1) The present invention performs a time-history analysis of flood discharge vibration by artificially inputting vibration waves, obtains the time and frequency signals of the vibration source and the slope, establishes an agent model based on a machine learning algorithm, performs a Sobol global sensitivity analysis, and obtains the influence index of each vibration source on the slope monitoring point, thereby solving the problem of multi-vibration source identification of flood discharge vibration.

[0068] (2) The present invention involves time domain and frequency domain characteristics, with less vibration signal loss, which is conducive to using numerical simulation methods and machine learning methods to solve the problem of multi-vibration source identification of flood discharge vibration, avoiding the data loss problem in traditional analysis methods, and providing reference value for slope engineering applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 This is a flow chart of the method for calculating the site impact index of multiple vibration sources of flood discharge vibration according to the present invention;

[0070] Figure 2 Comparison of the earthquake time history curve in the present invention and the modified artificial flood discharge vibration time history curve;

[0071] Figure 3 This is a diagram of the process of filtering high-frequency components in the present invention;

[0072] Figure 4 A vibration time history curve diagram of artificial flood discharge generated based on the earthquake motion time history of the present invention;

[0073] Figure 5 The numerical model of the flood discharge vibration site and the monitoring point diagram of the present invention;

[0074] Figure 6 Output displacement time history curves and spectrum analysis results for the numerical simulation of flood discharge vibration at the vibration source and monitoring points;

[0075] in, Figure 6 (a) is the displacement time history curve of the numerical simulation output of flood discharge vibration at the vibration source and monitoring point;

[0076] Figure 6 (b) The displacement spectrum of the numerical simulation output of flood discharge vibration at the vibration source and monitoring point;

[0077] Figure 7 This is the linear regression prediction result diagram of flood discharge vibration monitoring in time domain and frequency domain;

[0078] Figure 8 This is the analysis result of the influence index of the vibration source on the vibration of the slope measuring point. DETAILED DESCRIPTION

[0079] The method for calculating the impact index of multiple vibration sources of flood discharge vibration on the site of the present invention includes generating an artificial flood discharge vibration time history curve, flood discharge vibration numerical simulation and vibration source impact index analysis based on machine learning algorithm. The process is as follows Figure 1 The specific steps are as follows:

[0080] (1) Generate artificial flood discharge vibration time history curves. Based on the displacement, velocity or acceleration time history curves and amplitudes of the monitoring points in the flood discharge vibration in-situ observation or flood discharge vibration hydroelastic model test, the vibration main frequency of each measuring point is obtained through Fourier transform. The recorded flood discharge vibration waves are relatively few, so the earthquake record time history curve is processed according to the above vibration main frequency and curve amplitude, mainly including filtering high-frequency components and baseline correction. On the basis of the earthquake record time history curve, the displacement, velocity or acceleration time history curve vibration main frequency and amplitude are obtained through correction, which are consistent with the artificial flood discharge vibration time history curve of the flood discharge vibration process.

[0081] In step (1.1), unlike earthquakes, the site vibration caused by flood discharge is a narrow-band, low-frequency random vibration. Therefore, the time history curve of the seismic wave needs to be manually corrected to make it conform to the characteristics of flood discharge vibration. Based on the main vibration frequency range and the main vibration frequency of the vibration source obtained from experimental observations, a Butterworth filter is used to filter the high-frequency components of the vibration time history curve.

[0082] In step (1.2), a linear method is used to perform baseline correction on the vibration time history curve after filtering in step (1.1), and an artificial flood discharge vibration time history curve that conforms to the flood discharge vibration characteristics at different vibration source locations is generated, such as Figure 2-Figure 4 shown.

[0083] Step (2) is to numerically simulate the vibration of the flood discharge site. The study of the impact index of flood discharge vibration on the site requires the detailed modeling of the engineering site. The modeling focuses on the basic topography of the site, the distribution of strata, the distribution of structural surfaces such as faults, and the establishment of a numerical model of the slope flood discharge vibration site based on the topographic contours, geological profiles, strata and fault distributions given by engineering geological data. The flood discharge vibration source is determined based on the topography and the location of hydraulic structures. The artificial vibration time history curve is input at the location of the flood discharge vibration source. The static boundary and free field boundary are used to perform numerical simulation of the site flood discharge vibration. Monitoring points are set in the flood discharge vibration site numerical model to detect the displacement, velocity, and acceleration history response curves at the vibration source and the measuring points.

