Expansive soil landslide monitoring and early warning system

Abstract

The invention discloses an expansive soil landslide monitoring and early warning system, and relates to the technical field of geological monitoring, the expansive soil landslide is monitored in real time by using multiple types of sensors so as to obtain multiple environmental data and displacement data of the expansive soil landslide in real time, and the environmental data with high correlation with the environmental data are extracted at the same time; fusion of multi-source data is realized, and an accurate early warning grade of the expansive soil landslide is established; meanwhile, the displacement data is decomposed into trend term displacement and periodic term displacement, polynomial fitting is used for predicting the trend term displacement to obtain a trend term displacement predicted value, PSO-LSTM is used for predicting the periodic term displacement to obtain a periodic term displacement predicted value, and displacement data representing displacement long-term deformation rules and short-term fluctuation are captured at the same time; and adding the trend term displacement predicted value and the periodic term displacement predicted value to obtain an accurate displacement predicted value, and realizing high-precision landslide monitoring and early warning according to the deformation rate obtained by the displacement predicted value and the early warning level to which the tangent angle belongs.

Description

Technical Field

[0001] The present invention relates to the technical field of geological monitoring, and particularly relates to a monitoring and early warning system for expansive soil landslides. Background Art

[0002] Due to the swelling and shrinking properties, fissure properties, and overconsolidation properties of expansive soil, the deformation mechanism of expansive soil landslides is complex, the soil structure is easily damaged, landslides are extremely likely to be induced, and the suddenness is strong; therefore, it is necessary to conduct real-time and effective monitoring of expansive soil landslides, and reduce the life and property losses caused by landslides through monitoring and early warning.

[0003] Most of the current monitoring systems for expansive soil landslides use single sensors (such as only using GNSS to monitor displacement, etc.) and monitoring systems based on sensors and 4G / 5G communication devices. The single sensor ignores that expansive soil landslides are the result of the coupling of multiple factors, fails to achieve comprehensive monitoring of key parameters such as geological deformation, stress, temperature and humidity, greatly reduces the prediction accuracy rate, and at the same time, most systems only predict single factors during prediction, without considering various precursor factors of landslides, resulting in low early warning accuracy. Summary of the Invention

[0004] The embodiments of the present invention provide a monitoring and early warning system for expansive soil landslides, which can solve the problems in the prior art that single sensors at the current stage ignore that expansive soil landslides are the result of the coupling of multiple factors, fail to achieve comprehensive monitoring of key parameters such as geological deformation, stress, temperature and humidity, greatly reduce the prediction accuracy rate, and at the same time, most systems only predict single factors during prediction, without considering various precursor factors of landslides, resulting in low early warning accuracy.

[0005] The embodiments of the present invention provide a monitoring and early warning system for expansive soil landslides, including a landslide monitoring module, a data analysis module and a solution and alarm module;

[0006] The landslide monitoring module is used to collect various environmental data and displacement data of the expansive soil landslide in real time through multiple types of sensors arranged in the expansive soil landslide;

[0007] The data analysis module is used to analyze the correlation between various environmental data and displacement by using the Pearson correlation coefficient, and extract the environmental data with a correlation value higher than a preset threshold;

[0008] The displacement data is decomposed into a trend item displacement and a periodic item displacement by using the first-order moving average method; the trend item displacement is fitted by using the polynomial least squares method to capture the long-term deformation data within the trend item displacement, and a trend item displacement prediction value is obtained; the periodic item displacement is fitted by using the PSO-LSTM model to capture the short-term fluctuation data within the periodic item displacement, and a periodic item displacement prediction value is obtained; the trend item displacement prediction value and the periodic item displacement prediction value are added together to obtain a displacement prediction value;

[0009] The solution and alarm module is used to obtain the deformation rate and tangent angle of the expansive soil landslide based on the environmental data and displacement data with a correlation value higher than a preset threshold, and divide the warning level of the expansive soil landslide based on the deformation rate and tangent angle; according to the displacement prediction value, obtain the predicted deformation rate and tangent angle of the expansive soil landslide, and conduct early warning of the expansive soil landslide according to the warning level to which the predicted deformation rate and tangent angle belong.

[0010] Preferably, it further includes:

[0011] The data storage module is used to receive and classify and store various environmental data and displacement data collected by the landslide monitoring module;

[0012] The battery management module is used to monitor the device battery status in real time, set overcharge protection, under-voltage protection and energy-saving mode, and dynamically adjust the power supply strategy of the device;

[0013] The control and management module is used to configure the monitoring station parameters and display the monitoring data and warning information in real time through a visual interface.

[0014] Preferably, the landslide monitoring module includes an environmental data acquisition sub-module and a positioning sub-module;

[0015] The environmental data acquisition sub-module includes a rain gauge, a soil temperature and humidity meter, and an earth pressure gauge, and is used to collect rainfall, soil temperature and humidity, and earth pressure data;

[0016] The positioning sub-module uses a GNSS receiver and RTK positioning to obtain the three-dimensional displacement data of the expansive soil landslide body in real time.

[0017] Preferably, the acquisition of the displacement prediction value includes:

[0018] Using the first-order moving average method to analyze the displacement data of the expansive soil landslide, and decomposing the displacement data into a trend-term displacement and a periodic-term displacement; wherein, the trend-term displacement represents the long-term deformation law of the landslide displacement, and the periodic-term displacement represents the short-term fluctuation of the landslide displacement;

[0019] Using the polynomial least squares method to fit the trend-term displacement, capture the long-term deformation data in the trend-term displacement, and obtain the trend-term displacement prediction value, expressed as:

[0020]

[0021] Among them: y represents the trend-term prediction value; a i represents the polynomial coefficient; x i represents the time variable;

[0022] Optimize the number of neurons, learning rate, and number of iterations of LSTM through the Particle Swarm Optimization (PSO) algorithm. Combine the mean square error as the fitness function to train the PSO-LSTM model, and use the trained PSO-LSTM model to fit the periodic term displacement to obtain the predicted value of the periodic term displacement.

