Geological disaster risk intelligent early warning system

Abstract

The invention relates to the technical field of risk early warning, in particular to a geological disaster risk intelligent early warning system which comprises a multi-dimensional time sequence lagging resolving module, a response field space profile construction module, a permeability coefficient disturbance judgment module and a lagging field self-adaptive correction module. According to the invention, by revealing the internal time sequence correlation between the rainfall event and the displacement response of the soil bodies with different depths, the spatial profile reflecting the response characteristics of the whole geologic body is constructed, the displacement change can be monitored in real time, and the dynamic influence of the internal damage state of the rock-soil body on the permeability parameters can be reversely deduced according to the displacement change; the rainfall response relation and the early warning criterion are adjusted in a self-adaptive mode, the mechanism of comprehensively considering the multi-dimensional data coupling relation and the dynamic evolution of the catastrophe process accurately describes the physical process of disaster development, and the timeliness and accuracy of early warning information are greatly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of risk early warning, and in particular to a geological disaster risk intelligent early warning system. BACKGROUND

[0002] The technical field of risk early warning is a technology category related to the continuous state monitoring of a specific monitoring object, collecting various types of monitoring data through sensors and observation equipment, analyzing and processing the collected data, identifying the precursor information of disasters or accidents, and ultimately issuing warning information before the event occurs to reduce losses. The core items include monitoring data collection, data feature extraction, trend analysis, and threshold judgment. Among them, the traditional geological disaster risk intelligent early warning system refers to the technical items for early warning of disasters such as landslides, debris flows, and ground subsidence. The traditional system usually adopts the way of deploying rain gauges, ground fissure meters, deep displacement inclinometers, and groundwater level meters at disaster-prone points, and periodically transmitting the collected data through wireless communication. The main basis for its early warning judgment is to compare the monitoring data of a single element with the pre-set fixed threshold value to determine whether to issue a warning information.

[0003] The traditional early warning method relies on setting a fixed threshold for a single monitoring index. This method ignores the complex lag relationship between rainfall infiltration and internal displacement response, and cannot reflect the dynamic changes of the physical and mechanical parameters of the rock-soil mass during the accumulation of damage. Therefore, in the nonlinear disaster-causing process scenario, for example, if the internal structure has been deteriorated due to previous sustained rainfall, even if the subsequent rainfall does not exceed the limit, the system may fail to identify this cumulative effect, resulting in delayed or missed early warning. SUMMARY

[0004] The purpose of the present application is to solve the problems existing in the prior art and to provide a geological disaster risk intelligent early warning system.

[0005] In order to achieve the above-mentioned purpose, the present application adopts the following technical solution: a geological disaster risk intelligent early warning system comprises: A multi-dimensional time series lag calculation module acquires a rainfall intensity sequence and a multi-depth displacement sequence, calculates a displacement rate sequence, generates an optimal lag time set through cross-correlation operation, and transmits the optimal lag time set to a response field spatial profile construction module; The response field spatial profile construction module generates a lag response profile according to the optimal lag time set, and substitutes the node depth of a three-dimensional grid model into the lag response profile to construct an initial response field, and transmits the initial response field to a lag field adaptive correction module; The permeability coefficient disturbance judgment module monitors the multiple depth displacement sequences, calculates a displacement increment per unit time, and if the displacement increment per unit time is greater than a preset displacement threshold, substitutes a value of the corresponding displacement increment per unit time into a damage-permeability correlation function, calculates a permeability coefficient adjustment factor, and transmits the permeability coefficient adjustment factor and the depth to the lag field adaptive correction module. The lag field adaptive correction module adjusts a permeability coefficient value based on the initial response field, according to the permeability coefficient adjustment factor and the depth, updates the optimal lag time set, and generates a corrected response profile and a corrected response field.

[0006] As a further scheme of the present application, the optimal lag time set includes a rainfall-displacement response time difference, a monitoring point depth index, and an associated response intensity, the initial response field specifically refers to a warning grid unit, a node response prediction value, and a spatial distribution weight, the permeability coefficient adjustment factor specifically refers to a damage equivalent parameter, an adjustment direction, and an amplitude, and the corrected response field includes a dynamic risk level, a warning area range, and a disaster occurrence probability.

[0007] As a further scheme of the present application, the multi-dimensional time series lag calculation module includes: The time series data stream receiving submodule obtains a rainfall intensity sequence and a multiple depth displacement sequence, integrates the rainfall intensity sequence and the multiple depth displacement sequence into a structured data entity, and aligns and calibrates time stamps to generate an original time series data set; The displacement rate gradient calculation submodule calls the multiple depth displacement sequence in the original time series data set, performs a first-order difference operation on displacement data of each depth along a time axis to obtain multiple depth displacement rates, and performs a weighted average processing on the displacement rates of all depths according to their monitoring depths to obtain an average displacement rate sequence; The sequence cross-correlation operation submodule extracts the rainfall intensity sequence in the original time series data set, sets a preset time window according to the average displacement rate sequence, shifts the rainfall intensity sequence point by point within the time window, calculates the sum of the products of the corresponding data points of the two sequences after each shift, and generates a cross-correlation coefficient vector; The peak index positioning submodule traverses all coefficients in the cross-correlation coefficient vector to locate the coefficient with the maximum value as a peak point for the cross-correlation coefficient vector, converts the index value of the peak point in the cross-correlation coefficient vector into an offset time amount corresponding to the starting point of the preset time window, and constructs an optimal lag time set.

[0008] As a further scheme of the present application, the response field spatial profile construction module includes: The profile function fitting submodule executes a least square curve fitting operation on time values in the optimal lag time set and corresponding depth values as discrete data points, taking depth as the independent variable and time as the dependent variable, establishes a depth-time function, and generates a lag response profile; The node depth mapping submodule obtains three-dimensional mesh model node depths, and calls the lag response profile to input each of the three-dimensional mesh model node depths into the depth-time function represented by the lag response profile to calculate and obtain target time values at each mesh node, thereby obtaining a node lag time scalar set; The response field space construction submodule assigns each of the time values in the node lag time scalar set to a corresponding node in the three-dimensional mesh model according to the corresponding spatial coordinates, forms a discrete data field taking node coordinates as indexes and time values as attributes, and constructs an initial response field.

