Slope quality monitoring and early warning method and system based on seismic oscillation data

Abstract

The invention provides a slope quality monitoring and early warning method and system based on seismic oscillation data, and the method comprises the steps: firstly obtaining a seismic oscillation monitoring data set of a target slope region, covering seismic oscillation waveform data and environment correlation parameters of a plurality of monitoring nodes in a preset time period, and then carrying out the dynamic preprocessing of the seismic oscillation monitoring data set, comprising data alignment, noise filtering and dimension unification, a standardized data set is obtained, then a dynamic association feature set containing seismic oscillation energy distribution, slope response frequency and environment coupling fluctuation features is extracted based on the standardized data, and then a slope quality analysis model is called to carry out state prediction on the feature set. And generating a slope quality evaluation result according to the difference degree between the stable state characteristics and the abnormal fluctuation characteristics, and finally determining an early warning strategy according to the slope quality evaluation result and triggering early warning signal output, thereby realizing effective monitoring and early warning of the slope quality.

Description

Technical Field

[0001] The present invention relates to the technical field of geological disaster warning, and more specifically, to a slope quality monitoring and warning method and system based on seismic motion data. Background Art

[0002] The monitoring and warning of slope quality are crucial for ensuring the safety of infrastructure and people's lives and property. In the prior art, the monitoring of slope quality mainly relies on traditional monitoring means. Some methods only simply monitor the displacement changes of slopes through displacement sensors. This way can only obtain limited information about the displacement of slopes and cannot comprehensively reflect the comprehensive influence of internal and surrounding environmental factors of slopes on their stability.

[0003] In addition, when evaluating slope quality in the prior art, most existing technologies adopt simple threshold judgment methods and lack state prediction processing based on scientific models. They cannot accurately distinguish the stable state and abnormal fluctuations of slopes and are difficult to generate comprehensive and accurate slope quality evaluation results.

[0004] In the warning link, existing warning strategies are often relatively fixed and single, cannot be flexibly adjusted according to different slope quality evaluation results, resulting in insufficient timeliness and effectiveness of warnings, and cannot provide scientific and reasonable countermeasure suggestions for relevant personnel. Summary of the Invention

[0005] In view of the problems mentioned above, in combination with the first aspect of the present invention, embodiments of the present invention provide a slope quality monitoring and warning method based on seismic motion data, and the method includes:

[0006] Obtain a set of seismic motion monitoring data for a target slope area, where the set of seismic motion monitoring data includes seismic motion waveform data collected by multiple monitoring nodes within a preset time period and corresponding environmental correlation parameters;

[0007] Perform dynamic preprocessing operations on the set of seismic motion monitoring data to obtain a standardized set of seismic motion monitoring data, where the dynamic preprocessing operations include data alignment, noise filtering, and dimension unification processing;

[0008] Extract features based on the standardized set of seismic motion monitoring data to generate a set of dynamic correlation features for the target slope area, where the set of dynamic correlation features includes seismic motion energy distribution features, slope response frequency features, and environmental coupling fluctuation features;

[0009] Call a slope quality analysis model to perform state prediction processing on the set of dynamic correlation features to obtain stable state features and abnormal fluctuation features of the target slope area, and generate a slope quality evaluation result according to the difference degree between the stable state features and the abnormal fluctuation features;

[0010] Determine an early warning response strategy based on the slope quality assessment result, and trigger the warning signal output operation of the slope monitoring system according to the early warning response strategy.

[0011] On the other hand, an embodiment of the present invention further provides a slope quality monitoring and early warning system based on ground motion data, including a processor and a machine-readable storage medium. The machine-readable storage medium is connected to the processor. The machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0012] Based on the above aspects, the embodiment of the present invention obtains a ground motion monitoring data set including ground motion waveform data of multiple monitoring nodes in a target slope area and corresponding environmental correlation parameters within a preset time period, and performs dynamic preprocessing operations including data alignment, noise filtering and dimension unification on the ground motion monitoring data set, which can effectively improve the quality and consistency of the data, and avoid the adverse effects on subsequent analysis caused by problems such as data deviation, noise interference and inconsistent dimensions. Based on the standardized ground motion monitoring data set, a dynamic correlation feature set including ground motion energy distribution characteristics, slope response frequency characteristics and environmental coupling fluctuation characteristics is extracted, which describes the dynamic characteristics of the target slope area from multiple key aspects, and can more accurately capture the potential changes and abnormal conditions of the slope. Call the slope quality analysis model to perform state prediction processing on the dynamic correlation feature set, obtain the stable state characteristics and abnormal fluctuation characteristics of the target slope area, and generate a slope quality assessment result based on the difference between the two, realizing the scientific and accurate assessment of the slope quality. Finally, determine the early warning response strategy based on the slope quality assessment result and trigger the warning signal output operation, which can timely and effectively warn of potential dangerous situations of the slope, help relevant personnel take countermeasures in advance, reduce the losses caused by slope disasters, and improve the overall efficiency of slope monitoring and early warning. Description of the Drawings

[0013] Figure 1 It is a schematic execution flow diagram of the slope quality monitoring and early warning method based on ground motion data provided by an embodiment of the present invention.

[0014] Figure 2 It is a schematic diagram of exemplary hardware and software components of the slope quality monitoring and early warning system based on ground motion data provided by an embodiment of the present invention. Detailed Embodiments

[0015] The present invention will be specifically described below in conjunction with the accompanying drawings of the specification. Figure 1It is a schematic flowchart of a slope quality monitoring and early warning method based on ground motion data provided by an embodiment of the present invention. The slope quality monitoring and early warning method based on ground motion data will be introduced in detail below.

[0016] Step S110: Obtain a set of ground motion monitoring data for the target slope area. The set of ground motion monitoring data includes ground motion waveform data collected by multiple monitoring nodes within a preset time period and corresponding environmental correlation parameters.

[0017] In this embodiment, in order to effectively monitor and early warn the quality of the target slope area, the first step is to obtain a set of ground motion monitoring data for the target slope area. The target slope area usually has a complex geological structure and environmental conditions, and its stability is affected by the comprehensive influence of ground motion and various environmental factors. In order to collect relevant data comprehensively and accurately, multiple monitoring nodes can be reasonably distributed in the target slope area. The distribution of these monitoring nodes is determined based on factors such as the topography, geological structure, and potential risk areas that may exist in the target slope area, so as to ensure that the entire target slope area can be covered and key ground motion and environmental information can be captured.

[0018] Each monitoring node is equipped with a ground motion monitoring device for collecting ground motion waveform data and an environmental parameter monitoring sensor for collecting environmental correlation parameters. The ground motion monitoring device can record the vibration conditions generated in the target slope area under the action of ground motion in real time and generate ground motion waveform data. These ground motion waveform data are presented in the form of a time series and contain rich information, such as the amplitude, frequency, and duration of ground motion. The environmental parameter monitoring sensor is used to measure environmental factors closely related to the stability of the target slope area. Common environmental correlation parameters include temperature, humidity, air pressure, etc. These environmental factors will affect the geotechnical properties and mechanical states of the slope, and thus affect the stability of the slope.

[0019] The setting of the preset time period is to ensure that the obtained data is representative and timely. The selection of this time period needs to comprehensively consider factors such as the geological activity law, earthquake activity frequency, and actual monitoring requirements of the target slope area. Within the preset time period, each monitoring node continuously collects data according to the set sampling frequency. The determination of the sampling frequency should be based on the characteristics of the changes in ground motion and environmental parameters to ensure that the changes in the data can be accurately captured.

[0020] Suppose there are n monitoring nodes distributed in the target slope area, and the i-th monitoring node is denoted as Ni (i = 1, 2, …, n). During the preset time period T, the ground motion waveform data collected by each monitoring node Ni can be represented as a time series Di = {di1, di2, …, dim}, where m represents the number of sampling points during the time period T. The corresponding environmental correlation parameters can be represented as a vector Ei = {ei1, ei2, …, eik}, where k represents the number of types of environmental correlation parameters. For example, ei1 can represent temperature, ei2 can represent humidity, ei3 can represent air pressure, etc.

[0021] By summarizing the ground motion waveform data and environmental correlation parameters collected by all monitoring nodes during the preset time period, the ground motion monitoring data sets D = {D1, D2, …, Dn} and E = {E1, E2, …, En} of the target slope area are obtained. Combining these two ground motion monitoring data sets together constitutes the complete ground motion monitoring data set.

[0022] Step S120: Perform dynamic preprocessing operations on the ground motion monitoring data set to obtain a standardized ground motion monitoring data set, where the dynamic preprocessing operations include data alignment, noise filtering, and dimension unification processing.

[0023] After obtaining the ground motion monitoring data set of the target slope area, since there may be problems such as noise interference, data loss, and inconsistent dimensions between different data in these ground motion monitoring data, in order to improve the accuracy and reliability of subsequent analysis and processing, it is necessary to perform dynamic preprocessing operations on it to obtain a standardized ground motion monitoring data set. The dynamic preprocessing operations mainly include three aspects: data alignment, noise filtering, and dimension unification processing.

