A highway spoil site water environment protection detection system and a detection method thereof

Abstract

The application provides a highway spoil site ring water conservation detection system and a detection method thereof, which comprise a sensing and collecting module, which is used for collecting multi-source monitoring data from slope top displacement monitoring points, slope waist sliding surface monitoring points and slope foot stress monitoring points; a data transmission module, which is in communication connection with the sensing and collecting module and is used for sending the multi-source monitoring data to a remote monitoring center; a data processing and early warning platform, which is disposed in the remote monitoring center, is in communication connection with the data transmission module, is used for receiving the multi-source monitoring data and generating comprehensive environmental situation evaluation results and early warning information of the spoil site; and a user interaction module, which is in communication connection with the data processing and early warning platform. The application can realize the collaborative collection and intelligent fusion analysis of multi-dimensional parameters such as the stability of the spoil site, soil erosion and environmental factors, and can realize the transformation from passive response to active early warning and from experience judgment to data driving.

Description

Technical Field

[0001] This invention belongs to the technical field of the interdisciplinary field of civil engineering safety monitoring and environmental protection, specifically relating to an environmental protection and water conservation monitoring system and its monitoring method for highway spoil disposal sites. Background Technology

[0002] The construction of large-scale infrastructure projects such as highways and railways inevitably generates a large amount of excavated soil. This excavated soil is usually piled up in designated spoil heaps. As artificial loose deposits, spoil heaps are structurally unstable and are highly susceptible to geological disasters and environmental problems such as slope instability and landslides, and severe soil erosion under the influence of natural forces such as continuous rainfall, groundwater infiltration, and erosion. This not only directly threatens the safety of roads, bridges, buildings, and people's lives and property below, but also leads to serious consequences such as river siltation, water pollution, and ecosystem destruction.

[0003] Currently, environmental and soil and water conservation monitoring of spoil heaps still mainly relies on regular manual inspections. Inspectors assess the condition of spoil heaps through visual inspections and simple instrument measurements. This method has several inherent drawbacks:

[0004] Inefficient and narrow coverage: Manual inspections are time-consuming and labor-intensive, making it difficult to achieve comprehensive coverage of large-scale, multi-location spoil disposal sites, resulting in monitoring blind spots.

[0005] Data is discrete and lacks continuity: patrols are intermittent, making it impossible to obtain continuous time-series data, making it difficult to capture the gradual process and key precursor information before a disaster occurs, thus missing the best opportunity for early warning.

[0006] Poor real-time performance and delayed early warning: From the discovery of a problem to the reporting and decision-making process, the process is lengthy and time-consuming, which cannot meet the needs of rapid response to sudden emergencies.

[0007] Highly subjective and lacking in quantitative analysis: Inspection results rely heavily on personal experience, making it difficult to conduct objective and accurate quantitative assessments and trend predictions.

[0008] High safety risks: Patrol personnel need to go to the slope site where there may be risks, and their personal safety is threatened.

[0009] In recent years, although some solutions have emerged that attempt to use single sensors (such as single displacement monitoring or video surveillance) for monitoring, these solutions are often functionally limited and have failed to form an effective monitoring network. They are usually just a simple supplement to manual inspections and have the following shortcomings:

[0010] Isolated monitoring indicators: Focusing only on single-dimensional information such as displacement or video, it is impossible to conduct correlation analysis between slope deformation and inducing factors (such as rainfall and soil moisture content), making it difficult to reveal the mechanism of disaster occurrence.

[0011] Weak data fusion capabilities: There is a lack of platforms for effectively cleaning, integrating, and intelligently analyzing multi-source heterogeneous data, and the value of the data has not been deeply explored.

[0012] The level of intelligence in early warning is low: early warnings are mostly based on simple single-indicator threshold judgments, resulting in a high false alarm rate and an inability to achieve early risk identification and trend prediction.

[0013] Therefore, there is an urgent need in this field for a highway spoil disposal site environmental protection and water conservation monitoring system that can achieve all-weather, multi-parameter, automated monitoring and has intelligent data analysis and early warning capabilities, in order to overcome the shortcomings of existing technologies and improve the modernization level of spoil disposal site safety and environmental supervision. Summary of the Invention

[0014] To address the problems existing in the prior art, this invention provides a water and soil conservation monitoring system and method for highway spoil heaps. The purpose is to achieve collaborative collection and intelligent fusion analysis of multi-dimensional parameters such as spoil heap stability, soil erosion, and environmental factors by constructing an integrated technical architecture of "sensing-transmission-platform-application". This enables a shift from passive response to proactive early warning and from experience-based judgment to data-driven approaches, thereby solving problems such as low efficiency, high risk, and discontinuous data in manual monitoring, as well as the single monitoring indicators and low level of intelligent early warning in existing simple technical solutions.

[0015] To achieve the above objectives, the specific solution of the present invention is as follows:

[0016] A water and soil conservation monitoring system for highway spoil disposal sites includes:

[0017] The sensing and acquisition module is used to collect multi-source monitoring data from key stability monitoring points in the spoil disposal site. These key stability monitoring points include the slope top displacement monitoring point, the slope waist sliding surface monitoring point, and the slope toe stress monitoring point.

[0018] The data transmission module is communicatively connected to the sensing and acquisition module and is used to send the multi-source monitoring data to the remote monitoring center.

[0019] The data processing and early warning platform is deployed in the remote monitoring center and is communicatively connected to the data transmission module. It is used to receive the multi-source monitoring data and generate comprehensive environmental situation assessment results and early warning information for the spoil disposal site.

[0020] The user interaction module is communicatively connected to the data processing and early warning platform.

[0021] Furthermore, the sensing and acquisition module includes:

[0022] The slope stability monitoring unit is used to monitor the deformation and displacement information of the spoil disposal site slope; the slope stability monitoring unit includes a global navigation satellite system surface displacement monitor, a microelectromechanical system tilt accelerometer, and a high-definition PTZ camera for macroscopic image monitoring;

[0023] The soil and water loss monitoring unit is used to monitor the moisture content and sediment migration of the surface soil in the spoil disposal site; the soil and water loss monitoring unit includes a soil volume water content sensor based on the frequency domain reflectance principle and a runoff sediment content sensor based on optical or ultrasonic principles.

