Geological disaster early warning method and system based on dynamic data monitoring

Abstract

The invention relates to the technical field of geological disaster early warning, in particular to a geological disaster early warning method and system based on dynamic data monitoring, and the method comprises the following steps: obtaining geological displacement monitoring data, setting a time window range, respectively calculating a local extreme point and a local zero point of the monitoring data in a time window, and calculating the local extreme point and the local zero point of the monitoring data in the time window; obtaining a multi-scale reference value; and based on the multi-scale reference value, calculating a comprehensive statistical magnitude, and generating a data quality evaluation parameter. According to the method, a local extreme point and a zero point are dynamically calculated by setting a time window range, a multi-scale reference value is constructed to carry out comprehensive statistic generation and data evaluation, and collaborative discrimination of piecewise linear regression and threshold comparison are executed to form a critical index; and performing characteristic quantity recursive residual spectrum analysis and matrix decomposition operation by combining the power spectral density distribution and the spectral density matrix to establish an early warning characteristic sequence, and realizing multi-stage early warning indication output based on a grading rule.

Description

Technical Field

[0001] The present invention relates to the technical field of geological disaster warning, and particularly relates to a geological disaster warning method and system based on dynamic data monitoring. Background Art

[0002] The technical field of geological disaster warning is a comprehensive technical research direction, mainly focusing on the monitoring, identification, assessment and warning of geological disasters. This field involves multiple disciplines such as geology, hydrology, remote sensing technology, sensor technology, data analysis, etc. By establishing a geological disaster monitoring network, key parameters such as geological body displacement, groundwater level, rainfall, crack opening, etc. are collected in real time, and finally the early identification and warning of geological disasters such as landslides, collapses, debris flows, etc. are realized. However, in the prior art, simple trend judgment methods are used for deformation analysis, and a complete feature extraction process is not established, resulting in the omission of key warning information. In addition, the warning index system is too simple and lacks a hierarchical warning mechanism, making it difficult to implement differential management for different risk levels. Therefore, improvements are needed. Summary of the Invention

[0003] The purpose of the present invention is to solve the deficiencies existing in the prior art, and to propose a geological disaster warning method and system based on dynamic data monitoring.

[0004] To achieve the above purpose, the present invention adopts the following technical solutions. A geological disaster warning method based on dynamic data monitoring includes the following steps:

[0005] Obtain geological displacement monitoring data, set the time window range, calculate the local extreme points and local zero points of the monitoring data within the time window for the monitoring data within the time window respectively to obtain multi-scale reference values; based on the multi-scale reference values, calculate comprehensive statistics and generate data quality evaluation parameters;

[0006] Based on the data quality evaluation parameters, calculate the objective function values under each group of iteration parameters to obtain window optimization parameters; substitute the window optimization parameters into piecewise linear regression, perform collaborative discrimination and threshold comparison on the regression operation results, and generate deformation critical indicators;

[0007] Based on the deformation critical indicators, perform periodic analysis on the monitoring data sequences at each time scale, calculate the power spectral density distribution and spectral density matrix in the frequency domain to obtain frequency domain characteristic quantities, perform recursive residual spectrum analysis and matrix decomposition operations on the frequency domain characteristic quantities, accumulate the residual spectrum and matrix eigenvalues, and establish a warning feature sequence;

[0008] Based on the warning feature sequence, compare the feature sequences at each time scale with the corresponding thresholds, mark the time points exceeding the thresholds to obtain scale warning identifiers, count the distribution of the scale warning identifiers at each time scale, perform a matching operation on the distribution with the preset classification rules, and output a multi-scale warning indication.

[0009] Preferably, the steps for obtaining the multi-scale reference value are as follows:

[0010] Obtain geological displacement monitoring data, set time window range parameters and time window moving step parameters, analyze the variation characteristics of the monitoring data within the time window at different scales, and generate a multi-scale decomposition result;

[0011] Based on the multi-scale decomposition result, analyze the local extreme points and local zero points in the monitoring data to obtain local feature points;

[0012] Based on the local feature points, calculate the relative relationships and change trends between the feature points to form a multi-scale reference value.

[0013] Preferably, the steps for obtaining the data quality evaluation parameters are as follows:

[0014] Based on the multi-scale reference value, perform a difference calculation to obtain the difference results of the local extreme points and local zero points, and form a local difference value;

[0015] According to the local difference value, calculate a comprehensive statistic, and the calculation formula is:

[0016]

[0017] where Δx i is the i-th item in the difference result of the local extreme points, k is the number of local extreme points, Δy j is the j-th item in the difference result of the local zero points, m is the number of local zero points, and S is the comprehensive statistic;

[0018] Based on the comprehensive statistic, combined with the characteristic distribution of the comprehensive statistic, obtain the data quality evaluation parameters.

[0019] Preferably, the steps for obtaining the window optimization parameters are as follows:

[0020] Based on the data quality evaluation parameters, initialize the time window range parameters and time window moving step parameters, construct a recursive iteration process, gradually adjust the value ranges of the time window range and time window step parameters, and form an iteration parameter set;

[0021] According to the iteration parameter set, calculate the objective function values of each group of iteration parameters, and the calculation formula is:

[0022]

[0023] Among them, T i represents the squared value of the time window range parameter in the i-th iteration, R i represents the reference value of the time window range, n represents the total number of calculations of the time window range parameter, S j represents the squared value of the time window step parameter in the j-th iteration, M j represents the target reference value of the time window step, xm represents the total number of calculations of the time window step parameter, and F represents the objective function value under the current iteration parameter set;

[0024] Based on the objective function value, analyze the time window range parameter and the time window step parameter after recursive screening, and perform an optimal combination of all the screened parameters to generate window optimization parameters.

[0025] Preferably, the steps for obtaining the deformation critical index are as follows:

[0026] Based on the window optimization parameters, calculate the regression residuals and fitting coefficients in each segment respectively to generate a segmented regression analysis result;

[0027] According to the segmented regression analysis result, calculate the collaborative discrimination value, and the calculation formula is:

[0028]

[0029] Among them, R i is the fitting residual of the i-th segment, Q i is the corresponding window optimization parameter, p is the total number of segments, T j is the test value of the j-th order distribution in the rank sum test, q is the total number of test distributions, and I is the collaborative discrimination value;

[0030] Based on the collaborative discrimination value, by comparing with a set threshold, identify the deformation distribution and generate a deformation critical index.

[0031] Preferably, the steps for obtaining the frequency domain characteristic quantity are as follows:

[0032] Based on the deformation critical index, extract the monitoring data sequence at each time scale, perform Fourier transform to convert it into a frequency domain signal, and generate a frequency domain signal distribution;

[0033] According to the frequency domain signal distribution, calculate the power spectral density distribution value, and the calculation formula is:

[0034]

[0035] Wherein, P is the power spectral density distribution value, S(f) is the frequency response function of the frequency-domain signal, f1 and f2 are the upper and lower limits of the frequency range, and XM k is the amplitude of the k-th component in the frequency-domain signal, and N k is the total number of frequency points of this component, and tm is the total number of components;

[0036] Based on the power spectral density distribution value, a spectral density matrix is established, and through the correlation analysis between matrix elements, the characteristic frequency is extracted and the frequency-domain characteristic quantity is generated.

[0037] Preferably, the steps for obtaining the warning feature sequence are as follows:

[0038] Based on the frequency-domain characteristic quantity, the residual value of each frequency-domain data segment is calculated in segments, and the residual values are accumulated in sequence to form a recursive residual spectrum, and the recursive residual spectrum is generated;

[0039] According to the recursive residual spectrum, the residual spectrum matrix is decomposed, the eigenvalues of the matrix are extracted one by one, and all the eigenvalues are sorted in order and merged into a set to obtain the matrix eigenvalue set;

[0040] Based on the matrix eigenvalue set, all the residual values in the recursive residual spectrum and the extracted matrix eigenvalues are accumulated, the accumulated results are rearranged according to the time series, and a warning feature sequence is established.

[0041] Preferably, the steps for obtaining the scale warning identifier are as follows:

[0042] Based on the warning feature sequence, each feature sequence on each time scale is processed by data segmentation, arranged in time series, and the eigenvalue of each segment is compared with the corresponding threshold value, and all the feature points exceeding the threshold value are extracted to generate a super-threshold feature record;

[0043] According to the super-threshold feature record, the time points corresponding to the feature points are marked, all the recorded data within all time scales are called, and all the super-threshold time points are marked in segments in the order of time points to obtain the super-threshold time point identifier;

[0044] Based on the super-threshold time point identifier, the distribution of time point identifiers on all time scales is analyzed hierarchically, all the identifiers are merged according to the established identifier classification rules, and the merged results are integrated according to the time scale to generate the scale warning identifier.

[0045] Preferably, the steps for obtaining the multi-scale warning indication are as follows:

[0046] Based on the scale warning identifier, extract the identifier distribution from the data of each time scale, count the quantity and distribution characteristics of each identifier one by one in the order of time scale, and generate the time scale identifier distribution statistical result by combining the time span and distribution characteristics of the identifier;

[0047] According to the time scale identifier distribution statistical result, call the preset grading rules one by one, perform matching operations on the identifier distribution of each time scale, compare each group of identifier distributions with the grading rules item by item, and generate the grading matching result according to the grading standard;

[0048] Based on the grading matching result, integrate the matching categories on all time scales, analyze and merge the graded matching categories in chronological order, and uniformly output the merged result as a multi-scale warning indication.