[0023] Add the predicted value of the trend term displacement and the predicted value of the periodic term displacement to obtain the predicted value of the displacement.

[0024] Preferably, the division of the warning levels of the expansive soil landslide includes:

[0025] Calculate the deformation rate v and the tangent angle θ of the expansive soil landslide based on the environmental data and displacement data with correlation values higher than the preset threshold; the environmental data includes rainfall data, soil pressure data, velocity data, and humidity data.

[0026] Compare the deformation rate v with the preset landslide thresholds v a 、v b 、v c and compare the tangent angle θ with the preset tangent angle thresholds θ1 and θ2 to divide the warning levels.

[0027] If the deformation rate v of the monitored slope section i is less than the landslide threshold v a , and the displacement tangent angle θ is less than the tangent angle threshold θ1, then determine that the landslide warning level of the monitored slope section is a blue warning, and mark the corresponding monitored slope section as a blue warning slope section.

[0028] If the deformation rate v of the monitored slope section i is greater than or equal to the landslide threshold v a , less than the landslide threshold v b , and the displacement tangent angle θ is less than the tangent angle threshold θ1, then determine that the landslide risk of the monitored slope section is a yellow warning, and mark the corresponding monitored slope section as a yellow warning slope section.

[0029] If the deformation rate v of the monitored slope section i is greater than or equal to the landslide threshold v b , less than the landslide threshold vc, and the displacement tangent angle θ is less than the tangent angle threshold θ2 and greater than θ1, then determine that the landslide risk of the monitored slope section is an orange warning, and mark the corresponding monitored slope section as an orange warning slope section.

[0030] If the deformation rate v of the monitored slope section i is greater than or equal to the landslide threshold v c , and the displacement tangent angle θ is greater than the tangent angle threshold θ2, then determine that the landslide risk of the monitored slope section is a red warning, and mark the corresponding monitored slope section as a red warning slope section.

[0031] Preferably, the real-time display of the monitoring data and warning information through the visualization interface includes:

[0032] The hourly rainfall and daily rainfall obtained by the rain gauge are displayed using a bar chart or a line chart;

[0033] The sudden temperature and humidity obtained by the soil temperature and humidity gauge are displayed using a line chart;

[0034] The soil pressure data obtained by the soil pressure gauge is displayed using a water drop chart;

[0035] The displacement data obtained by the GNSS receiver is displayed using a three - point chart, an error ellipse chart or a three - dimensional bar chart.

[0036] Preferably, the management and control module supports multi - satellite system solution configuration, including BDS, GPS, GLONASS and GALILEO;

[0037] It supports dynamically adjusting the elevation angle, signal - to - noise ratio threshold and solution constraint mode of the monitoring station; it supports downloading historical monitoring data by time period.

[0038] The embodiment of the present invention provides an expansive soil landslide monitoring and early warning system. Compared with the prior art, its beneficial effects are as follows:

[0039] The present invention uses multiple types of sensors to monitor the expansive soil landslide in real time, so as to obtain various environmental data and displacement data of the expansive soil landslide in real time, and at the same time extracts environmental data with high correlation with the environmental data, realizes the fusion of multi - source data and establishes an accurate early warning level for the expansive soil landslide; at the same time, the displacement data is decomposed into a trend - term displacement and a periodic - term displacement. The polynomial fitting is used to predict the trend - term displacement to obtain the predicted value of the trend - term displacement, and the PSO - LSTM is used to predict the periodic - term displacement to obtain the predicted value of the periodic - term displacement, so as to capture the displacement data representing the long - term deformation law and short - term fluctuation of the displacement at the same time. This process uses multiple displacement factors to obtain high - precision displacement data, and adds the predicted value of the trend - term displacement and the predicted value of the periodic - term displacement to obtain an accurate displacement prediction value, and realizes high - precision landslide monitoring and early warning according to the warning level to which the deformation rate and tangent angle obtained from the displacement prediction value belong. Description of the Drawings

[0040] Figure 1 It is the overall logic schematic diagram of an expansive soil landslide monitoring and early warning system provided by the embodiment of the present invention;

[0041] Figure 2 It is the working process schematic diagram of an expansive soil landslide monitoring and early warning system provided by the embodiment of the present invention;

[0042] Figure 3 It is the logic schematic diagram of the landslide monitoring module of an expansive soil landslide monitoring and early warning system provided by the embodiment of the present invention;

[0043] Figure 4Logical schematic diagram of the data storage module of an expansive soil landslide monitoring and early warning system provided by an embodiment of the present invention;

[0044] Figure 5 Logical schematic diagram of the battery management module of an expansive soil landslide monitoring and early warning system provided by an embodiment of the present invention;

[0045] Figure 6 Logical schematic diagram of the data analysis module of an expansive soil landslide monitoring and early warning system provided by an embodiment of the present invention;

[0046] Figure 7 Logical schematic diagram of the solution and alarm module of an expansive soil landslide monitoring and early warning system provided by an embodiment of the present invention;

[0047] Figure 8 Logical schematic diagram of the control and management module of an expansive soil landslide monitoring and early warning system provided by an embodiment of the present invention. Detailed implementation manners

[0048] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following detailed description of the specific implementation manners of the present invention will be given in conjunction with the accompanying drawings. Many specific details are set forth in the following description to fully understand the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0049] See Figure 1 and Figure 2 , an embodiment of the present invention provides an expansive soil landslide monitoring and early warning system. Most of the current landslide monitoring systems use a single sensor (such as only using GNSS to monitor displacement, etc.) and a monitoring system based on sensors and 4G / 5G communication devices. The single sensor ignores that the expansive soil landslide is the result of the coupling of multiple factors and fails to achieve the comprehensive monitoring of key parameters such as geological deformation, stress, temperature and humidity, which will greatly reduce the prediction accuracy rate; while the monitoring system based on sensors and 4G / 5G communication devices lacks battery management, and the system is often limited by the power supply. Both the sensors and communication devices need to be powered continuously for a long time. There may be problems with imperfect power facilities in the landslide area, and in bad weather, there will be situations such as insufficient solar power supply and wire breakage, resulting in equipment communication interruption and affecting the continuity, integrity and real-time nature of the monitoring data. And many systems or platforms only predict single factors during prediction, without considering various precursor factors of the landslide, resulting in a reduction in the early warning accuracy.