[0009] As a further scheme of the present application, the permeability coefficient disturbance determination module comprises: The displacement increment calculation submodule monitors the multi-depth displacement sequence, divides the multi-depth displacement sequence into continuous time units according to a preset sampling frequency, extracts displacement measurement values at the start and end times in each of the time units, and obtains unit time displacement increments by performing subtraction operation on the displacement measurement values; The overrun event discrimination submodule obtains a preset displacement threshold matrix, calls the unit time displacement increments, compares each of the unit time displacement increments reported at each depth with a threshold value corresponding to the depth in the preset displacement threshold matrix, filters all increments that exceed the threshold value, and constructs an overrun increment event set; The permeability factor solving submodule collects a damage-permeability correlation function, and substitutes increment values recorded in the overrun increment event set as independent input variables into the damage-permeability correlation function expression to perform solving operation, obtains function outputs corresponding to multiple increments, and generates a permeability coefficient adjustment factor.

[0010] As a further scheme of the present application, the lag field adaptive correction module comprises: The permeability coefficient value adjustment submodule locates a spatial region corresponding to a depth in the initial response field based on the initial response field and the permeability coefficient adjustment factor and the depth, performs multiplication operation on a permeability coefficient value in the region and the permeability coefficient adjustment factor to generate an updated permeability coefficient matrix; The lag time set updating submodule calls the updated permeability coefficient matrix, recalculates flow permeation rates according to values of the updated permeability coefficient matrix for multiple depth levels, adjusts time delays between rainfall and displacement responses, and obtains an updated optimal lag time set; The correction profile fitting submodule combines the time data and corresponding depth coordinates in the updated optimal lag time set to form data point pairs, performs polynomial interpolation operation on all the data point pairs to establish a function relationship, and generates a corrected response profile; The correction field reconstruction submodule calls the corrected response profile, extracts the depth coordinates of the nodes of the three-dimensional mesh model, substitutes the depth coordinates into the function represented by the corrected response profile one by one to perform evaluation, and assigns the calculated lag time to the corresponding node to construct a corrected response field.

[0011] As a further scheme of the present application, the permeation factor solver submodule specifically performs the process of calling the calculation formula defined by the damage-permeation correlation function: ; substituting the unit time displacement increment in the over-limit incremental event set as a damage measure parameter , and obtaining the preset displacement threshold of the corresponding depth from the preset displacement threshold matrix as a threshold parameter ; wherein, is a permeation coefficient adjustment factor to be calculated, is a soil material damage sensitivity coefficient, is a dimensionless damage response index, is the unit time displacement increment, is the preset displacement threshold; According to the formula, the function output value corresponding to each over-limit increment is calculated, and all the function output values are combined to generate a permeation coefficient adjustment factor.

[0012] As a further scheme of the present application, the lag time set updating submodule specifically performs the process of obtaining the updated permeation coefficient matrix and the optimal lag time set, and according to the formula: ; calculating the updated lag time; wherein, is the updated lag time at depth , is the initial lag time at depth extracted from the optimal lag time set, is the initial permeation coefficient value at depth extracted from the initial response field, is the updated permeation coefficient value at depth extracted from the updated permeation coefficient matrix; Traverse all monitoring depths, calculate the updated lag time of all depths one by one, and combine the corresponding monitoring point depth index and the associated response intensity to build an updated optimal lag time set.

[0013] As a further scheme of the present application, the generation process of the dynamic risk level included in the modified response field, the pre-warning area range and the disaster occurrence probability comprises: A risk level determination step, a plurality of risk division thresholds are obtained, and the node response prediction value of each pre-warning grid unit in the modified response field is compared with the plurality of risk division thresholds one by one, an initial risk level is assigned to each pre-warning grid unit, and a grid risk distribution map is obtained; A pre-warning range identification step, the grid risk distribution map is called, a high risk level threshold is set, a density-based spatial clustering algorithm is used, all adjacent pre-warning grid units with a risk level higher than the high risk level threshold are identified and merged into one or more connected regions, and a pre-warning area range is generated; A disaster probability solving step, for each pre-warning area range, the average node response prediction value of all pre-warning grid units in the area, the area volume and the cumulative value of the rainfall intensity sequence are comprehensively calculated, and a preset probability conversion model is used for operation to obtain a disaster occurrence probability.

[0014] As a further scheme of the present application, the sequence cross-correlation operation sub-module first performs standardization processing on the data before calculating the cross-correlation coefficient vector, comprising: A mean standard deviation calculation step, the arithmetic mean and standard deviation of all data points in the rainfall intensity sequence and the average displacement rate sequence in the preset time window are independently calculated; A sequence standardization step, the arithmetic mean and the standard deviation are called, each data point in the rainfall intensity sequence and the average displacement rate sequence is traversed, the data point value is subtracted from the arithmetic mean of the sequence, and the difference is divided by the standard deviation of the sequence to generate a standardized rainfall sequence and a standardized displacement rate sequence; A correlation coefficient calculation step, the standardized rainfall sequence is shifted point by point according to the preset time window, the sum of the products of the corresponding data points of the standardized rainfall sequence and the standardized displacement rate sequence after each shift is calculated, and the sum of the products is divided by the sequence length to generate a cross-correlation coefficient vector.

[0015] Compared with the prior art, the present application has the advantages and positive effects that: In the present application, by revealing the internal time sequence correlation of rainfall events and displacement response of different depth soil bodies, a spatial profile reflecting the response characteristics of the entire geological body is constructed, and the displacement change can also be monitored in real time, and the dynamic influence of the internal damage state of the rock-soil body on the permeability parameter is deduced reversely, the response relationship to rainfall and the early warning criterion are adaptively adjusted, the mechanism of comprehensively considering the coupling relationship of multi-dimensional data and the dynamic evolution of disaster process is accurately described the physical process of disaster incubation and development, and the timeliness and accuracy of early warning information is greatly improved. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 The present application is a geological disaster risk intelligent early warning system overall flow chart; Figure 2 The present application is a multi-dimensional time sequence lag calculation flow chart; Figure 3 The present application is a response field spatial profile construction flow chart; Figure 4 The present application is a permeability coefficient disturbance judgment flow chart; Figure 5 The present application is a lag field adaptive correction flow chart. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical scheme and advantages of the present application clearer, the technical scheme realized based on software will be described in detail below combined with system architecture diagram and embodiment. It should be understood that the specific embodiments described herein are only used to explain the technical scheme of the present application, and do not constitute a limitation on the protection scope.