[0084] Step (2.1), establish the slope flood discharge vibration site numerical model based on the topographic contours, geological profiles, strata and fault distribution given by the engineering geological data.

[0085] Step (2.2): Input the artificial flood discharge vibration time history curve generated in step (1) into the corresponding vibration source position of the flood discharge vibration site numerical model in the form of surface load according to the test observation results, including displacement, acceleration or velocity time history curves, such as Figure 5 To ensure that the vibration wave propagates without distortion within the scope of the flood discharge vibration site model, a static boundary is locally imposed at the vibration source input position to absorb the incident flood discharge vibration wave. At the same time, a free-field boundary is used to simulate the infinite domain outside the model, so that the vibration wave is completely transmitted at the model boundary, reducing the influence of the boundary on the numerical calculation results, as shown in the figure. Figure 5 shown.

[0086] In step (2.3), the finite difference method is used to perform numerical simulation of the flood discharge vibration of the site slope, and monitoring points are set on the site model of the flood discharge site vibration numerical simulation, such as Figure 5 As shown in Figure 2, the monitoring points include multiple vibration source input locations and model site slope locations; the velocity, acceleration and displacement responses of the monitoring points are recorded during the flood discharge vibration simulation, as shown in Figure 2. Figure 6 As shown in (a).

[0087] Step (3) analyzes the vibration source influence index based on the machine learning algorithm. The displacement, velocity, and acceleration history response curves of the vibration source and the slope site monitoring points during the flood discharge vibration numerical simulation are output through step (2). The time history curves are subjected to spectral analysis through Fourier transform. A linear regression proxy model is established based on the machine learning algorithm, and a Sobol global sensitivity analysis is performed to analyze the influence index of each vibration source on the site monitoring point vibration in the time domain and frequency domain.

[0088] Step (3.1), after the simulation is completed, extract the horizontal and vertical acceleration time history curves at the vibration source and the slope monitoring point in the numerical simulation of the flood discharge site vibration, and perform spectrum analysis based on Fourier transform to obtain the amplitude-frequency curves at the flood discharge vibration source and the site slope monitoring point, as shown in Figure 3. Figure 6 (b) shown.

[0089] In step (3.2), based on the machine learning algorithm, a linear regression analysis is performed with the horizontal and vertical time history curves and frequency spectrum at the input vibration source as independent variables and the flood discharge vibration time history curve and frequency spectrum at the site slope measuring point as dependent variables, as shown in the following example: Figure 7 As shown in the figure, a vibration proxy model of site slope flood discharge is established.

[0090] In step (3.3), the flood discharge vibration proxy model of the site slope is used to obtain the flood discharge vibration sample data set, and the Sobol global sensitivity analysis based on variance contribution is performed on the independent variables of the flood discharge vibration. The influence index of the independent variables (horizontal and vertical time history curves and frequency spectrum at each vibration source) on the site slope monitoring point is output, as shown in Table 1. The influence index bar chart of each vibration source is drawn, as shown in Table 1. Figure 8 As shown in the figure, the influence index of each vibration source on the vibration of the site monitoring point is analyzed in the time domain and frequency domain respectively.

[0091] Example

[0092] Based on the test results, the seismic motion time history was corrected, the artificial flood discharge vibration time history curve was generated, a numerical model of the site slope was established, and numerical simulation of flood discharge vibration was carried out. A proxy model was established based on the machine learning regression algorithm, and a Sobol global sensitivity analysis was performed to obtain the influence index of each vibration source on the vibration of the slope monitoring point.