[0049] The present invention provides a geological disaster warning system, including:

[0050] A data acquisition and preliminary processing module, which collects geological displacement monitoring data, sets the time window range, calculates the local extreme points and local zero points of the monitoring data within the time window, and generates multi-scale reference values;

[0051] A data quality assessment module, which calculates comprehensive statistics based on the multi-scale reference values to obtain data quality assessment parameters;

[0052] A parameter optimization and regression analysis module, which uses the data quality assessment parameters to iteratively calculate the objective function value, obtains window optimization parameters, substitutes the window optimization parameters into the piecewise linear regression model, performs collaborative discrimination and threshold comparison, and generates deformation critical indicators;

[0053] A frequency domain feature analysis module, which performs periodic analysis on the monitoring data sequence according to the deformation critical indicators, including the calculation of power spectral density distribution and spectral density matrix, performs recursive residual spectrum analysis and matrix decomposition, accumulates the residual spectrum and matrix eigenvalues, and establishes a warning feature sequence;

[0054] A warning signal generation and output module, which, based on the warning feature sequence, compares the feature sequences on each time scale with the set threshold, marks the time points exceeding the threshold, generates scale warning identifiers, counts the distribution of scale warning identifiers on each time scale, performs matching operations with the preset grading rules, and outputs a multi-scale warning indication.

[0055] Compared with the prior art, the advantages and positive effects of the present invention are:

[0056] The present invention dynamically calculates local extreme points and zero points by setting a time window range, constructs multi-scale reference values for comprehensive statistic generation and data evaluation, performs collaborative discrimination and threshold comparison of piecewise linear regression to form critical indicators, combines power spectral density distribution and spectral density matrix for recursive residual spectrum analysis of characteristic quantities and matrix decomposition operations to establish an early warning characteristic sequence, and realizes multi-level early warning indication output based on hierarchical rules. The multi-scale time window is adaptively adjusted to eliminate external interference factors. The piecewise regression and collaborative discrimination mechanism accurately identifies the trend change of geological deformation. The spectral analysis and matrix operations deeply excavate the hidden characteristics of data, realizing the early prediction of risk factors. The recursive residual analysis of frequency domain characteristic quantities evaluates the periodic law of the monitoring sequence and accurately identifies the deformation evolution characteristics of geological bodies. The rule matching operation mechanism ensures the rationality of early warning output. The multi-level early warning indication system effectively improves the effectiveness of early warning, laying a data foundation for disaster prevention and control decisions. The early warning sequence constructed based on the accumulation of matrix eigenvalues can locate key time nodes and achieve positioning and rapid response to disaster precursor signals. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is a schematic diagram of the steps of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0058] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0059] Please refer to Figure 1 , the present invention provides a technical solution, a geological disaster early warning method based on dynamic data monitoring, including the following steps:

[0060] Obtain geological displacement monitoring data, set a time window range, and calculate the local extreme points and local zero points of the monitoring data within the time window for the monitoring data within the time window to obtain multi-scale reference values; based on the multi-scale reference values, calculate comprehensive statistics and generate data quality evaluation parameters;

[0061] Based on the data quality evaluation parameters, calculate the objective function values under each set of iteration parameters to obtain window optimization parameters; substitute the window optimization parameters into piecewise linear regression, perform collaborative discrimination and threshold comparison on the regression operation results, and generate deformation critical indicators;

[0062] Based on the deformation critical indicators, perform periodic analysis on the monitoring data sequences at each time scale, calculate the power spectral density distribution and spectral density matrix in the frequency domain to obtain frequency domain characteristic quantities, perform recursive residual spectrum analysis and matrix decomposition operations on the frequency domain characteristic quantities, and perform cumulative operations on the residual spectrum and matrix eigenvalues to establish an early warning characteristic sequence;

[0063] Based on the early warning feature sequence, compare the feature sequence at each time scale with the corresponding threshold, mark the time points exceeding the threshold to obtain the scale early warning identifier, count the distribution of the scale early warning identifier at each time scale, perform a matching operation on the distribution with the preset grading rules, and output a multi-scale early warning indication.

[0064] The steps for obtaining the multi-scale reference value are as follows:

[0065] Obtain geological displacement monitoring data, set the time window range parameter and the time window moving step parameter, analyze the variation characteristics of the monitoring data within the time window at different scales, and generate a multi-scale decomposition result;

[0066] Based on the multi-scale decomposition result, analyze the local extreme points and local zero points in the monitoring data to obtain local feature points;

[0067] Based on the local feature points, calculate the relative relationship and change trend between the feature points to form a multi-scale reference value.

[0068] Specifically, first set a number of displacement sensors in the target area and continuously collect geological displacement monitoring data. Obtain a sample every 10 seconds and use it as the basis for subsequent analysis. Subsequently, set the time window range parameter in combination with the observation results of surface displacement changes in historical data. For example, a 24-hour window can be set in areas with relatively frequent surface activities and a 48-hour window can be set in areas with relatively stable activities. Then, perform a rotational comparison with a 6-hour time window moving step. To determine whether the geological displacement amplitude within each time period is in an abnormal state, the monitoring value can be compared with the thresholds of different stratum types obtained through long-term observation and statistics in advance. For example, for loose geological areas, a displacement value recorded between 0 cm and 5 cm is considered a normal range, 5 cm to 10 cm is listed as a higher change range, and exceeding 10 cm is considered a significant fluctuation area. If the recorded value exceeds the corresponding threshold, it is marked as the target time period for subsequent in-depth analysis. After all time periods are recorded, split the monitoring values within each time period into the corresponding windows for multi-scale difference and trend analysis. Identify the significant change patterns by comparing the merging results of the high-frequency fluctuation area and the low-frequency fluctuation area, and summarize the fluctuation characteristics at these different scales together. The summarized result is regarded as the multi-scale decomposition result.

[0069] The monitoring data are extracted from each time period distinguished in the multi-scale decomposition result. To further locate the local extreme point, several adjacent sampling values ​​can be selected in the same time period and compared in sequence. When a sampling value is greater than its two adjacent sampling values ​​and the difference exceeds the reference difference (for example, 0.5 cm) calculated based on the natural variation range in past observations, it is marked as a local maximum point. When a sampling value is less than its two adjacent sampling values ​​and lower than the difference established by the same rule, it is marked as a local minimum point. To determine the local zero point, it is necessary to record the change of the monitoring value from positive to negative or negative to positive over time. The reference difference and range are derived from the average distribution and maximum and minimum interval evaluation of surface deformation data over many years (for example, 0.3 cm to 1 cm is set as the turning interval according to different stratum characteristics). After completing the above search for high and low values ​​and zero points in each time period, all the identified extreme points and zero points are listed in the same sequence and stored in the corresponding records. By arranging the positions of these two types of points in the time domain, a more intuitive fluctuation change trajectory can be obtained. Finally, this sequence is called a local feature point.

[0070] Taking local feature points as the main reference, we first calculate the displacement differences between adjacent extreme points and record the specific values. Then, we analyze the change trend by combining the order of the previous and next extreme points and the zero point on the time axis. The displacement increment that is greater than 1 cm and continuously spans at least 3 sampling periods can be regarded as an accelerated upward trend, while the displacement decrease that is less than -1 cm and spans the same sampling period can be regarded as a significant downward trend. The above specific values ​​and the values ​​spanning the period are all derived from the statistical results of many years of historical data and field experience settings. Then, the positive and negative intervals where the zero point is located are distinguished. By summarizing the distribution of these displacement differences and positive and negative intervals in each time period and combining the relative position relationship between the aforementioned extreme points and the zero point, a set of displacement increase and decrease patterns under multi-scales can be obtained in the time domain. The key parameters of these increase and decrease patterns are listed together and stored in a unified record. All these summarized parameter sets are regarded as the final multi-scale benchmark values.

[0071] The steps to obtain data quality assessment parameters are as follows:

[0072] Based on the multi-scale reference value, a differential calculation is performed to obtain the differential result between the local extreme point and the local zero point to form a local difference value;

[0073] According to the local difference value, the comprehensive statistics are calculated, and the calculation formula is:

[0074]

[0075] Where Δx i is the i-th item in the local extreme point difference result, k is the number of local extreme points, Δy jis the j-th item in the local zero-point difference result, m is the number of local zero points, and S is the comprehensive statistic;

[0076] Based on the comprehensive statistic and combined with the characteristic distribution of the comprehensive statistic, a data quality evaluation parameter is obtained.

[0077] Specifically, the multi-scale reference values have been obtained in the previous steps. Based on this reference value, adjacent sampling points in any section of the monitoring time series are compared and subtracted, so as to generate a difference result between each pair of sampling pairs. To accurately record the tiny changes in geological displacement in the actual scenario, first, the preliminary range applicable to the current formation characteristics is statistically obtained through the pre-accumulated geological sampling historical data (for example, regarding the single-displacement change within 0 cm to 1 cm as normal, exceeding 1 cm to 2 cm as medium amplitude, and exceeding 2 cm as high change). These ranges are divided into multiple intervals to form a one-to-one comparison process between the difference result and the monitoring moment. By reading the previously obtained multi-scale reference values, the interval where the difference is located is judged. If the difference result exceeds the set upper limit of the medium amplitude, it is directly marked as a potential anomaly, and the differences within the same time period are cumulatively summed during subsequent statistics. If the difference values of three consecutive sampling pairs all exceed the upper limit of the medium amplitude, it is upgraded to a high-change record. At the same time, it is judged whether there has been a maximum or minimum change trend at the sampling point positions corresponding to each difference. By continuously tracking these marks, the deviation of the local extreme points and zero points is located, so as to obtain the difference value set of each local extreme point and local zero point and integrate it to form the final local difference value.

[0078] The advantage of the formula is that by jointly operating on the sum of squares of the local extreme point differences and the average value of the local zero point differences, the comprehensive state of large-scale changes and zero-crossing distribution can be reflected in the same index, so as to summarize the overall characteristics of different types of differences in one value.