[0018] In the description of the present application, the system architecture relationship or data processing flow indicated by the terms "level", "module", "interface", "data flow", "client", "server" and the like are defined based on the corresponding architecture diagram or flow chart of the embodiment, and this description method is only used to clearly explain the logical relationship of each element in the technical scheme, but not to limit the physical deployment form. The "multiple" contains two or more technical units, including but not limited to multiple data nodes, processing threads, service instances or functional components, etc. Extensible elements, the specific number is determined according to the actual business scene.

[0019] Please refer to Figure 1 and Figure 2 , the present application provides a kind of technical scheme: a geological disaster risk intelligent early warning system includes: Multi-dimensional time sequence lag calculation module obtains rainfall intensity sequence and multi-depth displacement sequence, calculates displacement rate sequence, generates optimal lag time set by cross-correlation operation, and passes the optimal lag time set to response field spatial profile construction module; The optimal lag time set includes rainfall-displacement response time difference, monitoring point depth index and associated response intensity; The multi-dimensional time-lag solving module comprises: The time-series data stream receiving submodule obtains the rainfall intensity sequence and the multi-depth displacement sequence, integrates the rainfall intensity sequence and the multi-depth displacement sequence into a structured data entity, aligns and calibrates the time stamp, and generates an original time-series data set; The displacement rate gradient solving submodule calls the multi-depth displacement sequence in the original time-series data set, performs a first-order difference operation on the displacement data of each depth along the time axis, obtains a plurality of depth displacement rates, and performs a weighted average processing on the displacement rates of all depths according to their monitoring depths to obtain an average displacement rate sequence; The sequence cross-correlation operation submodule extracts the rainfall intensity sequence in the original time-series data set, sets a preset time window according to the average displacement rate sequence, shifts the rainfall intensity sequence point by point within the time window, calculates the sum of the products of the corresponding data points of the two sequences after each shift, and generates a cross-correlation coefficient vector; The sequence cross-correlation operation submodule performs standardization processing on the data before calculating the cross-correlation coefficient vector, including: The mean standard deviation calculation step independently calculates the arithmetic mean and the standard deviation of all data points in the rainfall intensity sequence and the average displacement rate sequence within the preset time window, respectively; The sequence standardization step calls the arithmetic mean and the standard deviation, traverses each data point in the rainfall intensity sequence and the average displacement rate sequence, subtracts the arithmetic mean of the sequence to which the data point belongs from the data point value, and then divides the obtained difference by the standard deviation of the sequence to generate a standardized rainfall sequence and a standardized displacement rate sequence; The correlation coefficient calculation step shifts the standardized rainfall sequence point by point according to the preset time window, calculates the sum of the products of the corresponding data points of the standardized rainfall sequence and the standardized displacement rate sequence after each shift, and divides the sum of the products by the sequence length to generate a cross-correlation coefficient vector; The peak index positioning submodule traverses all coefficients in the cross-correlation coefficient vector to locate the coefficient with the maximum value as the peak point, and converts the index value of the peak point in the cross-correlation coefficient vector into the offset time corresponding to the starting point of the preset time window to construct an optimal lag time set.

[0020] The time series data stream receiving submodule collects data from a sensor network deployed in the monitoring area, which includes surface rain gauges and deep displacement meters. In a specific implementation scenario, the monitoring system is deployed for a potential landslide, covering an area of 1 square kilometers. Among them, the rain gauge records the rainfall intensity at an interval of 5 minutes, with the unit of millimeters per hour (mm / h); the deep displacement meter is arranged at three different depths, 5 meters, 10 meters and 15 meters, and records the cumulative displacement at this depth at an interval of 1 hour, with the unit of millimeters (mm). This submodule receives these two types of data streams uninterruptedly within a continuous 72-hour monitoring period. After receiving, the system integrates the discrete data points into a structured data entity, specifically a data table in a time series database. In this process, the system performs timestamp alignment calibration, aggregates the 5-minute interval rainfall data, calculates the average rainfall intensity per hour, and strictly aligns the timestamp with the 1-hour interval displacement data. For example, the arithmetic mean of the 12 rainfall intensity data points from 08:00 to 08:55 is taken as the representative rainfall intensity value at 08:00. In this way, an original time series data set is generated, which takes hours as the minimum time unit, and each row of data contains a unique timestamp, the average rainfall intensity corresponding to that hour, and the cumulative displacement readings at three depths (-5m, -10m, -15m).

[0021] The displacement rate gradient solver submodule retrieves multi-depth displacement sequences from the original time series data set. For each monitoring depth, this submodule performs a first-order difference operation on the displacement measurements at consecutive time points along the time axis. Specifically, if the displacement at depth z at time t is , the displacement at time t-1 is , then the displacement rate at time t is calculated as . Taking the data at the monitoring depth of -10 meters as an example, if the cumulative displacement at t=10 (10th hour) is 22.5 mm and the cumulative displacement at t=9 (9th hour) is 22.1 mm, then the displacement rate at t=10 is mm / h. This operation traverses all time points and all depths, resulting in three independent depth displacement rate sequences. Subsequently, a weighted average processing is performed. The weights are set according to the geotechnical properties of each soil layer and their influence on landslide stability as reported in the geological exploration report. In a previous experimental verification, through triaxial shear tests and permeability tests on sampled soil, combined with numerical simulation analysis, the contribution of soil layers at different depths to the overall deformation was determined.

[0022] Table 1 Experimental verification data table of monitoring depth weight coefficients As shown in Table 1, based on experimental data and simulation results, the weights for depths of -5 meters, -10 meters, and -15 meters are set as follows: , , At t=10, if the displacement rates at the three depths are respectively mm / h, mm / h, mm / h, then the average displacement rate sequence value at that moment Calculated as mm / h. This process is applied to the entire time series to generate a single average displacement rate sequence.