[0093] The steps of the method for calculating the impact index of multiple sources of flood discharge vibration on the site are as follows:

[0094] (1) According to the results of the in-situ test of flood discharge vibration, the main vibration frequency range of the measuring point at the guide wall is 0.7~1.0Hz, and the main vibration frequency range of the measuring point at the stilling pool bottom plate is 0.5~2.0Hz. Based on the main vibration frequency at the guide wall and the stilling pool bottom plate, the high-frequency components of the general earthquake record time history are filtered and baseline correction is performed to generate the artificial flood discharge vibration time history curve, as shown in Figure 2. Figure 2-Figure 4 As shown;

[0095] (2) Establish a numerical model of the slope flood discharge vibration site, apply a static boundary locally at the corresponding vibration source location, input the artificial vibration time history curve as the load, set monitoring points, adopt a free field boundary, and use the finite difference method to simulate the site slope flood discharge vibration. The numerical model of the slope flood discharge vibration site and the location of the monitoring points are as follows: Figure 5 shown.

[0096] (3) Extract the vibration numerical simulation of the flood discharge site, the horizontal and vertical displacement time history curves at the vibration source and the slope site monitoring point, perform Fourier transform, and obtain the time domain and frequency domain characteristics of the vibration source and the monitoring point, such as Figure 6 As shown. Taking the flood discharge vibration time history curve and spectrum at the input vibration source as the independent variable and the flood discharge vibration time history curve and spectrum at the site slope measuring point as the dependent variable, a linear regression analysis was performed using machine learning methods to establish a site slope flood discharge vibration proxy model. The results are shown in Figure 7 As shown in Figure 1, the data set is obtained through the proxy model. The Sobol global sensitivity analysis method is used for flood discharge, and the importance analysis of each vibration source is output. As shown in Table 1, the sensitivity ratio bar chart of each vibration source is obtained. The influence index of each vibration source on the vibration of the site monitoring point is analyzed in the time domain and frequency domain respectively, as shown in Figure 1. Figure 8 shown.

[0097] Table 1 Impact index obtained by sobol global sensitivity analysis

[0098]

Claims

1. A method for calculating the impact index of multiple sources of flood discharge vibration on a site, characterized by: The steps include: Step (1) is to manually correct the seismic wave according to the amplitude and main vibration frequency of the displacement, velocity or acceleration time history curve of the monitoring point, and generate an artificial flood discharge vibration time history curve of displacement, velocity or acceleration that is consistent with the flood discharge vibration amplitude and frequency characteristics; the process of manually correcting the seismic wave is as follows: In step (1.1), use Butterworth to filter the high-frequency components of the vibration time history curve: Among them, f low is the normalized low-frequency cutoff frequency; f high is the normalized high frequency cutoff frequency; f s is the sampling frequency lowcut, highcut is the cutoff frequency of the actual filter; In step (1.2), the baseline of the vibration time history curve after Butterworth filtering is corrected using a linear method: x corr (t)=x(t)-(at+b)#(3) Where: x(t) is the original signal; a and b are trend terms fitted by the least squares method; Step (2) is to establish a numerical model of the flood discharge vibration site and add a static boundary at the vibration source to absorb the vibration wave: t n =-ρC p v n #(4) t s =-ρC s v s #(5) Among them, v n and v s are the normal and shear components of the velocity at the static boundary; ρ is the mass density; C p and C s is the wave velocity of p-wave and s-wave; The artificial flood discharge vibration time history curve generated in step (1) is input in the form of surface load, and the free field boundary is used as the boundary of the flood discharge vibration site numerical model. The free field boundary is: Where A is the influence area of the free-field grid; and is the velocity of the grid points in the main grid of the side boundary in the x, y and z directions; and is the velocity of the grid point in the side free field; and is the free-field grid point force under the stress in the free-field region surrounding the grid point; The finite difference method is used to perform numerical simulation of flood discharge vibration and obtain the historical time-history response curves of displacement, velocity or acceleration of the vibration source and monitoring points; Step (3) is to perform Fourier transform on the output displacement, velocity or acceleration historical time-history response curves of the vibration source and the monitoring point to obtain the amplitude-frequency curve A(f) and extract the main vibration frequency; based on the machine learning method, the time domain and frequency domain characteristic curves at the flood discharge vibration source are used as input and the time domain and frequency domain characteristic curves at the monitoring point are used as output to establish a flood discharge vibration linear regression proxy model: Where, is the predicted value, i.e., the time domain or frequency domain characteristic curve at the monitoring point; β0 is the bias term; β1~β n is the regression coefficient; x1~x n is the input variable; The Sobol global sensitivity analysis method is used to calculate the vibration impact index of the flood discharge vibration source on the site slope monitoring point in the time domain and frequency domain: Among them, ST i The x axis component of the horizontal and vertical axes of the curve at the input variable vibration source i The global sensitivity index, is the conditional expected value of the model output, ~x i is x i A vector of all input variables except .

