High slope stability monitoring and early warning system

Abstract

The embodiment of the invention discloses a high slope stability monitoring and early warning system. The system comprises a data fusion module, data acquisition is carried out through monitoring equipment such as a measurement robot, a GNSS, a tension wire type displacement meter, an inclination measurement hole, a micro-core pile, a water level gauge and a rain gauge, and normalization processing is carried out on the acquired data to generate preliminary slope monitoring data. According to the method and the device, the accuracy and the availability of data analysis are improved by integrating the multi-source data and applying normalization processing. Normalization processing ensures that data from different devices can be compared under the same standard, the data processing efficiency is enhanced, meanwhile, introduction of time sequence analysis allows recognition of normal and abnormal modes through historical data, the prediction capacity of an early warning system is enhanced, potential risks can be recognized earlier, and the timeliness and accuracy of early warning are improved. A dynamic algorithm evaluates the risk level in real time and compares the risk level with a safety threshold, so that risk management is more refined.

Description

Technical Field

[0001] One or more embodiments of the present specification relate to the technical field of safety monitoring and early warning, and in particular, to a high slope stability monitoring and early warning system. Background Art

[0002] Slope is a common engineering form in geotechnical engineering construction. High slope is the main engineering carrier in engineering construction. High slope is often located in areas with complex geological conditions. Its stability plays a vital role in ensuring the safety of engineering construction. Therefore, it is particularly important to monitor the stability of high slopes. By monitoring high slopes and evaluating feedback warning information, it is beneficial to ensure the long-term stability of high slopes, which is of great significance to ensuring engineering safety and personal and property safety.

[0003] Existing technologies usually rely on a single data source or simple monitoring technology, which often makes it difficult to provide comprehensive and accurate monitoring results under variable and complex geological conditions. Failure to identify data patterns in real time may lead to untimely or inaccurate responses to potential risks, increasing the risk of engineering safety accidents. Therefore, improvements are needed. Summary of the invention

[0004] The embodiment of this specification provides a high slope stability monitoring and early warning system, and its technical solution is as follows: In a first aspect, an embodiment of this specification provides a high slope stability monitoring and early warning system, the system comprising: A data fusion module is used to collect data through a measuring robot, GNSS, a tension-wire displacement meter, an inclinometer hole, a micro-core pile, a water level meter and a rain gauge, normalize the collected data to generate preliminary slope monitoring data, and integrate multi-source data and reduce noise on the preliminary slope monitoring data to obtain comprehensive monitoring data; A pattern recognition module is used to perform time series analysis on the comprehensive monitoring data, identify normal data patterns and abnormal patterns, generate pattern analysis results, and analyze the risk level of potential landslides and collapses based on the pattern analysis results to obtain early warning signal candidates; A risk assessment module is used to perform stability analysis on the current monitoring data and the historical monitoring data according to the early warning signal candidate, obtain the slope risk probability value corresponding to the current monitoring data, and compare the slope risk probability value with the set safety threshold, adjust the early warning level according to the comparison result, and obtain the slope stability assessment result; The automatic alarm module is used to receive the slope stability assessment result. When the slope stability assessment result exceeds the alarm threshold, emergency communication is triggered, warning information is sent to related personnel and institutions, and an alarm execution record is generated. The warning information includes the slope location and risk level.

[0005] As an option, the steps for obtaining the preliminary slope monitoring data are: The original monitoring data set was obtained by collecting surface deformation, deep underground deformation, tilt, vibration, groundwater level and rainfall data from high slopes through measurement robots, GNSS, tension-wire displacement meters, inclinometer holes, micro-core piles, water level gauges and rain gauges; Performing standardization processing on each original monitoring data in the original monitoring data set to generate a normalized monitoring data set; Each normalized monitoring data in the normalized monitoring data set is analyzed, and information from different monitoring devices is integrated through a data fusion method to obtain preliminary slope monitoring data.

[0006] As an option, the steps for obtaining the comprehensive monitoring data are: Based on the preliminary slope monitoring data, the data of the measuring robot, GNSS, tension-wire displacement meter, inclinometer hole, micro-core pile, water level meter and rain gauge are integrated by stepwise weighted average to form a data fusion intermediate result; According to the intermediate result of data fusion, the Kalman gain is calculated and noise reduction is performed to obtain denoised data, and the formula is: ; in, is the estimated state at time t, is the observed data, is the observation model, is the Kalman gain, is the noise adjustment factor; The denoised data are integrated to obtain comprehensive monitoring data.

[0007] As an option, the steps for obtaining the pattern analysis result are: Receive comprehensive monitoring data, use time recurrent neural network to perform time series analysis, process data at consecutive time points through the network recursive structure, and form time series analysis results; Classifying the time series analysis results, and identifying normal and abnormal patterns of data points in the sequence to obtain classified pattern recognition results; According to the pattern recognition results after the classification, the patterns of all data points are integrated and analyzed to obtain pattern analysis results.