[0023] The sequence cross-correlation submodule first sets a preset time window, the length of which is determined based on the typical time lag range from rainfall to response in historical geological disaster events in the region. For example, historical data shows that the time lag is usually between 0 and 24 hours, so the preset time window length is set to 25 data points (i.e., 25 hours). Before performing the cross-correlation operation, this submodule first standardizes the rainfall intensity sequence and the average displacement rate sequence within the selected time window. This process includes the calculation of the mean and standard deviation and the sequence standardization step. Taking a time window of 25 hours as an example, the arithmetic mean of the 25 rainfall intensity data points within the window is first calculated independently. with standard deviation and the arithmetic mean of 25 average displacement rate data points. with standard deviation Assuming the calculation yields... mm / h, mm / h, mm / h, mm / h. Next, iterate through each data point in both sequences, performing a standardization operation. For example, for the rainfall intensity value at a certain moment within the window... mm / h, its standardized value is Similarly, for displacement rate values mm / h, its standardized value is This process is applied to all data points within the window, generating standardized rainfall and standardized displacement rate sequences. Then, the correlation coefficient calculation step proceeds, shifting the standardized rainfall sequence relative to the standardized displacement rate sequence point by point, with the shift range from 0 to 24 time units. For each shift (let the shift amount be...), ... ), calculate the sum of the products of corresponding data points in the two sequences, and then divide the sum by the sequence length (25). For example, when the translation amount At that time, calculate When the translation amount At that time, calculate . And so on, finally a vector containing 25 cross-correlation coefficient values is generated.

[0024] The peak index locating sub-module receives the vector of cross-correlation coefficients and traverses all 25 coefficient values in the vector. For example, the resulting vector is . The system locates the coefficient with the largest value, which is the peak value 0.89. Then, the index value of the peak value in the vector is determined to be 3 (the index starts from 0). This index value is converted to the amount of offset time corresponding to the starting point of the preset time window, i.e. the lag time. Since each index represents a 1-hour shift, index 3 corresponds to a 3-hour offset (index 0 represents 1 hour, index 1 represents 2 hours, and so on, index n represents n+1 hours). Finally, the system constructs an optimal lag time set containing multiple information units, each unit recording the strongest response intensity, the associated monitoring depth index, and the cross-correlation coefficient value at the lag time. In this example, a core element of the generated optimal lag time set is {lag time: 4 hours, associated response intensity: 0.89}.

[0025] Please refer to Figure 1 and Figure 3 , the response field spatial profile construction module generates a lag response profile according to the optimal lag time set, and substitutes the node depth of the three-dimensional grid model into the lag response profile to construct an initial response field, and transfers the initial response field to the lag field adaptive correction module; The initial response field is specifically the warning grid unit, the node response prediction value, and the spatial distribution weight; The response field spatial profile construction module includes: The profile function fitting sub-module takes the time value in the optimal lag time set and the corresponding depth value as a pair of discrete data points, takes the depth as the independent variable and the time as the dependent variable, performs a least squares curve fitting operation, establishes a depth-time function, and generates a lag response profile; The node depth mapping sub-module obtains the node depth of the three-dimensional grid model and calls the lag response profile, takes each three-dimensional grid model node depth as an input value, substitutes it into the depth-time function represented by the lag response profile, calculates the target time value at each grid node, and obtains a node lag time scalar set; The response field spatial construction sub-module assigns each time value in the node lag time scalar set to the corresponding node in the three-dimensional grid model according to the corresponding spatial coordinates, forms a discrete data field with node coordinates as the index and time values as the attribute, and constructs an initial response field.

[0026] The response field spatial profile construction module receives the optimal lag time set transferred by the previous module and generates a lag response profile accordingly. Then, the node depth information of a pre-constructed three-dimensional geological grid model is substituted into the profile function to construct an initial response field.

[0027] The profile function fitting submodule first extracts discrete data point pairs from the optimal lag time set. Each data point pair consists of a monitoring depth value and a corresponding optimal lag time value. A core element of the optimal lag time set is {lag time: 4 hours, associated response intensity: 0.89}. This data point pair is: at a depth of -10 meters, the optimal lag time is 4 hours; at a depth of -5 meters, the time is 2.5 hours; at a depth of -15 meters, the time is 6.0 hours. These data point pairs are {(-5, 2.5), (-10, 4.0), (-15, 6.0)}. This submodule takes depth as the independent variable, and lag time as the dependent variable, and performs a least squares curve fitting operation. The goal of this operation is to find a function such that the sum of the squared residuals between the actual time values of all data points and the time values calculated by the function is minimized. Specifically, the system sets a polynomial function model, such as a quadratic polynomial . Substituting the above three sets of data points forms an overdetermined equation set about the coefficients . The optimal solution of is determined by solving the normal equation of the equation set. Based on the three data point sets {(-5, 2.5), (-10, 4.0), (-15, 6.0)}, through least squares curve fitting, the depth-time function obtained is . This function is the lag response profile, which describes the continuous distribution of the time required for rainfall influence to propagate to different depths in the vertical direction.

[0028] The node depth mapping submodule first loads the pre-established three-dimensional geological grid model. This model discretizes the entire landslide body into a large number of hexahedral elements, each consisting of eight nodes, each node having a unique three-dimensional spatial coordinate . This submodule extracts the depth coordinates, i.e. z values, of all nodes from the model. Then, it calls the lag response profile function generated in the previous step . The system iterates through each node in the three-dimensional grid model, taking its depth coordinate as the input value to substitute into the function. For example, for a node with coordinates (105.2, 210.5, -8.5), the depth is -8.5 meters. Substituting this into the function gives the target time value for this node: hours. This calculation process is applied to all nodes in the model, from several thousand to tens of thousands, ultimately obtaining a node lag time scalar set corresponding to all grid nodes.