2. The method for calculating the site impact index of flood discharge vibration from multiple sources according to claim 1, characterized in that: In step (1), based on the amplitude and vibration main frequency characteristics of the displacement, velocity or acceleration time history curve of the monitoring point in the in-situ observation of flood discharge vibration or the flood discharge vibration hydroelastic model test, the seismic wave is artificially corrected to generate an artificial flood discharge vibration time history curve of displacement, velocity or acceleration that is consistent with the flood discharge vibration amplitude and frequency characteristics.

3. The method for calculating the site impact index of flood discharge vibration from multiple sources according to claim 1 is characterized by: In step (1.1), the Butterworth transfer function is: Where s is a complex frequency variable; ω c is the filter cutoff frequency; n is the filter order.

4. The method for calculating the site impact index of flood discharge vibration from multiple sources according to claim 1 is characterized by: In step (2), a numerical model of the flood discharge vibration site is established based on the topographic contours, geological profiles, strata and fault distribution.

5. The method for calculating the site impact index of flood discharge vibration from multiple sources according to claim 1 is characterized by: In step (2), the flood discharge vibration source is determined according to the topography and the location of the hydraulic structure, and a static boundary is used to absorb the vibration wave at the flood discharge vibration source.

6. The method for calculating the site impact index of flood discharge vibration from multiple sources according to claim 1, characterized in that: In step (2), the Mohr-Coulomb constitutive model is used for the rock and soil materials, and the finite difference method is used to perform numerical simulation of the site slope flood discharge vibration.

7. The method for calculating the site impact index of flood discharge vibration from multiple sources according to claim 1, characterized in that: In step (3), the sampling frequency of the flood discharge vibration time history curve is calculated using Fourier transform: Among them, f s is the sampling frequency; Δt is the sampling time interval; Generate a Hanning window from the Fourier transform: Where: w(n) is the window function weight; N is the number of signal data points, and n is the index of the current data point. The vibration frequency components are obtained by performing discrete Fourier transform on the flood discharge vibration time history curve: Where: X(k) is the kth frequency component, x(n) is the nth time domain data point; N is the signal length; j is the imaginary unit; This yields the frequency axis of the Fourier transform: Where f(k) is the corresponding frequency; N is the number of data points; Δt is the sampling time interval; Extract the main vibration frequency of the amplitude-frequency curve A(f) obtained by processing the vibration time history curve using equations (7) to (10): f main =arg maxA(f)#(11) Among them, f main is the main frequency (Hz), and A(f) is the amplitude-frequency curve obtained by numerical simulation and Fourier transform of flood discharge vibration.

8. The method for calculating the site impact index of flood discharge vibration from multiple sources according to claim 4, characterized in that: In step (3), the flood discharge vibration frequency-amplitude curve A(f) is composed of the frequency f(k) and the corresponding vibration frequency component X(k).

9. The method for calculating the site impact index of flood discharge vibration from multiple sources according to claim 1, characterized in that: In step (3), the horizontal and vertical axis components of the time domain and frequency domain characteristic curves at the vibration source are used as independent variables x i , the horizontal and vertical components of the time domain or frequency domain characteristic curve at the monitoring point are the dependent variable y, and a linear regression proxy model is established for the dependent variable: Where, is the predicted value, i.e., the time domain or frequency domain characteristic curve at the monitoring point; β0 is the bias term; β1~β n is the regression coefficient; x1~x n is the input variable.

10. The method for calculating the site impact index of flood discharge vibration from multiple sources according to claim 1, characterized in that: In step (3), the least squares method is used to solve the linear regression proxy model coefficients: Where m is the number of samples and n is the number of features.

Images