[0050] Lack of battery management: At present, many early warning platforms can monitor landslides, but most of them lack a battery management module and are difficult to protect the battery. There are problems such as excessive power supply on sunny days that cannot be disconnected in time and insufficient power supply on rainy days that can easily cause the battery to be under-voltage. In the long run, this will reduce the battery life, resulting in a decrease in the endurance ability, affecting the discharge performance, and even damaging the equipment during the normal operation of the device.

[0051] Single model: Existing early warning models mostly adopt a single prediction index, such as rainfall, displacement, etc. However, there are many inducing factors for expansive soil landslides, and a single index is difficult to accurately reflect the complex genetic mechanism of expansive soil landslides, greatly affecting the accuracy and timeliness of landslide prediction, resulting in a high probability of false alarms and missed alarms.

[0052] Insufficient visualization: At present, traditional monitoring and early warning platforms lack an intuitive data visualization display interface. The data display of some platforms is too professional, making it difficult for non-professional personnel to accurately identify the signals of landslide early warning, increasing the professional barrier and making it difficult to provide comprehensive and intuitive early warning information for decision-makers.

[0053] Information lag: The information transmission efficiency of traditional monitoring and early warning platforms is low, resulting in a lag in emergency response. Expansive soil landslides have a certain gradualness. It often takes a long time from local instability to overall sliding, making it difficult to accurately capture the critical state. Moreover, landslides come suddenly and can cause several people to lose their lives in just a few seconds. Real-time monitoring and early warning are required to maintain the timeliness of information.

[0054] To address the above problems, the expansive soil landslide monitoring and early warning system designed by the present invention includes a landslide monitoring module, a data storage module, a battery management module, a data analysis module, a solution and alarm module, and a control and management module; the specific composition and functions of each module are as follows:

[0055] I. Landslide monitoring module.

[0056] The landslide monitoring module is jointly composed of an environmental data sensor acquisition sub-module and a positioning sub-module; it is used to monitor the data of the landslide body, usually set at the landslide monitoring point, number all the monitoring points, and collect landslide parameters in real time to obtain monitoring data, including data from rain gauges, GNSS receivers, soil temperature and humidity meters, earth pressure gauges, etc.; after the sensors at the monitoring points collect environmental data, they are transmitted to the cloud server through communication devices (4G / 5G antennas) and network facilities (HTTP protocol, TCP protocol) and stored in the database; among them, the environmental data sensor acquisition sub-module includes a rain gauge unit, a soil temperature and humidity meter unit, and an earth pressure gauge unit, and the positioning sub-module includes a GNSS receiver unit; each sub-module is as Figure 3 shown.

[0057] II. Data storage module.

[0058] The data storage module consists of a data reception sub-module and a data sharing sub-module, and its logic is as follows Figure 4 shown; the data reception sub-module first receives the data from the landslide monitoring module, and then stores the data in the data sharing sub-module to achieve data sharing of different stations and different sensors; at the same time, MySQL is used to store the parsed device data, which is stored by field classification (such as device ID, time, temperature, humidity, etc.); the obtained real-time data is stored in the database. After the data is stored, the back-end service will query the database when the user requests and dynamically return the latest data; the user can see the latest real-time data by refreshing the web page through the browser. Among them, due to the multi-source nature of the data, in order to avoid data storage chaos, rainfall, soil temperature and humidity, and soil pressure are stored in the same database, named the disaster monitoring database, which belongs to disaster monitoring data, and GNSS displacement is stored in another database, named the basic geographic database, which belongs to basic geographic data. The data return format is JSON format.

[0059] III. Battery management module.

[0060] In the battery management module, the power, current, voltage, power generation, power consumption, power supply situation of the solar panel, etc. of the battery can be viewed, which is convenient to master the power consumption situation of the battery. The on-lighting voltage, on-lighting delay, overcharge voltage, under-voltage protection, under-voltage return, working mode, energy-saving start point, energy-saving end point, low-temperature charge prohibition, low-temperature discharge prohibition, etc. of the battery can also be set to protect and manage the battery, and the real-time situation can be displayed on the platform page and adjusted according to needs to avoid damage to the battery; its logic is as follows Figure 5 shown.

[0061] IV. Data analysis module.

[0062] The data analysis module conducts real-time data visualization. It uses bar charts and line charts to display the hourly rainfall and daily rainfall obtained by the rain gauge, three-dimensional dot plots, error ellipsoids, three-dimensional bar charts, etc. to display the displacement data obtained by the GNSS receiver, line charts to display the temperature and humidity obtained by the soil temperature and humidity sensor, and water droplet charts to display the soil pressure data obtained by the soil pressure sensor. It will also visualize the information of the subsequent calculation and warning module, such as: displaying point numbers, warning times, displacements, rates, etc.; in this module, the data will first be subjected to simple data cleaning, removing invalid data, empty data, and abnormal data to ensure data reliability, and aligning the timestamps and node positions of different devices to ensure data consistency; during data analysis, simple numerical calculations will be performed, such as: correlation analysis between multi-source data (Pearson correlation coefficient method), and whether there is an obvious time mutation response relationship between them and the displacement, etc. When the device power supply is insufficient or communication is impossible under harsh conditions, data can be inferred based on the time relationship to avoid false detection and missed detection. Then, time series analysis of single-site and multi-site data will be carried out, and the data will be imported into the model for the next warning analysis, and its logic is as Figure 6 shown.