[0079] Δx i The acquisition steps of are as follows: segment the geological displacement data recorded by the sensor during 72 hours of continuous displacement observation, select every 1 hour as the sampling interval, statistically obtain the positions of possible local extreme points within each interval, and calculate the difference values between adjacent extreme points. Each difference value is denoted as Δx i , for example, during the monitoring of the same formation, the displacements are recorded as 2.2 cm and 3.0 cm at the 5th hour and the 6th hour respectively, and the difference is 0.8 cm. This difference is denoted as Δx1 and stored in the sequence. If the displacement of 2.7 cm is recorded again at the 7th hour at this measuring point, the difference value between the 6th hour and the 7th hour is approximately -0.3 cm and is denoted as Δx2. Repeat the same method to process all adjacent extreme point differences within 72 hours and form a sequence {Δx1, Δx2,..., Δx k}, where k is determined by the number of extreme points identified during the monitoring.

[0080] The steps to obtain k are as follows: During the above-mentioned 72-hour observation period, based on the displacement sampling values monitored hourly, all the time-series data are traversed to identify the maximum or minimum values, and the total number of such extreme points is counted to form the corresponding number k. If 12 extreme points are confirmed during the entire monitoring process, then k is equal to 12. This quantity is taken from the complete geological displacement sequence uploaded by the sensor and, combined with the interactive inspection results of the technician for the time of each maximum or minimum peak, forms the final set of extreme point numerical values, thereby obtaining the actual value of k.

[0081] Δy j The steps to obtain it are as follows: During the same 72-hour observation, to identify the position where the zero point appears, the point where the adjacent sampling value changes from positive to negative or from negative to positive is recorded as the zero point, and the difference between the samplings before and after this point is called Δy. j Specifically, the positive and negative signs of adjacent sampling values are compared and the numerical magnitudes are recorded. If the sampling value in the previous period is 0.6 cm and the sampling value in the next period is -0.1 cm, then a zero point is identified between these two samplings, and the absolute value of the difference before and after this zero point is recorded as Δy. j After collecting all such zero-point differences, {Δy1, Δy2,..., Δy m} is formed, where m corresponds to the number of zero points found.

[0082] The steps to obtain m are as follows: In the previous process, whenever a change in sampling value from positive to negative or from negative to positive is found, that position is recorded as a zero point. After a full-scale detection of the entire 72-hour data, all the zero-point positions are marked one by one, and then the total number of zero points is summarized and used to represent m. For example, if 9 zero points are finally calibrated, then m is 9. The final confirmation of this value requires comparing the time stamp of the sensor with the actual situation of the displacement direction change, and each zero point is matched with the sampling record at a specific moment to ensure the accuracy of the statistics.

[0083] The steps to obtain S are as follows: S represents the final comprehensive statistic obtained by combining the sum of the squares of the differences of local extreme points and the average of local zero-point differences and performing a logarithmic operation. To facilitate comparison in the output, its numerical range is set in a logarithmic mapping space. By performing operations on the previously obtained Δx i 、Δy j as well as k and m, S can be made an index to measure the distribution of differences between each extreme value and zero point. When combined with subsequent data quality analysis, S needs to be written into the same record together with other statistical parameters.

[0084] Calculation process:

[0085] The first step is to calculate For example, when k = 12, if the difference value sequence {Δx1, Δx2,...Δx 12} are {0.8, -0.3, 0.5, 1.2, -0.9, 1.1, 0.6, -0.4, 0.2, 1.4, -0.8, 0.7} cm respectively, and their squared values are {0.64, 0.09, 0.25, 1.44, 0.81, 1.21, 0.36, 0.16, 0.04, 1.96, 0.64, 0.49} respectively. Summing up these values gives:

[0086]

[0087] In the second step, using k = 12 as the denominator, we get:

[0088]

[0089] In the third step, calculate and divide by m. If m = 9 and the zero - point difference sequence {Δy1, Δy2,... Δy9} are {0.5, 0.2, -0.3, 0.6, -0.7, 1.0, 0.9, -0.4, 0.3} cm respectively, then their absolute values are {0.5, 0.2, 0.3, 0.6, 0.7, 1.0, 0.9, 0.4, 0.3} respectively. Summing up these values:

[0090]

[0091] Then divide by m = 9 to get:

[0092]

[0093] In the fourth step, take the natural logarithm of the second term in the product of the above two terms. Thus:

[0094] S=(0.674)·ln(0.544)≈(0.674)·(-0.608)

[0095] S≈ - 0.410

[0096] This result indicates that when S is negative, it means that the squared value of the local extreme - point difference accounts for a relatively high proportion and the average value of the zero - point difference is relatively small. If more extreme - point differences show large - magnitude values during sampling and the overall zero - point difference is not significant, then the value of S may further decrease. If S is greater than a certain defined limit (such as 0.5), it means that both the extreme values and the zero - point differences in this area show greater fluctuations, and the discreteness of the monitoring data can be further confirmed subsequently.

[0097] The comprehensive statistics have been calculated above. By combining their distributions in different time periods, the overall fluctuations of the monitoring data can be analyzed. To further form data quality assessment parameters, it is necessary to find out whether there are multiple concentrated distribution intervals in the characteristic distribution. The specific approach is to first establish a set of discrimination criteria based on the magnitude of the S value. For example, the range of -1.0 to 0.3 for the S value is regarded as the normal interval, 0.3 to 0.7 is listed as the critical interval, and greater than 0.7 is listed as the abnormal interval. The above interval ranges are determined by technicians summarizing the minimum and maximum records of the S values obtained from previous monitoring and after multiple tests. By traversing the S values corresponding to all monitoring times, determine the interval they belong to and record the distribution status that appears at each time point. At the same time, record the frequency of the S values within each interval. If the cumulative time period within the critical interval exceeds 5 hours and the S value is greater than 0.3 three times in a row, it will be marked for key attention. For the situation where the S value is greater than 0.7, it will be marked as abnormal. After the determination is completed, associate all the marks with the original time series and the geological displacement situation, and organize the set of mark results into data quality assessment parameters.

[0098] The steps to obtain the window optimization parameters are as follows:

[0099] Based on the data quality assessment parameters, initialize the time window range parameter and the time window moving step size parameter, construct a recursive iteration process, and gradually adjust the value ranges of the time window range and the time window step size parameter to form an iteration parameter set;

[0100] According to the iteration parameter set, calculate the objective function value of each group of iteration parameters. The calculation formula is:

[0101]

[0102] where, T i represents the square value of the time window range parameter in the i-th iteration, R i represents the reference value of the time window range, n represents the total number of calculations of the time window range parameter, S j represents the square value of the time window step size parameter in the j-th iteration, M j represents the target reference value of the time window step size, xm represents the total number of calculations of the time window step size parameter, and F represents the objective function value under the current iteration parameter set;

[0103] Based on the objective function value, analyze the time window range parameter and the time window step size parameter after recursive screening, and optimize the combination of all the screened parameters to generate window optimization parameters.

[0104] Specifically, the data quality assessment parameters have been obtained in advance. When setting the time window range parameter and the time window moving step parameter based on these parameters, first, make a preliminary induction by referring to the periods when the geological displacement changes are relatively concentrated in previous monitoring records, and compare these monitoring records with the pre-established effective ranges one by one. For example, compare the geological displacement rate with the range of 0 cm / hour to 2 cm / hour, and compare the cumulative displacement with the range of 0 cm to 20 cm. For the periods with relatively frequent geological activities, the time window range parameter can be shortened to 6 hours and the time window moving step can be appropriately increased to 3 hours. For relatively stable periods, the time window range parameter can be extended to 24 hours and the time window moving step can be set to 6 hours. By batch-processing all the recorded data in different periods, a multi-group iterative parameter candidate list is formed. For each candidate list, recursive iteration will be performed in the current environment to test its adaptability to the displacement distribution in each period. If the record corresponding to the time window range parameter exceeds 2 cm / hour in a certain section, it is marked as a high-activity period and the time window range parameter in this group of candidate lists is retained for subsequent statistics. If there are no significant fluctuations in the displacement amplitude in multiple sections defined by the time window moving step parameter, it is determined that this parameter combination is applicable to relatively stable areas, and it will continue to be compared with other candidate combinations in the next recursive link. When all candidate lists have undergone multiple rounds of statistical comparison, a complete set of iterative parameter sets covering the geological activity characteristics of each period will be formed.

[0105] The advantage of the formula is that it measures the deviation degree of the time window range parameter by comparing the square difference of the first part with the reference value, and at the same time combines the second part to perform a product average operation with the product of the square value of the step parameter and the target reference value, so that both the time window range and the time window moving step parameters can be comprehensively evaluated in the same index.

[0106] T i The acquisition steps of

[0107] T i represents the square value of the time window range parameter in the i-th iteration. The time window range parameter involved comes from the historical data records of surface displacement monitoring. Technicians divide it in multiple different time slicing methods during a monitoring period of up to 360 hours (about 15 days), and mark the geological activity periods collected daily to form a series of candidate periods as the preliminary time window. After quantifying these candidate periods in hours, an original window sequence is formed, and then the effective windows are selected and numbered according to the distribution characteristics of the geological activity peak periods and low-activity periods. For example, different ranges such as 6 hours, 12 hours, 18 hours, 24 hours, etc. are selected as the core windows, denoted as T1, T2, T3, T4, etc. In the iterative process, square operations will be performed on these window ranges to obtain T iFor the value of , if 6 hours is used as the time window range in the first iteration, then T1 = 36; if 12 hours is used as the time window range in the second iteration, then T2 = 144, and so on for T3, T4, etc.

