Soil heavy metal pollution identification system

Abstract

The invention relates to the technical field of soil pollution identification, and discloses a soil heavy metal pollution identification system. A multispectral remote sensing sensing module of the system obtains surface reflectance data and soil in-situ spectral data through a satellite load and a vehicle-mounted mobile platform respectively, and the surface reflectance data and the soil in-situ spectral data are processed by a heterogeneous data fusion gateway to generate multiband spectral response signals. In the pollution risk assessment module, a spatial distribution analysis unit outputs a heavy metal spatial distribution map, a migration risk prediction unit generates a pollution migration probability cloud map in combination with meteorological and hydrological data, and a pollution threshold defining unit outputs a soil remediation safety threshold. In the treatment decision execution module, an in-situ remediation execution unit adjusts passivator injection parameters, a pollution source management and control unit regulates pollution source blocking equipment and collects monitoring signals, and a three-dimensional dynamic early warning platform generates a comprehensive pollution risk index. According to the system, the cooperative operation of soil heavy metal pollution identification, evaluation and treatment is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of soil pollution identification, and in particular to a soil heavy metal pollution identification system. Background Art

[0002] There are numerous limitations to identifying and assessing heavy metal contamination in soil. Traditional soil sampling and analysis methods rely on manual collection of soil samples followed by laboratory testing to obtain contamination data. This approach is not only time-consuming and labor-intensive, but also limited in sampling locations. This makes it difficult to fully reflect the contamination status of large areas and can easily miss areas of severe contamination. In remote sensing monitoring, existing technologies often rely on single satellite remote sensing data. While this allows for large-scale monitoring, it is limited by spatial and spectral resolution, making it difficult to accurately identify the types and concentrations of heavy metals in soil. Furthermore, ground-based spectral data is not well integrated with satellite remote sensing data. The two data sets are often independent and unable to effectively complement each other, resulting in inefficient utilization of spectral information. In pollution risk assessment, most methods offer only a simple description of the spatial distribution of heavy metals, lacking in-depth analysis of their migration patterns, making it difficult to predict pollution spread trends. Furthermore, pollution thresholds are often defined based on empirical values ​​or a single factor, failing to dynamically adjust them based on multi-source spectral data, compromising their accuracy and applicability. In terms of governance decision-making, the existing system often operates independently of remediation execution and pollution source control, lacking a coordinated mechanism. Adjustments to remediation measures fail to promptly incorporate real-time data on pollution migration, and pollution source control struggles to precisely adjust to the actual thresholds for soil remediation. This results in suboptimal governance outcomes and makes it difficult to establish a closed-loop pollution governance system. Summary of the Invention

[0003] The purpose of the present invention is to provide a soil heavy metal pollution identification system to solve the problems raised in the above background technology.

[0004] To achieve the above objectives, the present invention provides a soil heavy metal pollution identification system, the system comprising: A multispectral remote sensing perception module is configured with a high-resolution imaging unit and a ground spectrum acquisition unit. The high-resolution imaging unit acquires surface reflectance data via a satellite payload, and the ground spectrum acquisition unit collects in-situ soil spectrum data using a vehicle-mounted mobile platform. The multispectral remote sensing perception module is integrated with a heterogeneous data fusion gateway, which receives surface reflectance data and in-situ soil spectrum data and generates a multi-band spectral response signal. The pollution risk assessment module includes a spatial distribution analysis unit, a migration risk prediction unit, and a pollution threshold definition unit. The spatial distribution analysis unit processes the multi-band spectral response signal and outputs a heavy metal spatial distribution map. The migration risk prediction unit combines the heavy metal spatial distribution map with meteorological and hydrological data to generate a pollution migration probability cloud map. The pollution threshold definition unit defines ecological safety constraints based on the multi-band spectral response signal and outputs a soil remediation safety threshold. The governance decision-making execution module includes an in-situ remediation execution unit, a pollution source control unit and a three-dimensional dynamic early warning platform. The in-situ remediation execution unit receives the pollution migration probability cloud map and adjusts the passivation agent injection parameters. The pollution source control unit regulates the operating status of the pollution source blocking equipment based on the soil remediation safety threshold and the multi-band spectral response signal, and simultaneously collects the real-time monitoring signal of soil pore water. The three-dimensional dynamic early warning platform integrates the heavy metal spatial distribution map and the pollution migration probability cloud map to generate a comprehensive pollution risk index.