[0024] Step S121: Perform noise component filtering and elimination operations and time-frequency conversion processing on the ground motion waveform data to generate the ground motion spectrum data of each monitoring node, and perform band energy normalization processing on the ground motion spectrum data to obtain band energy standardized data.

[0025] For the ground motion waveform data, first, noise component filtering and elimination operations are required. The ground motion waveform data may be interfered by various external factors during the collection process, such as electromagnetic interference, mechanical vibration, etc. These interferences will introduce noise components, affecting the data quality and the accuracy of subsequent analysis. To eliminate these noise components, filtering algorithms can be used, such as low-pass filtering, high-pass filtering, or band-pass filtering, etc. Which filtering algorithm to choose specifically needs to be determined according to the frequency characteristics of the noise and the frequency range of the ground motion signal.

[0026] Assume that the ground motion waveform data is Di = {di1, di2, …, dim}, and the filtered ground motion waveform data Di' = {di1', di2', …, dim'} is obtained after being processed by the filtering algorithm.

[0027] Next, perform time-frequency conversion processing on the filtered ground motion waveform data. Time-frequency conversion can convert the ground motion waveform data in the time domain to the frequency domain to better analyze the frequency characteristics of the ground motion signal. Commonly used time-frequency conversion methods include Fourier transform, wavelet transform, etc. Through time-frequency conversion, the ground motion waveform data of each monitoring node is converted into ground motion spectrum data Si = {si1, si2, …, sinf}, where nf represents the number of points in the spectrum data.

[0028] Then, perform band energy normalization processing on the ground motion spectrum data. Since there may be large differences in the energy of different frequency bands, in order to facilitate subsequent analysis and comparison, it is necessary to perform normalization processing on the band energy. The specific steps of band energy normalization processing are as follows: First, divide the ground motion spectrum data into multiple frequency bands according to the frequency range. Assume that it is divided into nb frequency bands, and the energy of each frequency band can be obtained by summing the spectrum data within that frequency band. Let the energy of the j-th frequency band be Ej, then Ej = ∑si (si belongs to the j-th frequency band). Then, calculate the total energy Et of all frequency bands, Et = ∑Ej (j = 1, 2, …, nb). Finally, perform normalization processing on the energy of each frequency band to obtain the band energy standardized data Sj' = Ej / Et (j = 1, 2, …, nb).

[0029] Step S122: Perform data integrity verification processing on the environment-related parameters, perform interpolation filling operations on the missing parameters based on the verification results, and perform numerical range normalization processing on the filled environment-related parameters to obtain environment parameter standardized data.

[0030] In this embodiment, for the environment-related parameters, data integrity verification processing should be performed first. Since environmental parameter monitoring sensors may malfunction, communication may be interrupted, etc., resulting in missing values in the collected environment-related parameters. Data integrity verification is to check whether there are missing values in the environment-related parameter vector Ei = {ei1, ei2, …, eik} of each monitoring node. A judgment condition can be set, such as when a certain parameter value is an invalid value (such as NaN), it is determined that the parameter value is missing.

[0031] Based on the verification result, perform interpolation filling operation on the missing parameters. The purpose of interpolation filling is to estimate the missing parameter values based on the existing valid data to ensure the integrity of the data. Common interpolation methods include linear interpolation, spline interpolation, etc. Suppose the j-th environment-related parameter eij of the i-th monitoring node is missing. The interpolation filling can be performed according to other valid environment-related parameters of this monitoring node and the relevant parameters of adjacent monitoring nodes. Let the filled environment-related parameter vector be Ei' = {ei1', ei2', …, eik'}.

[0032] Considering that different environment-related parameters have different physical dimensions, for example, temperature may be in degrees Celsius, humidity is expressed as a percentage, air pressure is in kilopascals, etc., directly performing numerical range standardization on these parameters will lead to incorrect results. Because the numerical ranges of parameters with different dimensions vary greatly, if the influence of dimensions is not eliminated, some parameters may be over-compressed or expanded during the standardization process, thus affecting the accuracy of subsequent analysis. Therefore, before performing numerical range standardization, it is necessary to first perform non-dimensionalization on the filled environment-related parameters.

[0033] Specifically, the Z-score standardization method is used here for non-dimensionalization. First, for each environment-related parameter, calculate its mean and standard deviation in the historical data. For each parameter eij' in the filled environment-related parameter vector Ei', subtract the historical mean of this parameter from eij', and then divide the obtained difference by the historical standard deviation of this parameter to obtain the standardized parameter eij”. In this way, each environment-related parameter is converted into a non-dimensional numerical value, eliminating the influence of dimensions on the data, enabling different types of environment-related parameters to be analyzed and processed on a unified scale, and finally obtaining the environment parameter standardized data Ei” = {ei1”, ei2”, …, eik”}.

[0034] Step S123: Further perform global standardization on the band energy standardized data and the environment parameter standardized data, then perform timestamp alignment processing to generate standardized data units corresponding to each monitoring node one by one, and merge each standardized data unit to generate the standardized ground motion monitoring data set.

[0035] After obtaining the band energy standardized data and the environment parameter standardized data, further global standardization needs to be performed on them. The purpose of global standardization is to ensure that different types of data have the same scale and distribution in the entire data set for subsequent comprehensive analysis. A similar standardization method as in step S122 can be used to perform unified standardization on the band energy standardized data and the environment parameter standardized data.

[0036] Suppose the frequency band energy standardized data is S = {S1', S2', …, Sn'}, and the environmental parameter standardized data is E” = {E1”, E2”, …, En”}. After global standardization processing, S' = {S1”, S2”, …, Sn”} and E”' = {E1”', E2”', …, En”'} are obtained respectively.

[0037] Next, perform timestamp alignment processing. Since there may be time differences in the acquisition of ground motion waveform data and environmental correlation parameters, it is necessary to align their timestamps to ensure that the data at the same time point can correspond. For example, by finding the common part of the timestamps, different data can be arranged and matched in chronological order.

[0038] After timestamp alignment processing, standardized data units Ui = {Si”, Ei”'}(i = 1, 2, …, n) corresponding to each monitoring node are generated. Finally, the individual standardized data units are merged, that is, all Uis are combined together to obtain the standardized ground motion monitoring data set U = {U1, U2, …, Un}.

[0039] Step S130: Extract features based on the standardized ground motion monitoring data set to generate a dynamic correlation feature set of the target slope area, where the dynamic correlation feature set includes ground motion energy distribution features, slope response frequency features, and environmental coupling fluctuation features.

[0040] After obtaining the standardized ground motion monitoring data set, in order to deeply analyze the stability of the target slope area, it is necessary to extract relevant features from these data to generate a dynamic correlation feature set of the target slope area. The dynamic correlation feature set mainly includes ground motion energy distribution features, slope response frequency features, and environmental coupling fluctuation features.

[0041] Step S131: Perform energy integration processing on the frequency band energy standardized data of each monitoring node to obtain the ground motion energy distribution sequence of each monitoring node, and calculate the energy difference degree between adjacent monitoring nodes based on the ground motion energy distribution sequence to generate energy distribution difference features.

[0042] First, perform energy integration processing on the frequency band energy standardized data of each monitoring node. The frequency band energy standardized data reflects the energy distribution of different frequency bands. Through energy integration, the change of the total energy of each monitoring node in the entire frequency band range over time can be obtained, that is, the ground motion energy distribution sequence. Suppose the frequency band energy standardized data of the i-th monitoring node is Si” = {si1”, si2”, …, sinb”}, and perform energy integration processing on it to obtain the ground motion energy distribution sequence Ei” = ∑sij”(j = 1, 2, …, nb).

[0043] Then, based on the seismic ground motion energy distribution sequence, calculate the energy difference degree between adjacent monitoring nodes. The energy difference degree between adjacent monitoring nodes can reflect the uneven distribution of seismic ground motion energy in space, which is of great significance for analyzing the stability of the slope. Let two adjacent monitoring nodes be Ni and Ni+1, and their seismic ground motion energy distribution sequences be Ei” and Ei+1” respectively. The energy difference degree can be obtained by calculating the absolute value of the difference between the two, that is, ΔEi = |Ei” - Ei+1”|.

[0044] Combining the energy difference degrees between all adjacent monitoring nodes generates the energy distribution difference feature ΔE = {ΔE1, ΔE2, …, ΔEn-1}.

[0045] Step S132: Perform main frequency component extraction processing on the frequency band energy normalized data to obtain the dominant frequency set of each monitoring node, and calculate the resonance response probability of the slope structure according to the dominant frequency set, generating the slope response frequency feature.