[0108] R i The obtaining steps of are as follows:

[0109] R i represents the reference value of the time window range. Each reference value is formed by comparing and analyzing the stable area and the active area in the actual monitoring environment according to geological activities. Technicians split the 360-hour monitoring record by the whole day and extract the sections where high displacement rates (such as exceeding 4 cm / hour) occur. Then, the start and end times of these sections are used to measure the size of the time window in the ideal state, so as to statistically obtain the average time span in each section. For example, if it is found that multiple sections with high displacement rates often last about 12 hours, then the value 12 is recorded as one of the i R. After that, through classifying and summarizing more monitoring records, several reference value sequences are formed. For example, R1 = 12, R2 = 18, R3 = 24 hours, and they are converted into the same order of magnitude (i.e., hours) as required for comparison with i T for storage, so as to correspond to the square difference operation of i T in each iteration.

[0110] The obtaining steps of n are as follows:

[0111] n represents the total number of calculations of the time window range parameter. During the 360-hour monitoring period, technicians divide the time window according to several spans such as 6 hours, 12 hours, 18 hours, 24 hours, etc. and conduct multiple rounds of recursive attempts. Each round of attempt will generate a specific candidate for the window range and square it for comparison with the corresponding reference value. If four possible time window ranges are allocated in one iteration, then n = 4. In subsequent iterations, more sub-intervals can be introduced. For example, 12 hours can be split into 9 hours and 15 hours. Finally, if there are 8 feasible window ranges in this round of iteration, then n = 8. This method generates the final value of n by comparing the window range parameters in the iteration order and, together with the aforementioned i T and i R, completes the subsequent accumulation and averaging calculations.

[0112] S j The obtaining steps of S are as follows:

[0113] S jDenote the square value of the time window step parameter in the j-th iteration. The acquisition of the time window step parameter also relies on the observation of geological activity characteristics. Technicians will check whether there is a continuous large displacement difference between the geological sections with obvious activities and the relatively gentle geological sections within a certain time interval, so as to determine the appropriate step interval. For example, the step length in the area with relatively frequent geological activities is set to 2 hours, and the step length in the area with relatively stable activities is set to 4 hours or 6 hours, etc. A number of step candidate sequences are statistically obtained and squared successively to form the value of S j If the step length selected for the first time is 2 hours, then S1 = 4; if the step length selected for the second time is 4 hours, then S2 = 16. Different step lengths are successively constructed into a candidate parameter set for iteration.

[0114] M j The acquisition steps of M are as follows:

[0115] M j Denote the target reference value of the time window step length. Its determination process lies in comprehensively considering the dynamic changes in geological displacement monitoring and the time period information of key sections confirmed manually. The average step lengths of these key sections are statistically calculated. If it is found that the intervals of most displacement mutations are about 3 hours, then the value 3 is listed in the candidate reference value list as M j At the same time, technicians may also list hour values such as 4 or 6 according to the activity records in different regions over a long period. If three target reference values of 3, 4, and 6 are finally selected, then j can be incremented from 1 to 3 in the iteration. Each time, the specific step length extracted is matched with the target reference value to quantify their deviation degree.

[0116] The acquisition steps of xm are as follows:

[0117] xm represents the total number of calculations of the time window step parameter, which is similar to the concept of n, but is statistically calculated for the independent dimension of the step length. In each iteration, a step length is selected from multiple options such as 2 hours, 4 hours, 6 hours, etc. and squared. If a total of 3 feasible step lengths are confirmed, then xm = 3. When technicians think that a finer step length granularity is needed, intermediate values such as 3 hours or 5 hours can also be inserted into the above-mentioned step lengths, so that xm can be increased to 5 or more in the same round of iteration, in order to multiply with S j and M j one by one and finally realize the comprehensive evaluation of all step candidates.

[0118] The acquisition steps of F are as follows:

[0119] Let \(F\) denote the objective function value under the current set of iteration parameters. The formula consists of two parts. The first part is the measurement of the difference between the squared value of the time window range parameter and the reference value. The second part is the product evaluation obtained by combining the squared value of the step size parameter with the target reference value. \(F\) combines the results of these two parts through addition to express the suitability of the time window range and the step size within the same quantization metric. To obtain the specific value of \(F\), \(T\) as described previously is required. i 、 \(R\) i 、 \(S\) j 、 \(M\) j and \(n\) and \(x_m\) need to be determined first.

[0120] Calculation process:

[0121] In the first step, calculate

[0122] In one iteration, the time window ranges are squared after taking 6, 12, 18, and 24 hours, resulting in 36, 144, 324, and 576 respectively. The corresponding reference values are recorded as 6, 12, 24, and 24. The values 6, 12, 24, and 24 are all extracted from the actual geological activity records using the method described above. At this time, \(n = 4\). Calculate the absolute value of each difference: \(\{|36 - 6|\) , \(|144 - 12|\) , \(|324 - 24|\) , \(|576 - 24|\} = \{30, 132, 300, 552\}\). Denote their sum as 1014 and divide by \(n = 4\) to obtain:

[0123]

[0124] Take the square root of this value and round to approximately 15.93 to form the result of the first part.

[0125] In the second step, for the time window step size, set it to 2, 4, and 6 and square them to get 4, 16, and 36 respectively, corresponding to the target reference values 3, 4, and 6. The three groups of multiplications are \((4 + 3)\), \((16 + 4)\), \((36 + 6)=\{7, 20, 42\}\). Their product is \(7×20×42 = 5880\). Then divide by \(x_m = 3\) to approximately get 1960 as the intermediate result of the second part.

[0126] In the third step, add the result of the first part (approximately 15.93) to the result of the second part (1960) to get \(F≈1975.93\). This is the objective function value when the combination of 6, 12, 18, 24 - hour window ranges and 2, 4, 6 - hour step sizes is selected in this round of iteration.

[0127] The result shows that the smaller the value of 1975.93, the better the matching degree between the time window range and the step size for the monitoring data. When this value is greater than 2000, it indicates that there are obvious parameter differences in the current combination, and it may be necessary to conduct another screening or introduce a more appropriate combination of time window range and step size. When the value falls between 1500 and 2000, it is usually regarded as a medium deviation, and technicians can evaluate whether further refinement or local adjustment is needed. If the value is less than 1500, it means that the parameter combination fits well with the coverage of the monitoring period and the step size segmentation.

[0128] The calculation of the current objective function value has been completed. To compare a series of time window range parameters and time window step size parameters after recursive screening, the F values of each group of parameters need to be sorted by size and recorded in the same sequence. Then, check whether there are values higher than 2000 in the sorted list. If so, mark this group of parameters as the type to be excluded in the subsequent screening. Then, select the parameter combinations with F values between 1500 and 2000 respectively and compare them with the existing combinations with values less than 1500, and record the specific distributions of the two types of combinations in each monitoring period. Through this inductive screening, a set of parameter sets that are considered to best meet the monitoring requirements and have good adaptability in both coverage and step size can be obtained. For combinations with repeated or similar F values, the decision to keep or discard them is determined based on the comparison judgment of the geological activity distribution. Finally, after comprehensively analyzing all parameter combinations that meet the screening criteria, window optimization parameters are generated.

[0129] The steps to obtain the deformation critical index are as follows:

[0130] Based on the window optimization parameters, calculate the regression residuals and fitting coefficients for each segment respectively to generate the segmented regression analysis results;

[0131] According to the segmented regression analysis results, calculate the collaborative discrimination value. The calculation formula is:

[0132]

[0133] where, R i is the fitting residual of the i-th segment, Q i is the corresponding window optimization parameter, p is the total number of segments, T j is the test value of the j-th order distribution in the rank sum test, q is the total number of test distributions, and I is the collaborative discrimination value;

[0134] Based on the collaborative discrimination value, by comparing it with the set threshold, identify the deformation distribution and generate the deformation critical index.

[0135] Specifically, after obtaining the window optimization parameters, it is necessary to extract the observed data at different time periods. First, determine the start and end time points of each segment and compare them with the previously recorded geological displacement values. Arrange the sampling values within each segment in order and use the time and displacement amount as the independent and dependent variables for regression. Then, perform a fitting operation on each segment using a unary linear method, and extract the corresponding slope and intercept as fitting coefficients. The slope can be obtained by dividing the displacement difference between two selected monitoring times within the segment by the monitoring time difference, and the intercept is calculated from the minimum deviation between all sampling values and the fitting line through least squares operation. After completing the linear fitting for each segment, record the difference between the predicted value and the measured value to obtain the regression residual. If the residual value at any time is greater than 2 cm or exceeds 1 cm continuously three times, mark this segment as an abnormal segment. These thresholds are determined by the long-term statistical results of the historical observed data and corrected in combination with the on-site measured situation. The regression residuals and fitting coefficients of all segments are recorded in sequence and centrally sorted after traversing all monitoring periods. During this process, the residual distribution of each segment can be further identified, and the segments with continuously large errors are marked as high-priority areas to be verified. After all records are summarized, the segmented regression analysis results are formed.

[0136] The benefit of the formula is that after multiplying and summing the squared residual amounts in each segment with the window optimization parameters and taking the average, and then calculating the difference with the square root of the sum of the squared values of the rank sum test distribution values, it can comprehensively measure the squared scale of the regression residuals and the overall deviation degree reflected by the test distribution in the same index, which is conducive to realizing the correlation evaluation of multi-dimensional information with fewer calculation steps.