[0024] An environmental factor monitoring unit is used to monitor external meteorological and internal water quality parameters that drive environmental changes at the spoil disposal site. The environmental factor monitoring unit includes a tipping bucket rain gauge and a multi-parameter water quality sensor. The multi-parameter water quality sensor includes at least a pH electrode for measuring acidity and alkalinity and an optical sensor for measuring turbidity.

[0025] The data collected by the slope stability monitoring unit, the soil erosion monitoring unit, and the environmental factor monitoring unit together constitute multi-source monitoring data.

[0026] The key stability monitoring points also include drainage outlet water quality monitoring points and open high-altitude meteorological monitoring points.

[0027] Furthermore, the data transmission module includes:

[0028] 4G / 5G data transmission unit, used to transmit data within the coverage area of ​​public mobile network signals;

[0029] LoRaWAN / ZigBee self-organizing network unit is used to aggregate monitoring point data to a gateway node with public network signal;

[0030] The BeiDou satellite communication unit is used to transmit data via short message communication;

[0031] The data transmission module switches its working unit based on the on-site signal strength.

[0032] Furthermore, the data processing and early warning platform includes:

[0033] The data preprocessing unit is configured to perform outlier identification and smoothing on the multi-source monitoring data; the data preprocessing unit includes a sliding window statistics module for calculating the mean and standard deviation of data within a local window, a standardization transformation module for standardizing each data point within the window, and an outlier determination module for comparing the standardized data with a preset threshold.

[0034] A multi-source data fusion unit, connected to the data preprocessing unit, is configured to perform spatiotemporal registration and feature fusion on the multi-source monitoring data after outlier processing to generate a comprehensive stability index for the spoil disposal site.

[0035] The intelligent early warning decision unit is connected to the multi-source data fusion unit, has built-in multi-level early warning thresholds, and is configured to trigger an early warning of the corresponding level based on the comprehensive stability index;

[0036] The machine learning prediction submodule is connected to the data preprocessing unit and the multi-source data fusion unit, respectively. It is configured to acquire the time series of historical monitoring data and train a long short-term memory network model to predict the displacement change trend or stability state of the spoil disposal site within a preset period in the future, and input the prediction results into the intelligent early warning decision unit.

[0037] The environmental and water conservation monitoring method for highway spoil heaps using the aforementioned system includes the following steps:

[0038] Step 1: Collect multi-source monitoring data from the slope top displacement monitoring point, the slope waist sliding surface monitoring point, and the slope toe stress monitoring point according to the preset sampling period using the sensing and acquisition module. The multi-source monitoring data includes at least displacement, tilt angle, soil moisture content, and rainfall.

[0039] Step 2: The collected multi-source monitoring data is compressed and encrypted by the data transmission module and then transmitted to the data processing and early warning platform for cleaning, noise reduction and outlier removal. Spatiotemporal registration and feature-level fusion are performed to calculate a comprehensive index reflecting the overall stability of the spoil disposal site.

[0040] Step 3: Based on the comprehensive index and key single indicator thresholds, the built-in dual early warning mechanism determines whether an early warning is triggered and the level of the early warning. When an early warning is triggered, an early warning message containing the early warning level, risk location, risk description, and recommended measures is generated and published through the user interaction module and third-party communication interface. The user interaction module displays monitoring data charts, a 3D model of the spoil disposal site, and the status of early warning processing in real time.

[0041] The next step also includes the following steps:

[0042] Step 4: Establish a historical database through the data processing and early warning platform to store the time series data of the multi-source monitoring data processed in Step 2;

[0043] Step 5: Input the latest time series data from the historical database into the pre-trained LSTM prediction model to obtain the displacement prediction sequence for the next K time steps.

[0044] Step 6: Calculate the slope of the displacement prediction value sequence. If it is greater than the preset critical slope threshold, then increase the urgency of the warning based on the current warning level.

[0045] Furthermore, the training method for the LSTM prediction model described in step 5 includes the following steps:

[0046] Step 51: Extract historical monitoring data sequences containing N time points from the historical database as a training set. The historical monitoring data should include at least the time series of displacement, tilt angle, soil moisture content, and rainfall.

[0047] Step 52: Initialize the LSTM network parameters and define the loss function as mean squared error, as follows:

[0048] (7),

[0049] In the formula, The loss function; The number of samples; For the first The actual observations at each time step The first prediction of the LSTM model The predicted value at each time step;

[0050] Step 53: Iteratively optimize the network parameters using the backpropagation algorithm and the Adam optimizer until the loss function converges or the maximum number of training epochs is reached.

[0051] Furthermore, the cleaning, noise reduction, and outlier removal described in step 2 employ a formula combining sliding window and Z-score normalization, as follows:

[0052] (3),

[0053] In the formula, These are the original data points; This represents the mean of the data within the sliding window. For standard deviation, when When the value exceeds a preset threshold (usually 3), the data point is identified as an outlier and smooth interpolation is performed.

[0054] Furthermore, the spatiotemporal registration and feature-level fusion method described in step 2 includes the following steps:

[0055] Step 21: For different types of multi-source monitoring data, perform time series alignment and spatial location matching based on their physical meaning and monitoring frequency; Step 22: Calculate the displacement change rate and cumulative effective rainfall within a specific time window. The displacement change rate is calculated using the following formula:

[0056] (4),

[0057] In the formula: For the current time The rate of displacement change within; For time Cumulative displacement obtained from GNSS monitoring at any given time; This represents the cumulative displacement at the start of the time window. The length of the time window; This represents the displacement increment within the time window;

[0058] The formula for calculating the cumulative effective rainfall is as follows:

[0059] (5),

[0060] In the formula: This represents the cumulative effective rainfall within the current time window; For the first Rainfall at each sampling time; Indicates that it is within the time window The discrete-time index within the window is used to enumerate each sampling period within the window; This is the effective rainfall coefficient (dimensionless), used to account for losses such as evaporation and surface runoff. This refers to the discrete-time index or current sampling number corresponding to the current calculation moment; This represents the cumulative time window length.