[0063] The sensor data obtained by the present invention is diverse, but not all data needs to be input into the prediction model. Some data requires time accumulation to be related to the landslide instability height. Therefore, the data of multiple sensors is first preprocessed and screened, and the factors with a greater correlation with landslide instability are selected and used as the input values of the prediction model, and then the next deformation prediction is carried out; in the present invention, four parameters, namely GNSS displacement value, rainfall value, soil pressure value, and soil humidity value, are selected for preprocessing and correlation analysis, so as to construct a landslide prediction model for multi-source heterogeneous data fusion; the calculation process in the data analysis module includes:

[0064] First, data preprocessing is carried out. Multi-source heterogeneous data fusion predicts with a day as the step length, so it is necessary to obtain the monitoring data of each device every day; for rainfall data, the hourly rainfall of each day is accumulated to obtain the daily rainfall data of that day; for soil pressure data and soil humidity data, the average value of the monitoring data of each day needs to be calculated to obtain the daily soil pressure data and soil humidity data; there are certain random errors in the GNSS monitoring data, and if not processed, it will reduce the model accuracy, so it is also necessary to perform a certain degree of smoothing on the monitoring data. Therefore, for GNSS displacement data, the moving window method is selected for smoothing processing, and the window size is set to an appropriate value.

[0065] Next, the screening of model input factors is carried out; the candidate input factors of the model include the sum of rainfall, earth pressure, and soil moisture for the previous 1 day, previous 2 days, previous 3 days, previous 7 days, and previous 15 days; it is necessary to use the Pearson correlation coefficient to screen the input factors to obtain the factors most correlated with displacement; the value range of the Pearson correlation coefficient is [-1, 1], where a coefficient value of 1 proves a perfect positive correlation between the two variables; a coefficient value of -1 proves a perfect negative correlation between the two variables; a coefficient value of 0 proves that there is no linear relationship between the two variables; it should be noted that the Pearson correlation coefficient can only measure the degree of linear correlation between two variables, and for non-linear relationships, the results of its correlation coefficient may not be accurate enough; the Pearson coefficient will be affected by factors such as the sample size and data distribution.

[0066] Calculating the Pearson correlation coefficient with GNSS displacement using the above data shows that the correlation between the earth pressure data and the soil moisture data gradually decreases with the accumulation of time. Therefore, the two factors of the earth pressure data and the soil moisture data for the previous 1 day are selected as inputs; the correlation between the rainfall data and displacement increases with the accumulation of time. Considering the change in the length of the data and the increasing trend of the correlation, the sum of rainfall for the previous 7 days is selected as the input factor for the model.

[0067] In the next step of inputting the factors into the model for displacement prediction, if the selected GNSS displacement data, rainfall data, soil moisture, and earth pressure data are directly input into the model, it will lead to situations such as reduced prediction accuracy, slower model training speed, and abnormal weights due to the different dimensions of these data. To avoid the occurrence of the above situations, it is necessary to perform normalization and inverse normalization processing before inputting the data into the model. The normalization uses the Min-Max method, and the expression is as follows:

[0068]

[0069] where: X represents the original data; X min and X max represent the minimum and maximum values of this feature respectively.

[0070] The purpose of inverse normalization is to convert the normalized data back to the original data range after the prediction is completed in order to obtain the true prediction value. Its expression is as follows:

[0071] X orig =X norm ×(X max -X min )+X min .

[0072] where: X norm represents the normalized data; X max and Xmin Represents the maximum and minimum values in the characteristic factor training data (here, GNSS data).

[0073] V. Solution and Alarm Module.

[0074] The solution and alarm module consists of a solution operator module and an alarm sub-module, as Figure 7 shown; the early warning sub-module processes all data from the data analysis module. First, it determines the location of the monitoring point through the positioning sub-module, then establishes a deformation displacement prediction model PSO-LSTM for the data of this monitoring point. According to the landslide judgment criteria, it classifies the warning levels of the landslide body into four levels: blue warning, yellow warning, orange warning, and red warning, for the alarm sub-module to give alarms in a timely manner; in the alarm sub-module, warning information is alarmed. According to the positioning, the real-time warning level of the monitoring point is displayed in the 3D map, and alarm notifications are sent through text messages, emails, etc.; it ensures that remote users, including management departments, can real-time grasp the monitoring information of the expansive soil landslide through a web browser and formulate corresponding measures according to the monitoring information; among them, the warning level of the solution operator module is determined according to the displacement rate and displacement tangent angle of the landslide; after the monitoring station experiences multiple severe wet-dry alternations or complex and extreme weather, the cracks of the landslide will fully develop, significantly reducing the stability of the landslide, the slope gradually becomes unstable, and the deformation rate increases. However, considering only the deformation rate cannot accurately judge the actual state of the landslide, and sometimes it will cause misjudgment. The factor of the tangent angle also needs to be considered to achieve the joint discrimination of the landslide state and improve the accuracy of prediction; among them, the displacement data WYi is the displacement value of the monitoring slope section i, the rainfall data JYi is the rainfall value of the monitoring slope section i, the soil pressure data TYi is the soil pressure value of the monitoring slope section i, the temperature data WDi is the soil temperature value of the monitoring slope section i, and the humidity data SDi is the soil humidity value of the monitoring slope section i; the process of marking the monitoring slope section as blue warning, yellow warning, orange warning, or red warning includes: through the warning analysis module, calculating the deformation rate and tangent angle through the displacement data WYi, rainfall data JYi, soil pressure data Tyi, temperature data WDi, and humidity data SDi, and classifying the landslide according to the warning criteria in Table 1.