[0137] R i The acquisition steps of

[0138] R i represents the fitting residual of the i-th segment, which refers to the difference between the regression predicted value and the measured value within the segment. When technicians perform linear regression on each segment, they compare the numerical differences between the measured data and the fitting line point by point, and then average or weight all these differences within the same segment. Finally, the average residual or weighted residual of this segment is obtained. To maintain consistency in subsequent calculations, this residual is often recorded in centimeters and saved in the sequence {R1, R2, …, R p}, where p is the total number of segments, and each R iAll come from the regression process of the corresponding segment. If the time is divided into four segments every 10 hours in a 40-hour geological displacement monitoring sequence, then p = 4 can be obtained. The regression residuals of the corresponding segments can be recorded as R1, R2, R3, and R4 respectively. In actual implementation, if a certain segment contains 60 sampling data, the 60 sampling values will be compared with the regression line of this segment one by one to calculate the differences, and then the differences will be aggregated according to the established weighting method to finally form R i 。

[0139] Q i The acquisition steps of are as follows:

[0140] Q i is the corresponding window optimization parameter. In the previous process of screening and combining the time window range and step size parameters, each regression corresponds to a finally confirmed time window size or the matching value of the window moving step size. The technical personnel quantify the time window or step size data accordingly to form a Q corresponding to the segment number i , for example, in a calculation, the time window is optimized to 8 hours and the step size is optimized to 4 hours. These combined values obtained after multiple rounds of recursive iteration screening in the previous text can be marked on each segment regression task to form a sequence such as {Q1, Q2,..., Q p}. Each Q i is recorded in hours, and in some scenarios, the window parameters with a longer time span will be converted to correspond to the residual unit, so as to perform multiplication operations in the same dimension system. If the window optimization parameter corresponding to the first segment is 8, then Q1 = 8; if the window optimization parameter corresponding to the second segment is 6, then Q2 = 6, and so on.

[0141] The acquisition steps of p are as follows:

[0142] p is the total number of segments. When the data monitoring period is divided into several time blocks by the technical personnel, each time block will perform a regression analysis to form an R i and the corresponding Q i . When the entire monitoring period is divided into p segments, there will be p residual values paired with p window optimization parameters. The value of p is related to the monitoring duration and the segmentation strategy. For example, in a 40-hour observation period, if each 10 hours is taken as a segment, then p = 4 can be obtained; if each 5 hours is divided into a segment in a more detailed segmentation method, then p = 8. The technical personnel will determine the final number of segments according to the window optimization parameter range obtained above and combined with the actual characteristics of the monitoring object. Once p is determined, it can be used in iterative operations and the statistics of subsequent discriminant indicators.

[0143] T j The acquisition steps of are as follows:

[0144] T jis the test value for the j-th order distribution in the rank-sum test. This test value comes from the process of comparing the multiple distributions of the observed data. Technicians will sort the residual data in each segment and compare it with the rank information of the known distribution, and assign the test statistic one by one. In this process, the historical geological displacement records or the results of repeated sampling in the same area will also be referred to, obtaining a set of T1, T2, …, T q and other test values. These values are usually presented as dimensionless numbers. For the convenience of subsequent calculations, they can be extended into a multinomial sequence that is not directly related to the number of segments but can reflect the dispersion degree of the error distribution. If q = 3 different levels of distribution are adopted in a rank-sum test session, then T1, T2, and T3 can be obtained according to the statistical ranking. The specific values depend on the ranking of the residual in each segment among all the observed samples.

[0145] The steps to obtain q are as follows:

[0146] q is the total number of test distributions. In a complete rank-sum test process, technicians may perform multiple distribution comparisons on the monitoring samples. Each time a comparison is performed, a T related to a specific ranking is obtained j , when k comparison distributions are selected for the test, q = k Ts can be obtained j , if the geological activities in a certain area are relatively complex, three to five dispersion levels may be set, and then q = 3 or q = 5 can be obtained in the test. The specific value of q is determined by the on-site observation scale and accuracy requirements, and is related to the selected statistical strategy and the range of the target geological interval. All Ts j are saved in the same sequence during the test and the corresponding order j is marked.

[0147] Calculation process:

[0148] The first step is to calculate For example, when p = 4, let the segmented regression residual sequence {R1, R2, R3, R4} be {1.2, 0.8, 1.0, 1.5} cm in turn, and the window optimization parameters {Q1, Q2, Q3, Q4} be {8, 6, 8, 4} hours in turn, then Sum up the above values to get 32.36.

[0149] The second step is to divide the above result by p = 4 to get the average value

[0150] The third step is for the test value of the j-th order distribution in the rank-sum test. Let q = 3, {T1, T2, T3} = {2.0, 3.0, 1.5}, and find the sum of squares, that is

[0151] The fourth step is to take the square root of the sum of squares 15.25 to get

[0152] The fifth step is to calculate the difference between the two results and take the absolute value, that is, |8.09-3.905|=4.185, so as to obtain I=4.185.

[0153] The result shows that the difference between the current segmented regression residual square and the average value and rank sum test distribution value reflected by the window optimization parameters is 4.185. When I exceeds 5, it means that there is a greater distribution deviation between the regression residual level and the rank sum test value. When I is between 2 and 5, it means that the degree of deviation is in a medium range. If it is lower than 2, it means that the distribution is relatively close. Technicians can use this to judge the degree of fit between the segmented regression and rank sum test rankings and combine with subsequent steps to further identify the deformation distribution.

[0154] The collaborative discrimination value has been obtained. In order to realize the identification of deformation distribution, it is necessary to read the I value at each moment in the monitoring sequence one by one, and compare the I at all moments with the threshold value formed based on the statistics of the same type of geological environment in the historical monitoring. The technicians will first distinguish the segments with a value lower than 2.0 and mark them as low deviation intervals, the segments between 2.0 and 5.0 are recorded as moderate deviation intervals, and the segments greater than 5.0 are marked as high deviation intervals. These thresholds are derived from the distribution characteristic range obtained by summarizing the surface monitoring results during multiple geological surveys. When I is greater than 5.0 for three consecutive moments, the segment is regarded as a potential significant deformation segment. In subsequent operations, it is necessary to compare the displacement curves corresponding to these moments with the historical case library, and check whether the monitoring instruments deployed on the ground have failed or whether they exceed the specified safety redundancy value. When the moderate deviation interval accumulates for more than 6 hours in a row, it will also be additionally marked and included in the priority investigation scope. Through the above comparison records, it is possible to quickly confirm which time points or deformation amplitudes in the segments deserve special attention, and finally integrate all deviation intervals with timestamps to generate deformation critical indicators.

[0155] The steps to obtain the frequency domain feature quantity are:

[0156] Based on the deformation critical index, the monitoring data sequence at each time scale is extracted, and Fourier transform is performed to convert it into frequency domain signal to generate frequency domain signal distribution;

[0157] According to the frequency domain signal distribution, the power spectrum density distribution value is calculated. The calculation formula is:

[0158]

[0159] Where P is the power spectrum density distribution value, S(f) is the frequency response function of the frequency domain signal, f1 and f2 are the upper and lower limits of the frequency range, and XM k is the amplitude of the kth component in the frequency domain signal, N kis the total number of frequency points of this component, and tm is the total number of components;

[0160] Based on the power spectral density distribution values, a spectral density matrix is established. Through the correlation analysis between matrix elements, characteristic frequencies are extracted and frequency-domain characteristic quantities are generated.

[0161] Specifically, the deformation critical index has been obtained through the previous steps. It is necessary to extract the monitoring data sequences for different time scales and discretely organize these sequences. First, the number of sampling points and the sampling time interval are statistically counted within each time scale, and the moments corresponding to the deformation critical index are marked. By summarizing the displacement records of ground monitoring instruments at multiple measuring points, the possible ranges of different frequency components are distinguished. Subsequently, the segmented time series are read and the Fourier transform is performed for each sequence to extract the amplitude and phase information of the monitoring values in the frequency domain. Since the monitoring frequencies under different locations and geological conditions may be distributed in the range of 0.001 Hz to 10 Hz or higher, it is necessary to first divide several characteristic frequency bands according to the on-site observation history and the established geological activity classification criteria. For example, 0.001 Hz to 1 Hz is regarded as the low-frequency band, 1 Hz to 5 Hz is regarded as the medium-frequency band, and 5 Hz to 10 Hz is regarded as the higher-frequency band. During the analysis process, the monitoring sequences are projected into the corresponding frequency intervals one by one. In order to record the correspondence between the amplitude information and the frequency, it is necessary to sort according to the frequency values after the Fourier transform is completed and list the main peak amplitudes of each frequency component for subsequent identification of whether there are characteristic peaks. When the amplitude exceeds the reference threshold determined in advance according to the monitoring rules over the years (for example, taking the conversion amount with an amplitude higher than 5 cm as a significant fluctuation), it is marked in the frequency-domain result. At the same time, the amplitude intervals adjacent to this threshold are also additionally recorded to finely distinguish the distribution of frequency components at the critical boundary. For multiple sampling sequences of the same time scale, several frequency-domain curves can be obtained through multiple Fourier transforms. By comparing the amplitude differences of each curve at the same frequency, if the amplitude of a curve is continuously greater than the reference threshold in the high-frequency band, it is determined as a suspicious fluctuation segment. If the amplitudes of multiple curves remain similar in the low-frequency band and there are no large-amplitude mutations, the low-frequency band is marked as a relatively stable region. All classification, comparison, and marking information during the processing is recorded in a frequency distribution table. Finally, based on this distribution table, the frequency-domain signal distributions under different time scales can be comprehensively obtained.

[0162] The advantage of the formula is that by integrating the energy of the frequency response function obtained by the Fourier transform and combining the amplitude and the number of frequency points of each frequency-domain component into a square root operation of the sum of squares, it can simultaneously reflect the power accumulation of the frequency-domain signal in the main frequency band and multiple sub-components.