[0029] The function of the response field space construction submodule is to convert the discrete node lag time scalar set calculated in the previous step into a structured and continuous initial response field. This submodule iterates through each time value in the node lag time scalar set and assigns the time value as an attribute to the node in the three-dimensional mesh model based on the spatial coordinates of the node corresponding to the time value. The specific operation is to add a "lag time" attribute field to each node in the node data structure of the three-dimensional model, and fill in the calculated value. For example, for node (105.2, 210.5, -8.5), its "lag time" attribute is assigned a value of 3.11375. After completing the assignment operation of all nodes, a discrete data field is formed with node spatial coordinates as the index and lag time values as the attributes. This data field intuitively shows the time required for the rainfall infiltration wave front to reach any location in three-dimensional space, i.e., the initial response field. The initial response field contains the warning grid elements (i.e., hexahedral elements of the three-dimensional model), the response prediction value of each node (i.e., lag time), and a spatial distribution weight calculated based on the node spacing and interpolation function. The weight is calculated by the formula where is the coordinate of node , is its neighbor node set, and is used for subsequent field value smoothing and gradient calculation. In the initial stage, all node weights are equal, . Finally, the initial response field data structure constructed by this module is passed to the lag field adaptive correction module.

[0030] Please refer to Figure 1 and Figure 4 , the permeability coefficient disturbance judgment module monitors the multi-depth displacement sequence, calculates the displacement increment per unit time, and if the displacement increment per unit time is greater than the preset displacement threshold, the value of the corresponding displacement increment per unit time is substituted into the damage-permeability correlation function to calculate the permeability coefficient adjustment factor, and the permeability coefficient adjustment factor and the depth are passed to the lag field adaptive correction module. The permeability coefficient adjustment factor specifically refers to the damage equivalent parameter, adjustment direction and amplitude; The permeability coefficient disturbance judgment module includes: The displacement increment calculation submodule monitors the multi-depth displacement sequence, divides the multi-depth displacement sequence into continuous time units according to the preset sampling frequency, extracts the displacement measurement values at the start and end times in each time unit, and performs subtraction operation on the displacement measurement values to obtain the displacement increment per unit time; The over-limit event discrimination sub-module acquires a preset displacement threshold matrix, and calls a unit time displacement increment, compares the unit time displacement increment reported by each depth with the threshold value corresponding to the depth in the preset displacement threshold matrix, screens all increments exceeding the threshold value, and constructs an over-limit increment event set; The permeability factor solver module collects a damage-permeability correlation function, takes the increment values recorded in the over-limit increment event set as independent input variables, substitutes them into the damage-permeability correlation function expression, performs a solving operation, obtains function outputs corresponding to multiple increments, and generates a permeability coefficient adjustment factor; The specific execution process of the permeability factor solver module is to call the calculation formula defined by the damage-permeability correlation function: ; The unit time displacement increment in the over-limit increment event set is taken as a damage measurement parameter , and the preset displacement threshold corresponding to the depth is taken as a threshold parameter from the preset displacement threshold matrix; Among them, is the permeability coefficient adjustment factor to be calculated, is the damage sensitivity coefficient of the soil material, is the dimensionless damage response index, is the unit time displacement increment, is the preset displacement threshold; According to the formula, the function output value corresponding to each over-limit increment is obtained, all function output values are combined, and the permeability coefficient adjustment factor is generated.

[0031] The permeability coefficient disturbance judgment module continuously monitors the multi-depth displacement sequence to calculate the displacement increment in unit time. When the increment is greater than the preset displacement threshold, the system starts calculation, substitutes the increment value into a damage-permeability correlation function, thereby solving the permeability coefficient adjustment factor, and transmits it together with the corresponding depth information.

[0032] The displacement increment calculation sub-module continuously acquires the multi-depth displacement sequence from the time series data stream. According to a preset sampling frequency, the continuous displacement sequence is divided into non-overlapping time units by the sub-module. The sampling frequency is set according to the monitoring sensitivity requirement, for example, 1 hour. For each 1-hour time unit, the sub-module extracts the displacement measurement values at the start and end times. Taking the-10-meter depth as an example, the recorded cumulative displacement is 24.8mm at the 15th hour of the monitoring period, and the recorded cumulative displacement is 26.1mm at the 16th hour. The sub-module performs subtraction operation on the two displacement measurement values to obtain the displacement increment in the 1-hour unit time: mm. This calculation process is performed synchronously at all monitoring depths and generates a set of unit-time displacement increments corresponding to each depth for each time unit.

[0033] The over-limit event discrimination submodule first loads a preset displacement threshold matrix. This matrix is ​​established based on long-term creep and shear failure tests on soil at a specific site. The experimental procedure involves applying different confining pressures and shear stresses to undisturbed soil samples at different depths and continuously monitoring their deformation process. The critical strain rate at which the soil sample transitions from the steady creep stage to the accelerated creep stage is recorded and converted into displacement per unit time.

[0034] Table 2. Experimental Determination of Preset Displacement Threshold As shown in Table 2, precise displacement thresholds were set for each monitoring depth through systematic laboratory testing. For example, the threshold for a depth of -10 meters... The value is set to 0.75 mm / h. The submodule calls the displacement increment calculated in the previous step and compares it with the threshold value for the corresponding depth in the threshold matrix. For a depth of -10 meters, the calculated increment... mm / h, because The event is determined to be out of bounds. The system filters out all increments that exceed their corresponding depth thresholds and, together with information such as the time of occurrence, depth, increment value, and threshold, constructs an out-of-bounds increment event set.

[0035] Upon receiving the set of excess incremental events, the permeability factor calculation submodule calls a pre-defined damage-permeability correlation function. The specific expression of this function is: This formula is used to quantify the change in permeability of soil caused by excessive deformation (damage). The advantage of this formula lies in the introduction of a damage sensitivity coefficient. and damage response index It employs an exponential function to accurately simulate the physical process by which the expansion of internal microcracks in soil beyond its elastic deformation limit leads to a rapid, nonlinear increase in the permeability coefficient. Compared to traditional linear models, this more accurately reflects the nonlinear growth of the permeability coefficient after soil damage.