[0061] Step 23: Calculate the comprehensive stability index of the spoil heap using a weighted fusion algorithm; the formula for calculating the comprehensive stability index of the spoil heap is:

[0062] (6),

[0063] In the formula, The cumulative displacement monitored by GNSS; The change in tilt angle monitored by the tilt sensor; This refers to the soil volumetric water content. This refers to the cumulative effective rainfall. These are the preset empirical thresholds for the corresponding parameters; Let be the weight coefficients of each parameter, and satisfy . The weighting coefficients are determined by expert experience or principal component analysis. The larger the comprehensive stability index value of the spoil disposal site, the higher the stability risk of the spoil disposal site. Step 24: Normalize the comprehensive stability index of the spoil disposal site to the range of 0 to 1, and divide it into three levels: low risk 0 to 0.3, medium risk 0.3 to 0.7, and high risk 0.7 to 1.0.

[0064] 10. The method according to claim 5, wherein the warning levels in step 3 include a first-level warning, a second-level warning, and a third-level warning;

[0065] The triggering conditions for a Level 1 warning are: any key single indicator exceeds its first threshold, or the composite index is in the medium-risk range for T1 consecutive hours.

[0066] The triggering conditions for a Level 2 warning are: at least two key single indicators simultaneously exceed their first threshold, or the comprehensive index is in the high-risk range for two consecutive hours (T2 hours).

[0067] The triggering conditions for a Level 3 warning are: any key single indicator exceeds its second threshold, or the instantaneous value of the comprehensive stability index of the spoil disposal site exceeds the alarm threshold.

[0068] Warning information is sent to the designated administrator through at least two of the following methods: highlighted on the platform interface, SMS, email, and mobile app push notification.

[0069] Advantages of the present invention

[0070] 1. The highway spoil heap environmental protection and water conservation monitoring system and method of the present invention, by deploying various intelligent sensing devices, realizes real-time automatic collection of key parameters (such as displacement, rainfall, soil moisture, etc.), achieving automation and unmanned operation of monitoring work and improving supervision efficiency. It solves the problem that traditional manual inspection methods require technicians to frequently enter spoil heap areas with complex terrain and high potential risks, which is inefficient and threatens personal safety.

[0071] 2. This invention achieves real-time and continuous monitoring data through 4G / 5G / LoRa / BeiDou wireless communication networks, providing a solid data foundation for accurate early warning. It enables managers to remotely and in real-time monitor the status of spoil heaps from the monitoring center, overcoming the shortcomings of discrete and delayed data from manual inspections. Continuous, high-frequency time-series data provides indispensable data support for analyzing slope deformation patterns, soil erosion dynamics, and constructing accurate predictive models.

[0072] 3. This invention breaks through the limitations of single monitoring indicators, innovatively conducting synergistic monitoring of multi-source and heterogeneous parameters such as slope stability (displacement, tilt), soil erosion (soil moisture, sediment content), and environmental driving factors (rainfall, water quality). The systematic layout reveals the inherent connections in the disaster chain of "rainfall-infiltration-deformation-instability," providing a comprehensive information dimension for the integrated analysis of spoil disposal site health, achieving a leap from "single-point perception" to "global insight."

[0073] 4. The core advantage of this invention lies in its embedded intelligent data analysis and early warning capabilities. It not only supports threshold-based primary alarms but also performs comprehensive status assessments through multi-source data fusion algorithms (such as the Comprehensive Stability Index (CSI)) and trend prediction using machine learning models (such as LSTM). This enables the system to identify slow deformation trends that are difficult to detect with the naked eye, issuing early warnings before potential hazards occur, thus achieving true "prevention before the event," greatly enhancing the initiative of safety management, and realizing intelligent and forward-looking risk warnings, shifting from "passive response" to "proactive prevention."

[0074] 5. This invention designs a flexible networking and power supply scheme suitable for spoil disposal sites located in remote areas with poor communication conditions and difficult power supply. The data transmission module can adaptively select the optimal communication method from the public network / self-organizing network / satellite, ensuring the long-term stability and reliability of the system in harsh field environments. It has strong environmental adaptability and system reliability, ensuring stable operation under complex field conditions. Attached Figure Description

[0075] Figure 1 This is a schematic diagram of the module principle of the highway spoil disposal site environmental protection and water conservation monitoring system of the present invention.

[0076] Figure 2 This is a flowchart of the environmental protection and water conservation testing method for highway spoil disposal sites according to the present invention. Detailed Implementation

[0077] The present invention will be further explained and described below with reference to the accompanying drawings and specific embodiments. It should be noted that the specific embodiments are not intended to limit the scope of the present invention.

[0078] like Figure 1 As shown in the figure, this specific embodiment provides an environmental protection and water conservation monitoring system for highway spoil disposal sites, which includes a sensing and acquisition module, a data transmission module, a data processing and early warning platform, and a user interaction module.

[0079] 1. Specific implementation and deployment of the sensing and acquisition module

[0080] The deployment scheme of the sensing and acquisition module is crucial to the success of this embodiment. The sensing and acquisition module is used to collect multi-source monitoring data from the slope crest displacement monitoring point, the slope mid-slope sliding surface monitoring point, and the slope toe stress monitoring point; specifically, based on the topographic map of the spoil heap, geological survey report, and stability analysis, the following key stability monitoring points are identified:

[0081] Slope crest (point A): The area of ​​maximum displacement and deformation of the slope.

[0082] Slope waist (points B and C): Potential sliding surface area of ​​the slope.

[0083] Slope toe (point D): The area where slope stress is concentrated and water flows out.

[0084] Drainage outlet (point E): The final outlet for soil erosion at the spoil disposal site.

[0085] Open, high vantage point (Point F): Deploy meteorological monitoring equipment, avoiding obstruction.

[0086] The sensing and acquisition module includes a slope stability monitoring unit, a soil erosion monitoring unit, and an environmental factor monitoring unit. The specific equipment design and installation are as follows:

[0087] 1.1 Implementation of Slope Stability Monitoring Units

[0088] The slope stability monitoring unit is designed to accurately capture the macroscopic and microscopic deformation of slopes and is used to monitor the deformation and displacement information of spoil disposal site slopes. The slope stability monitoring unit includes a Global Navigation Satellite System (GNSS) surface displacement monitor, a microelectromechanical system (MEMS) tilt accelerometer, and a high-definition PTZ camera for macroscopic image monitoring.