[0046] The purpose of performing main frequency component extraction processing on the frequency band energy normalized data is to find out the frequency components where the energy of the seismic ground motion signal mainly concentrates at each monitoring node, that is, the dominant frequencies. By sorting the frequency band energy normalized data, the frequencies corresponding to the frequency bands with larger energy values can be found, and these frequencies are used as the dominant frequencies. Suppose the frequency band energy normalized data of the i-th monitoring node is Si” = {si1”, si2”, …, sinb”}, sort it in descending order of energy value, and select the frequencies corresponding to the first m frequency bands with larger energy values to obtain the dominant frequency set Fi = {fi1, fi2, …, fim} of this monitoring node.

[0047] Then, calculate the resonance response probability of the slope structure according to the dominant frequency set. The resonance response probability of the slope structure is closely related to the natural frequency of the slope and the dominant frequency of the seismic ground motion signal. When the dominant frequency of the seismic ground motion signal is close to the natural frequency of the slope, the slope is prone to resonance, thereby increasing the risk of slope instability. A probability model can be established to calculate the resonance response probability. Suppose the natural frequency set of the slope is F0 = {f01, f02, …, f0p}. For the dominant frequency set Fi of each monitoring node, calculate the frequency matching degree between it and the natural frequency set F0 of the slope. The frequency matching degree can be measured by calculating the absolute value of the frequency difference, that is, |fi - f0j| (i = 1, 2, …, m; j = 1, 2, …, p). When the frequency difference is less than a certain threshold, it is considered that there is a frequency match. Count the number of frequency matches, and then calculate the resonance response probability Pi according to the conventional probability calculation formula in the related technology.

[0048] Combining the resonance response probabilities of all monitoring nodes generates the slope response frequency feature P = {P1, P2, …, Pn}.

[0049] Step S133: Perform dynamic correlation analysis on the standardized environmental parameter data to determine the coupling correlation degree between environmental parameter changes and the distribution of ground motion energy, and generate an environmental coupling fluctuation feature based on the coupling correlation degree.

[0050] Perform dynamic correlation analysis on the standardized environmental parameter data to determine the coupling correlation degree between environmental parameter changes and the distribution of ground motion energy. Changes in environmental factors can affect the geotechnical properties and mechanical state of the slope, thereby affecting the distribution of ground motion energy. Therefore, analyzing the coupling correlation degree between environmental parameter changes and the distribution of ground motion energy is of great significance for understanding the stability of the slope.

[0051] Step S1331: Perform sliding window slicing on the standardized environmental parameter data according to a preset time granularity to generate a set of environmental parameter time series slices corresponding to each monitoring node, where each environmental parameter time series slice contains a standardized numerical sequence of temperature, humidity, and air pressure parameters at consecutive time steps.

[0052] First, perform sliding window slicing on the standardized environmental parameter data according to a preset time granularity. The selection of the preset time granularity should be determined according to the characteristics of environmental parameter changes and the needs of analysis. Assume the preset time granularity is Δt and the size of the sliding window is W. For the standardized environmental parameter data Ei”' = {ei1”', ei2”', …, eik”'} of the i-th monitoring node, perform sliding window slicing in chronological order. Each sliding window contains a standardized numerical sequence of environmental parameters at consecutive time steps, generating a set of environmental parameter time series slices Ci = {ci1, ci2, …, cinw} corresponding to each monitoring node, where nw represents the number of slices, and each environmental parameter time series slice cij contains a standardized numerical sequence of parameters such as temperature, humidity, and air pressure at consecutive time steps.

[0053] Step S1332: Slice the ground motion energy distribution sequence according to the same time granularity, and allow the environmental parameter time series slices to slide and align within a preset lag time range to generate a set of ground motion energy time series slices considering the lag effect of environmental parameters.

[0054] Slice the ground motion energy distribution sequence according to the same preset time granularity Δt. Assume the ground motion energy distribution sequence of the i-th monitoring node is Ei”, and perform slicing in chronological order to obtain a set of ground motion energy time series slices Di = {di1, di2, …, dinw}.

[0055] Considering that there may be a lag effect in the influence of environmental parameter changes on the seismic ground motion energy distribution, the time series slices of environmental parameters are allowed to slide and align within a preset lag time range. The preset lag time range can be set according to the actual situation, for example, set to ±Δt'. By adjusting the relative positions of the time series slices of environmental parameters and the time series slices of seismic ground motion energy within the preset lag time range, the best alignment method is found, and a set of time series slices of seismic ground motion energy considering the lag effect of environmental parameters Di'={di1', di2',…, dinw'} is generated.

[0056] Step S1333: Perform dynamic correlation analysis on the time series slices of environmental parameters and the time series slices of seismic ground motion energy of the same monitoring node, and combine the linear Pearson correlation coefficient and the non - linear mutual information index to generate a comprehensive correlation strength matrix for each parameter dimension and the energy distribution dimension. Among them, the element value in the comprehensive correlation strength matrix is the weighted average of the linear and non - linear correlation strengths.

[0057] When performing dynamic correlation analysis on the time series slices of environmental parameters and the time series slices of seismic ground motion energy of the same monitoring node, it is necessary to combine the linear Pearson correlation coefficient and the non - linear mutual information index. Considering that the linear Pearson correlation coefficient and the non - linear mutual information index have different dimensions and value ranges, the value range of the linear Pearson correlation coefficient is from - 1 to 1, which mainly measures the linear correlation between two variables; while the value of the non - linear mutual information index is greater than or equal to 0 and is used to capture the non - linear relationship between variables. Therefore, before calculating the comprehensive correlation strength, this embodiment needs to normalize the linear Pearson correlation coefficient and the non - linear mutual information index respectively. For the linear Pearson correlation coefficient, a mapping method is used to normalize it to the range of 0 to 1. The specific method is to add 1 to the linear Pearson correlation coefficient and then divide by 2, so that the linear Pearson correlation coefficient originally between - 1 and 1 is mapped to the interval of 0 to 1, denoted as the normalized linear Pearson correlation coefficient ri'jl. For the non - linear mutual information index, each index value is divided by the maximum value of the index in all data to normalize it to the range of 0 to 1, denoted as the normalized non - linear mutual information index ri'jn.

[0058] After normalization, the weighted average of the normalized linear Pearson correlation coefficient ri'jl and the normalized non - linear mutual information index ri'jn is performed. A weighting coefficient α is set, and its value range is between 0 and 1, which is used to adjust the proportion of the linear and non - linear correlation strengths in the comprehensive correlation strength. Multiply α by the normalized linear Pearson correlation coefficient ri'jl, then subtract α multiplied by the normalized non - linear mutual information index ri'jn from 1, and add these two results to obtain the comprehensive correlation strength rij.

[0059] Combining the comprehensive correlation strengths between each environmental parameter dimension and the energy distribution dimension of the same monitoring node generates the comprehensive correlation strength matrix \(R_i = [r_{ij}]\) (\(i = 1, 2, \ldots, k\); \(j = 1, 2, \ldots, n_w\)) of each parameter dimension and the energy distribution dimension, where each element value \(r_{ij}\) in the matrix is the weighted average of the linear and non-linear correlation strengths obtained after normalization and weighted average processing.

[0060] Step S1334: Perform a max-pooling operation across parameter dimensions on the real-time correlation strength matrix to extract the maximum coupling correlation strength between environmental parameters and ground motion energy distribution at each time step, generating a preliminary coupling correlation degree sequence.

[0061] Perform a max-pooling operation across parameter dimensions on the comprehensive correlation strength matrix \(R_i\). The purpose of the max-pooling operation is to extract the maximum coupling correlation strength between environmental parameters and ground motion energy distribution at each time step. For each column of the comprehensive correlation strength matrix \(R_i\), find the maximum value among them, i.e., \(r_{ij\_max}=\max(r_{ij})\) (\(i = 1, 2, \ldots, k\)).

[0062] Combine the maximum coupling correlation strengths at all time steps to generate a preliminary coupling correlation degree sequence \(R_i'=\{r_{i1\_max}, r_{i2\_max}, \ldots, r_{in_w\_max}\}\).

[0063] Step S1335: Calculate the regional propagation consistency index of the coupling correlation strength based on the spatial gradient change of the preliminary coupling correlation degree sequence between adjacent monitoring nodes, and perform spatial smoothing and correction processing on the preliminary coupling correlation degree sequence to generate a corrected dynamic coupling correlation degree set.

[0064] In this embodiment, the coupling correlation strengths between adjacent monitoring nodes should have a certain continuity and consistency. If there are large differences, it may be caused by local sensor anomalies or other interference factors. Let two adjacent monitoring nodes be \(N_i\) and \(N_{i + 1}\), and their preliminary coupling correlation degree sequences be \(R_i'\) and \(R_{i + 1}'\) respectively. Calculating the spatial gradient between them can be achieved by calculating the difference in the coupling correlation degrees at corresponding time steps. For each time step \(t\), calculate the gradient value \(\Delta r_{it}=|r_{it}'-r_{i + 1t}'|\), where \(r_{it}'\) and \(r_{i + 1t}'\) are the coupling correlation degree values of \(R_i'\) and \(R_{i + 1}'\) at time step \(t\) respectively. Combine the gradient values at all time steps to obtain the spatial gradient sequence \(\Delta R_i=\{\Delta r_{i1}, \Delta r_{i2}, \ldots, \Delta r_{in_w}\}\).