[0008] As an option, the step of obtaining the early warning signal candidate is: Receiving the pattern analysis result, analyzing normal pattern data and abnormal pattern data, determining an initial risk classification of the comprehensive monitoring data, and obtaining a preliminary classification result; Based on the preliminary classification results, the potential landslide and collapse risk levels are calculated to obtain the risk score, which is calculated as follows: ; in, represents the risk score of data point i, is the monitoring value of data point i under condition j, is the weight coefficient, is the average value of the monitoring value, is the abnormal probability, is the adjustment coefficient, is the number of data points; The risk scores of all points are comprehensively evaluated to form early warning signal candidates.

[0009] As an option, the step of obtaining the risk probability value is: Based on the early warning signal candidate, current monitoring data and historical monitoring data are subjected to data fusion processing to obtain a comprehensive monitoring data set; Based on the comprehensive monitoring data set, the current risk probability value of the slope is calculated using the following formula: ; in, represents the risk probability value of the slope, represents the warning signal strength of the ith data point, represents the deviation between the candidate warning signal and the reference warning signal, Represents the total number of data points.

[0010] As an option, the steps for obtaining the slope stability assessment result are: The warning level is calculated based on the slope risk probability value and the set safety threshold, and the calculation formula is: ; in, Indicates the warning level. is the risk probability value, θ is the safety threshold; A slope stability assessment is performed according to the warning level to obtain a slope stability assessment result.

[0011] As an option, the steps for obtaining the alarm execution record are: receiving the slope stability assessment result, and triggering an emergency communication signal when the slope stability assessment result exceeds a set alarm threshold; Based on the emergency communication signal, an emergency response is activated, warning information is sent to all related personnel and institutions, and an alarm execution record is generated, wherein the warning information includes the slope location and risk level.

[0012] The beneficial effects brought by the technical solutions provided by some embodiments of this specification include at least: In one or more embodiments of the present specification, the accuracy and availability of data analysis are improved by integrating multi-source data and applying normalization. Normalization ensures that data from different devices can be compared under the same standard, enhancing data processing efficiency. At the same time, the introduction of time series analysis allows the identification of normal and abnormal patterns through historical data, enhances the predictive ability of the early warning system, can identify potential risks earlier, and improve the timeliness and accuracy of early warning. The dynamic algorithm evaluates the risk level in real time and compares it with the safety threshold to make risk management more refined. This adaptive adjustment mechanism improves response speed and flexibility. The design of the automatic alarm module ensures that once the risk level exceeds the standard, relevant personnel and institutions can quickly receive early warning information, enhance emergency response capabilities, and have high practicality. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0014] Figure 1 A schematic diagram of the principle flow of a high slope stability monitoring and early warning system provided in an embodiment of this specification. DETAILED DESCRIPTION

[0015] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0016] The terms "first", "second", "third", etc. in the description and claims of this specification and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices.

[0017] The following description provides examples and does not limit the scope, applicability or examples set forth in the claims. Changes may be made to the functions and arrangements of the elements described without departing from the scope of the present specification. Various processes or components may be appropriately omitted, substituted or added to each example. For example, the described method may be performed in an order different from the order described, and various steps may be added, omitted or combined. In addition, features described with respect to some examples may be combined in other examples.

[0018] See also Figure 1 , Figure 1 The invention shows the principle of a high slope stability monitoring and early warning system provided by an embodiment of this specification, including: A data fusion module is used to collect data through a measuring robot, GNSS, a tension-wire displacement meter, an inclinometer hole, a micro-core pile, a water level meter and a rain gauge, normalize the collected data to generate preliminary slope monitoring data, and integrate multi-source data and reduce noise on the preliminary slope monitoring data to obtain comprehensive monitoring data; A pattern recognition module is used to perform time series analysis on the comprehensive monitoring data, identify normal data patterns and abnormal patterns, generate pattern analysis results, and analyze the risk level of potential landslides and collapses based on the pattern analysis results to obtain early warning signal candidates; A risk assessment module is used to perform stability analysis on the current monitoring data and the historical monitoring data according to the early warning signal candidate, obtain the slope risk probability value corresponding to the current monitoring data, and compare the slope risk probability value with the set safety threshold, adjust the early warning level according to the comparison result, and obtain the slope stability assessment result; The automatic alarm module is used to receive the slope stability assessment result. When the slope stability assessment result exceeds the alarm threshold, emergency communication is triggered, warning information is sent to related personnel and institutions, and an alarm execution record is generated. The warning information includes the slope location and risk level.