[0089] GNSS surface displacement monitoring instrument: manufactured by Leica Geosystems, model Leica GS18T. One high-precision GNSS receiver is installed at each of the following locations: point A (slope top), point B (slope midpoint), and the off-site stable reference point (point G). Its horizontal positioning accuracy is better than ±(2.5 + 0.5×10⁻⁶). -6 ×D) mm, the elevation positioning accuracy is better than ±(5.0 + 0.5×10 -6 ×D) mm (D is the baseline length in km). The equipment is mounted on a sturdy concrete observation pier, powered by solar panels, and collects data once per second. The platform calculates and outputs a smoothed three-dimensional coordinate result every 10 minutes.

[0090] MEMS tilt accelerometer: Manufacturer: Analog Devices, Model: ADIS16488A. Deep inclinometer tubes were drilled at points B, C, and D, with a MEMS tilt sensor fixed every 0.5 meters depth inside each tube. The range is ±30°, and the accuracy is 0.01% FS. This sensor array can monitor the displacement profile of deep soil, accurately determining the location and depth of potential slip surfaces. The sensors are connected in series via cables to a data acquisition box located at the toe of the slope.

[0091] High-definition intelligent PTZ camera: Manufacturer: Hikvision, Model: DS-2DF8836IX-AELW(T5). A 2-megapixel high-definition network PTZ camera with 30x optical zoom and thermal imaging capabilities is installed at point F. Through preset positions and cruise scanning, it enables real-time video monitoring and timed image capture of the entire spoil heap, key slope areas at points A and B, and can utilize AI algorithms to identify abnormal phenomena such as slope cracks and water seepage.

[0092] 1.2. Implementation of Soil and Water Loss Monitoring Units

[0093] The soil and water conservation monitoring unit aims to quantify the soil and water conservation status of spoil heaps, and is used to monitor the moisture content and sediment migration of the topsoil. The unit includes a soil volumetric water content sensor based on the frequency domain reflectance (FDR) principle and a runoff sediment content sensor based on optical or ultrasonic principles. The specific implementation is as follows:

[0094] Soil volumetric water content sensor: Manufacturer is Campbell Scientific (USA), model CS659 soil three-parameter sensor. Based on the FDR principle, soil three-parameter sensors are buried in the shallow layer (depth 0-30cm) of the slope at points A, B, and C to simultaneously measure volumetric water content, electrical conductivity, and temperature. Water content measurement range is 0-100%, with an accuracy of ±3%. Data is collected hourly.

[0095] Runoff sediment content sensor: Manufacturer: Hach, Model: OBS-3+ Online Turbidity / Suspended Solids Sensor. Online turbidimeters based on the Optical Backscattering (OBS) principle are installed at points D (upstream of the drainage ditch at the slope toe) and E (drainage outlet). The measured turbidity values ​​are converted to sediment content using a preset empirical formula: C = k * Turbidity + b (where C is sediment content, Turbidity is turbidity, and k and b are calibration coefficients). This runoff sediment content sensor can monitor the sediment concentration in runoff in real time, directly reflecting the intensity of soil erosion.

[0096] 1.3 Implementation of Environmental Factor Monitoring Units

[0097] An environmental factor monitoring unit is used to monitor external meteorological and internal water quality parameters that drive environmental changes at the spoil heap. The environmental factor monitoring unit includes a tipping bucket rain gauge and a multi-parameter water quality sensor. The multi-parameter water quality sensor includes at least a pH electrode for measuring acidity and alkalinity and an optical sensor for measuring turbidity. Specific implementation is as follows:

[0098] Tipping bucket rain gauge: Manufacturer is Campbell Scientific (USA), model TE525. A standard 0.5mm resolution tipping bucket rain gauge is installed at point F to record the rainfall process and cumulative rainfall in real time.

[0099] Multi-parameter water quality sensor: Manufacturer is YSI (Xylem), USA, model EXO2 MultiparameterSonde multi-parameter water quality monitoring probe. A multi-parameter water quality monitor is installed at point E (drainage outlet), integrating sensors for pH, dissolved oxygen, turbidity, conductivity, etc., to monitor in real time whether the quality of the discharged water meets standards.

[0100] The data collected by the slope stability monitoring unit, soil erosion monitoring unit, and environmental factor monitoring unit together constitute multi-source monitoring data.

[0101] 2. Specific implementation of the data transmission module

[0102] The data transmission module is communicatively connected to the sensing and acquisition module. It is used to encapsulate and convert the multi-source monitoring data according to protocols, and then transmit it to the remote monitoring center via a wireless communication network. It also switches working units based on the on-site signal strength. The data transmission module includes a 4G / 5G data transmission unit, a LoRaWAN / ZigBee self-organizing network unit, and a BeiDou satellite communication unit. The 4G / 5G data transmission unit is used to transmit data within the coverage area of ​​a public mobile network. The LoRaWAN / ZigBee self-organizing network unit is used to aggregate monitoring point data to a gateway node with a public network signal. The BeiDou satellite communication unit is used to transmit data via short message communication.

[0103] Given the uneven 4G signal coverage in the spoil disposal area, this embodiment adopts a "hybrid networking" strategy.

[0104] For points E and F with good signal: industrial-grade 4G DTUs are used directly to send data to the cloud platform via the MQTT protocol.

[0105] For points A, B, C, and D with weak or no signal: a LoRaWAN self-organizing network solution is adopted. Each point is configured with a LoRa data acquisition terminal to aggregate sensor data to a LoRa gateway (e.g., the Weichuan Technology Storm5 model) deployed in the center of the site where the signal is relatively good. This gateway then uploads all aggregated data via a 4G network.

[0106] As a backup in extreme cases: An additional Beidou communication terminal is equipped at point A (the most important point). In the event of a complete public network outage, critical data (such as displacement and equipment status) can be sent once a day in the form of Beidou short messages.

[0107] 3. Construction of a data processing and early warning platform

[0108] The data processing and early warning platform in this embodiment adopts a cloud-based three-tier B / S architecture (browser / server), is developed using the Java language, and is deployed on an Alibaba Cloud ECS server with CentOS 7.9 as the operating system.