[0005] Preferably, the multispectral remote sensing perception module also includes an unmanned aerial vehicle near-ground detection unit and a groundwater quality sensing network. The unmanned aerial vehicle near-ground detection unit obtains vegetation stress characteristic data through a multi-rotor flight platform equipped with a hyperspectral sensor. The groundwater quality sensing network is deployed in the monitoring well of the contaminated site and collects groundwater ion concentration data. The heterogeneous data fusion gateway receives the vegetation stress characteristic data and the groundwater ion concentration data and performs spatiotemporal alignment processing to form a multi-band spectral response signal containing spectral characteristic signals, vegetation response signals and ion diffusion signals. The multi-band spectral response signal is specifically divided into spatially resolved spectral signals, migration prediction spectral signals and threshold-defined spectral signals.

[0006] Preferably, the spatial distribution analysis unit uses a convolutional neural network to construct a geostatistical inversion model, and the spatial analysis spectral signal includes visible light-near infrared band reflectance data, short-wave infrared absorption characteristic data and historical pollution census data. The geostatistical inversion model extracts spectral spatial features through a deep separable convolution layer, and inputs the visible light-near infrared band reflectance data, short-wave infrared absorption characteristic data and historical pollution census data into a three-dimensional convolution kernel to generate a heavy metal spatial distribution map, and at the same time, fuses the vegetation response signal and the ion diffusion signal through a feature pyramid network to output the pollution source tracing result.

[0007] Preferably, the migration risk prediction unit includes a pollution migration model and a Monte Carlo simulator, the migration prediction spectral signal includes soil texture parameters, groundwater flow rate data and rainfall erosion intensity data, the pollution migration model processes the migration prediction spectral signal by coupling the convection-diffusion equation and the adsorption-desorption kinetic equation, and outputs the heavy metal migration flux prediction value, the Monte Carlo simulator performs thousands of random sampling operations based on the heavy metal migration flux prediction value to generate a pollution migration probability cloud map, and the pollution migration probability cloud map is fed back to the geostatistical inversion model to update the pollution source tracing results.

[0008] Preferably, the pollution threshold definition unit processes the threshold definition spectral signal through an ecotoxicological constraint model, and the ecotoxicological constraint model is integrated with a species sensitivity distribution analysis module. The species sensitivity distribution analysis module calls the biological half-lethal concentration data in the local biological database to define the ecological safety constraint conditions, outputs the soil remediation safety threshold, and simultaneously adopts the model predictive control algorithm to dynamically adjust the soil remediation safety threshold to not be lower than the statutory limit standard.

[0009] Preferably, the in-situ remediation execution unit is configured with a pollution migration model and a passivator optimization library. The pollution migration model constructs a three-dimensional diffusion equation based on Fick's second law. The input parameters of the three-dimensional diffusion equation include soil redox potential data, organic matter content data and the pollution migration probability cloud map, and outputs an optimal passivator dosage signal. The passivator optimization library stores the spectral response characteristic data of historical remediation materials, and generates a passivator type selection instruction by matching the spectral response characteristic data with the optimal passivator dosage signal.

[0010] Preferably, the pollution source control unit includes a pollution flux monitoring submodule and a blocking device regulation submodule. The pollution flux monitoring submodule receives the real-time monitoring signal of soil pore water and the spatial distribution map of heavy metals, and calculates the heavy metal leakage flux in the pollution source area. The blocking device regulation submodule generates a vertical barrier wall depth adjustment instruction based on the difference between the heavy metal leakage flux and the soil remediation safety threshold. The vertical barrier wall depth adjustment instruction is transmitted to the three-dimensional diffusion equation to update the optimal passivator dosage signal.

[0011] Preferably, the three-dimensional dynamic early warning platform is deployed with a pollution heat map generation engine and a remediation scheme deduction module. The pollution heat map generation engine renders the heavy metal spatial distribution map into a three-dimensional pollution heat map through the Kriging interpolation algorithm. The remediation scheme deduction module calls the historical remediation case library based on the comprehensive pollution risk index to generate a multi-scheme remediation path. The multi-scheme remediation path is input into the in-situ remediation execution unit to trigger the dynamic adjustment of the passivation agent parameters.