[0019] The steps for obtaining preliminary slope monitoring data are as follows: The original monitoring data set was obtained by collecting surface deformation, deep underground deformation, tilt, vibration, groundwater level and rainfall data from high slopes through measurement robots, GNSS, tension-wire displacement meters, inclinometer holes, micro-core piles, water level gauges and rain gauges; Performing standardization processing on each original monitoring data in the original monitoring data set to generate a normalized monitoring data set; Each normalized monitoring data in the normalized monitoring data set is analyzed, and information from different monitoring devices is integrated through a data fusion method to obtain preliminary slope monitoring data.

[0020] Specifically, based on the monitoring equipment such as surveying robots, Global Navigation Satellite System (GNSS), tension-wire displacement meters, inclinometers, micro-core piles, water level meters and rain gauges deployed on the high slopes, the specific installation position of each monitoring equipment is first determined and the unified coordinates are marked. Continuous and periodic observation operations are performed on each device, and the multi-source data output by the monitoring equipment are preliminarily collected and recorded in time series. After confirming that the monitoring equipment is working properly through on-site inspection, the obtained data is quickly compared for consistency to eliminate abnormal information that may be caused by failure of the monitoring equipment, and finally a complete original monitoring data set is obtained.

[0021] For the original monitoring data set obtained earlier, first read each data content and extract its central value and standard deviation, and then use the in-line formula to calculate its central value Standardization is performed, where Represents the original data, represents the mean of the batch data. Represents the standard deviation of the batch of data. Through this process, the data of different dimensions and numerical ranges can be kept in a relatively uniform scale range, and then all the standardized results are summarized to obtain the corresponding normalized monitoring data set.

[0022] After obtaining the normalized monitoring data set, the key timestamps of each monitoring device channel are synchronized and aligned, and the monitoring values ​​at the same time point are mapped to a unified index structure. Then, the values ​​of different monitoring devices are integrated according to the pre-set weighting strategy. For example, a higher weight can be set for the historical monitoring data of the measuring robot, and a relatively low weight can be set for the monitoring data of other monitoring instruments. By superimposing the weighted values ​​and calculating the comprehensive results of each time slice, a unified monitoring value sequence is output at the same time. To ensure that the fusion result does not produce an absolute proportion of a single data channel, the total value of all weighted coefficients needs to be set to a fixed value greater than 1. This value is usually given based on experience or previous calibration experiments, and then repeated numerical smoothing and outlier checking are performed to exclude unexpected monitoring equipment interference or obviously abnormal data information. Finally, all fused sequences are summarized and annotated in a unified index table to obtain the final preliminary slope monitoring data.

[0023] The steps for obtaining comprehensive monitoring data are as follows: Based on the preliminary slope monitoring data, the data of the measuring robot, GNSS, tension-wire displacement meter, inclinometer hole, micro-core pile, water level meter and rain gauge are integrated by stepwise weighted average to form a data fusion intermediate result; According to the intermediate result of data fusion, the Kalman gain is calculated and noise reduction is performed to obtain denoised data, and the formula is: ; in, is the estimated state at time t, is the observed data, is the observation model, is the Kalman gain, is the noise adjustment factor; The denoised data are integrated to obtain comprehensive monitoring data.

[0024] Specifically, based on the preliminary slope monitoring data obtained earlier, we first retrieve the observation items within the same recording time period from the surveying robot and GNSS, and at the same time, match the reading values ​​of monitoring equipment such as tension-line displacement meters, inclinometers, micro-core piles, water level gauges and rain gauges point by point. These monitoring data are processed accordingly according to a unified time index, and the corresponding monitoring values ​​are aligned at each time node and listed uniformly. Then, the data are grouped and marked in combination with the equipment installation area division information obtained earlier, and the continuity of instantaneous events captured by the surveying robot and GNSS is included in a weight reference respectively. Separate weight bases are also set for the monitoring range of monitoring instruments such as tension-line displacement meters, inclinometers, micro-core piles, water level gauges and rain gauges and their performance in slope deformation and seepage estimation. The specific values ​​of these weight parameters It can be obtained through the preliminary on-site calibration and historical observation comparison process, and then according to the pre-set weighting strategy, the data output by different monitoring devices at the same time node are superimposed and corrected in turn, among which the monitoring devices with longer observation time and lower failure rate will be given higher coefficients. If it is detected that a certain type of monitoring equipment has a range drift problem in an individual period, the relevant values ​​will be dynamically adjusted downward during the weighting process of that period. The accumulation and correction of weights will continue to follow up the progress of batch processing of data. If it is found that there is a mutation between the monitoring values ​​that exceeds the on-site experience threshold, it is necessary to additionally check the monitoring data of the corresponding period. The on-site experience threshold is formulated by the monitoring personnel based on the slope deformation amplitude observed in previous times and the corresponding mutation range of multi-source data. After completing the weighted processing, all processed data are summarized to obtain the intermediate result of data fusion.