[0075] Table 1 Warning Criteria

[0076]

[0077] The warning criteria need to compare these two factors of the deformation rate and the tangent angle, obtain the deformation rate v of the monitoring slope section i, and compare the deformation rate v of the monitoring slope section i with the landslide thresholds v a 、v b 、v cCompare to obtain the displacement tangent angle θ of the monitored slope, and compare the tangent angle θ with the tangent angle thresholds θ1 and θ2 (the above thresholds need to be adjusted according to the corresponding on-site conditions, and different landslides, altitudes, and different climate conditions, etc. will affect the numerical values of these thresholds), and obtain: If the deformation rate v of the monitored slope section i is less than the landslide threshold v a , and the displacement tangent angle θ is less than the tangent angle threshold θ1, then determine that the landslide warning level of the monitored slope section is a blue warning, and mark the corresponding monitored slope section as a blue warning slope section; if the deformation rate v of the monitored slope section i is greater than or equal to the landslide threshold v a , less than the landslide threshold v b , and the displacement tangent angle θ is less than the tangent angle threshold θ1, then determine that the landslide risk of the monitored slope section is a yellow warning, and mark the corresponding monitored slope section as a yellow warning slope section; if the deformation rate v of the monitored slope section i is greater than or equal to the landslide threshold v b , less than the landslide threshold vc, and the displacement tangent angle θ is less than the tangent angle threshold θ2 and greater than θ1, then determine that the landslide risk of the monitored slope section is an orange warning, and mark the corresponding monitored slope section as an orange warning slope section; if the deformation rate v of the monitored slope section i is greater than or equal to the landslide threshold v c , and the displacement tangent angle θ is greater than the tangent angle threshold θ2, then determine that the landslide risk of the monitored slope section is a red warning, and mark the corresponding monitored slope section as a red warning slope section. The landslide monitoring module sends the monitored slope section and the warning signal to the monitoring and warning platform. After receiving the monitored slope section and the warning signal, the monitoring and warning platform sends the monitored slope section and the warning signal to the mobile terminal of the management personnel.

[0078] The prediction process of the landslide deformation displacement prediction model PSO-LSTM is as follows:

[0079] For the prediction of landslide displacement, first use the first-order moving average method to analyze the total displacement data, decompose the total displacement of the landslide into a trend term and a periodic term displacement, construct different models for them respectively for prediction, and then superimpose the prediction results to obtain the final landslide displacement prediction value; its decomposition formula is as follows:

[0080] S(t) = φ(t) + η(t).

[0081] Where: S(t) represents the displacement time series; φ(t) represents the trend term displacement; η(t) represents the periodic term displacement.

[0082] Among them, the trend term refers to the influence of the long-term evolution of the landslide's own geological conditions, and the periodic term refers to the influence of external factors such as rainfall on the landslide. For the trend term displacement, the present invention uses the polynomial least squares method for fitting prediction. For the periodic term displacement, a model combining the particle swarm optimization (PSO) algorithm and the LSTM model is used. Finally, the accuracy of the displacement prediction result is evaluated, and the prediction results are accumulated to obtain the total prediction value and error analysis, so as to obtain an accurate landslide displacement prediction result.

[0083] In the present invention, the correlation coefficient (R) and the root mean square error (RMSE) are selected to evaluate the accuracy of the prediction model; the formula is:

[0084]

[0085] Where: x i represents the true value; represents the predicted value; represents the average value of the true values; represents the average value of the predicted values; N represents the number of samples.

[0086] For the prediction of the trend term, first use the first-order moving average method to extract the trend term, and then use the polynomial fitting method for prediction. The first-order moving average method means that the data in the observation period is averaged once from far to near according to a certain span period (N periods), which can better reflect the trend and changes of the data. The calculation formula is as follows:

[0087] D i =[D1, D2,..., D t ,..., D n .

[0088]

[0089] Where: D i represents a set of time series data; represents the trend term corresponding to D extracted from the original sequence t , t = N, N + 1,..., n; the number of segments N represents the number of spans, that is, the number of data used for calculation from far to near, and generally N = 6 - 200 is selected.

[0090] The polynomial fitting is expressed as:

[0091]

[0092] Where: a represents the polynomial coefficient.

[0093] For the prediction of periodic terms, Long Short-Term Memory (LSTM) is adopted. Before using the LSTM algorithm, it is necessary to first define a fitness function to evaluate the performance of the LSTM algorithm. Usually, the Mean Squared Error (MSE) or the Mean Absolute Error (MAE) is selected to evaluate the model accuracy. Next, it is necessary to define the parameter space of the LSTM model, including the number of neurons, learning rate, number of iterations, batch size, etc., and use them as the search space of the particle swarm algorithm. Then, initialize the position and velocity of each particle in the particle swarm. The position of the particle represents the parameter combination of the LSTM model, and the velocity represents the moving speed of the particle in the parameter space. Next, update the particle swarm to make the particles move towards the optimal solution. The formula for updating the particle swarm is:

[0094]

[0095] xi(t + 1) = xi(t) + vi(t + 1).

[0096] where: v i represents the velocity of the i-th particle at time t; x i represents the position of the i-th particle at time t; ω represents the inertia weight; cl and c2 represent the learning factors; pbest i represents the best position in the history of the i-th particle; gbest represents the global best position.

[0097] By updating the particle swarm, a set of optimal LSTM parameter combinations can be obtained, and the LSTM model is trained using these parameters. The training process includes steps such as forward propagation, backward propagation, and parameter update. Use the trained LSTM model to predict the test data, and use the fitness function to evaluate the performance. After obtaining the predicted displacement value, calculate the deformation rate (displacement / time) and the tangent angle (the angle between the tangent of the monitoring site and the abscissa), and then the landslide warning level can be determined according to the corresponding criteria.

[0098] VI. Control and Management Module.