[0012] Preferably, the three-dimensional dynamic early warning platform implements a multi-level risk response mechanism. When the comprehensive pollution risk index reaches the primary warning threshold, the pollution source area production deceleration instruction is initiated. When the comprehensive pollution risk index reaches the intermediate warning threshold, the contaminated site isolation protection instruction is activated and the in-situ remediation execution unit is triggered to urgently add a passivating agent. When the comprehensive pollution risk index reaches the highest warning threshold, an emergency site closure signal is sent to the governance decision execution module.

[0013] Preferably, the in-situ repair execution unit specifically includes an intelligent grouting system and a passivator reaction monitoring device. The intelligent grouting system analyzes the optimal passivator dosage signal and dynamically adjusts the grouting pressure and grouting flow rate. The passivator reaction monitoring device uses laser induced breakdown spectroscopy technology to detect the heavy metal chelation rate of the passivator reaction product in real time to generate a passivator efficiency feedback signal. The passivator efficiency feedback signal is transmitted to the pollution migration model to update the three-dimensional diffusion equation parameters.

[0014] Compared with the prior art, the present invention has the following beneficial effects: The system integrates satellite remote sensing and ground-based spectral data through a multispectral remote sensing module. Surface reflectance data acquired by a high-resolution imaging unit and in-situ soil spectral data collected by a vehicle-mounted mobile platform are processed by a heterogeneous data fusion gateway to generate a multi-band spectral response signal. This integrates the strengths of spectral data from different sources, providing a more comprehensive and enriched spectral profile of heavy metal contamination in soil.

[0015] In the pollution risk assessment module, the spatial distribution analysis unit processes multi-band spectral response signals to produce a clear spatial distribution map of heavy metals, visually demonstrating the spatial pattern of pollution. The migration risk prediction unit, combined with meteorological and hydrological data, generates a pollution migration probability cloud map that clearly reflects the migration trends of heavy metals in the soil. The pollution threshold definition unit defines ecological safety constraints based on multi-band spectral response signals, ensuring that the output soil remediation safety thresholds are more closely aligned with actual soil environmental conditions and better adapted to the pollution characteristics of different regions. The governance decision-making and execution module combines in-situ remediation, pollution source control, and dynamic early warning to form a complete governance chain. The in-situ remediation execution unit adjusts passivation agent injection parameters based on the pollution migration probability cloud map, allowing remediation measures to flexibly adapt to pollution migration. The pollution source control unit regulates equipment operating status based on soil remediation safety thresholds and multi-band spectral response signals, while simultaneously collecting soil pore water monitoring signals for more targeted pollution source blocking. The three-dimensional dynamic early warning platform integrates multiple types of maps to generate a comprehensive pollution risk index, which comprehensively presents the pollution risk situation and facilitates timely monitoring of the overall pollution situation.

[0016] Data is transmitted smoothly between modules, forming a complete process from data perception, risk assessment to decision-making execution, realizing the integrated operation of soil heavy metal pollution identification, assessment and treatment, and improving the comprehensive management and control capabilities of soil heavy metal pollution. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a working principle diagram of the soil heavy metal pollution identification system of the present invention; Figure 2 This is the workflow diagram of the multispectral remote sensing perception module; Figure 3 This is the workflow diagram of the migration risk prediction unit; Figure 4 This is the workflow diagram of the pollution source control unit. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0019] See also Figure 1 The present invention provides a soil heavy metal pollution identification system, the system comprising: Accurate pollution identification and remediation are achieved through closed-loop control of remote sensing perception, risk assessment, and governance execution. The system hardware consists of a satellite payload, a vehicle-mounted mobile platform, an underground sensor network, and remediation equipment. The software algorithm includes a heterogeneous data fusion engine, a deep learning model, and a dynamic early warning mechanism. When the system is in operation, the high-resolution imaging unit obtains surface reflectance data in the 400-2500nm band through a satellite multispectral sensor. The ground spectrum acquisition unit uses a vehicle-mounted ASDFieldSpec spectrometer to collect in-situ soil spectra with a resolution of 1nm. The two are radiometrically corrected and geometrically aligned through a heterogeneous data fusion gateway to generate a multi-band spectral response signal containing visible light, near-infrared, and short-wave infrared characteristics.