[0025] The formula is useful because it can be used to analyze the observed data. The normalized correction is used to update the estimated state , and using the Kalman gain Dynamically correct the noise factor to achieve real-time adjustment of the slope status; The acquisition step is to initialize and estimate the state quantity at each moment according to the fusion intermediate result obtained previously. The acquisition step is to read the comprehensive value of multi-source observation data at the corresponding time. The acquisition step is to obtain the observation matrix calibration value in the slope environment through actual measurement and system modeling before deployment. The acquisition step is to combine the noise covariance and the observation error covariance to perform Kalman filtering. The acquisition step is to use multiple measurements to monitor the noise level of the equipment and take the stable value; Calculation process: when When is 2.15 (displacement in millimeters), is 2.30 (obtained by fusing multi-source observations at this moment), Take 1.00 (obtained through early modeling calibration), is 0.60 (calculated by the Kalman filter at the fifth iteration), is 0.05 (measured by the noise adjustment experiment), first calculate the denominator The value of , then calculate The value of , and then calculate the difference , then multiplied by the Kalman gain Available , and finally compare the result with Add together and get , take two decimal places to get 1.49; The results show that when the value fluctuates between 2.00 and 2.50, it often indicates that the monitoring status is in a relatively stable range. When the value is greater than 2.50, it indicates that the slope displacement has increased significantly and requires in-depth investigation. When the value is lower than 1.50, it indicates that the displacement has dropped to a certain limit range in a short time and the relevant monitoring equipment should be reviewed. The above results can be used for state estimation at the next moment and continuously updated in subsequent iterations.

[0026] After obtaining the noise-reduced data, first retrieve the data entries of all monitoring points in each time period, and align these noise-corrected values ​​according to the unified index used previously. Merge the monitoring values ​​of multiple monitoring devices at the same time. For the missing values ​​of individual time periods, use the fitting method of adjacent moments to supplement and mark the source. Then, check the noise reduction results of each time point repeatedly. If the difference compared with adjacent moments or adjacent positions exceeds the project experience threshold, it is necessary to check the observation information of the time period again. The project experience threshold is obtained by collecting and selecting the maximum displacement difference of the previous monitoring in the same area. The value is determined by the critical value at the upstream end of the statistical distribution. Its value will be updated regularly after combining the measuring range of the monitoring equipment and the local geological activity level. Then, after all records have been checked for difference, the noise reduction results corresponding to all moments are merged into a unified index structure. Subsequently, the fusion intermediate results obtained earlier are used as a reference for data integration. The numerical conflicts that occurred during the integration process are located to specific moments and positions and checked one by one. If some measuring point locations have been marked as loose rock areas in the field report, the data collection of the monitoring equipment at that location can be compared separately. Finally, all monitoring records are processed in sequence to obtain comprehensive monitoring data.

[0027] The steps to obtain the pattern analysis results are: Receive comprehensive monitoring data, use time recurrent neural network to perform time series analysis, process data at consecutive time points through the network recursive structure, and form time series analysis results; Classifying the time series analysis results, and identifying normal and abnormal patterns of data points in the sequence to obtain classified pattern recognition results; According to the pattern recognition results after the classification, the patterns of all data points are integrated and analyzed to obtain pattern analysis results.

[0028] Specifically, after receiving the comprehensive monitoring data, all the data in the form of time series are read into a time recurrent neural network model in a unified time sequence. The model is composed of a number of hidden layer units and is cyclically trained according to a fixed number of iterations. During training, the previously sorted monitoring data are first put into the network in batches, and the number of each batch is set according to the field monitoring period, and a fixed iteration step is set for each batch. Then, the weights of the network are reversely updated in each round of iteration, and the actual monitoring results collected are used as a comparison basis. The loss value is calculated to measure the network's fit to the monitoring trend at each time node, and the recursive coefficient within the network is continuously adjusted. If it is found that the loss value is difficult to decrease or the speed of improving the fit of the monitoring trend is significantly slowed down after several consecutive rounds of training, the training process can be stopped at this moment, the final converged weights in the network are fixed, and the preliminary time series analysis of the monitoring data is completed. Then, the prediction sequence output by the network is summarized and stored in chronological order to obtain the preliminary time series analysis results.