[0099] The control and management module is responsible for the management of monitoring stations and epoch times, and can realize the parameter configuration of monitoring points and the data statistics of monitoring points, such as Figure 8As shown in the figure; it includes: selecting a solution system, constraint mode, modifying the elevation angle, signal-to-noise ratio threshold, enabling or disabling a certain monitoring station, selecting an alarm type, downloading data within a certain time period, etc. It can also manage the battery to achieve battery protection; users can select different parameter configurations according to specific monitoring points to achieve the best monitoring results; in order to display the monitoring information of the landslide in real time, the above battery management module, data analysis module, and early warning and alarm module are all placed on the web server, and accurate location positioning and information reading can be achieved.

[0100] In specific implementation, multiple-source sensors are arranged near the landslide monitoring station, including: earth pressure gauges, rain gauges, soil temperature and humidity gauges, GNSS displacement monitoring, etc., all of which are based on the TCP protocol, to construct a landslide monitoring module to monitor the humidity data SDi, temperature data WDi, earth pressure data TYi, displacement data WYi, and rainfall data JYi of the landslide in real time; among them, the earth pressure gauge is an important sensor for measuring the internal stress change of the soil body. A vibrating wire earth pressure gauge is used to read relevant parameters, and the earth pressure data TYi is obtained through platform calculation; the rain gauge is a hydrological and meteorological instrument used to measure the rainfall in nature, and at the same time converts the rainfall into digital information output in the form of a switching quantity. A tipping bucket rain gauge is used, and the internal structure is a mechanical tipping bucket structure. The tipping bucket is triggered by the weight of the precipitation to calculate the rainfall data JYi; the soil humidity gauge uses a direct-insert type soil temperature and humidity gauge as the soil temperature and humidity measurement device for the landslide monitoring system. By measuring the dielectric constant of the soil, it can directly and stably reflect the true moisture content and temperature of various soils to obtain the humidity data SDi and temperature data WDi; the GNSS monitoring device is used to monitor the three-dimensional displacement change of the landslide body. The reference station + rover station mode is adopted. One GNSS reference station and multiple mobile monitoring stations are arranged in the landslide area. Data processing is carried out between the reference station and the rover station through RTK (real-time kinematic) technology to obtain the monitoring station location and displacement data WYi.

[0101] After the data receiving sub-module receives the above data, the obtained data is split into basic geographic data (GNSS displacement data) and disaster monitoring data (rainfall data, soil temperature and humidity data, earth pressure data), and stored in the corresponding database according to field classification, which is convenient for the subsequent data sharing sub-module to share the data of different stations; after the data is stored, when the user clicks on the corresponding web page to access, the latest data can be obtained, and the return format of the data is JSON format.

[0102] The collected data will first undergo some simple data cleaning in the data analysis module and then be visually displayed. Among them, the rainfall data JYi is represented by a bar chart and a line chart, the earth pressure data Tyi is represented by a water droplet chart, the soil temperature WDi and humidity SDi are represented by a line chart, and the displacement data WYi is represented by a 3D map and a 3D point chart, etc.

[0103] First, preprocess all the data. Accumulate the rainfall data to obtain daily data, take the daily average of the earth pressure data and soil humidity data as daily data, perform smoothing processing on the GNSS data using the moving window method, set the window size to 3, and select N as 12 to better smooth the trend term. Select data from different time periods to calculate the Pearson correlation coefficient. Select the earth pressure data and soil humidity data of the previous 1 day, and the rainfall and data of the previous 7 days as input factors. Their correlation coefficients with displacement are 0.84, -0.47, and 0.53 respectively. Normalize them for uniformity.

[0104] Then use the moving average method again to decompose the displacement into a trend term and a periodic term, and use different models for prediction; in this embodiment, the displacement of the landslide trend term component increases monotonically with time. Therefore, regard the displacement of the trend term component as a time series of a single variable, and use the least squares criterion polynomial fitting for prediction. Select different orders for polynomial fitting, and finally determine the order to be 5th order by comparing the accuracy; the polynomial fitting formula is expressed as:

[0105] y = 1.3e -9 -7.7e -7 x + 1.3e -4 x 2 -1.2e -3 x 3 -1.4x 4 -868.7x 5 .

[0106] For the periodic term, use the PSO algorithm to optimize the number of units, the number of training epochs, and the batch size of the LSTM layer. Finally, the best number of iterations of the LSTM is calculated to be 50 times, and the number of particles is 20.

[0107] Finally, the root mean square error was used to evaluate the accuracy of the displacement prediction results. The calculated root mean square error (RMSE) was 5.3425 mm. The lower this value, the higher the accuracy of the model in predicting the target variable, and the better the model performance. The calculated correlation coefficient R was 0.9687. The closer this value is to 1, the higher the correlation between the predicted value and the true value. Considering the above two indicators, it is concluded that the model has extremely high accuracy and reliability in predicting the target variable, and the predicted value of the model is very reliable, enabling the calculation of the deformation rate and tangent angle in the next step, and determining the warning level according to the criteria.

[0108] Meanwhile, various conditions of the battery are also displayed in the battery management module. The lighting-on voltage is set to 5V, the lighting-on delay is set to 5s, the overcharge voltage is set to 14.4V, the under-voltage protection is set to 11V, the under-voltage return is set to 12.5V, the working mode is energy-saving, the energy-saving starting point is 12.5V, and the energy-saving ending point is 11V to protect the battery from damage.