[0109] The data processing and early warning platform is deployed in the remote monitoring center and is connected to the data transmission module. The data processing and early warning platform is used to receive the multi-source monitoring data and generate comprehensive environmental situation assessment results and early warning information for the spoil disposal site.

[0110] The data processing and early warning platform includes a data preprocessing unit, a multi-source data fusion unit, an intelligent early warning decision-making unit, a user interaction module, and a machine learning prediction sub-module.

[0111] 2.1 Implementation of the Data Preprocessing Unit

[0112] The data preprocessing unit is configured to identify and smooth outliers in multi-source monitoring data. As the first processing step in the data processing and early warning platform, the data preprocessing unit receives data streams from various DTUs and gateways via an MQTT server (EMQX). Its core responsibility is to identify and smooth outliers in multi-source monitoring data.

[0113] The data preprocessing unit includes a sliding window statistics module, a standardization transformation module, and an outlier detection module. The sliding window statistics module is used to calculate the mean and standard deviation of the data within a local window. The standardization transformation module is used to standardize each data point within the window. The outlier detection module is used to compare the standardized data with a preset threshold. The processing flow is as follows:

[0114] (1) Format parsing and validation: Parse the JSON format data packet and validate the integrity and validity of the device ID, timestamp, and data fields.

[0115] (2) Outlier detection and removal: Real-time filtering is performed using a sliding window-based Z-score normalization method. Taking GNSS displacement data as an example, the sliding window size is set to 10 consecutive data points (i.e., 100 minutes) through the sliding window statistics module. The current data point X_t is Z-score normalized through the normalization transformation module, the mean μ and standard deviation σ of the data within the window are calculated, and then the Z value is calculated: Z = (X_t - μ) / σ (1).

[0116] The outlier detection module sets a threshold Z_threshold = 3. If |Z| > 3, X_t is considered an outlier (e.g., a jump caused by satellite signal loss) and is replaced with the median of the window; otherwise, the data point is retained.

[0117] (3) Data standardization: To facilitate subsequent fusion analysis, data of different dimensions are normalized to the interval [0, 1]. For displacement data, minimum-maximum normalization is used: D_norm = (D - D_min) / (D_max - D_min), where D_min and D_max are set according to historical data or experience.

[0118] 2.2 Implementation of Multi-Source Data Fusion Unit

[0119] The multi-source data fusion unit is connected to the data preprocessing unit and is configured to perform spatiotemporal registration and feature fusion on the multi-source monitoring data after outlier processing to generate a comprehensive stability index for the spoil disposal site.

[0120] This embodiment selects four key indicators: hourly displacement rate V_d (mm / h), tilt angle change Δθ, soil volumetric water content SM (%), and current hourly rainfall intensity I_r (mm / h).

[0121] The formula for calculating the Comprehensive Stability Index (CSI) of spoil heaps is as follows:

[0122] (2),

[0123] In the formula: The preset thresholds for each parameter are set as follows, based on specifications and experience: . The weights are determined using the analytic hierarchy process (AHP). Five experts were invited to score the results, a judgment matrix was constructed, and the weights were calculated as follows: (Displacement weight is the largest) And satisfy The theoretical range of CSI is [0, +∞), but in practice it is mapped to [0, 10] for display and judgment. Risk levels are set as follows: low risk [0, 3), medium risk [3, 7), and high risk [7, 10].

[0124] 2.3 Implementation of Intelligent Early Warning Decision Unit

[0125] The intelligent early warning decision unit is connected to the multi-source data fusion unit, and has built-in multi-level early warning thresholds, configured to trigger the corresponding level of early warning based on the comprehensive stability index.

[0126] The intelligent early warning decision-making unit adopts a "dual triggering" mechanism, namely, rule triggering and model triggering in parallel.

[0127] Rule triggering mechanism (based on threshold and CSI index):

[0128] Level 1 Warning (Blue):

[0129] Condition 1: Hourly rainfall intensity I_r > 15mm.

[0130] Condition 2: The CSI index is in the range of [3,5] for 2 consecutive hours.

[0131] Action: Pop up a reminder on the platform interface and send an internal message to the administrator.

[0132] Level II Warning (Yellow):

[0133] Condition 1: Hourly displacement rate V_d > 0.3 mm / h and hourly rainfall intensity I_r > 10 mm.

[0134] Condition 2: The CSI index remains in the range of [5,7] for one consecutive hour.

[0135] Action: In addition to the blue alert, send SMS notifications to the mobile phones of on-duty personnel.

[0136] Level 3 Warning (Red):

[0137] Condition 1: Hourly displacement rate V_d > 0.8 mm / h (critical threshold).

[0138] Condition 2: The instantaneous value of the CSI index is greater than or equal to 7.

[0139] Condition 3: The video AI identifies obvious longitudinal cracks on the slope.

[0140] Action: Based on the yellow alert, trigger the platform's audible and visual alarm and automatically dial the main person in charge.

[0141] 2.4: Implementation of the User Interaction Module

[0142] The user interaction module communicates with the data processing and early warning platform. The front end of the data processing and early warning platform is developed using the Vue.js framework, combined with the ECharts and Cesium 3D globe engine for visualization.

[0143] Integrated Cockpit: The homepage displays all sensor locations on a two-dimensional map or oblique photogrammetry 3D model in the form of a "single image", and uses colors to indicate their real-time status (green for normal, yellow for warning, and red for alarm).

[0144] Data charts: Clicking on any sensor will pop up a window to display its historical data curve, supporting multi-parameter comparison analysis (such as overlaying displacement curves with rainfall curves).

[0145] Early Warning Center: Centrally displays early warning information for all activities, including level, time, location, details, and processing status (pending, in progress, closed loop).

[0146] Report Management: Automatically generates daily, weekly, and monthly reports, and supports PDF export.

[0147] 2.5 Implementation of the Machine Learning Prediction Submodule

[0148] The machine learning prediction submodule is an advanced feature used for trend prediction. This example uses an LSTM model to predict displacement trends over the next 24 hours.