[0065] Next, calculate the regional propagation consistency index of the coupling correlation strength according to the spatial gradient sequence. The statistical analysis of the spatial gradient sequence can be carried out, such as calculating the mean μΔRi and the standard deviation σΔRi of the spatial gradient sequence. The regional propagation consistency index γi can be defined as a function related to the mean and the standard deviation. For example, γi = f(μΔRi, σΔRi), and the function f here can be selected according to the actual situation. A common way is γi = 1 / (1 + μΔRi + σΔRi). The closer the index value is to 1, the more consistent the coupling correlation strength between adjacent monitoring nodes is, and the better the continuity of regional propagation is.

[0066] In an actual slope monitoring scenario, for example, slope nodes in mountainous areas may be linearly distributed. If the Euclidean distance is used to calculate the topology, the actual physical paths may vary greatly due to complex terrain, which may lead to the selection of K-nearest neighbor nodes including irrelevant nodes, making the mean replacement operation based on K-nearest neighbor nodes ineffective.

[0067] To solve this problem, this embodiment uses the geographically weighted distance to replace the Euclidean distance. The geographically weighted distance takes into account actual factors such as terrain undulation and can more accurately reflect the actual physical distance between monitoring nodes. The geographically weighted distance can be calculated by obtaining the geographical location information (such as longitude and latitude) of the monitoring nodes and the elevation data of the terrain, and combining with the geographical information system (GIS) technology. For example, using the Dijkstra algorithm or the A* algorithm, considering factors such as the slope and obstacles of the terrain, calculate the actual path length between the monitoring nodes as the geographically weighted distance.

[0068] At the same time, in order to adapt to the changes in the density of monitoring nodes in different regions, the K value is dynamically adjusted. In the region with a large node density, the K value is appropriately increased to obtain more information of the nearest neighbor nodes and improve the accuracy of mean replacement; in the region with a small node density, the K value is decreased to avoid introducing irrelevant nodes. A node density threshold can be preset according to the distribution of the monitoring nodes. When the node density in a certain region is greater than the threshold, the K value is increased; when the node density is less than the threshold, the K value is decreased.

[0069] Based on the adjusted topological distance and K value, perform spatial smoothing correction processing on the preliminary coupling correlation degree sequence. First, perform a first-order difference calculation on the preliminary coupling correlation degree sequence in the time dimension to generate a coupling correlation degree change trend sequence for each monitoring node. For the preliminary coupling correlation degree sequence Ri' = {ri1_max, ri2_max,..., rinw_max} of the i-th monitoring node, its first-order difference sequence ΔRi' = {Δri1', Δri2',..., Δrinw-1'}, where Δrit' = rit+1_max - rit_max (t = 1, 2,..., nw - 1).

[0070] Then, based on the cosine similarity between the change trend sequences of the coupling correlation degrees of adjacent monitoring nodes, the regional propagation consistency index is calculated. Let the change trend sequences of the coupling correlation degrees of adjacent monitoring nodes Ni and Ni+1 be ΔRi' and ΔRi+1' respectively. The cosine similarity cosθi between them can be obtained by calculating the dot product of vectors and the vector norm, that is, cosθi = (ΔRi'·ΔRi+1') / (||ΔRi'||×||ΔRi+1'||). When the cosine similarity is lower than the preset threshold θ0, the corresponding node pair is marked as a spatial abnormal fluctuation region.

[0071] For the initial coupling correlation degree of the spatial abnormal fluctuation region, perform a mean replacement operation based on the K nearest neighbor nodes to eliminate the coupling correlation degree jump noise caused by local sensor anomalies. For the node Ni marked as a spatial abnormal fluctuation region, find its K nearest neighbor nodes. Let the initial coupling correlation degree sequences of these nearest neighbor nodes be Rj' (j ∈ Nk, Nk represents the set of K nearest neighbor nodes of Ni). For each value rit' in the initial coupling correlation degree sequence Ri' of Ni, replace it with the mean of the coupling correlation degrees of its K nearest neighbor nodes at the same time step, that is, rit” = (1 / K)×∑rjt' (j ∈ Nk).

[0072] Next, use the Gaussian kernel function to perform spatial neighborhood weighted smoothing on the corrected initial coupling correlation degree, where the kernel bandwidth is adaptively adjusted according to the topological distance of the monitoring nodes. Let the corrected initial coupling correlation degree sequence be Ri” = {ri1”, ri2”, …, rinw”}. For the coupling correlation degree value rit” at each time step t, its spatially neighborhood weighted smoothed result rit”' can be calculated as follows: First, determine the spatial neighborhood node set Nj of Ni (including Ni itself). For each neighborhood node Nj, calculate its geographical weighted distance dij from Ni. Then, calculate the Gaussian kernel function value wij = exp(-dij2 / (2×h2)) according to the geographical weighted distance dij, where h is the kernel bandwidth, and h can be adaptively adjusted according to the topological distance distribution of the monitoring nodes. For example, h can be taken as the mean of the geographical weighted distances of the neighborhood nodes. Finally, rit”' = (∑wij×rjt”) / (∑wij) (j ∈ Nj).

[0073] Finally, perform exponential weighted moving average filtering on the smoothed dynamic coupling correlation degree in the time dimension to suppress high-frequency fluctuation noise and retain the trend component. Let the smoothed dynamic coupling correlation degree sequence be Ri''' = {ri1''', ri2''', …, rinw'''}, and the result after exponential weighted moving average filtering be Ri'''' = {ri1'''',' ri2''''', …, rinw''''}, where ri1'''' = ri1'''. For t > 1, rit'''' = β × rit''' + (1 - β) × rit-1'''', where β is the smoothing coefficient with a value range of (0, 1) and can be adjusted according to the actual situation. Combining the corrected dynamic coupling correlation degree sequences of all monitoring nodes together generates the corrected dynamic coupling correlation degree set R'''' = {R1'''', R2'''', …, Rn''''}.

[0074] Step S1336: Input the dynamic coupling correlation degree set into a preset coupling strength grading model, and map the coupling correlation degrees of each monitoring node to a unified feature dimension through a fully connected layer to generate an environmental coupling fluctuation feature vector matching the seismic ground motion energy distribution feature dimension.

[0075] Input the corrected dynamic coupling correlation degree set R'''' into a preset coupling strength grading model. The coupling strength grading model is a trained AI model, and its main function is to process the coupling correlation degrees of each monitoring node so that they can better reflect the coupling relationship between environmental parameters and seismic ground motion energy distribution.

[0076] The input of the coupling strength grading model is the corrected dynamic coupling correlation degree set R'''', where the dynamic coupling correlation degree sequence Ri'''' of each monitoring node is a multi-dimensional vector. Through the fully connected layer in the coupling strength grading model, the coupling correlation degrees of each monitoring node are mapped to a unified feature dimension. The role of the fully connected layer is to perform linear transformation and non-linear activation on the input coupling correlation degrees so that the coupling correlation degrees of different monitoring nodes can be represented in the same feature space. Let the weight matrix of the fully connected layer be W and the bias vector be b. For the dynamic coupling correlation degree sequence Ri'''' of the i-th monitoring node, the result after passing through the fully connected layer is Zi = f(W × Ri'''' + b), where f is a non-linear activation function, such as the ReLU function.

[0077] Combine the results of all monitoring nodes after passing through the fully connected layer together to generate a new feature vector Z = {Z1, Z2, …, Zn}. Then, adjust the dimension of the feature vector Z to match the seismic ground motion energy distribution feature dimension. This can be achieved through operations such as interpolation and dimensionality reduction, and finally generate an environmental coupling fluctuation feature vector Rf matching the seismic ground motion energy distribution feature dimension.

[0078] Step S134: Perform spatial topological association processing on the energy distribution difference feature, slope response frequency feature, and environmental coupling fluctuation feature to generate a dynamic association feature set for the target slope area.

[0079] Perform spatial topological association processing on the energy distribution difference feature ΔE, slope response frequency feature P, and environmental coupling fluctuation feature vector Rf. The purpose of spatial topological association processing is to consider the spatial position relationship between monitoring nodes and integrate different types of features to more comprehensively reflect the dynamic characteristics of the target slope area.

[0080] First, according to the actual physical positions of the monitoring nodes, construct a spatial topological relationship graph G = (V, E), where V represents the set of monitoring nodes and E represents the connection relationship between the monitoring nodes. The connection relationship can be determined based on factors such as the distance between monitoring nodes and signal correlation.

[0081] For the energy distribution difference feature ΔE, slope response frequency feature P, and environmental coupling fluctuation feature vector Rf, associate them with the spatial topological relationship graph G respectively. Each feature can be corresponded to the nodes in the spatial topological relationship graph according to the monitoring node numbers to obtain the feature values corresponding to each node. For example, for the energy distribution difference feature ΔE, correspond ΔEi to the node Ni in the spatial topological relationship graph.