[0036] The detailed explanations and value determination processes for each parameter in this formula are as follows: It is the permeability coefficient adjustment factor to be calculated, which is a dimensionless multiplier. This is the damage sensitivity coefficient of soil materials, characterizing the degree to which a material's permeability is sensitive to damage. This coefficient is determined through a graded loading pressure permeability test on soil samples. In the test, the axial strain is gradually increased while the change in permeability coefficient is measured, and a permeability coefficient growth curve is fitted to obtain the coefficient. The value of the permeability coefficient is determined by the slope of the line connecting the origin and the point of intersection of the two lines. For the sandy clay at a depth of -10 meters, the experimental value is . is a dimensionless damage response index that controls the degree of nonlinearity in the damage effect. This value is also obtained by fitting the pressure infiltration test data described above, and it reflects the degree of steepness in the growth of the permeability coefficient beyond the threshold strain. For this sandy clay, the fitting yields . is the incremental displacement per unit time, which is directly obtained from the set of threshold-exceeded increments. In this example, mm / h. is the preset displacement threshold value, which is obtained from the preset displacement threshold matrix for the corresponding depth. In this example, the value for a depth of -10 meters is mm / h.

[0037] The sub-module substitutes these specific values into the formula to solve it: first, calculate the relative threshold-exceeded degree in the parentheses: . Then, calculate the exponential part: . Finally, calculate the complete expression: . The result indicates that, near a depth of -10 meters, the permeability coefficient of the soil in this region should be adjusted to 4.4151 times the original value due to the occurrence of a threshold-exceeded displacement of 1.3 mm / h. The system performs this solving operation for each event in the set of threshold-exceeded increments, generating a set of permeability coefficient adjustment factors, each associated with a specific depth and threshold-exceeded event.

[0038] Please refer to Figure 1 and Figure 5 , the lag field adaptive correction module, based on the initial response field, adjusts the permeability coefficient value according to the permeability coefficient adjustment factor and the depth, updates the optimal lag time set, and generates a corrected response profile and a corrected response field; The corrected response field includes dynamic risk level, warning area range, and disaster occurrence probability; The lag field adaptive correction module includes: The permeability coefficient value adjustment sub-module, based on the initial response field, locates the spatial region corresponding to the depth in the initial response field according to the permeability coefficient adjustment factor and the depth, performs a multiplication operation between the permeability coefficient value in the region and the permeability coefficient adjustment factor, and generates an updated permeability coefficient matrix; The lag time set update sub-module calls the updated permeability coefficient matrix, recalculates the water flow permeation rate for multiple depth levels based on the values of the updated permeability coefficient matrix, and adjusts the time delay between rainfall and displacement response to obtain an updated optimal lag time set; The specific execution process of the lag time set update sub-module is to obtain the updated permeability coefficient matrix and the optimal lag time set, and according to the formula: ; calculating the updated lag time; wherein, is the initial lag time at depth is the updated lag time at depth is the initial lag time at depth extracted from the initial response field, is the initial permeability value at depth extracted from the initial response field, is the updated permeability value at depth extracted from the updated permeability matrix; iterating through all the monitoring depths, calculating the updated lag time at each depth one by one, and combining the corresponding monitoring point depth index and the associated response intensity to construct the updated optimal lag time set; the profile fitting correction submodule combines the time data in the updated optimal lag time set with the corresponding depth coordinates as data point pairs, performs polynomial interpolation operation on all data point pairs to establish a functional relationship, and generates a corrected response profile; the corrected field reconstruction submodule calls the corrected response profile, extracts the depth coordinates of the nodes of the three-dimensional grid model, substitutes the depth coordinates into the function represented by the corrected response profile one by one for evaluation, and assigns the calculated lag time to the corresponding nodes to construct a corrected response field; the generation process of the dynamic risk level, the warning area range, and the disaster occurrence probability included in the corrected response field includes: a risk level determination step, obtaining a multi-level risk division threshold, and comparing the node response prediction value of each warning grid element in the corrected response field with the multi-level risk division threshold one by one to assign an initial risk level to each warning grid element, obtaining a grid-based risk distribution map; a warning range identification step, calling the grid-based risk distribution map, setting a high risk level threshold, and using a density-based spatial clustering algorithm to identify and merge all adjacent warning grid elements with a risk level higher than the high risk level threshold into one or more connected regions to generate a warning area range; a disaster probability calculation step, for each warning area range, comprehensively calculating the average node response prediction value of all warning grid elements in the area, the area volume, and the cumulative value of the rainfall intensity sequence, and performing operation through a preset probability conversion model to obtain the disaster occurrence probability.

[0039] The lag field adaptive correction module receives the initial response field data and the permeability adjustment factor generated by the permeability coefficient disturbance determination module, dynamically corrects the initial response field, and finally generates a corrected response field containing risk level, warning range, and disaster probability.

[0040] The permeability coefficient value adjustment submodule locates the spatial region corresponding to the depth in the initial response field according to the permeability coefficient adjustment factor and the depth, performs a multiplication operation on the permeability coefficient value in the region and the permeability coefficient adjustment factor, and generates an updated permeability coefficient matrix; The core task of the lag time set updating submodule is to recalculate the time delay between rainfall and displacement response according to the updated permeability coefficient. The submodule calls the updated permeability coefficient matrix and the initial optimal lag time set, and applies the following formula for calculation: This formula is based on the basic physical principle that the seepage velocity of water in porous media is proportional to the permeability coefficient (Darcy's law). The lag time is mainly determined by the time required for rainwater to infiltrate to a certain depth, so the lag time is inversely proportional to the permeability rate, i.e., inversely proportional to the permeability coefficient. This formula accurately expresses this physical relationship. The parameter descriptions and assignments in the formula are as follows: is the updated lag time at depth , which is the calculation target of this step. According to the previous embodiment, the initial lag time at -10 meters is hours. is the initial permeability coefficient value at depth extracted from the initial response field, i.e., cm / s. is the updated permeability coefficient value at depth extracted from the updated permeability coefficient matrix, i.e., cm / s.

[0041] wherein, represents the updated lag time at depth , which is the calculation target of this step; represents the initial lag time at depth , whose value is extracted from the initial optimal lag time set; represents the initial permeability coefficient value at depth , whose value is extracted from the initial response field; represents the updated permeability coefficient value at depth , whose value is extracted from the updated permeability coefficient matrix.

[0042] Substitute the numerical values into the formula for calculation: hours. This result shows that due to the significant increase in permeability coefficient, the lag time for rainwater to infiltrate to a depth of -10 meters is sharply shortened from the original 4.0 hours to about 0.906 hours. The submodule iterates through all monitoring depths where the permeability coefficient adjustment occurs, and calculates the updated lag time at all depths one by one, and combines the corresponding monitoring point depth index and associated response intensity to construct an updated optimal lag time set.