[0149] The machine learning prediction submodule, connected to both the data preprocessing unit and the multi-source data fusion unit, is configured to acquire time series data from historical monitoring data and train a long short-term memory (LSTM) network model to predict the displacement trend or stability state of the spoil heap within a preset future time period. The prediction results are then input to the intelligent early warning decision unit. The machine learning prediction submodule uses historical monitoring data to train the prediction model, which includes at least time series data on displacement, tilt angle, soil moisture content, and rainfall. The prediction model is used to predict the displacement trend or stability state of the spoil heap within a specific future time period (e.g., the next 24 hours). The prediction model is a long short-term memory (LSTM) network model, and its core computation process involves an input gate. Forgotten Gate Output gate and cell state The update will be implemented as follows:

[0150] (1) Data preparation: Feature data from the past 72 hours were extracted from the historical database as input X, including: displacement, cumulative rainfall, soil moisture content, and temperature. The corresponding actual displacement for the next 24 hours was used as label Y. The data was then normalized.

[0151] (2) Model construction: The LSTM model was constructed using the Keras library in Python. The model structure is as follows: one input layer (input shape is (72, 4), that is, 72 time steps and 4 features), two LSTM layers (the number of neurons is 50 and 25 respectively), one Dropout layer (dropout rate of 0.2 to prevent overfitting), and one fully connected output layer (24 neurons, corresponding to the predicted value of the next 24 hours).

[0152] (3) Model training: The loss function is mean squared error (MSE), the optimizer is Adam, and the learning rate is set to 0.001. The model is trained using data from the past year, with a batch size of 32 and 100 epochs.

[0153] (4) Prediction and Early Warning: The model runs every 6 hours. The latest 72-hour sequence is input into the model to obtain the displacement prediction sequence Y_pred for the next 24 hours. The linear regression slope K_pred of this prediction sequence is calculated. If K_pred > 0.1 (indicating a significant trend of accelerated deformation), a "trend warning" will be generated regardless of the current CSI index, indicating that the risk may be escalating and suggesting that patrols be strengthened or preliminary measures be taken.

[0154] A complete working cycle of the above system is as follows:

[0155] Data acquisition and uploading: Each sensor collects data at a preset frequency (displacement: 1Hz -> 10 minutes output; soil moisture: 1 / hour; rainfall: real-time) and uploads it to the cloud platform via 4G / LoRa network.

[0156] Data stream processing: After receiving the data, the platform immediately enters the preprocessing pipeline (parsing -> Z-score filtering -> normalization), and then stores it in the time series database InfluxDB.

[0157] Real-time calculation and judgment: The preprocessed data triggers two processes simultaneously: a) It is sent to the rule engine to calculate the CSI index and determine whether to trigger the threshold warning; b) It updates the latest value in the real-time database Redis for front-end display.

[0158] Scheduled prediction task: Every 6 hours, the data service extracts the latest 72-hour data from InfluxDB, calls the pre-trained LSTM model to make predictions, and stores the prediction results and trend analysis in the database.

[0159] Warning Issuance and Display: If a warning is triggered, the platform immediately sends an event to the message queue, and the notification service will simultaneously send various push notifications, including interface pop-ups, SMS messages, and emails. All information is centrally managed in the warning center.

[0160] Human interaction and decision-making: Administrators log in to the platform to view warning details, review videos to confirm the on-site situation, handle warnings and record measures, forming a closed management loop.

[0161] For projects with excellent communication conditions and ample budgets, all transmission modules can utilize 5G CPEs to achieve lower latency and greater bandwidth, supporting real-time streaming of high-definition video. For smaller spoil disposal sites where cost is a greater concern, the sensing layer can be simplified, deploying only core sensors such as GNSS and rain gauges. The platform can also be deployed on a local server to save on cloud service costs. These variations all fall within the scope of this invention.

[0162] like Figure 2As shown, the environmental and water conservation monitoring method for highway spoil heaps using the above system includes the following steps:

[0163] Step 1: Collect multi-source monitoring data from the slope top displacement monitoring point, the slope waist sliding surface monitoring point, and the slope toe stress monitoring point according to the preset sampling period using the sensing and acquisition module. The multi-source monitoring data includes at least displacement, tilt angle, soil moisture content, and rainfall.

[0164] Step 2: The collected multi-source monitoring data is compressed and encrypted by the data transmission module and then transmitted to the data processing and early warning platform for cleaning, noise reduction and outlier removal. Spatiotemporal registration and feature-level fusion are performed to calculate a comprehensive index reflecting the overall stability of the spoil disposal site.

[0165] The cleaning, noise reduction, and outlier removal processes employ a formula combining sliding window and Z-score normalization, as follows:

[0166] (3),

[0167] In the formula, These are the original data points; This represents the mean of the data within the sliding window. For standard deviation, when When the value exceeds a preset threshold (usually 3), the data point is identified as an outlier and smooth interpolation is performed.

[0168] The spatiotemporal registration and feature-level fusion method includes the following steps:

[0169] Step 21: For different types of multi-source monitoring data, perform time series alignment and spatial location matching based on their physical meaning and monitoring frequency; Step 22: Calculate the displacement change rate and cumulative effective rainfall within a specific time window. The displacement change rate is calculated using the following formula:

[0170] (4),

[0171] In the formula: For the current time The rate of displacement change within; For time Cumulative displacement obtained from GNSS monitoring at any given time; This represents the cumulative displacement at the start of the time window. The length of the time window; This represents the displacement increment within the time window;

[0172] The formula for calculating the cumulative effective rainfall is as follows:

[0173] (5),

[0174] In the formula: This represents the cumulative effective rainfall within the current time window. For the first Rainfall at each sampling time; Indicates that it is within the time window The discrete-time index within the window is used to enumerate each sampling period within the window; It is the effective rainfall coefficient, which is dimensionless and is used to account for loss effects such as evaporation and surface runoff. This refers to the discrete-time index or current sampling number corresponding to the current calculation moment; The cumulative time window length; Step 23: Calculate the Comprehensive Stability Index (CSI) of the spoil heap using a weighted fusion algorithm; the formula for calculating the Comprehensive Stability Index of the spoil heap is:

[0175] (6),

[0176] In the formula, The cumulative displacement monitored by GNSS; The change in tilt angle monitored by the tilt sensor; This refers to the soil volumetric water content. This refers to the cumulative effective rainfall. These are the preset empirical thresholds for the corresponding parameters; Let be the weight coefficients of each parameter, and satisfy . The weighting coefficients are determined by expert experience or principal component analysis. The larger the comprehensive stability index value of the spoil disposal site, the higher the stability risk of the spoil disposal site. Step 24: Normalize the comprehensive stability index of the spoil disposal site to the range of 0 to 1, and divide it into three levels: low risk 0 to 0.3, medium risk 0.3 to 0.7, and high risk 0.7 to 1.0.