[0082] Then, use a graph convolutional network (GCN) to process the associated features. The graph convolutional network can effectively capture the spatial topological information between monitoring nodes, and propagate and aggregate the features. Let the input feature matrix be X = [ΔE; P; Rf], the adjacency matrix of the graph convolutional network be A, and the weight matrix be W. After one layer of processing by the graph convolutional network, the output feature matrix X' can be calculated as follows: X' = f(A × X × W), where f is a non-linear activation function.

[0083] After multiple processes of the graph convolutional network, different types of features are fully fused and propagated, and finally a dynamic association feature set F = {F1, F2,..., Fn} for the target slope area is generated, where Fi represents the dynamic association feature vector corresponding to the i-th monitoring node.

[0084] Step S140: Invoke the slope quality analysis model to perform state prediction processing on the dynamic association feature set, obtain the stable state feature and abnormal fluctuation feature of the target slope area, and generate a slope quality assessment result according to the difference degree between the stable state feature and the abnormal fluctuation feature.

[0085] After obtaining the dynamic association feature set of the target slope area, call the slope quality analysis model to perform state prediction processing on it. The slope quality analysis model is a trained AI model, and its purpose is to predict the stable state and abnormal fluctuation conditions of the target slope area based on the input dynamic association feature set.

[0086] Step S141: Input the dynamic association feature set into the spatio-temporal encoding network of the slope quality analysis model, and model the global dependence relationship of the spatial topological association data in the dynamic association feature set through the multi-head self-attention mechanism to generate a context encoding vector containing cross-node spatio-temporal interaction information.

[0087] Input the dynamic association feature set F into the spatio-temporal encoding network of the slope quality analysis model. The main role of the spatio-temporal encoding network is to process the spatial topological association data in the dynamic association feature set and capture the global dependence relationship and spatio-temporal interaction information between monitoring nodes.

[0088] First, perform time dimension slicing processing on the dynamic association feature set to generate a spatio-temporal feature slice set divided by a preset time window. Let the size of the preset time window be T, and the length of the dynamic association feature set F in the time dimension be L, then it can be divided into N = L / T spatio-temporal feature slice sets Ft = {F1t, F2t,..., Fnt} (t = 1, 2,..., N), where Fit represents the dynamic association feature vector of the i-th monitoring node in the t-th time window.

[0089] Based on the actual physical position distance and historical ground motion signal correlation of the monitoring nodes in each spatio-temporal feature slice, construct the weighted connection relationship between the monitoring nodes to generate a node adjacency matrix and a feature node vector reflecting spatial proximity and signal association strength. For each spatio-temporal feature slice Ft, calculate the distance matrix D = [dij] according to the actual physical position distance of the monitoring nodes, where dij represents the physical distance between nodes Ni and Nj. At the same time, calculate the correlation matrix C = [cij] according to the historical ground motion signal correlation, where cij represents the ground motion signal correlation between nodes Ni and Nj. Then, fuse the distance matrix and the correlation matrix to obtain the weighted connection relationship matrix W = [wij], where wij can be calculated through a fusion function f(dij, cij), for example, wij = α×(1 / dij)+(1-α)×cij, and α is the weighting coefficient.

[0090] According to the weighted connection relationship matrix W, generate the node adjacency matrix A = [aij], where aij = 1 indicates that there is a connection relationship between nodes Ni and Nj, and aij = 0 indicates that there is no connection relationship. At the same time, use the dynamic association feature vector Fit of each monitoring node as the feature node vector.

[0091] Input the node adjacency matrix A and the eigen-node vector into the multi-head self-attention mechanism. The multi-head self-attention mechanism can calculate multiple attention heads in parallel, and each attention head can capture dependencies in different aspects. For each attention head, first obtain the query vector Q, key vector K, and value vector V by linearly transforming the eigen-node vector. Let the eigen-node vector be X, the query vector Q = X × WQ, the key vector K = X × WK, and the value vector V = X × WV, where WQ, WK, and WV are learnable weight matrices.

[0092] Then, calculate the attention scores. For each node pair (i, j), the attention score sij can be obtained by the dot product of the query vector and the key vector, i.e., sij = Qi · Kj. To avoid the dot product result from being too large, scale the attention scores, i.e., sij' = sij / √d, where d is the dimension of the query vector and the key vector.

[0093] Next, normalize the attention scores through the softmax function to obtain the attention weights aij = softmax(sij'). Finally, perform weighted aggregation on the value vectors according to the attention weights to obtain the output vector Oi of each node, i.e., Oi = ∑aij × Vj.

[0094] Concatenate the output vectors of multiple attention heads to obtain the output of the multi-head self-attention mechanism. Then, perform linear transformation and residual connection on the output of the multi-head self-attention mechanism to obtain the final node-level attention weight matrix.

[0095] Based on the node-level attention weight matrix, perform weighted aggregation on the eigen-node vector to generate the spatial aggregation feature vector within each time window. For each time window t, perform matrix multiplication on the node-level attention weight matrix and the eigen-node vector to obtain the spatial aggregation feature vector St = [S1t, S2t, …, Snt], where Sit represents the spatial aggregation feature vector of the i-th monitoring node in the t-th time window.

[0096] Perform positional encoding and time dimension convolution fusion on the spatial aggregation feature vectors of multiple time windows. The purpose of positional encoding is to introduce information in the time dimension so that the model can distinguish features at different time steps. The sine-cosine positional encoding method can be used to add positional encoding information to the spatial aggregation feature vector of each time window. Then, perform convolution fusion on the spatial aggregation feature vector after adding positional encoding through the convolution layer in the time dimension to capture the feature changes in the time dimension. Finally, generate the context encoding vector C = {C1, C2, …, Cn} containing the temporal evolution law, where Ci represents the context encoding vector corresponding to the i-th monitoring node.

[0097] Step S142: Input the context encoding vector into the prediction branch of the slope quality analysis model. Extract the seismic motion evolution patterns at different time scales through the dilated temporal convolutional network, output the predicted sequence of energy distributions for multiple future time steps, and calculate the smoothness index characterizing the slope structure stability based on the fluctuation consistency at each time step in the predicted energy distribution sequence, thereby generating the stable state feature.

[0098] Input the context encoding vector C into the prediction branch of the slope quality analysis model. The main role of the prediction branch is to predict the seismic motion energy distribution for multiple future time steps based on the context encoding vector and evaluate the stable state of the slope.

[0099] First, extract the seismic motion evolution patterns at different time scales through the dilated temporal convolutional network (TCN). The dilated temporal convolutional network can capture feature information at different time scales through different dilation factors. Assume the input context encoding vector is C, the convolutional kernel of the dilated temporal convolutional network is K, and the dilation factor is d. Then the output Y after one layer of the dilated temporal convolutional network processing can be calculated as follows: Y = f(∑K[i] × C[t - i×d]), where f is the non-linear activation function, i represents the index of the convolutional kernel, and t represents the time step.

[0100] Through the processing of multiple layers of the dilated temporal convolutional network, extract the seismic motion evolution patterns at different time scales. Then, input the extracted features into the fully connected layer to output the predicted sequence of energy distributions for multiple future time steps E' = {E1', E2', …, Em'}, where m represents the number of future prediction time steps, and Et' represents the predicted energy distribution vector at the t-th time step.

[0101] Calculate the smoothness index characterizing the slope structure stability based on the fluctuation consistency at each time step in the predicted energy distribution sequence. The fluctuation consistency can be measured by calculating the difference between the energy distribution vectors at adjacent time steps in the predicted energy distribution sequence. Assume Et' and Et+1' are the predicted energy distribution vectors at the t-th and (t + 1)-th time steps respectively, and the difference between them can be calculated by the Euclidean distance, i.e., ΔEt = ||Et' - Et+1'||.

[0102] Conduct statistical analysis on the difference values for all adjacent time steps, such as calculating the mean μΔE and the standard deviation σΔE. The smoothness index S can be defined as a function related to the mean and the standard deviation. For example, S = 1 / (1 + μΔE + σΔE). The closer the value of this index is to 1, the smoother the fluctuation of the predicted energy distribution sequence, and the more stable the slope structure. Take the smoothness index S as the stable state feature.

[0103] Step S143: Input the context encoding vector into the anomaly detection branch of the slope quality analysis model. Reconstruct the context encoding vector in the latent space through a variational autoencoder to generate a reconstructed feature vector that matches the historical normal state distribution. Then, use the wavelet packet decomposition algorithm to calculate the residual gradient of the energy distribution of the original encoding vector and the reconstructed feature vector in each frequency band, and extract the frequency band components and corresponding spatial positions where the residual gradient exceeds the historical fluctuation threshold to generate the anomaly fluctuation feature.

[0104] Input the context encoding vector C into the anomaly detection branch of the slope quality analysis model. The main function of the anomaly detection branch is to detect whether there are abnormal fluctuations in the target slope area.