[0109] In the solution and alarm module, the deformation rate v and the displacement tangent angle are calculated through formulas. The deformation rate is a value that measures the speed of landslide deformation in the slope section of the monitoring station. The larger the deformation rate, the higher the landslide risk in the monitored slope section. The displacement tangent angle is a value that represents the magnitude of the displacement angle in the slope section of the monitoring station. The larger this value, the more intense the displacement in the monitored slope section and the higher the landslide risk. Compare the deformation rate with the landslide thresholds v a 、v b 、v c and compare the tangent angle with the tangent angle threshold. If the deformation rate v of the monitored slope section i is less than the landslide threshold v a , and the displacement tangent angle is less than the tangent angle threshold, then the landslide warning level of the monitored slope section is determined as a blue warning. If the deformation rate v of the monitored slope section i is greater than or equal to the landslide threshold v a , less than the landslide threshold v b , and the displacement tangent angle is less than the tangent angle threshold, then the landslide risk of the monitored slope section is determined as a yellow warning. If the deformation rate v of the monitored slope section i is greater than or equal to the landslide threshold v b , less than the landslide threshold v c , and the displacement tangent angle is greater than the tangent angle threshold and less than, then the landslide risk of the monitored slope section is determined as an orange warning. If the deformation rate v of the monitored slope section i is greater than or equal to the landslide threshold v c , and the displacement tangent angle is greater than the tangent angle threshold, then the landslide risk of the monitored slope section is determined as a red warning. The solution and alarm module sends the slope section and the warning signal to the monitoring and warning platform. After receiving the slope section and the warning signal, the monitoring and warning platform sends them to the mobile terminal of the management personnel.

[0110] In the control and management module, each monitoring point can be managed. Areas with good satellite conditions can be set with the BDS, GPS, GLONASS, and GALILEO positioning systems for the solution system, the constraint mode can be set to unconstrained, the fixed mode can be set to single-epoch fixing, the elevation angle can be set to 15°, the SNR threshold can be set to 50, etc., to achieve the best early warning effect. At the same time, different monitoring stations can be enabled, and corresponding data can be downloaded according to time periods.

[0111] The present invention uses the PSO-LSTM model to decompose displacement prediction into a trend term and a periodic term, selects different models and methods for displacement prediction respectively, and then accumulates the displacements of the two to accurately predict landslide displacement. A battery management module is provided to solve the problems of unprotected batteries and easy overvoltage and undervoltage: the battery management module can view the power generation, power consumption, voltage, current, etc. of the battery in real time, and can set the working mode, delay, energy-saving end point, etc. of the battery to protect the battery and extend its service life. An intuitive and user-friendly visualization display interface is provided: the data of this platform uses various forms such as text, line charts, bar charts, real-scene maps, 3D maps, and 3D stereograms to intuitively display early warning data and early warning results, break down professional barriers, provide concise and easy-to-read early warning information, and provide intuitive and scientific support for decision-makers. A variety of sensors are used to provide multi-dimensional data: the platform supports a variety of communication protocols, can access various monitoring devices, realize multi-source heterogeneous data, comprehensively grasp the deformation trend of landslides, and improve the accuracy of prediction. Realize real-time dynamic early warning and efficient information transmission: it can receive data in real time for prediction, and push the prediction results to the user side, realize real-time processing and analysis of data and dynamic release of early warning information, ensure the timeliness of early warning information, and make preparations for emergency response in advance.

[0112] The present invention uses the Pearson correlation coefficient to perform correlation analysis on candidate input factors, then uses the moving average method once to decompose the displacement into the trend term and the periodic term displacement, uses polynomial fitting to predict the trend term displacement, uses PSO-LSTM to predict the periodic term displacement, and finally combines the deformation rate and the tangent angle to judge the landslide warning level; Battery management and remote scheduling function: An integrated battery management module can monitor the power status of the device in real time and remotely schedule power management strategies (such as automatically protecting the power supply when the battery is low, remotely restarting the device, etc.) through the cloud platform; Multi-level warning mechanism of the solution and alarm module: Through the solution and alarm module, the landslide risk is divided into four levels: blue warning, yellow warning, orange warning, and red warning, and warning information is promptly released through methods such as text messages and emails; Modular structure and system integration: Adopting a modular design, the system is divided into multiple functional modules (such as landslide monitoring module, data transmission module, data analysis module, solution and warning module, etc.), and data interaction is carried out between modules through standardized interfaces to ensure the efficient operation and flexible expansion of the system; Real-time data visualization and interactive display: Provide a rich visual display interface, including various forms such as line charts, bar charts, 3D maps, and three-dimensional dot plots, to intuitively display monitoring data and warning information.

[0113] Through the landslide monitoring module and the data storage module, the present invention can monitor the landslide in all aspects and multi-dimensions, improve the accuracy of early warning, classify and store the data, comprehensively analyze various factors, and the positioning module can accurately obtain the location of the monitoring station. When a certain landslide becomes unstable, it can be quickly located for easy emergency response; Through the data analysis module, invalid data can be eliminated to ensure the consistency of data in time, and then the data is visually displayed to facilitate users to view in real time and grasp the overall state and movement trend of the landslide; Through the battery management module, the power consumption of the battery can be quickly mastered, parameters can be set in advance. When the battery is under-voltage or over-voltage, the battery can be protected and its lifespan can be extended. It can also avoid the problem that other devices are damaged due to unstable current and voltage of the battery. When the battery is unstable, the power supply plan can be adjusted in time to further improve the stability and adaptability of the device and the system; Through the solution and alarm module, the landslide warning level can be comprehensively evaluated. Through various data and mathematical models, according to the landslide judgment criteria, the warning level is divided into four levels: blue warning, yellow warning, orange warning, and red warning, and users are warned through various methods to facilitate quickly grasping the landslide situation and formulating specific measures; Through the control and management module, the overall device can be set to be enabled or disabled, and the use and solution of the device can be dynamically adjusted according to the actual situation of different monitoring points to achieve the purpose of optimizing the monitoring of the landslide.

[0114] The above-described embodiments merely represent several implementation manners of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all fall within the protection scope of the present invention. Therefore, the protection scope of the present invention patent shall be subject to the appended claims.