[0029] Based on the above preliminary time series analysis results, it is necessary to classify the time series data points. First, compare the predicted values ​​of all moments output by the network with the corresponding monitoring records, count the degree of difference of each data point and use the degree of difference as a reference for the classification process. Then, a difference interval will be set as the judgment standard in the classification process. The upper and lower limits of the interval can be determined based on the error distribution obtained from multiple analyses of the same area in the early stage. If the degree of difference is lower than the lower limit of the interval, it is regarded as a normal mode data point. If the degree of difference falls between the upper and lower limits of the interval, further check whether other judgment conditions are met, such as whether the local feature changes observed by the monitoring equipment remain stable. If so, it is still classified as normal, otherwise it is an abnormal mode. When the degree of difference exceeds the upper limit of the interval, it is directly judged as an abnormal mode. Finally, the classified labels are mapped to the corresponding time nodes to obtain the classified pattern recognition results.

[0030] Based on the pattern recognition results after classification, the integration of the patterns of all time series data points requires that the marked normal and abnormal labels be grouped and summarized in chronological order, and then the time difference and spatial distribution of adjacent data points are analyzed in each group. Similar labels at adjacent positions or adjacent moments in the sequence are merged, and when necessary, the data points across sites are synchronized and compared according to the previously obtained monitoring site layout information. If conflicting label assignments are found between multiple monitoring values ​​in adjacent areas, the corresponding difference records must be reviewed to confirm the accuracy of the current labels. After completing the inspection of all groups, the grouping results of normal and abnormal patterns are mapped to all monitoring moments, and the classification labels and their time indexes are uniformly numbered and recorded to finally obtain complete pattern analysis results.

[0031] The steps for obtaining early warning signal candidates are as follows: Receiving the pattern analysis result, analyzing normal pattern data and abnormal pattern data, determining an initial risk classification of the comprehensive monitoring data, and obtaining a preliminary classification result; Based on the preliminary classification results, the potential landslide and collapse risk levels are calculated to obtain the risk score, which is calculated as follows: ; in, represents the risk score of data point i, is the monitoring value of data point i under condition j, is the weight coefficient, is the average value of the monitoring value, is the abnormal probability, is the adjustment coefficient, is the number of data points; The risk scores of all points are comprehensively evaluated to form early warning signal candidates.

[0032] Specifically, after receiving the pattern analysis results obtained above, the normal pattern data is first separated from the abnormal pattern data and compared moment by moment. Then, in the comparison process, the key values ​​of each time node are extracted, such as the difference of surrounding monitoring points and the local fluctuation record of the measuring robot, etc. These values ​​are summarized into a unified index structure, and then the difference degree of each time node is retrieved according to the field experience threshold. The field experience threshold is mainly formulated by summarizing the maximum displacement deviation range collected during the previous slope monitoring processes and combining it with the on-site confirmation of geological professionals. Nodes with a difference degree far higher than the threshold will be marked as high-risk initial classification, and nodes with a difference degree close to the threshold will be marked as high-risk initial classification. The node with the highest value will call the abnormal mode data obtained earlier for verification, and check whether there are interruptions or interference factors in the continuous observation period of the monitoring equipment. If there is no interruption and the value is significantly abnormal, it will be classified as a high-risk initial classification; otherwise, it will be classified as a lower-risk initial classification. Subsequently, the time series curve of each node marked as high-risk will be retrieved one by one and compared with the changing trend of adjacent monitoring points. If the comparison result shows an abnormal trend throughout an observation period, the original classification conclusion will be maintained. If completely different abnormal marks are found in the same monitoring period during the comparison process, the working status of the monitoring equipment will be checked again. After completing the above retrieval and classification operations, the preliminary classification results will be obtained.

[0033] The benefit of the formula is that it comprehensively measures the potential landslide and collapse risks by considering both the fluctuation of monitoring values ​​and the probability of abnormality. It uses the measurement item of the degree of fluctuation and the logarithmic correction item of the probability of abnormality to make the risk measurement more comprehensive. Indicates The risk score of a data point is obtained by comparing the monitoring data of the point in the same time window and introducing the abnormal probability for weighting. It represents the value of the i-th data point under the j-th monitoring condition, which can be decomposed and extracted through the multi-source fusion data obtained above. The weight coefficients for different monitoring conditions are set based on the evaluation of the accuracy and coverage of each monitoring device during previous on-site calibrations. Indicates the average value of the monitoring value, which is obtained by taking the arithmetic mean of all data points in the current monitoring window. It represents the probability of abnormality, which can be counted and quantified based on the abnormal pattern matching results obtained previously. It represents the adjustment coefficient, which is determined by the relationship between risk change and displacement amplitude in the records of several previous monitoring of the slope by geological professionals and remains unchanged during the monitoring period. is the number of data points, which is determined by the number of all valid data points in this monitoring window; Calculation process: For example, for the 5th data point in a slope area, set the total number of data points n=50, and calculate the value of the data point under five monitoring conditions They are 2.10, 1.95, 2.00, 2.15 and 2.05 respectively. The weight coefficients are obtained through historical data comparison and expert evaluation. They are 0.25, 0.20, 0.20, 0.20, and 0.15 respectively, using the average value of all measurement points in this monitoring window. , abnormal probability , adjustment coefficient , first calculate : Divide by And take the square root of , and then calculate: Finally, add the two parts together to get ; The results show that when When it is between 2.50 and 4.00, it can be classified as medium risk. If it is greater than 4.00, it means that the data point needs to be closely monitored in the near future. If it is lower than 2.00, it means that the risk score in the near future is low.