[0043] The correction profile fitting sub-module and the correction field reconstruction sub-module perform similar operations as the response field space profile construction module, but use the updated optimal lag time set. First, the updated data point pairs, such as {(-5, 2.8), (-10, 0.906), (-15, 6.0)}, are subjected to polynomial interpolation or curve fitting operations to establish a new depth-time function relationship and generate a corrected response profile. Then, the depth coordinates of all nodes of the three-dimensional grid model are extracted again, and these coordinates are substituted into the new corrected response profile function one by one for evaluation. The new lag time obtained by calculation is assigned to the corresponding node, thereby constructing the final corrected response field.

[0044] On the basis of the corrected response field, the system performs risk assessment. The risk level determination step first loads a multi-level risk division threshold table, which is based on the length of the lag time. For example, a first-level risk (blue): hours; a second-level risk (yellow): hours; a third-level risk (orange): hours; and a fourth-level risk (red): hours. The system compares the lag time of each warning grid cell node in the corrected response field with the threshold table one by one. For nodes near a depth of -10 meters, the lag time is 0.906 hours, which is less than 2 hours, so it is assigned the highest fourth-level (red) risk level. This process traverses all grid cells to generate a grid-based risk distribution map.

[0045] The warning range identification step calls the risk distribution map and sets a high risk level threshold, for example, defining areas with risk levels of third-level (orange) and fourth-level (red) as high-risk areas. The system uses a density-based spatial clustering algorithm to search for all adjacent warning grid cells with risk levels higher than the threshold in three-dimensional space and identifies and merges these cells into one or more connected three-dimensional regions. These connected regions constitute the specific warning area range.

[0046] The disaster probability calculation step calculates for each identified warning area range. The system first calculates the average node response prediction value (average lag time) of all warning grid cells in the area and calculates the total volume of the area. At the same time, the rainfall intensity sequence is called to calculate the cumulative rainfall of the rainfall event that triggered the warning. These three parameters (average lag time, area volume, and cumulative rainfall) are input into a pre-set probability conversion model. The model is a classification model based on gradient boosting decision trees (GBDT), with average lag time, area volume, and cumulative rainfall as input features, and uses - Fold cross-validation for training and parameter optimization. For example, for a red early warning area with an average lag time of 1.5 hours, a volume of 50000 cubic meters, and a cumulative rainfall of 120mm, the model output after operation is a disaster occurrence probability of 85%. Finally, the response field is corrected to output in the form of dynamic risk level (grid risk distribution map), early warning area range (three-dimensional space area) and disaster occurrence probability (specific numerical value), providing support for disaster prevention and mitigation decision-making.

[0047] The above examples demonstrate the preferred embodiments of the present application, and any equivalent adjustment of the technical solutions based on software engineering methods is within the protection scope, including but not limited to: implementing algorithm logic in different programming languages, service reconstruction of functional modules, adjusting data interaction protocols, optimizing resource scheduling strategies, etc. Any implementation derived by reasonable modification of the data processing flow, service calling link or system architecture level without deviating from the technical core of the present application should be considered within the protection scope defined by the claims of the present application.