[0177] Step 3: Based on the comprehensive index and key single indicator thresholds, the built-in dual early warning mechanism determines whether an early warning is triggered and the level of the early warning. When an early warning is triggered, an early warning message containing the early warning level, risk location, risk description and recommended measures is generated and published through the user interaction module and third-party communication interface. The user interaction module displays monitoring data charts, 3D model status map of the spoil disposal site and early warning processing status in real time.

[0178] The warning levels include Level 1, Level 2, and Level 3.

[0179] The first-level warning is blue, which is an alert level. It is triggered when any key single indicator (such as hourly rainfall intensity) exceeds its first-level threshold, or when the CSI index is in the medium-risk range for T1 consecutive hours (such as T1=2).

[0180] The second-level warning is yellow, which is an alert level. It is triggered when at least two key single indicators exceed their first-level threshold at the same time, or when the CSI index is in the high-risk range for T2 consecutive hours (e.g., T2=1).

[0181] The third-level warning is red, which is an alarm level. The triggering conditions are: when any key single indicator (such as displacement rate) exceeds its secondary threshold (critical threshold), or when the instantaneous value of the CSI index exceeds the red alarm threshold.

[0182] Warning information is sent to the designated administrator through at least two of the following methods: highlighted on the platform interface, SMS, email, and mobile app push notification.

[0183] Step 4: Establish a historical database through the data processing and early warning platform to store the time series data of the multi-source monitoring data processed in Step 2;

[0184] Step 5: Input the latest time series data from the historical database into the pre-trained LSTM prediction model to obtain the displacement prediction sequence for the next K time steps.

[0185] The training method for the LSTM prediction model includes the following steps:

[0186] Step 51: Extract historical monitoring data sequences containing N time points from the historical database as a training set. The historical monitoring data should include time series of displacement, tilt angle, soil moisture content and rainfall.

[0187] Step 52: Initialize the LSTM network parameters and define the loss function as mean squared error, as follows:

[0188] (7),

[0189] In the formula, The loss function; The number of samples; For the first The actual observations at each time step The first prediction for the LSTM model The predicted value at each time step;

[0190] Step 53: Iteratively optimize the network parameters using the backpropagation algorithm and the Adam optimizer until the loss function converges or the maximum number of training epochs is reached.

[0191] Step 6: Calculate the slope of the displacement prediction value sequence. If it is greater than the preset critical slope threshold, then increase the urgency of the warning based on the current warning level.

[0192] Assume the displacement prediction value sequence is , , ..., , ..., ,in For the predicted first The displacement prediction values ​​at each time point. The slope of the displacement prediction value sequence is calculated using the following formula:

[0193] (8),

[0194] In the formula: The predicted displacement value for the last time point in the sequence; This is the predicted displacement value for the first time point in the sequence. The slope represents the number of data points in the time series; the slope represents the rate of change in the predicted sequence, reflecting the trend of displacement prediction over time; when the slope is greater than the preset critical threshold, it means that the displacement change rate is abnormally accelerated, and the early warning system will increase the urgency of the early warning based on this.

[0195] The above embodiments are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention. For example, replacing the LSTM model with a GRU or Transformer model, or replacing LoRa communication with ZigBee communication, are all equivalent substitutions of the present invention.

Claims

1. A water and soil conservation monitoring system for highway spoil disposal sites, characterized in that, include: The sensing and acquisition module is used to collect multi-source monitoring data from key stability monitoring points in the spoil disposal site. These key stability monitoring points include the slope top displacement monitoring point, the slope waist sliding surface monitoring point, and the slope toe stress monitoring point. The data transmission module is communicatively connected to the sensing and acquisition module and is used to send the multi-source monitoring data to the remote monitoring center. The data processing and early warning platform is deployed in the remote monitoring center and is communicatively connected to the data transmission module. It is used to receive the multi-source monitoring data and generate comprehensive environmental situation assessment results and early warning information for the spoil disposal site. The user interaction module is communicatively connected to the data processing and early warning platform.

2. The system according to claim 1, characterized in that, The sensing and acquisition module includes: The slope stability monitoring unit is used to monitor the deformation and displacement information of the spoil disposal site slope; the slope stability monitoring unit includes a global navigation satellite system surface displacement monitor, a microelectromechanical system tilt accelerometer, and a high-definition PTZ camera for macroscopic image monitoring; The soil and water loss monitoring unit is used to monitor the moisture content and sediment migration of the surface soil in the spoil disposal site; the soil and water loss monitoring unit includes a soil volume water content sensor based on the frequency domain reflectance principle and a runoff sediment content sensor based on optical or ultrasonic principles. An environmental factor monitoring unit is used to monitor external meteorological and internal water quality parameters that drive environmental changes at the spoil disposal site. The environmental factor monitoring unit includes a tipping bucket rain gauge and a multi-parameter water quality sensor. The multi-parameter water quality sensor includes at least a pH electrode for measuring acidity and alkalinity and an optical sensor for measuring turbidity. The data collected by the slope stability monitoring unit, the soil erosion monitoring unit, and the environmental factor monitoring unit together constitute multi-source monitoring data. The key stability monitoring points also include drainage outlet water quality monitoring points and open high-altitude meteorological monitoring points.

3. The system according to claim 1, characterized in that, The data transmission module includes: 4G / 5G data transmission unit, used to transmit data within the coverage area of ​​public mobile network signals; LoRaWAN / ZigBee self-organizing network unit is used to aggregate monitoring point data to a gateway node with public network signal; The BeiDou satellite communication unit is used to transmit data via short message communication; The data transmission module switches its working unit based on the on-site signal strength.