[0034] Based on the above risk scores, it is necessary to perform batch retrieval of the values ​​of all monitoring points, associate the score corresponding to each data point with its geographical location and monitoring period record, and then classify and mark the regional points with higher scores, so that points with the same high score range in the same monitoring period and adjacent areas are included in the same group, and then superimpose the group information with the geological structure zoning of the slope to identify the concentrated areas of score distribution. If it is found that the scores of some areas show continuous high values ​​in the same period, they are marked as high-priority monitoring objects, and check the abnormal probability distribution corresponding to each monitoring point again to determine whether the corresponding preset threshold is triggered. The preset threshold is mainly formulated by combining the evaluation of geological experts and the upper limit of the error of the monitoring equipment in the early stage. When the threshold is exceeded, more detailed observation data need to be called for the location point and its monitoring timeline is continuously tracked. Finally, after all the score markings are completed, the group index and slope position are uniformly sorted and recorded to obtain early warning signal candidates.

[0035] The steps to obtain the risk probability value are: Based on the early warning signal candidate, current monitoring data and historical monitoring data are subjected to data fusion processing to obtain a comprehensive monitoring data set; Based on the comprehensive monitoring data set, the current risk probability value of the slope is calculated using the following formula: ; in, represents the risk probability value of the slope, represents the warning signal strength of the ith data point, represents the deviation between the candidate warning signal and the reference warning signal, Represents the total number of data points.

[0036] Specifically, after integrating the previously obtained warning signal candidates with the current and past geological monitoring data, a unified time index is first set for each data source in the data management platform to ensure that all records in the same time period are aligned with each other. Then, based on the spatial location of the monitoring site, the data points in different regions are merged into the corresponding index structure, and these records are reviewed one by one to confirm that the value range read by the monitoring equipment is consistent with the previous annotation. If a large deviation is found from the field experience threshold, it is necessary to check the record status at that moment. The field experience threshold is a value obtained by combining the maximum displacement difference measured in multiple monitorings and the drift range of the monitoring equipment. In the merging process, the possible time interval difference of the data source is processed, and necessary comparisons and markings are performed using interpolation or interval filling. If it is found that the difference between certain time periods and previous monitoring data exceeds the predetermined standard continuously, it indicates that the observation results of the period need additional attention. The predetermined standard is based on the average difference range between adjacent time periods and the previous and subsequent records in historical monitoring. After completing the integration of all data, these fusion results are sorted and recorded in chronological order to obtain a comprehensive monitoring data set.

[0037] The formula is useful in that it integrates and accumulates the warning signal strength of each monitoring point in the form of a product, and uses the deviation term to correct the denominator, so as to take into account the relative differences of different data points in the overall evaluation; represents the risk probability value of the slope. The acquisition step is to match the fused multi-source monitoring data with the early warning signal candidates under a unified index and extract the corresponding signal strength and deviation information. It represents the warning signal strength of the ith data point. The value is confirmed by searching the warning signal candidates generated in the previous step one by one and combining the recorded peak value of the monitoring equipment and other factors. The deviation between the candidate warning signal and the benchmark warning signal can be determined by comparing the benchmark value in time and space. The benchmark value is usually obtained by repeated measurements and statistics on site by geological engineers. Represents the total number of data points, which is defined according to the number of valid data points actually collected in the current monitoring window; Calculation process: Taking the data points at five key positions in the monitoring window as an example, let , and their corresponding warning signal strength The values ​​are 2.10, 1.80, 2.50, 0.90, and 1.30 respectively. The numerical range is determined by comparing the peak value of the monitoring equipment output and combining the signal amplitude recorded previously. The deviation between the candidate warning signal and the benchmark warning signal is They are set to 0.50, 0.65, 1.00, 0.30, and 0.40 respectively. These deviation values ​​are obtained by checking the difference point by point and considering the calibration results of the monitoring equipment under the premise that the on-site measurement reference value is 2.00. First, calculate each factor And multiply them successively: Multiplying the above five results gives: ; Then calculate , so: ; The results show that within the current monitoring window, based on the collected data, the risk probability value of the slope is approximately 0.9910. When the result value is close to 1.0000, it means that the slope is in a very high risk state, and when the result value is significantly lower than 0.5000, it means that the risk is relatively low.