[0105] First, reconstruct the context encoding vector in the latent space through a variational autoencoder (VAE). The variational autoencoder consists of an encoder network and a decoder network. Input the context encoding vector C into the encoder network of the variational autoencoder. The encoder network maps the input features to the mean vector μ and variance vector σ2 in the latent space through a multi-layer perceptron. Let the weight matrix of the encoder network be W1 and the bias vector be b1, then the mean vector μ = f1(W1×C + b1), and the variance vector σ2 = f2(W1×C + b1), where f1 and f2 are non-linear activation functions.

[0106] Based on the mean vector μ and variance vector σ2, perform random sampling to generate a latent space vector z that follows a Gaussian distribution. The sampling process can be achieved through the reparameterization trick, that is, z = μ + ε×√σ2, where ε is a random vector sampled from the standard normal distribution.

[0107] Input the latent space vector z into the decoder network of the variational autoencoder. The decoder network gradually restores the feature dimension through deconvolution layers and fully connected layers to generate a reconstructed feature vector C'. Let the weight matrix of the decoder network be W2 and the bias vector be b2, then the reconstructed feature vector C' = f3(W2×z + b2), where f3 is a non-linear activation function.

[0108] Unfold the original context encoding vector C and the reconstructed feature vector C' into time series signals according to the time step, and perform multi-scale wavelet packet decomposition on the time series signals. Wavelet packet decomposition can decompose the time series signal into components in different frequency bands, which is convenient for analyzing the energy distribution of the signal in different frequency bands. Let the scale of wavelet packet decomposition be S, and there are multiple frequency bands at each scale. For the time series signal Cc after unfolding the original context encoding vector C and the time series signal Cr after unfolding the reconstructed feature vector C', perform S-layer wavelet packet decomposition respectively.

[0109] In the wavelet packet decomposition of the s-th layer (s = 1, 2, …, S), the time series signal is decomposed into multiple band signals. Let the signal of Cc in the k-th band of the s-th layer be Ccsk, and the signal of Cr in the k-th band of the s-th layer be Crsk. Calculate the energy of each band signal, and the calculation of energy can be achieved by the sum of squares of the band signal. Let Ecsk be the energy of Ccsk, and Ecrsk be the energy of Crsk, that is, Ecsk is equal to the sum of the squares of each element in Ccsk, and Ecrsk is equal to the sum of the squares of each element in Crsk.

[0110] Next, calculate the residual gradient of the energy distribution of the original coding vector and the reconstructed feature vector in each band. The residual gradient is used to measure the difference degree between the original signal and the reconstructed signal in the energy distribution of each band. For the k-th band of the s-th layer, the residual gradient Gsk can be obtained by calculating the absolute value of the energy difference, that is, Gsk = |Ecsk - Ecrsk|.

[0111] To determine the band components of abnormal fluctuations, it is necessary to set a historical fluctuation threshold. Based on the sliding window, statistically analyze the distribution boundary of the residual gradient in the historical normal dataset, so as to determine the dynamic fluctuation threshold. In the historical normal dataset, select a sliding window with a set length in chronological order, and calculate the statistical characteristics of the residual gradient of each band in each window, such as the mean and standard deviation. Let the mean of the residual gradient of the k-th band of the s-th layer in the sliding window in the historical normal dataset be μsk, and the standard deviation be σsk. The dynamic fluctuation threshold Tsk can be determined according to the mean and standard deviation. For example, Tsk = μsk + α × σsk, where α is an adjustable coefficient used to control the strictness of the threshold.

[0112] Mark the band components and corresponding spatial positions that exceed the dynamic fluctuation threshold as abnormal fluctuation features. For each band, if its residual gradient Gsk is greater than the corresponding dynamic fluctuation threshold Tsk, it is considered that there is an abnormal fluctuation in this band. Record the numbers (s, k) of these abnormal fluctuation bands and the corresponding monitoring node numbers i. Since the context coding vector C corresponds to the monitoring nodes one by one, the spatial position where the abnormal fluctuation occurs can be determined. Combine the numbers of all abnormal fluctuation bands and the corresponding monitoring node numbers to generate an abnormal fluctuation feature set A = {(i1, s1, k1), (i2, s2, k2), …}, where (ij, sj, kj) indicates that there is an abnormal fluctuation in the kj-th band of the sj-th layer of the ij-th monitoring node.

[0113] After obtaining the stable state feature S of the target slope area and the abnormal fluctuation feature set A, generate a slope quality assessment result according to the difference degree between the stable state feature and the abnormal fluctuation feature. The calculation of the difference degree can comprehensively consider multiple aspects of the stable state feature and the abnormal fluctuation feature.

[0114] The steady-state feature S is an indicator reflecting the stability of the slope structure. The closer the value is to 1, the more stable it is. The abnormal fluctuation feature set A contains the frequency band and spatial location information where abnormal fluctuations occur. A difference degree function D can be defined to measure the difference between the two. First, for each abnormal fluctuation point (ij, sj, kj) in the abnormal fluctuation feature set A, a weight wij can be assigned to it. The size of the weight can be determined according to the severity of the abnormal fluctuation. For example, it is related to the magnitude of the residual gradient Gskj, and wij can be a monotonically increasing function of Gskj.

[0115] Then, calculate the weighted sum WA = ∑wij ((ij, sj, kj) ∈ A) of all abnormal fluctuation points. The difference degree function D can be defined as D = f(S, WA), where f is a function that comprehensively considers the steady-state feature and the weighted sum of abnormal fluctuations. For example, D can be in the form of a linear combination, D = β×(1 - S)+(1 - β)×WA, where β is a weighting coefficient with a value range between 0 and 1, which is used to adjust the importance of the steady-state feature and the weighted sum of abnormal fluctuations in the calculation of the difference degree.

[0116] According to the value of the difference degree D, the slope quality assessment results are divided into different grades. For example, when D is greater than a relatively high threshold D1, it is considered that the slope quality is poor and there is a relatively high risk of instability; when D is within an intermediate threshold range (between D2 and D1), it is considered that the slope quality is average and requires close attention; when D is less than a lower threshold D2, it is considered that the slope quality is good and in a relatively stable state.

[0117] Step S150: Determine the early warning response strategy based on the slope quality assessment result, and trigger the warning signal output operation of the slope monitoring system according to the early warning response strategy.

[0118] After obtaining the slope quality assessment result, it is necessary to determine the corresponding early warning response strategy based on this result and trigger the warning signal output operation of the slope monitoring system to ensure the safety of the target slope area.

[0119] Step S151: Divide the early warning levels according to the magnitude of the difference degree. Among them, when the difference degree exceeds the first threshold, a first-level early warning is triggered; when the difference degree is within the second threshold range, a second-level early warning is triggered; when the difference degree is lower than the third threshold, the monitoring state is maintained, where the first threshold > the upper limit of the second threshold range > the lower limit of the second threshold range > the third threshold.

[0120] In this embodiment, the early warning levels are divided according to the magnitude of the difference degree D. Three key thresholds are set, namely the first threshold T1, the upper limit T2u of the second threshold range, the lower limit T2l of the second threshold range, and the third threshold T3, and T1 > T2u > T2l > T3 is satisfied.

[0121] The extreme value theory (EVT) is used to dynamically set the threshold.

[0122] The extreme value theory is a statistical method specifically used to deal with extreme events. It can estimate the probability and threshold of future extreme events based on extreme values in historical data. First, collect the historical difference data of the target slope area, which includes the stability state and abnormal fluctuations of the slope in different time periods.

[0123] Conduct extreme value analysis on the historical difference data and select an appropriate extreme value model, such as the generalized extreme value distribution (GEV) model or the generalized Pareto distribution (GPD) model. Taking the generalized extreme value distribution model as an example, it can describe the extreme value distribution characteristics of the data, and its distribution function includes location parameters, scale parameters, and shape parameters.

[0124] Estimate the parameters of the historical difference data through methods such as maximum likelihood estimation to determine the specific parameter values of the generalized extreme value distribution model. Based on the estimated parameters, calculate different quantiles of the historical difference data. For example, to determine the first threshold T1, the 99% quantile of the historical difference data can be selected as this threshold. This means that in the historical data, only 1% of the difference values will exceed this threshold. When the currently calculated difference exceeds this threshold, it indicates that the abnormal situation of the slope has reached a very serious level, and a first-level warning needs to be triggered.

[0125] For the second threshold range (from T2l to T2u), the 80% quantile and 90% quantile of the historical difference data can be selected as the lower limit T2l and upper limit T2u of the second threshold range respectively. When the difference is within this range, it indicates that there are certain abnormal fluctuations in the slope, but it has not reached the serious level of the first-level warning. At this time, a second-level warning is triggered to remind relevant personnel to closely monitor the state of the slope.

[0126] The third threshold T3 can be selected as the 20% quantile of the historical difference data. When the difference is lower than this threshold, it indicates that the slope is in a relatively stable state, and normal monitoring can be maintained.