4. The system according to claim 1, characterized in that, The data processing and early warning platform includes: The data preprocessing unit is configured to perform outlier identification and smoothing on the multi-source monitoring data; the data preprocessing unit includes a sliding window statistics module for calculating the mean and standard deviation of data within a local window, a standardization transformation module for standardizing each data point within the window, and an outlier determination module for comparing the standardized data with a preset threshold. A multi-source data fusion unit, connected to the data preprocessing unit, is configured to perform spatiotemporal registration and feature fusion on the multi-source monitoring data after outlier processing to generate a comprehensive stability index for the spoil disposal site. The intelligent early warning decision unit is connected to the multi-source data fusion unit, has built-in multi-level early warning thresholds, and is configured to trigger an early warning of the corresponding level based on the comprehensive stability index; The machine learning prediction submodule is connected to the data preprocessing unit and the multi-source data fusion unit, respectively. It is configured to acquire the time series of historical monitoring data and train a long short-term memory network model to predict the displacement change trend or stability state of the spoil disposal site within a preset period in the future, and input the prediction results into the intelligent early warning decision unit.

5. A method for detecting environmental and water conservation at highway spoil heaps using the system described in any one of claims 1 to 4, characterized in that, Includes the following steps: Step 1: Collect multi-source monitoring data from the slope top displacement monitoring point, the slope waist sliding surface monitoring point, and the slope toe stress monitoring point according to the preset sampling period using the sensing and acquisition module. The multi-source monitoring data includes at least displacement, tilt angle, soil moisture content, and rainfall. Step 2: The collected multi-source monitoring data is compressed and encrypted by the data transmission module and then transmitted to the data processing and early warning platform for cleaning, noise reduction and outlier removal. Spatiotemporal registration and feature-level fusion are performed to calculate a comprehensive index reflecting the overall stability of the spoil disposal site. Step 3: Based on the comprehensive index and key single indicator thresholds, the built-in dual early warning mechanism determines whether an early warning is triggered and the level of the early warning. When an early warning is triggered, an early warning message containing the early warning level, risk location, risk description, and recommended measures is generated and published through the user interaction module and third-party communication interface. The user interaction module displays monitoring data charts, a 3D model of the spoil disposal site, and the status of early warning processing in real time.

6. The method according to claim 5, characterized in that, It also includes the following steps: Step 4: Establish a historical database through the data processing and early warning platform to store the time series data of the multi-source monitoring data processed in Step 2; Step 5: Input the latest time series data from the historical database into the pre-trained LSTM prediction model to obtain the displacement prediction sequence for the next K time steps. Step 6: Calculate the slope of the displacement prediction value sequence. If it is greater than the preset critical slope threshold, then increase the urgency of the warning based on the current warning level.

7. The method according to claim 6, characterized in that, The training method for the LSTM prediction model described in step 5 includes the following steps: Step 51: Extract historical monitoring data sequences containing N time points from the historical database as a training set. The historical monitoring data should include at least the time series of displacement, tilt angle, soil moisture content, and rainfall. Step 52: Initialize the LSTM network parameters and define the loss function as mean squared error, as follows: （7）， In the formula, The loss function; The number of samples; For the first The actual observations at each time step The first prediction for the LSTM model The predicted value at each time step; Step 53: Iteratively optimize the network parameters using the backpropagation algorithm and the Adam optimizer until the loss function converges or the maximum number of training epochs is reached.

8. The method according to claim 5, characterized in that, Step 2, which involves cleaning, denoising, and outlier removal, uses a formula combining sliding window and Z-score normalization, as follows: （3）， In the formula, These are the original data points; This represents the mean of the data within the sliding window. For standard deviation, when When the value exceeds a preset threshold (usually 3), the data point is identified as an outlier and smooth interpolation is performed.

9. The method according to claim 5, characterized in that, The spatiotemporal registration and feature-level fusion method described in step 2 includes the following steps: Step 21: For different types of multi-source monitoring data, perform time series alignment and spatial location matching based on their physical meaning and monitoring frequency; Step 22: Calculate the displacement change rate and cumulative effective rainfall within a specific time window. The displacement change rate is calculated using the following formula: （4）， In the formula: For the current time The rate of displacement change within; For time Cumulative displacement obtained from GNSS monitoring at any given time; This represents the cumulative displacement at the start of the time window. The length of the time window; This represents the displacement increment within the time window; The formula for calculating the cumulative effective rainfall is as follows: （5）， In the formula: This represents the cumulative effective rainfall within the current time window; For the first Rainfall at each sampling time; Indicates that it is within the time window The discrete-time index within the window is used to enumerate each sampling period within the window; This is the effective rainfall coefficient, which is dimensionless and used to account for losses such as evaporation and surface runoff. This refers to the discrete-time index or current sampling number corresponding to the current calculation moment; This is the cumulative time window length; Step 23: Calculate the comprehensive stability index of the spoil heap using a weighted fusion algorithm; the formula for calculating the comprehensive stability index of the spoil heap is: （6）， In the formula, The cumulative displacement monitored by GNSS; The change in tilt angle monitored by the tilt sensor; This refers to the soil volumetric water content. This refers to the cumulative effective rainfall. These are the preset empirical thresholds for the corresponding parameters; Let be the weight coefficients of each parameter, and satisfy . The weighting coefficients are determined by expert experience or principal component analysis. The larger the comprehensive stability index value of the spoil disposal site, the higher the stability risk of the spoil disposal site. Step 24: Normalize the comprehensive stability index of the spoil disposal site to the range of 0 to 1, and divide it into three levels: low risk 0 to 0.3, medium risk 0.3 to 0.7, and high risk 0.7 to 1.

0.

10. The method according to claim 5, characterized in that, The warning levels in step 3 include Level 1, Level 2, and Level 3 warnings; The triggering conditions for a Level 1 warning are: any key single indicator exceeds its first threshold, or the composite index is in the medium-risk range for T1 consecutive hours. The triggering conditions for a Level 2 warning are: at least two key single indicators simultaneously exceed their first threshold, or the comprehensive index is in the high-risk range for two consecutive hours (T2 hours). The triggering conditions for a Level 3 warning are: any key single indicator exceeds its second threshold, or the instantaneous value of the comprehensive stability index of the spoil disposal site exceeds the alarm threshold. Warning information is sent to the designated administrator through at least two of the following methods: highlighted on the platform interface, SMS, email, and mobile app push notification.

Images