[0038] The steps to obtain the slope stability assessment results are as follows: The warning level is calculated based on the slope risk probability value and the set safety threshold, and the calculation formula is: ; in, Indicates the warning level. is the risk probability value, θ is the safety threshold; A slope stability assessment is performed according to the warning level to obtain a slope stability assessment result.

[0039] Specifically, the formula is useful in that By performing logarithmic operations on the ratio to θ and rounding it, we can use the exponential amplification or reduction mechanism to set the warning level to avoid being too smooth or too sensitive; L represents the warning level, and the acquisition step is to read the value calculated in the previous step. And it is associated with the safety threshold θ set in the field operation. represents the risk probability value, which can be uniformly calculated based on the noise reduction data obtained above and the strength of the warning signal and other factors. θ represents the safety threshold, and its value is determined by comparing the maximum displacement amplitude statistically obtained in previous on-site observations with the engineering safety specifications. Calculation process: For example, the data obtained during the current monitoring period , and geological experts determined θ=0.30 based on the historical deformation statistics of the slope and local construction safety standards. , and then perform Calculation , take the integer , and then compare it with 0 and select the maximum value, that is , so we can get L=2; The result shows that when L=2, it means that the current slope warning level is classified into the second level range. Further increase may cause the value of L to rise to a higher level. If the calculated result is lower than 1 and less than 0 after rounding, L=0, indicating the lowest alert level.

[0040] When evaluating the above-mentioned warning level, first compare the L value just obtained with the historical records of on-site observations, read the warning level status of the same slope in the past multiple monitoring cycles, and mark the partitions in combination with the geological structure information of the corresponding time period, and then classify the level values ​​that have been evaluated in each time period. For the parts where there is a sudden jump in the level change between adjacent time periods, the continuity of the monitoring data and the completeness of the monitoring equipment records will be separately checked, and the difference values ​​will be listed together in the classification process. If it is found that the warning levels of two or more consecutive time periods are high and the values ​​coincide with the risk range corresponding to the on-site experience threshold, the on-site operators are required to review the relevant monitoring equipment and geological conditions. The on-site experience threshold is determined by professionals using the alarm comparison table of historically large displacement situations and referring to monitoring cases of similar projects. If it is confirmed that the change in the warning level is consistent with the change trend of the data itself, a higher or lower warning level will be maintained in the final summary, and no additional manual intervention is required. Otherwise, it is necessary to compare the self-test records of the monitoring equipment to eliminate the impact of the fault. After all classification and comparison operations are completed, the warning level of the slope is adjusted and the partition mark information is retained simultaneously to obtain the slope stability assessment result.

[0041] The steps to obtain the alarm execution record are as follows: receiving the slope stability assessment result, and triggering an emergency communication signal when the slope stability assessment result exceeds a set alarm threshold; Based on the emergency communication signal, an emergency response is activated, warning information is sent to all related personnel and institutions, and an alarm execution record is generated, wherein the warning information includes the slope location and risk level.

[0042] Specifically, after receiving the slope stability assessment results obtained previously, the numerical information of the results is compared with the alarm threshold set on site when the monitoring equipment is deployed. This threshold is determined by the maximum slope displacement amplitude summarized in multiple historical monitorings and geological engineering related standards. If the system determines that any value in the assessment result exceeds the alarm threshold, it is compared and marked in the software to confirm that the trigger condition has been met. Then the confirmation results are centrally processed according to the time index and the corresponding geographical location, and the information marked with the alarm mark in each data record is summarized. If the value of a certain location or time period has been marked as exceeding the threshold, a signal triggering emergency communication is automatically generated. The signal has an independent identifier in the internal system and is forwarded to the subsequent program to focus on the environment of the relevant location and complete the preparation for starting communication. All relevant records in the time period where the threshold alarm occurs are checked separately to confirm and record these values ​​that exceed the preset threshold. Then the summary results are associated with the corresponding risk level and a signal triggering emergency communication is generated.

[0043] Based on the signal that triggers the emergency communication, the information of all pre-registered related personnel and institutions is queried to obtain the contact information of the relevant parties. The corresponding information includes specific mobile phone numbers, email addresses, and instant messaging accounts, etc. The slope location and current risk level are packaged and sent in order. The slope location will be marked in the map system before sending, and the corresponding longitude and latitude are obtained through the previously obtained monitoring equipment installation location and geographic coordinate conversion. Then, the alarm information is generated one by one in the program and matched to the corresponding contact entry. At the same time, in order to avoid duplicate alarms, a special de-duplication mark will be set in the system. Only one sending record is generated for the same risk level in the same time period and the same geographical range. If it is confirmed through inspection that the new triggering of the emergency communication does not overlap with the previous sending period, it is allowed to be sent again. Finally, all the alarm content and its time index are concentrated for future inspection to complete the dissemination of early warning information to all related parties.