[0127] According to the thresholds determined based on the extreme value theory above, when the difference D exceeds the first threshold T1, it indicates that the slope quality is extremely poor and there is a very high risk of instability. At this time, a first-level warning is triggered. The first-level warning means that the most urgent and strict measures need to be taken to deal with the possible slope instability.

[0128] When the difference D is within the second threshold range (between T2l and T2u), it indicates that the slope quality is average and there is a certain risk of instability, and it needs to be closely monitored. At this time, a second-level warning is triggered. The second-level warning requires taking appropriate measures to strengthen the monitoring and protection of the slope.

[0129] When the difference degree D is lower than the third threshold T3, it indicates that the slope quality is good and in a relatively stable state. At this time, the monitoring state is maintained, and the target slope area continues to be monitored according to the normal monitoring frequency and process.

[0130] Step S152: For a first-level warning, adjust the data acquisition frequency of the monitoring nodes to the first preset frequency, and synchronously update the time granularity of the sliding window and the timestamp alignment rule in the preprocessing.

[0131] When a first-level warning is triggered, in order to more timely and accurately grasp the state change of the target slope area, it is necessary to adjust the data acquisition frequency of the monitoring nodes to the first preset frequency. The first preset frequency is usually higher than the data acquisition frequency during normal monitoring, so that more real-time data can be obtained to more finely analyze the dynamic changes of the slope.

[0132] At the same time, it is necessary to synchronously update the time granularity of the sliding window and the timestamp alignment rule in the preprocessing. Due to the increase in the data acquisition frequency, the original time granularity of the sliding window and the timestamp alignment rule may no longer be applicable. The time granularity of the sliding window needs to be correspondingly reduced to adapt to the higher-frequency data acquisition and ensure that more subtle changes can be captured. The timestamp alignment rule also needs to be updated to ensure that the data of different monitoring nodes can be accurately corresponding to the same time point for subsequent analysis and processing.

[0133] For example, during normal monitoring, the data acquisition frequency is once per minute, the time granularity of the sliding window is 10 minutes, and the timestamp alignment rule is aligned according to the whole minute. After a first-level warning is triggered, the data acquisition frequency is adjusted to once every 30 seconds. Correspondingly, the time granularity of the sliding window is adjusted to 5 minutes, and the timestamp alignment rule is updated to be aligned every 30 seconds.

[0134] Step S153: For a second-level warning, generate the structural reinforcement suggestion parameters for the local area, and send the structural reinforcement suggestion parameters to the monitoring terminal for visual display.

[0135] When a second-level warning is triggered, it indicates that there is a certain risk of slope instability, and it is necessary to structurally reinforce the local area. According to the spatial position information of the abnormal fluctuations recorded in the abnormal fluctuation feature set A, determine the local area that needs to be structurally reinforced.

[0136] By analyzing information such as the severity and occurrence frequency band of the abnormal fluctuations, and combining factors such as the geological conditions and geotechnical mechanical properties of the target slope area, generate the structural reinforcement suggestion parameters for the local area. The structural reinforcement suggestion parameters can include the reinforcement method (such as bolt reinforcement, retaining wall reinforcement, etc.), the selection of reinforcement materials, the specific location and scope of reinforcement, the strength requirements for reinforcement, etc.

[0137] Send the generated structural reinforcement suggestion parameters to the monitoring terminal for visual display. The monitoring terminal can be a computer device or a mobile terminal device installed in the monitoring center. On the monitoring terminal, the structural reinforcement suggestion parameters can be displayed in the form of graphs, charts, text descriptions, etc., enabling relevant personnel to intuitively understand the areas that need to be reinforced and the specific reinforcement requirements.

[0138] Step S154: Generate an early warning signal trigger instruction according to the early warning level, and output an audible and visual alarm signal and risk location identification information through the slope monitoring system.

[0139] For example, for a first-level early warning, the early warning signal trigger instruction requires the output of a more intense and obvious audible and visual alarm signal to attract the high attention of relevant personnel. The audible and visual alarm signal can include a high-decibel alarm sound and a flashing warning light. At the same time, on the display screen of the slope monitoring system, the risk location information is marked with eye-catching colors and marks, such as marking the area where instability may occur in red.

[0140] For a second-level early warning, the early warning signal trigger instruction requires the output of a relatively weak audible and visual alarm signal, such as a medium-volume alarm sound and a slowly flashing warning light. The risk location information is also marked on the display screen, but the eye-catching degree of the color and mark is relatively low, such as marking the area that needs attention in yellow.

[0141] When maintaining the monitoring state, the audible and visual alarm signal is not triggered, but the slope monitoring system will continue to operate normally and display the monitoring data and status information of the target slope area in real time.

[0142] Through the above steps, the monitoring and early warning of the quality of the target slope area are realized based on the ground motion data, which can timely detect the abnormal fluctuations of the slope and take corresponding early warning response measures to ensure the safety of the target slope area.

[0143] Figure 2 FIG. shows a schematic diagram of exemplary hardware and software components of a slope quality monitoring and early warning system 100 based on ground motion data that can implement the idea of the present application. For example, the processor 120 can be used on the slope quality monitoring and early warning system 100 based on ground motion data and is used to execute the functions in the present application.

[0144] The slope quality monitoring and early warning system 100 based on ground motion data can be a general-purpose server or a special-purpose server, both of which can be used to implement the slope quality monitoring and early warning method based on ground motion data of the present application. Although only one server is shown in the present application, for convenience, the functions described in the present application can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0145] For example, the slope quality monitoring and early warning system 100 based on ground motion data may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and different forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the slope quality monitoring and early warning system 100 based on ground motion data may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The methods of the present application can be implemented according to these program instructions. The slope quality monitoring and early warning system 100 based on ground motion data further includes an I / O interface 150 between the computer and other input / output devices.

[0146] For ease of explanation, only one processor is described in the slope quality monitoring and early warning system 100 based on ground motion data. However, it should be noted that the slope quality monitoring and early warning system 100 based on ground motion data in the present application may also include multiple processors. Therefore, the steps executed by one processor described in the present application can also be jointly executed or separately executed by multiple processors. For example, if the processor of the slope quality monitoring and early warning system 100 based on ground motion data executes step A and step B, it should be understood that step A and step B can also be jointly executed by two different processors or separately executed in one processor. For example, the first processor executes step A, the second processor executes step B, or the first processor and the second processor jointly execute steps A and B.

[0147] In addition, an embodiment of the present invention further provides a readable storage medium, in which computer-executable instructions are preset. When the processor executes the computer-executable instructions, the slope quality monitoring and early warning method based on ground motion data as described above is implemented.

[0148] It should be noted that, in order to simplify the description of the present invention disclosure and thus help the understanding of one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, sometimes multiple features are merged into one embodiment, drawing, or description thereof.

Claims

1. A slope quality monitoring and early warning method based on ground motion data, characterized in that The method includes: Obtaining a set of ground motion monitoring data for a target slope area, where the set of ground motion monitoring data includes ground motion waveform data collected by multiple monitoring nodes within a preset time period and corresponding environmental correlation parameters; Performing a dynamic preprocessing operation on the set of ground motion monitoring data to obtain a standardized set of ground motion monitoring data, where the dynamic preprocessing operation includes data alignment, noise filtering, and dimension unification processing; Performing feature extraction based on the standardized set of ground motion monitoring data to generate a set of dynamic correlation features for the target slope area, where the set of dynamic correlation features includes ground motion energy distribution features, slope response frequency features, and environmental coupling fluctuation features; Invoking a slope quality analysis model to perform state prediction processing on the set of dynamic correlation features to obtain the stable state features and abnormal fluctuation features of the target slope area, and generating a slope quality assessment result based on the difference degree between the stable state features and the abnormal fluctuation features; Determining an early warning response strategy based on the slope quality assessment result, and triggering an early warning signal output operation of the slope monitoring system according to the early warning response strategy.

2. The slope quality monitoring and early warning method based on ground motion data according to claim 1, characterized in that, The performing a dynamic preprocessing operation on the set of ground motion monitoring data to obtain a standardized set of ground motion monitoring data includes: Performing a noise component filtering and elimination operation and a time-frequency conversion processing on the ground motion waveform data to generate ground motion spectrum data for each monitoring node, and performing a frequency band energy normalization processing on the ground motion spectrum data to obtain frequency band energy standardized data; Performing a data integrity verification processing on the environmental correlation parameters, performing an interpolation filling operation on the missing parameters based on the verification result, and performing a numerical range normalization processing on the filled environmental correlation parameters to obtain environmental parameter standardized data; Further performing a global normalization processing on the frequency band energy standardized data and the environmental parameter standardized data and then performing a timestamp alignment processing to generate standardized data units corresponding one by one to the monitoring nodes, and combining each standardized data unit to generate the standardized set of ground motion monitoring data.