[0163] The steps for obtaining P are as follows:

[0164] P represents the power spectral density distribution value, which is specifically composed of two parts. One part comes from the integral value of |S(f)| 2 in the frequency range from f1 to f2, and the other part comes from taking the square root after accumulating the product terms formed by the amplitude of all components and the total number of their corresponding frequency points. Technicians obtain S(f) by performing a Fourier transform on the monitoring sequence, and divide the frequency range into at least 10 small intervals. For example, from 0.001 Hz to 1 Hz is one interval, and from 1 Hz to 2 Hz is the next interval, etc. Each interval has corresponding f1 and f2 values. Subsequently, when integrating, the frequency domain energy of these intervals is calculated separately. For the amplitude peak part, the amplitude value and its corresponding total number of frequency points {N k} are additionally recorded in the {XM k} sequence. tm represents the total number of components, and each component is composed of an amplitude and a frequency point.

[0165] The steps to obtain S(f) are as follows:

[0166] S(f) is the frequency response function of the frequency domain signal. Technicians first collect a multi-channel displacement data sequence and sample the data at a frequency of once per second for 48 hours, thus obtaining 172,800 original data points. Then, these data are windowed in the time domain. The data within each window undergoes a fast Fourier transform to generate complex frequency domain coefficients. After calculating the real part and the imaginary part of the frequency domain coefficients, the amplitude and phase are formed, and then the corresponding values of S(f) at each frequency point are obtained. If the real part value at f = 2 Hz is identified as 30 and the imaginary part value is 40 in a certain transformation, then its amplitude is This amplitude can be directly used for the subsequent integral calculation of |S(f)| 2 The S(f) values obtained at all frequency points are combined into a complete frequency response curve.

[0167] The steps to obtain f1 and f2 are as follows:

[0168] f1 and f2 are the upper and lower limits of the frequency range, which need to be defined according to the periodicity of geological activities and the sampling frequency in the actual monitoring environment. Technicians will accumulate sufficient data after multiple on-site samplings to estimate the minimum and maximum frequencies that the monitoring system can cover. For example, in the low-frequency geological activity interval, it can start from 0.001 Hz until 10 Hz, and then be adjusted when matching the range of on-site equipment. Record 0.001 Hz as f1 and 10 Hz as f2. When integrating, this interval will be further divided into several segments, and corresponding integral operations can be performed on each segment. If higher frequency components also appear during the monitoring, f2 needs to be re-statistical and adjusted. In some seismic monitoring, it can be up to dozens of Hz or even more than a hundred Hz at the highest.

[0169] XM k The steps to obtain it are as follows:

[0170] XM k The amplitude of the kth component in the frequency domain signal is the local peak or significant energy point extracted from the fast Fourier transform result. The technician will traverse the value of S(f) at each frequency point, find the value whose amplitude exceeds a certain reference threshold (such as 3 cm conversion) or is located in the local peak area, and record these values ​​together with their corresponding frequencies to form {XM1,XM2,…,XM tm In this type of sequence, tm represents the total number of components. If five peaks are identified in a geological monitoring sequence, tm = 5 can be obtained. The high-amplitude component and the medium-amplitude component can be further distinguished based on the peak amplitude.

[0171] N k The steps to obtain are:

[0172] N k Indicates the total number of frequency points corresponding to the kth component. In order to more accurately measure the actual energy around a frequency peak, technicians will classify several sampling points with amplitudes close to the peak as the same component, thereby obtaining the sampling number N of the component. k For example, if there are 5 adjacent sampling points around 2 Hz, all of which have a high amplitude between 40 and 50, then these 5 points are regarded as the same component and recorded as N. k =5, each component corresponds to a peak center and several adjacent points. The technician can determine whether the adjacent points are included in the same component based on the smoothing coefficient, and finally mark these components with serial numbers k=1, 2, ..., tm and record their N k .

[0173] The steps to obtain tm are:

[0174] tm represents the total number of components, which refers to the number of peaks or significant energy segments confirmed in a frequency domain analysis. The technicians browse the Fourier transform results by threshold detection and amplitude continuity detection, and regard each peak or continuous high amplitude point segment as a single component. If the entire S(f) curve has 8 obvious peak segments, then tm = 8, and the subsequent result will be {XM k} and {N k The supporting sequence of}, tm varies depending on the characteristics of the monitoring data. It may be large when the data sampling is dense enough and the geological activities are rich, or it may be small due to low-amplitude changes in stable areas.

[0175] Calculation process:

[0176] The first step is to calculate

[0177] Assume that the frequency range is from 0.001 Hz to 10 Hz. The technician divides it into 20 segments and uniformly approximates it as the average value of |S(f)| within each segment multiplied by the bandwidth. If the bandwidth of the first segment is 0.5 Hz and the average |S(f)| 2 is 1600, then the integral of the first segment is 0.5 × 1600 = 800. By analogy, the sum of all segments is added to obtain the total integral value T 2 。 sum 。

[0178] Step 2: Processing Assume that tm = 3, XM1 = 40, XM2 = 30, XM3 = 50, corresponding to N1 = 5, N2 = 3, N3 = 4. Then:

[0179]

[0180] Step 3: Take the square root of 20700 to get

[0181] Step 4: Add T sum to 143.88 to form P. If T sum = 2000, then:

[0182] P = 2000 + 143.88 = 2143.88

[0183] This result indicates that 2143.88 can be regarded as the sum of the frequency-domain energy within the integration range of 0.001 Hz to 10 Hz and the contribution generated by multiple significant peaks. When the value of P is greater than 3000, it often means that there are more high-energy components and the peak amplitudes are significant. When the value of P is around 1500, it indicates that the overall energy is at a medium level. The technician will file different P values and track the energy level differences at certain specific time periods in combination with geological monitoring records.

[0184] After obtaining the power spectral density distribution values, it is necessary to arrange the energy values within the corresponding frequency intervals in matrix form. First, the monitoring data is split into several time periods. Within each time period, the frequency distribution and amplitude peak information are calculated in the same way. Then, for each time period, the corresponding energy levels are marked at multiple frequency points, and the frequency-energy correspondence relationships of different time periods are concentrated to form a spectral density matrix. Next, the arrangement values of each frequency point in different time periods in the matrix are read. When the amplitude continuously exceeds the threshold determined in advance according to historical records (for example, the conversion amount corresponding to an amplitude higher than 6 cm), it is recorded as an object of special attention, and the distribution of these matrix elements in the rows and columns and whether there are repeated peaks are counted. By comparing the correlation degrees between matrix elements, it is judged whether the characteristic frequencies in several time periods are consistent or whether new frequency components appear. If it is found that certain characteristic frequencies have relatively obvious amplitude distributions in multiple time periods, the frequency values are extracted and compared with the monitoring information of different measuring points. All the extracted characteristic frequencies are recorded as key frequency domain information. Finally, after comparing with the existing geological movement patterns, the extraction results are recorded as frequency domain characteristic quantities.

[0185] The steps for obtaining the early warning characteristic sequence are as follows:

[0186] Based on the frequency domain characteristic quantities, the residual values of each frequency domain data segment are calculated in segments, and the residual values are accumulated in sequence to form a recursive residual spectrum, generating a recursive residual spectrum;

[0187] According to the recursive residual spectrum, the residual spectrum matrix is decomposed, the eigenvalues of the matrix are extracted one by one, and all the eigenvalues are sorted in order and merged into a set to obtain the matrix eigenvalue set;

[0188] Based on the matrix eigenvalue set, all the residual values in the recursive residual spectrum and the extracted matrix eigenvalues are accumulated, and the accumulated results are rearranged according to the time series to establish an early warning characteristic sequence.

[0189] Specifically, based on the frequency domain characteristic quantities, the frequency domain data segments within each time period are split according to a fixed time interval and frequency range, and then the residual values of each data segment are statistically recorded one by one, so as to find the difference between each data segment and the established monitoring reference value. Then, the residual sequence within this data segment is obtained through point-by-point comparison. In order to correlate the residual sequences between different data segments under the same coordinate, the start and end times of the data segment are first compared with the applicable range of the reference value. For example, the time from 0 to 24 hours is divided into multiple intervals, and several sampling points are selected in each interval. These sampling points are compared item by item with the pre-set reasonable displacement change interval. If a single record exceeds the empirical threshold, it is marked as a high-discrete value in the residual sequence. Then, these residual values are accumulated in sequence and the time point information where large discrete values occur is retained. If three consecutive records exceed the empirical threshold (for example, in a certain formation activity, the reference value is agreed that a displacement exceeding 2 cm is considered a large discrete), a segment mark is added in the subsequent processing link. This threshold is generally obtained by comprehensively calculating the mean and variance of the displacement changes obtained from multiple monitoring in the same area, forming an empirical range that conforms to the geological characteristics. When the residual values of all segments are recorded, these differences are arranged in sequence in the same sequence and accumulated within the same segment, and the accumulated curve is mapped to the corresponding time series, so as to construct a complete recursive residual spectrum.

[0190] According to the generated recursive residual spectrum, in its matrix representation, the time axis is used as the row index and the identifiers of different frequency domain data segments are used as the column index to form a two-dimensional residual spectrum matrix. The residual value corresponding to each matrix element is read, and then the residual spectrum matrix is disassembled by means of numerical decomposition. Technicians will check each row and column of the matrix one by one, and aggregate the parts with higher similarity in the row vector or column vector into sub-matrices for subsequent extraction of their respective eigenvalues. In order to determine the decomposition dimension, a minimum sub-matrix size is set according to past monitoring experience (for example, at least containing 10 time segments or 10 frequency segments). After statistically organizing the values of each sub-matrix, the key position values are obtained respectively through linear decomposition or matrix segmentation methods, and then the corresponding eigenvalues are calculated and all eigenvalues are arranged in ascending or descending order. Then, the distribution of these values within each sub-matrix is statistically analyzed, the similar eigenvalues are merged into one category and the row and column ranges where they are located are recorded. If an eigenvalue continuously appears in multiple sub-matrices, it is included in the merging set. These merged results will correspond to the corresponding time periods or frequency segments, thus forming the final matrix eigenvalue set.