Claims

1. An expansive soil landslide monitoring and early warning system, characterized in that, Including: A landslide monitoring module, a data analysis module and a solution and warning module; The landslide monitoring module is used to collect various environmental data and displacement data of the expansive soil landslide in real time through multiple types of sensors arranged in the expansive soil landslide; The data analysis module is used to analyze the correlation between various environmental data and displacement by using the Pearson correlation coefficient, and extract the environmental data with a correlation value higher than a preset threshold; The displacement data is decomposed into a trend term displacement and a periodic term displacement by using the first-order moving average method; the trend term displacement is fitted by using the polynomial least square method to capture the long-term deformation data in the trend term displacement, and the predicted value of the trend term displacement is obtained; The PSO-LSTM model is used to fit the periodic term displacement to capture the short-term fluctuation data in the periodic term displacement, and the predicted value of the periodic term displacement is obtained; The predicted value of the trend term displacement and the predicted value of the periodic term displacement are added to obtain the predicted value of the displacement; The solution and warning module is used to obtain the deformation rate and tangent angle of the expansive soil landslide according to the environmental data and displacement data with a correlation value higher than the preset threshold, and divide the warning level of the expansive soil landslide based on the deformation rate and tangent angle; according to the predicted value of the displacement, obtain the predicted deformation rate and tangent angle of the expansive soil landslide, and issue a warning for the expansive soil landslide according to the warning level to which the predicted deformation rate and tangent angle belong.

2. The monitoring and early warning system for expansive soil landslides according to claim 1, wherein It also includes: A data storage module, which is used to receive and classify and store various environmental data and displacement data collected by the landslide monitoring module; A battery management module, which is used to monitor the battery status of the device in real time, set overcharge protection, under-voltage protection and energy-saving mode, and dynamically adjust the power supply strategy of the device; A control management module, which is used to configure the monitoring station parameters and display the monitoring data and warning information in real time through a visual interface.

3. The monitoring and early warning system for expansive soil landslides according to claim 2, wherein, The landslide monitoring module includes an environmental data acquisition sub-module and a positioning sub-module; The environmental data acquisition sub-module includes a rain gauge, a soil temperature and humidity gauge and an earth pressure gauge, which are used to collect rainfall, soil temperature and humidity and earth pressure data; The positioning sub-module uses a GNSS receiver and RTK positioning to obtain the three-dimensional displacement data of the expansive soil landslide body in real time.

4. The monitoring and early warning system for expansive soil landslides according to claim 1, characterized in that, The obtaining of the predicted value of the displacement includes: The displacement data of the expansive soil landslide is analyzed by using the first-order moving average method, and the displacement data is decomposed into a trend term displacement and a periodic term displacement; wherein, the trend term displacement represents the long-term deformation law of the landslide displacement, and the periodic term displacement represents the short-term fluctuation of the landslide displacement; The trend term displacement is fitted by using the polynomial least square method to capture the long-term deformation data in the trend term displacement, and the predicted value of the trend term displacement is obtained, which is expressed as: Among them: y represents the predicted value of the trend term; a i represents the polynomial coefficient; x i represents the time variable; The number of neurons, learning rate and number of iterations of the LSTM are optimized by the particle swarm algorithm PSO, and the PSO-LSTM model is trained by combining the mean square error as the fitness function. The trained PSO-LSTM model is used to fit the periodic term displacement to obtain the predicted value of the periodic term displacement; The predicted value of the trend term displacement and the predicted value of the periodic term displacement are added to obtain the predicted value of the displacement.

5. The monitoring and early warning system for expansive soil landslides according to claim 1, characterized in that, The division of the warning level of the expansive soil landslide includes: Calculate the deformation rate v and the tangent angle θ of the expansive soil landslide based on the environmental data and displacement data with correlation values higher than a preset threshold; the environmental data includes rainfall data, earth pressure data, velocity data, and humidity data; Compare the deformation rate v with the preset landslide threshold v a , v b , v c Compare the tangent angle θ with the preset tangent angle thresholds θ1 and θ2, and divide the warning levels; If the deformation rate v of the monitored slope section i is less than the landslide threshold v a , and the displacement tangent angle θ is less than the tangent angle threshold θ1, then the landslide warning level of the monitored slope section is determined to be a blue warning, and the corresponding monitored slope section is marked as a blue warning slope section; If the deformation rate v of the monitored slope section i is greater than or equal to the landslide threshold v a , and less than the landslide threshold v b , and the displacement tangent angle θ is less than the tangent angle threshold θ1, then the landslide risk of the monitored slope section is determined to be a yellow warning, and the corresponding monitored slope section is marked as a yellow warning slope section; If the deformation rate v of the monitored slope section i is greater than or equal to the landslide threshold v b , less than the landslide threshold vc, and the displacement tangent angle θ is less than the tangent angle threshold θ2 and greater than θ1, then the landslide risk of the monitored slope section is determined to be an orange warning, and the corresponding monitored slope section is marked as an orange warning slope section; If the deformation rate v of the monitored slope section i is greater than or equal to the landslide threshold v c , and the displacement tangent angle θ is greater than the tangent angle threshold θ2, then the landslide risk of the monitored slope section is determined to be a red warning, and the corresponding monitored slope section is marked as a red warning slope section.

6. The monitoring and early warning system for expansive soil landslides according to claim 3, characterized in that, The monitoring data and early warning information are displayed in real time through the visualization interface, including: Display the hourly rainfall and daily rainfall obtained by the rain gauge using a bar chart or a line chart; Display the sudden temperature and humidity obtained by the soil temperature and humidity sensor using a line chart; Display the earth pressure data obtained by the earth pressure gauge using a water drop chart; Display the displacement data obtained by the GNSS receiver using a three-point diagram, an error ellipse diagram, or a three-dimensional bar chart.

7. The monitoring and early warning system for expansive soil landslides according to claim 2, wherein, The management and control module supports the solution configuration of multi-satellite systems, including BDS, GPS, GLONASS, and GALILEO; Support dynamic adjustment of the elevation angle, signal-to-noise ratio threshold, and solution constraint mode of the monitoring station; support downloading historical monitoring data by time period.

Images