3. The slope quality monitoring and early warning method based on ground motion data according to claim 2, characterized in that, The performing feature extraction based on the standardized set of ground motion monitoring data to generate a set of dynamic correlation features for the target slope area includes: Performing an energy integration processing on the frequency band energy standardized data of each monitoring node to obtain a ground motion energy distribution sequence for each monitoring node, and calculating an energy difference degree between adjacent monitoring nodes based on the ground motion energy distribution sequence to generate an energy distribution difference feature; Performing a main frequency component extraction processing on the frequency band energy standardized data to obtain a set of dominant frequencies for each monitoring node, and calculating a resonance response probability of the slope structure based on the set of dominant frequencies to generate a slope response frequency feature; Performing a dynamic correlation analysis processing on the environmental parameter standardized data to determine a coupling correlation degree between the environmental parameter change and the ground motion energy distribution, and generating an environmental coupling fluctuation feature based on the coupling correlation degree; Perform spatial topological correlation processing on the energy distribution difference feature, the slope response frequency feature, and the environmental coupling fluctuation feature to generate a dynamic correlation feature set of the target slope area.

4. The slope quality monitoring and early warning method based on ground motion data according to claim 3, characterized in that Performing dynamic correlation analysis on the standardized environmental parameter data to determine the coupling correlation degree between environmental parameter changes and ground motion energy distribution, and generating an environmental coupling fluctuation feature based on the coupling correlation degree, includes: Perform sliding window slicing on the standardized environmental parameter data according to a preset time granularity to generate a set of environmental parameter time series slices corresponding to each monitoring node, where each environmental parameter time series slice contains a standardized numerical sequence of temperature, humidity, and air pressure parameters at consecutive time steps; Slice the ground motion energy distribution sequence according to the same time granularity, and allow the environmental parameter time series slices to slide and align within a preset lag time range to generate a ground motion energy time series slice set considering the lag effect of environmental parameters; Perform dynamic correlation analysis on the environmental parameter time series slice and the ground motion energy time series slice of the same monitoring node, and combine the linear Pearson correlation coefficient and the non-linear mutual information index to generate a comprehensive correlation intensity matrix of each parameter dimension and the energy distribution dimension, where the element value in the comprehensive correlation intensity matrix is the weighted average of the linear and non-linear correlation intensities; Perform a max pooling operation across parameter dimensions on the real-time correlation intensity matrix to extract the maximum coupling correlation intensity between environmental parameters and ground motion energy distribution at each time step, and generate a preliminary coupling correlation degree sequence; Based on the spatial gradient change of the preliminary coupling correlation degree sequence between adjacent monitoring nodes, calculate the regional propagation consistency index of the coupling correlation intensity, and perform spatial smoothing correction processing on the preliminary coupling correlation degree sequence to generate a corrected dynamic coupling correlation degree set; Input the dynamic coupling correlation degree set into a preset coupling intensity classification model, and map the coupling correlation degrees of each monitoring node to a unified feature dimension through a fully connected layer to generate an environmental coupling fluctuation feature vector matching the ground motion energy distribution feature dimension.

5. The slope quality monitoring and early warning method based on ground motion data according to claim 4, characterized in that Performing spatial smoothing correction processing on the preliminary coupling correlation degree sequence to generate a corrected dynamic coupling correlation degree set, includes: Perform first-order difference calculation on the preliminary coupling correlation degree sequence in the time dimension to generate a coupling correlation degree change trend sequence of each monitoring node; Calculate the regional propagation consistency index based on the cosine similarity between the coupling correlation degree change trend sequences of adjacent monitoring nodes, where the node pairs with cosine similarity lower than the preset threshold are marked as spatial abnormal fluctuation regions; Perform a mean replacement operation based on K nearest neighbor nodes on the preliminary coupling correlation degree of the spatial abnormal fluctuation region to eliminate the coupling correlation degree jump noise caused by local sensor anomalies; Perform spatial neighborhood weighted smoothing processing on the corrected preliminary coupling correlation degree using a Gaussian kernel function, where the kernel bandwidth is adaptively adjusted according to the topological distance of the monitoring nodes. Perform exponential weighted moving average filtering on the smoothed dynamic coupling correlation in the time dimension to suppress high-frequency fluctuation noise and retain the trend component, and generate the set of the corrected dynamic coupling correlations.

6. The slope quality monitoring and early warning method based on ground motion data according to claim 1, characterized in that Call the slope quality analysis model to perform state prediction processing on the dynamic association feature set, and obtain the stable state features and abnormal fluctuation features of the target slope area, including: Input the dynamic association feature set into the spatio-temporal encoding network of the slope quality analysis model, and perform global dependency modeling on the spatial topological association data in the dynamic association feature set through the multi-head self-attention mechanism to generate a context encoding vector containing cross-node spatio-temporal interaction information; Input the context encoding vector into the prediction branch of the slope quality analysis model, extract the seismic motion evolution patterns at different time scales through the dilated temporal convolutional network, output the energy distribution prediction sequence for multiple future time steps, and calculate the smoothness index characterizing the slope structure stability based on the fluctuation consistency at each time step in the energy distribution prediction sequence to generate the stable state features; And input the context encoding vector into the anomaly detection branch of the slope quality analysis model, perform latent space reconstruction on the context encoding vector through the variational autoencoder to generate a reconstructed feature vector matching the historical normal state distribution, and use the wavelet packet decomposition algorithm to calculate the residual gradient of the energy distribution of the original encoding vector and the reconstructed feature vector in each frequency band, and extract the frequency band components and corresponding spatial positions where the residual gradient exceeds the historical fluctuation threshold to generate the abnormal fluctuation features.

7. The slope quality monitoring and early warning method based on ground motion data according to claim 6, characterized in that, Perform global dependency modeling on the spatial topological association data in the dynamic association feature set through the multi-head self-attention mechanism to generate a context encoding vector, including: Perform time dimension slicing processing on the dynamic association feature set to generate a set of spatio-temporal feature slices divided according to a preset time window; Based on the actual physical position distance between the monitoring nodes in each spatio-temporal feature slice and the historical seismic motion signal correlation, construct the weighted connection relationship between the monitoring nodes to generate a node adjacency matrix and a feature node vector reflecting spatial proximity and signal association strength; Input the node adjacency matrix and the feature node vector into the multi-head self-attention mechanism, calculate the dynamic association strength between each node through the learnable attention weights to generate a node-level attention weight matrix; Based on the node-level attention weight matrix, perform weighted aggregation on the feature node vectors to generate a spatial aggregation feature vector within each time window; Perform position encoding and time dimension convolutional fusion on the spatial aggregation feature vectors of multiple time windows to generate the context encoding vector containing the temporal evolution law.

8. The slope quality monitoring and early warning method based on ground motion data according to claim 6, characterized in that, Perform latent space reconstruction on the context encoding vector through the variational autoencoder to generate a reconstructed feature vector matching the historical normal state distribution, including: Input the context encoding vector into the encoder network of the variational autoencoder, and map the input features to the mean vector and variance vector in the latent space through the multi-layer perceptron; Random sampling is performed based on the mean vector and variance vector to generate a latent space vector that follows a Gaussian distribution; The latent space vector is input into the decoder network of the variational autoencoder, and the feature dimension is gradually restored through the deconvolution layer and fully connected layer to generate the reconstructed feature vector; The original context encoding vector and the reconstructed feature vector are unfolded into time series signals according to the time step, and multi-scale wavelet packet decomposition is performed on the time series signals to calculate the relative error of the band energy distribution at each scale; Based on a sliding window, the distribution boundary of the relative error in the historical normal dataset is statistically determined to obtain a dynamic fluctuation threshold; The spatial positions and time points corresponding to the band energy errors exceeding the dynamic fluctuation threshold are marked as the abnormal fluctuation features.

9. The slope quality monitoring and early warning method based on ground motion data according to claim 1, characterized in that Determining an early warning response strategy based on the slope quality assessment result, and triggering an early warning signal output operation of the slope monitoring system according to the early warning response strategy, including: Dividing the early warning levels according to the magnitude of the difference degree. Specifically, when the difference degree exceeds the first threshold, a first-level early warning is triggered; when the difference degree is within the second threshold range, a second-level early warning is triggered; when the difference degree is lower than the third threshold, the monitoring state is maintained, where the first threshold > the upper limit of the second threshold range > the lower limit of the second threshold range > the third threshold; For a first-level early warning, adjust the data acquisition frequency of the monitoring node to a first preset frequency, and synchronously update the sliding window time granularity and timestamp alignment rule in the preprocessing; For a second-level early warning, generate structural reinforcement recommendation parameters for a local area, and send the structural reinforcement recommendation parameters to the monitoring terminal for visual display; Generate an early warning signal trigger instruction according to the early warning level, and output an audible and visual alarm signal and risk location identification information through the slope monitoring system.

10. A slope quality monitoring and early warning system based on seismic motion data, characterized in that, It includes a processor and a memory. The memory is connected to the processor. The memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the slope quality monitoring and early warning method based on ground motion data according to any one of claims 1-9 above.

Images