Claims

1. A geological disaster risk intelligent early warning system, characterized in that, The system comprises: A multi-dimensional time series lag solution module acquires a rainfall intensity sequence and a multi-depth displacement sequence, calculates a displacement rate sequence, generates an optimal lag time set through cross-correlation operation, and delivers the optimal lag time set to a response field spatial profile construction module; The response field spatial profile construction module generates a lag response profile according to the optimal lag time set, and substitutes a three-dimensional grid model node depth into the lag response profile to construct an initial response field, and delivers the initial response field to a lag field adaptive correction module; A permeability coefficient disturbance judgment module monitors the multi-depth displacement sequence, calculates a unit time displacement increment, and if the unit time displacement increment is greater than a preset displacement threshold, substitutes the numerical value of the corresponding unit time displacement increment into a damage-permeability correlation function to calculate a permeability coefficient adjustment factor, and delivers the permeability coefficient adjustment factor and depth to the lag field adaptive correction module; The lag field adaptive correction module adjusts the permeability coefficient value based on the initial response field according to the permeability coefficient adjustment factor and the depth, updates the optimal lag time set, and generates a corrected response profile and a corrected response field. 2.The intelligent early warning system for geological disaster risk according to claim 1, characterized in that, The optimal lag time set includes a rainfall-displacement response time difference, a monitoring point depth index, and an associated response intensity, the initial response field specifically refers to a warning grid unit, a node response prediction value, and a spatial distribution weight, the permeability coefficient adjustment factor specifically refers to a damage equivalent parameter, an adjustment direction, and an amplitude, and the corrected response field includes a dynamic risk level, a warning area range, and a disaster occurrence probability. 3.The intelligent early warning system for geological disaster risk according to claim 2, characterized in that, The multi-dimensional time series lag solution module comprises: A time series data stream receiving sub-module acquires a rainfall intensity sequence and a multi-depth displacement sequence, integrates the rainfall intensity sequence and the multi-depth displacement sequence into a structured data entity, and aligns and calibrates the time stamp to generate an original time series data set; A displacement rate gradient solution sub-module calls the multi-depth displacement sequence in the original time series data set, performs a first-order difference operation on the displacement data of each depth along the time axis to obtain a plurality of depth displacement rates, and performs weighted average processing on the displacement rates of all depths according to their monitoring depths to obtain an average displacement rate sequence; A sequence cross-correlation operation sub-module extracts the rainfall intensity sequence in the original time series data set, sets a preset time window according to the average displacement rate sequence, shifts the rainfall intensity sequence point by point within the time window, calculates the sum of the products of the corresponding data points of the two sequences after each shift, and generates a cross-correlation coefficient vector; A peak index positioning sub-module locates the numerical maximum coefficient in the vector as a peak point by traversing all coefficients in the vector for the cross-correlation coefficient vector, and converts the index value of the peak point in the cross-correlation coefficient vector into an offset time amount corresponding to the starting point of the preset time window to construct an optimal lag time set. 4.The intelligent early warning system for geological disaster risk according to claim 3, characterized in that, The response field spatial profile construction module comprises: The profile function fitting submodule executes a least square curve fitting operation on time values and corresponding depth values in the optimal lag time set as discrete data pairs with depth as the independent variable and time as the dependent variable, establishes a depth-time function, and generates a lag response profile; The node depth mapping submodule obtains three-dimensional mesh model node depths, and calls the lag response profile to calculate target time values at each mesh node by substituting each of the three-dimensional mesh model node depths into the depth-time function represented by the lag response profile, thereby obtaining a node lag time scalar set; The response field space construction submodule assigns each time value in the node lag time scalar set to a corresponding node in the three-dimensional mesh model according to the corresponding spatial coordinates, forms a discrete data field with node coordinates as the index and time values as the attributes, and constructs an initial response field. 5.The intelligent early warning system for geological disaster risk according to claim 4, characterized in that, The permeability coefficient perturbation determination module comprises: The displacement increment calculation submodule monitors the multi-depth displacement sequence, divides the multi-depth displacement sequence into continuous time units according to a preset sampling frequency, extracts the displacement measurement values at the start and end times in each time unit, and obtains the unit time displacement increment by performing a subtraction operation on the displacement measurement values; The overrun event discrimination submodule obtains a preset displacement threshold matrix, calls the unit time displacement increment, and performs a numerical comparison between the unit time displacement increment reported at each depth and the threshold value corresponding to the depth in the preset displacement threshold matrix to filter all increments that exceed the threshold value and construct an overrun increment event set; The permeability factor solving submodule collects a damage-permeability correlation function, substitutes the increment values recorded in the overrun increment event set as independent input variables into the damage-permeability correlation function expression to perform a solving operation, obtains function outputs corresponding to multiple increments, and generates a permeability coefficient adjustment factor. 6.The intelligent early warning system for geological disaster risk according to claim 5, characterized in that, The lag field adaptive correction module comprises: The permeability coefficient value adjustment submodule locates a spatial region corresponding to a depth in the initial response field based on the initial response field and the permeability coefficient adjustment factor and the depth, performs a multiplication operation between the permeability coefficient value in the region and the permeability coefficient adjustment factor to generate an updated permeability coefficient matrix; The lag time set updating submodule calls the updated permeability coefficient matrix, recalculates the water flow permeation rate according to the numerical values of the updated permeability coefficient matrix for multiple depth levels, adjusts the time delay between rainfall and displacement response, and obtains an updated optimal lag time set; The corrected profile fitting submodule combines time data in the updated optimal lag time set and corresponding depth coordinates as data point pairs, performs a polynomial interpolation operation on all the data point pairs to establish a functional relationship, and generates a corrected response profile; The corrected field reconstruction submodule calls the corrected response profile, extracts depth coordinates of three-dimensional mesh model nodes, substitutes the depth coordinates one by one into the function represented by the corrected response profile for evaluation, assigns the calculated lag time to the corresponding nodes, and constructs a corrected response field. 7.The intelligent early warning system for geological disaster risk according to claim 5, characterized in that, The permeation factor solver module specifically performs the process of calling the calculation formula defined by the damage-permeation correlation function: ； taking the unit time displacement increment in the set of ultra-limit incremental events as a damage measure parameter , and obtaining a preset displacement threshold value corresponding to the depth from the preset displacement threshold matrix as a threshold parameter substitute wherein, is a permeability coefficient adjustment factor to be calculated, is a soil material damage sensitivity coefficient, is a dimensionless damage response index, is the displacement increment of the unit time, is the preset displacement threshold value; According to the formula, the function output value corresponding to each over-limit increment is calculated, and all function output values are combined to generate a permeability adjustment factor. 8.The intelligent early warning system for geological disaster risk according to claim 6, characterized in that, The updated lag time set updating submodule specifically performs the process of obtaining the updated permeability coefficient matrix and the optimal lag time set, according to the formula: ； Calculate the updated lag time; wherein, is an updated lag time at a depth , is an initial lag time at the depth extracted from the set of optimal lag times, is an initial permeability value at the depth extracted from the initial response field, is an updated permeability value at the depth extracted from the updated permeability matrix. Traverse all monitoring depths, and calculate the updated lag time of all depths one by one, and combine the corresponding monitoring point depth index and the associated response intensity to construct an updated optimal lag time set. 9.The intelligent early warning system for geological disaster risk according to claim 6, characterized in that, The generation process of the dynamic risk level included in the corrected response field, the pre-warning area range and the disaster occurrence probability includes: Risk level determination step, obtain multi-level risk division threshold, and compare the node response prediction value of each pre-warning grid unit in the corrected response field with the multi-level risk division threshold one by one, and give each pre-warning grid unit an initial risk level, to obtain a grid risk distribution map; The pre-warning range identification step calls the grid risk distribution map, sets a high risk level threshold, uses a density-based spatial clustering algorithm to identify and merge all adjacent pre-warning grid units with a risk level higher than the high risk level threshold into one or more connected regions, and generates a pre-warning area range; Disaster probability solving step, for each pre-warning area range, the average node response prediction value of all pre-warning grid units in the area, the area volume and the cumulative value of the rainfall intensity sequence are comprehensively calculated, and a preset probability conversion model is used to obtain the disaster occurrence probability. 10.The intelligent early warning system for geological disaster risk according to claim 3, characterized in that, The sequence cross-correlation operator module first performs standardization processing on the data before calculating the cross-correlation coefficient vector, including: Mean standard deviation calculation step, for the rainfall intensity sequence and the average displacement rate sequence in the preset time window, the arithmetic mean and standard deviation of all data points in the sequence are calculated independently; Sequence standardization step, call the arithmetic mean and the standard deviation, traverse each data point in the rainfall intensity sequence and the average displacement rate sequence, subtract the data point value from the arithmetic mean of the sequence, and then divide the difference by the standard deviation of the sequence to generate a standardized rainfall sequence and a standardized displacement rate sequence; Correlation coefficient calculation step, according to the preset time window, the standardized rainfall sequence is shifted point by point, the sum of the products of the corresponding data points of the standardized rainfall sequence and the standardized displacement rate sequence after each shift is calculated, and the sum of the products is divided by the sequence length to generate a cross-correlation coefficient vector.

Images