[0044] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A high slope stability monitoring and early warning system, characterized in that: The system comprises: A data fusion module is used to collect data through a measuring robot, GNSS, a tension-wire displacement meter, an inclinometer hole, a micro-core pile, a water level meter and a rain gauge, normalize the collected data to generate preliminary slope monitoring data, and integrate multi-source data and reduce noise on the preliminary slope monitoring data to obtain comprehensive monitoring data; A pattern recognition module is used to perform time series analysis on the comprehensive monitoring data, identify normal data patterns and abnormal patterns, generate pattern analysis results, and analyze the risk level of potential landslides and collapses based on the pattern analysis results to obtain early warning signal candidates; A risk assessment module is used to perform stability analysis on the current monitoring data and the historical monitoring data according to the early warning signal candidate, obtain the slope risk probability value corresponding to the current monitoring data, and compare the slope risk probability value with the set safety threshold, adjust the early warning level according to the comparison result, and obtain the slope stability assessment result; The automatic alarm module is used to receive the slope stability assessment result. When the slope stability assessment result exceeds the alarm threshold, emergency communication is triggered, warning information is sent to related personnel and institutions, and an alarm execution record is generated. The warning information includes the slope location and risk level.

2. The system according to claim 1, characterized in that The steps for obtaining the preliminary slope monitoring data are as follows: The original monitoring data set was obtained by collecting surface deformation, deep underground deformation, tilt, vibration, groundwater level and rainfall data from high slopes through measurement robots, GNSS, tension-wire displacement meters, inclinometer holes, micro-core piles, water level gauges and rain gauges; Performing standardization processing on each original monitoring data in the original monitoring data set to generate a normalized monitoring data set; Each normalized monitoring data in the normalized monitoring data set is analyzed, and information from different monitoring devices is integrated through a data fusion method to obtain preliminary slope monitoring data.

3. The system according to claim 1, characterized in that The steps for obtaining the comprehensive monitoring data are as follows: Based on the preliminary slope monitoring data, the data of the measuring robot, GNSS, tension-wire displacement meter, inclinometer hole, micro-core pile, water level meter and rain gauge are integrated by stepwise weighted average to form a data fusion intermediate result; According to the intermediate result of data fusion, the Kalman gain is calculated and noise reduction is performed to obtain denoised data, and the formula is: ； in, is the estimated state at time t, is the observed data, is the observation model, is the Kalman gain, is the noise adjustment factor; The denoised data are integrated to obtain comprehensive monitoring data.

4. The system according to claim 1, characterized in that The steps for obtaining the pattern analysis results are: Receive comprehensive monitoring data, use time recurrent neural network to perform time series analysis, process data at consecutive time points through the network recursive structure, and form time series analysis results; Classifying the time series analysis results, and identifying normal and abnormal patterns of data points in the sequence to obtain classified pattern recognition results; According to the pattern recognition results after the classification, the patterns of all data points are integrated and analyzed to obtain pattern analysis results.

5. The system according to claim 1, characterized in that The steps of obtaining the early warning signal candidates are: Receiving the pattern analysis result, analyzing normal pattern data and abnormal pattern data, determining an initial risk classification of the comprehensive monitoring data, and obtaining a preliminary classification result; Based on the preliminary classification results, the potential landslide and collapse risk levels are calculated to obtain the risk score, which is calculated as follows: ； in, represents the risk score of data point i, is the monitoring value of data point i under condition j, is the weight coefficient, is the average value of the monitoring value, is the abnormal probability, is the adjustment coefficient, is the number of data points; The risk scores of all points are comprehensively evaluated to form early warning signal candidates.

6. The system according to claim 1, characterized in that The steps for obtaining the risk probability value are: Based on the early warning signal candidate, current monitoring data and historical monitoring data are subjected to data fusion processing to obtain a comprehensive monitoring data set; Based on the comprehensive monitoring data set, the current risk probability value of the slope is calculated using the following formula: ； in, represents the risk probability value of the slope, represents the warning signal strength of the ith data point, represents the deviation between the candidate warning signal and the reference warning signal, Represents the total number of data points.

7. The system according to claim 1, characterized in that The steps for obtaining the slope stability assessment result are as follows: The warning level is calculated based on the slope risk probability value and the set safety threshold, and the calculation formula is: ； in, Indicates the warning level. is the risk probability value, is the safety threshold; A slope stability assessment is performed according to the warning level to obtain a slope stability assessment result.

8. The system according to claim 1, characterized in that The steps for obtaining the alarm execution record are: receiving the slope stability assessment result, and triggering an emergency communication signal when the slope stability assessment result exceeds a set alarm threshold; Based on the emergency communication signal, an emergency response is activated, warning information is sent to all related personnel and institutions, and an alarm execution record is generated, wherein the warning information includes the slope location and risk level.

Images