[0020] Example 1: See Figure 2, involving the detailed construction and operation mechanism of the multispectral remote sensing perception module, focusing on the collaborative workflow of the UAV near-ground detection unit, the groundwater quality sensor network and the heterogeneous data fusion gateway. The system hardware configuration uses the DJI M300RTK six-rotor UAV platform as the air carrier. The platform is equipped with a dual-frequency GNSS receiver and an RTK positioning module, which can achieve centimeter-level positioning accuracy. The flight control system presets a gridded route, the flight altitude is stably maintained at 100 meters above the ground, and the flight speed is controlled at 3 meters per second to ensure that the hyperspectral sensor obtains stable data collection conditions. The HySpexVNIR-1800 imaging spectrometer carried by the drone has an operating band covering the range of 400-1000nm, a spectral resolution of 3.5nm, and a spatial resolution of 5 cm. The sensor has a built-in three-axis stabilized gimbal, which automatically compensates for image offsets caused by posture changes during flight.

[0021] The groundwater quality monitoring network consists of 12 monitoring wells distributed around the contaminated site. The depth of the wells is determined according to the direction of groundwater flow, with 3 wells arranged upstream, 6 wells arranged downstream, and 3 wells arranged laterally. Each monitoring well is equipped with a YSIEXO2 multi-parameter water quality sensor. The sensor probe contains six measurement channels, which respectively detect pH value, dissolved oxygen, conductivity, temperature, redox potential and turbidity. The concentration of heavy metal ions is detected using an independent voltammetric analysis module, and the content levels of elements such as Cd, Pb, and As are determined by anodic stripping voltammetry. Multi-level sampling ports are set up in the monitoring wells, and sensor nodes are arranged at depths of 5 meters, 10 meters, and 15 meters to form a three-dimensional monitoring network. All sensor nodes are connected to the wellhead data collector via an RS-485 bus. The collector has a built-in 4G communication module to realize remote data transmission.

[0022] The heterogeneous data fusion gateway is deployed on an edge computing server equipped with an Intel Xeon E5 processor and 64GB of memory, running the Ubuntu operating system. Four independent data receiving channels are configured: a satellite data channel receives Landsat 9 surface reflectance products updated daily at 12:00 PM via FTP; a vehicle-mounted spectral data channel uses TCP / IP to transmit real-time measurements from the ASD FieldSpec spectrometer; a drone data channel receives hyperspectral cube data transmitted back from flight missions via WiFi6; and a groundwater data channel obtains real-time monitoring values ​​from the sensor network via MQTT. Spatiotemporal alignment processing uses a PostgreSQL database to store all raw data and establishes a spatiotemporal index table to record the acquisition time and spatial coordinates of each data point.

[0023] During data preprocessing, satellite imagery data undergoes radiometric calibration and atmospheric correction, converting digital quantized values ​​into surface reflectance. Vehicle-mounted spectral data undergoes smoothing filtering to eliminate instrument noise and reflectance calibration using a standard whiteboard. UAV hyperspectral data undergoes geometric correction to eliminate lens distortion and is then stitched together to create a continuous orthophoto. Groundwater data undergoes quality control to remove outliers that exceed the sensor's range. After preprocessing, the data is uniformly converted to the UTM coordinate system, with timestamps set to UTC.

[0024] The spatiotemporal fusion algorithm utilizes a three-tiered processing architecture. The first tier implements spatial registration, using satellite imagery as a benchmark and aligning other data sources to the same spatial reference frame through feature point matching. The second tier performs temporal interpolation, using cubic spline interpolation to generate synchronized time series for data sources with different sampling frequencies. The third tier performs data fusion, using principal component analysis to reduce the dimensionality of spectral data and fuse hyperspectral and multispectral data into a data cube containing 20 characteristic bands. Kriging spatial interpolation is also performed on water quality data to generate a continuously distributed groundwater quality parameter field.