[0191] Based on the obtained set of matrix eigenvalues, find the residual points in the recursive residual spectrum that fall within the same range or have an associated relationship with these eigenvalues, and add the two numerically to obtain new sequence elements. To avoid the addition of interference terms, it is necessary to first check the label information of each residual value in the time series and reconfirm the timestamps of the paragraphs or sampling points where it is located. For example, check the start time order of different paragraphs to ensure that the final accumulated result can be accurately connected. In this process, the eigenvalue interval division obtained earlier will be used. For example, if a certain eigenvalue region is mainly concentrated from the 2nd hour to the 3rd hour and the amplitude distribution is in a relatively high range, it is necessary to focus on weighting or directly accumulating all the residual values within this range. If the residual value is below the preset threshold in some time periods, the accumulation operation is still completed but marked as a low amplitude segment. After all the accumulation processes are completed, the new accumulated sequence will be rearranged in chronological order, and the moments with marked high amplitude or high eigenvalue superposition parts will be listed in advance to form an index. Finally, the warning feature sequence will be sorted out and the accumulated results of each time point will be collected as the basic data for subsequent interpretation.

[0192] The steps for obtaining the scale warning identifier are as follows:

[0193] Based on the warning feature sequence, perform data segmentation processing on the feature sequences at each time scale one by one, arrange them in time series, compare the eigenvalue of each segment with the corresponding threshold, extract all the feature points that exceed the threshold, and generate a record of feature points exceeding the threshold;

[0194] According to the record of feature points exceeding the threshold, mark the time points corresponding to the feature points, call the recorded data within all time scale ranges, and mark all the time points exceeding the threshold in segments in the order of time points to obtain the identifier of time points exceeding the threshold;

[0195] Based on the identifier of time points exceeding the threshold, perform a hierarchical analysis on the distribution of time point identifiers at all time scales, merge all the identifiers according to the established identifier classification rules, and integrate the merged results according to the time scale to generate the scale warning identifier.

[0196] Specifically, based on the early warning feature sequence, each feature sequence at each time scale is disassembled into several segments item by item. First, the segment boundaries are determined by combining the previously obtained monitoring duration and sampling frequency, and the feature values are retrieved one by one within each segment. These feature values are compared item by item with the thresholds obtained by referring to the geological activity records of previous years. The thresholds may be set to specific values such as the surface displacement exceeding 2 cm or the displacement growth rate being greater than 1 cm / hour. These values are statistically obtained by technicians through continuous observations in the same area. When any feature value within a segment exceeds the threshold, it is marked as an abnormal point, and the corresponding time index and the magnitude of the feature value are recorded. If multiple adjacent sampling points within a segment exceed the threshold, these points are grouped together and a high-priority mark is appended. To avoid missing instantaneous peaks or small-range mutations, an interactive check is also performed on the connection interval between the previous and the next segments after each segment is processed to confirm whether there are abnormal values across segments exceeding the threshold. Then, all this abnormal feature value information is merged and summarized, and finally a list of exceeded feature points is formed, which is called the over-threshold feature record.

[0197] According to the over-threshold feature record, the exact moments corresponding to the feature points at different time scales are located one by one. For example, when mapping the minute-level records to the hour-level or day-level, it is necessary to first read the previously generated multi-scale time series data and match the occurrence periods of the feature points according to the intervals of each time scale. Subsequently, these marked moments are sequentially found in the recorded data within all time scale ranges. To keep the time order consistent, the marks can be inserted segment by segment according to the absolute time sequence of the feature points. When setting a minimum interval (such as every 2 hours as a segment) for segmentation, the sampling frequency and the geological area activity rate can be referred to. If there are multiple feature points from different time scales within the same time slice, they are listed in the mark list of this time slice. When the feature value corresponding to the feature point exceeds the critical setting (such as a displacement increment greater than 2 cm) and appears continuously, a high-frequency early warning label can be additionally noted. After all the segmentation marking operations are completed, these marked time slices and feature point information are integrated in order, and finally the over-threshold time point identification for subsequent analysis is obtained.

[0198] Based on the obtained over-threshold time point identification, the distribution status of each time slice in all time scales is classified and summarized, and the classification is determined by setting different categories of identification thresholds. For example, in statistical historical activities, if the displacement amplitude exceeds 5 cm, it is an extremely high level, and the range of 2 cm to 5 cm is identified as a medium level. Then, when merging, the situation where records of over-2 cm appear in multiple time scales and are continuously distributed for more than 4 hours can be marked as a medium-level warning, and the continuous displacement exceeding 5 cm can be recorded as an extremely high-level warning. These values ​​are also determined by many years of observation data and actual on-site assessments. Then, the cumulative number and distribution range of identifications are checked in each time scale. Identifications with a high degree of temporal overlap will be judged as concentrated and written into high-priority categories. At the same time, identifications distributed on different time scales but pointing to the same time point will be merged. If over-threshold conditions appear in three consecutive scales in the same time window, a merged record will be made. Finally, all merged identifications are arranged in sequence according to the time scale to obtain the final scale warning identification.

[0199] The steps to obtain multi-scale warning indicators are:

[0200] Based on the scale warning signs, the sign distribution is extracted from the data of each time scale, and the number and distribution characteristics of each sign are counted one by one according to the time scale sequence. The time span and distribution characteristics of the signs are combined to generate the statistical results of the time scale sign distribution.

[0201] According to the statistical results of the time scale identification distribution, the preset classification rules are called one by one, the identification distribution of each time scale is matched, each group of identification distribution is compared with the classification rules one by one, and the classification matching results are generated according to the classification standards;

[0202] Based on the hierarchical matching results, the matching categories on all time scales are integrated, the hierarchical matching categories are analyzed and merged in chronological order, and the merged results are uniformly output as multi-scale early warning indicators.

[0203] Specifically, based on the scale warning signs, the distribution of all signs is split from the data of each time scale. By reading the previously obtained time period division and sign information, the occurrence time of the signs is corresponding processed with the corresponding time scale. Technicians will first select record entries in different time periods such as hourly or daily levels and arrange them in the same interval manner, and then count the number of signs appearing within each interval. If a sign appears continuously more than 3 times within a certain time period and the interval is less than 2 hours, it is regarded as a high-density sign distribution and is marked separately in the subsequent steps. These settings are all from the previous monitoring experience in the same area and the analysis results of on-site observation data. For example, if obvious geological activity signs often appear when the hourly sign density in this area exceeds 3 in the past two years, it is set as the reference threshold. On this basis, the distribution characteristics are summarized into a time period and sign statistical table one by one, and the number of signs, intervals, and overlapping sign situations in adjacent time periods counted for each time period are recorded accordingly. If it is found that the signs under a single time scale are mostly concentrated in a specific time period, that time period is additionally marked in the statistical table as a potential high-distribution area. If the number of signs in some time periods under the same time scale is less than 1, it is marked as a low-distribution area. After completing the individual statistics one by one, the sign number distributions corresponding to all time scales are summarized and the time scale sign distribution statistical results are output.

[0204] According to the time scale sign distribution statistical results, the sign distribution obtained for each time scale is matched with the existing grading rules. The setting of the grading rules usually refers to multiple previous monitoring cases and is based on the displacement amplitude or frequency threshold. For example, it is stipulated that an hourly distribution higher than 3 signs per hour is recorded as a high level, a medium distribution is 1 to 3 signs per hour, and a low distribution is no more than 1 sign per hour. In the analysis, the number of signs in the statistical table will be compared item by item with the above division thresholds. When the distribution quantity falls on the interval boundary, it is necessary to supplement the judgment of the sign trend in the previous and subsequent time periods. For example, if the hourly sign has reached 3 in the previous time period and reaches 3 again in the current time period, it is directly regarded as a high level and the continuous situation of high-level time periods is recorded. Taking this as a reference, the distribution records are graded and matched. If the records of a certain time scale are all at a high level in 3 adjacent time periods, the cumulative high-level signs are displayed in the list. When the grading is completed, the corresponding grading and matching results are generated for each time scale.

[0205] Based on the hierarchical matching results, combine the matching categories at all time scales in chronological order. First, arrange the hour-level and day-level identifiers on the same time axis and compare the time periods when high-level or medium-level occurrences appear. If there are consecutive high-level occurrences at the hour level and corresponding high-distribution records also appear at the day level during the corresponding time period, merge the two into a high-level category and uniformly label the time period to which they belong. If the day-level identifier shows high distribution within multiple consecutive day segments while the hour level only reaches the high level in a few time periods, only record those time points that simultaneously meet the high level as the highest category identifier during the merging process. For the remaining time periods that do not overlap or have lower levels, classify them according to the results of their respective scales as lower category identifiers. Finally, output this multi-level merging result as a multi-scale early warning indicator and present it in a chronological list.