[0025] The generation process of multi-band spectral response signals involves two stages: feature extraction and signal classification. The feature extraction module identifies three key features from the fused data cube: vegetation green peak reflectance at 550 nm as an indicator of vegetation stress; clay mineral absorption depth at 2200 nm as an indicator of soil contamination; and iron oxide reflectance at 650 nm as a marker of groundwater contamination diffusion. The signal classifier, using a random forest algorithm, divides the input data stream into three output categories based on feature importance scores: spatially resolved spectral signals primarily encompass surface reflectance characteristics and soil property parameters; migration prediction spectral signals focus on groundwater quality changes and permeability coefficients; and threshold-defined spectral signals integrate biotoxicity data and environmental standard limits.

[0026] The implementation of the depthwise separable convolutional network is based on the TensorFlow framework. The network input layer is set as a three-dimensional tensor of 256×256×20, corresponding to the spatial distribution of 20 characteristic bands. The first convolutional layer uses 32 3×3×5 convolution kernels to extract spatial features, each of which slides independently in three dimensions. The batch normalization layer standardizes each feature channel, and the ReLU activation function introduces nonlinear transformations. Subsequent network layers gradually increase the number of convolution kernels to 128, while the spatial resolution is reduced through max pooling layers. Shortwave infrared absorption feature data is fused with the main network via skip connections in the fourth convolutional layer. The historical pollution census data is converted into 64-dimensional vectors by an embedding layer and connected to the fully connected layer.

[0027] The feature pyramid network constructs a top-down feature transfer path, and the high-resolution features extracted by the bottom network are integrated with the semantic features of the high-level network step by step. The vegetation response signal processing branch uses a two-layer cascaded LSTM network to analyze the temporal variation characteristics of NDVI. The ion diffusion signal processing branch constructs a graph convolutional network, and establishes a topological structure with monitoring wells as nodes and groundwater flow paths as edges. The final output of the heavy metal spatial distribution map is stored in GeoTIFF format, which contains five pollution level classification results and the spatial concentration distribution of three heavy metals. The pollution source tracing algorithm is based on the back-propagation particle tracking model, combining groundwater flow field data and pollutant diffusion coefficients to calculate the possible pollution source locations, and marks the high-probability pollution source areas with red polygons in the map.

[0028] Example 2: See Figure 3 , involving a collaborative computing system for pollution migration risk prediction and ecological safety threshold definition, focusing on describing a dynamic assessment method that combines physical process modeling with stochastic simulation. The migration risk prediction unit is deployed on a high-performance server equipped with an NVIDIA Tesla V100 computing card, running a numerical simulation environment based on multi-physics field coupling. During the system initialization phase, three types of basic data need to be loaded: soil texture parameters are measured using a Malvern laser particle size analyzer, and after the samples are sieved and dispersed, the measurement results are converted to the USDA soil classification standard; groundwater flow rate data comes from a distributed fiber optic sensing system, and the temperature field changes monitored by the DTS are inverted using the thermal pulse method to obtain the flow rate distribution; rainfall erosion intensity data comes from the precipitation intensity observations of the millimeter-wave radar at the meteorological station, and the spatial distribution of ground rainfall is obtained through ZR relationship conversion.

[0029] The pollution migration model adopts a modular architecture and includes six main functional components. The solute transport component implements the numerical solution of the convection-diffusion equation. It uses an unstructured tetrahedral grid for spatial discretization to accommodate complex geological conditions, and the Crank-Nicolson format for time advancement to ensure computational stability. The adsorption-desorption component integrates multiple constitutive relationships, including the linear partition model, the Freundlich isotherm, and the Langmuir equation, automatically selecting the appropriate model based on the type of pollutant. The groundwater flow component solves the modified Richards equation to handle coupled flow in the unsaturated and saturated zones. The chemical reaction component includes basic reaction types such as redox, precipitation-dissolution, and surface complexation, and describes multicomponent interactions using a stoichiometric matrix. The parameter estimation component uses the adjoint state method to automatically calibrate model parameters, reducing errors caused by initial condition uncertainty. The visualization component generates dynamic simulation results, displaying the spatiotemporal evolution of pollutant concentrations as color contour maps.

[0030] Built on the R language platform, the Monte Carlo simulator consists of three core modules: a parameter perturbation generator, a parallel task scheduler, and a result statistical analyzer. The parameter perturbation generator uses Latin hypercube sampling to perform spatially filling sampling of 12 key input parameters, including porosity, permeability, and distribution coefficient. Each parameter is set within a reasonable physical range, and the number of sampling points is set to 1000 to ensure probabilistic coverage. The parallel task scheduler allocates computing resources through the SLURM job management system, and each sampling point is submitted to the computing cluster as an independent task. The result statistical analyzer collects the result files of all simulation runs and calculates pollutant concentration statistics at each spatial location, including mean, standard deviation, and percentile.