[0206] The present invention provides a geological disaster early warning system, including:

[0207] A data acquisition and preliminary processing module, which collects geological displacement monitoring data, sets a time window range, calculates the local extreme points and local zero points of the monitoring data within the time window, and generates multi-scale reference values;

[0208] A data quality assessment module, which calculates comprehensive statistical quantities based on the multi-scale reference values to obtain data quality assessment parameters;

[0209] A parameter optimization and regression analysis module, which uses the data quality assessment parameters to iteratively calculate the objective function value, obtains window optimization parameters, substitutes the window optimization parameters into a piecewise linear regression model, performs collaborative discrimination and threshold comparison, and generates deformation critical indicators;

[0210] A frequency domain feature analysis module, which performs periodic analysis on the monitoring data sequence according to the deformation critical indicators, including the calculation of power spectral density distribution and spectral density matrix, conducts recursive residual spectrum analysis and matrix decomposition, accumulates the residual spectrum and matrix eigenvalues, and establishes an early warning feature sequence;

[0211] An early warning signal generation and output module, which compares the feature sequences at each time scale with the set thresholds according to the early warning feature sequence, marks the time points that exceed the thresholds, generates scale early warning identifiers, statistically analyzes the distribution of the scale early warning identifiers at each time scale, performs matching operations with the preset classification rules, and outputs a multi-scale early warning indicator.

[0212] The above are only the preferred embodiments of the present invention, and do not limit the present invention in other forms. Any person skilled in the relevant art may use the technical content disclosed above to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A geological disaster early warning method based on dynamic data monitoring, characterized in that: The following steps are involved: Acquire geological displacement monitoring data, set a time window range, calculate local extreme points and local zero points of the monitoring data in the time window respectively, and obtain multi-scale benchmark values; calculate comprehensive statistics based on the multi-scale benchmark values ​​to generate data quality assessment parameters; Based on the data quality assessment parameters, the objective function value under each set of iteration parameters is calculated to obtain the window optimization parameters; Substituting the window optimization parameters into the piecewise linear regression, performing collaborative discrimination and threshold comparison on the regression operation results, and generating a deformation critical index; Based on the deformation critical index, a periodic analysis is performed on the monitoring data sequence at each time scale, the power spectrum density distribution and the spectrum density matrix in the frequency domain are calculated, and the frequency domain feature quantity is obtained. Recursive residual spectrum analysis and matrix decomposition operations are performed on the frequency domain feature quantity, and the residual spectrum and the matrix eigenvalue are accumulated to establish an early warning feature sequence; Based on the warning feature sequence, the feature sequence on each time scale is compared with the corresponding threshold, the time point exceeding the threshold is marked, and the scale warning mark is obtained. The distribution of the scale warning mark on each time scale is counted, the distribution is matched with the preset classification rules, and a multi-scale warning indication is output.

2. The geological disaster early warning method based on dynamic data monitoring according to claim 1 is characterized in that: The steps for obtaining the multi-scale reference value are as follows: Obtain geological displacement monitoring data, set time window range parameters and time window moving step parameters, analyze the change characteristics of monitoring data at different scales within the time window, and generate multi-scale decomposition results; Based on the multi-scale decomposition result, analyzing the local extreme points and local zero points in the monitoring data to obtain local feature points; Based on the local feature points, the relative relationship and change trend between the feature points are calculated to form a multi-scale reference value.

3. The geological disaster early warning method based on dynamic data monitoring according to claim 1 is characterized in that: The steps for obtaining the data quality assessment parameters are as follows: Based on the multi-scale reference value, performing a differential calculation to obtain a differential result between a local extreme point and a local zero point to form a local difference value; According to the local difference value, the comprehensive statistics are calculated, and the calculation formula is: Where Δx i is the i-th item in the local extreme point difference result, k is the number of local extreme points, Δy j is the jth item in the local zero difference result, m is the number of local zeros, and S is the comprehensive statistic; Based on the comprehensive statistics and in combination with the characteristic distribution of the comprehensive statistics, a data quality assessment parameter is obtained.

4. The geological disaster early warning method based on dynamic data monitoring according to claim 1 is characterized in that: The steps for obtaining the window optimization parameters are as follows: Based on the data quality assessment parameters, a time window range parameter and a time window moving step parameter are initialized, a recursive iterative process is constructed, and the value ranges of the time window range and the time window step parameter are gradually adjusted to form an iterative parameter set; According to the set of iteration parameters, the objective function value of each set of iteration parameters is calculated, and the calculation formula is: Among them, T i represents the square value of the time window range parameter in the i-th iteration, R i represents the reference value of the time window range, n represents the total number of calculations of the time window range parameter, S j represents the square value of the time window step parameter in the jth iteration, M j represents the target reference value of the time window step, xm represents the total number of calculations of the time window step parameter, and F represents the objective function value under the current iteration parameter set; Based on the objective function value, the time window range parameter and the time window step parameter after recursive screening are analyzed, all screened parameters are optimized and combined to generate window optimization parameters.

5. The geological disaster early warning method based on dynamic data monitoring according to claim 1 is characterized in that: The steps for obtaining the critical deformation index are as follows: Based on the window optimization parameters, respectively calculating the regression residual and fitting coefficient in each segment to generate segmented regression analysis results; According to the segmented regression analysis results, the collaborative discriminant value is calculated, and the calculation formula is: Among them, R i is the fitting residual of the i-th segment, Q i is the corresponding window optimization parameter, p is the total number of segments, T j is the test value of the j-th order distribution in the rank sum test, q is the total number of test distributions, and I is the co-discriminant value; Based on the collaborative discrimination value, the deformation distribution is identified by comparing it with a set threshold value, and a deformation critical index is generated.

6. The geological disaster early warning method based on dynamic data monitoring according to claim 1 is characterized in that: The steps for obtaining the frequency domain feature quantity are as follows: Based on the deformation critical index, extract the monitoring data sequence on each time scale, perform Fourier transform to convert it into frequency domain signal, and generate frequency domain signal distribution; According to the frequency domain signal distribution, the power spectrum density distribution value is calculated, and the calculation formula is: Where P is the power spectrum density distribution value, S(f) is the frequency response function of the frequency domain signal, f1 and f2 are the upper and lower limits of the frequency range, and XM k is the amplitude of the kth component in the frequency domain signal, N k is the total number of frequency points of the component, and tm is the total number of components; Based on the power spectrum density distribution value, a spectrum density matrix is ​​established, and through correlation analysis between matrix elements, characteristic frequencies are extracted and frequency domain feature quantities are generated.

7. The geological disaster early warning method based on dynamic data monitoring according to claim 1 is characterized in that: The steps of obtaining the early warning feature sequence are: Based on the frequency domain feature quantity, the residual value of each frequency domain data segment is calculated segment by segment, and the residual values ​​are accumulated in sequence to form a recursive residual spectrum, thereby generating a recursive residual spectrum; According to the recursive residual spectrum, the residual spectrum matrix is ​​decomposed, the eigenvalues ​​of the matrix are extracted one by one, all the eigenvalues ​​are arranged in order and set merged to obtain a matrix eigenvalue set; Based on the matrix eigenvalue set, all residual values ​​in the recursive residual spectrum and the extracted matrix eigenvalues ​​are accumulated, and the accumulated results are rearranged according to the time series to establish an early warning feature sequence.

8. The geological disaster early warning method based on dynamic data monitoring according to claim 1 is characterized in that: The steps for obtaining the scale warning mark are: Based on the warning feature sequence, the feature sequence on each time scale is segmented one by one, arranged in time series and the feature value of each segment is compared with the corresponding threshold value, all feature points exceeding the threshold value are extracted, and an over-threshold feature record is generated; According to the super-threshold feature record, the time point corresponding to the feature point is marked, the record data within all time scales is called, and all super-threshold time points are segmented and marked according to the sequence of the time points to obtain the super-threshold time point identifier; Based on the above-threshold time point identification, the distribution of time point identifications on all time scales is analyzed in a hierarchical manner, all identifications are merged according to the established identification classification rules, and the merged results are integrated according to the time scale to generate a scale warning identification.

9. The geological disaster early warning method based on dynamic data monitoring according to claim 1 is characterized in that: The steps of obtaining the multi-scale early warning indication are: Based on the scale warning signs, extract the sign distribution from the data of each time scale, count the number and distribution characteristics of each sign one by one according to the time scale sequence, and generate the time scale sign distribution statistics result by combining the time span and distribution characteristics of the sign; According to the statistical results of the time scale identification distribution, the preset classification rules are called one by one, the identification distribution of each time scale is matched, each group of identification distribution is compared with the classification rules one by one, and a classification matching result is generated according to the classification standard; Based on the hierarchical matching results, the matching categories on all time scales are integrated, the hierarchical matching categories are analyzed and merged in chronological order, and the merged results are uniformly output as multi-scale warning indications.

10. The geological disaster early warning system according to any one of claims 1 to 9, characterized in that: include: The data acquisition and preliminary processing module collects geological displacement monitoring data, sets the time window range, calculates the local extreme points and local zero points of the monitoring data within the time window, and generates multi-scale benchmark values; The data quality assessment module calculates comprehensive statistics based on multi-scale benchmark values ​​to obtain data quality assessment parameters; The parameter optimization and regression analysis module uses data quality assessment parameters to iteratively calculate the objective function value, obtain the window optimization parameters, substitute the window optimization parameters into the piecewise linear regression model, perform collaborative discrimination and threshold comparison, and generate deformation critical indicators; The frequency domain feature analysis module performs periodic analysis on the monitoring data sequence according to the deformation critical index, including the calculation of power spectrum density distribution and spectrum density matrix, recursive residual spectrum analysis and matrix decomposition, accumulation of residual spectrum and matrix eigenvalues, and establishment of early warning feature sequence; The warning signal generation and output module compares the feature sequence on each time scale with the set threshold based on the warning feature sequence, marks the time point that exceeds the threshold, generates a scale warning mark, counts the distribution of the scale warning mark on each time scale, performs matching operations with the preset classification rules, and outputs multi-scale warning indications.

Images