[0031] The generation process of the pollution migration probability cloud map includes four processing steps. The first step is to implement data spatialization, interpolate the concentration field output by the simulation to a regular grid, and set the grid resolution to 10 meters to match the scale of remote sensing data. The second step is to perform probability calculation, count the number of simulations that exceed the specified threshold for each grid cell, and convert it into a probability value. The third step is to implement smoothing, using anisotropic Gaussian filters to eliminate numerical noise and maintain the spatial continuity of the main pollution plumes. The fourth step is to perform coordinate conversion, converting the calculation results from the local model coordinate system to the WGS84 geographic coordinate system, so as to facilitate overlay display with other spatial data. The final cloud map uses a graded coloring scheme, with warm colors representing high-risk areas and cold colors representing low-risk areas, and is displayed in a semi-transparent overlay on the three-dimensional geographic information system base map.

[0032] The architectural design of the ecotoxicological constraint model follows the standard process of species sensitivity distribution analysis. The local biological database stores strictly screened toxicity data, covering 112 species such as fish, crustaceans, algae and soil invertebrates. Quality control is implemented in the data preprocessing stage to eliminate data records with unclear experimental conditions or non-standard statistical methods. For multiple groups of test data for the same pollutant, the arithmetic mean method is used for integration. The model fitting process uses the maximum likelihood estimation method to solve the log-logistic distribution parameters, and the bootstrap resampling technique is used to evaluate parameter uncertainty. The hazard concentration calculation module implements four statistical extrapolation methods, including HC5 calculation, confidence interval estimation, model goodness of fit test and distribution function visualization.

[0033] The operational process of the Species Sensitivity Distribution Analysis module consists of four stages: data import, model fitting, result verification, and report generation. The data import stage supports input files in CSV and Excel formats, automatically identifying concentration units and test species information. The model fitting stage offers a choice of five distribution functions, with the log-logistic distribution used by default. The result verification stage uses residual analysis and QQ plot tests to assess model goodness of fit. The report generation stage outputs a PDF document containing a fitted curve graph, a parameter estimate table, and statistical test results. Data integrity checkpoints are set during the analysis process, and the system automatically switches to safety threshold calculation mode when fewer than 10 input data sets are present.

[0034] The implementation of the model predictive control algorithm is based on the Python control library and includes three components: a state observer, a dynamic model, and an optimization solver. The state observer monitors the calculation process of the soil remediation safety threshold in real time and records the historical trajectory of each parameter update. The dynamic model uses a first-order lag link to describe the inertial characteristics of threshold changes and sets a reasonable time constant to reflect the ecosystem response delay. The optimization solver constructs a constrained nonlinear programming problem, using the statutory limit standards as hard constraints and iteratively solving the optimal threshold through the interior point method. A double-check mechanism is set during the algorithm operation, and the manual review process is automatically triggered when the calculated value approaches the limit standard.

[0035] The dynamic threshold adjustment process is based on a feedforward-feedback composite control strategy. The feedforward phase predicts the threshold adjustment direction based on changes in pollution source intensity, while the feedback phase eliminates system deviations through error integration. Control parameters are tuned using the Ziegler-Nichols method, with trial-and-error fine-tuning achieving stable control results. The adjustment command generation module outputs standardized control signals, which are transmitted to the remediation equipment actuators via the OPC UA protocol. The system sets a threshold change rate limit to prevent large fluctuations in a short period of time from impacting remediation effectiveness. Historical threshold data is stored in a time series database, supporting queries by time range and trend analysis.

[0036] The output format of soil remediation safety thresholds uses the JSON standard and includes fields such as threshold value, effective date, applicable area, and review flag. The threshold application module implements spatially differentiated management, dividing risk areas into different zones based on the spatial distribution map of heavy metals and setting tiered thresholds for each. When monitoring data exceeds the current threshold, the system automatically triggers an early warning signal and logs the violation. The threshold review mechanism establishes a regular assessment cycle, recalculates safety thresholds based on the latest monitoring data, and maintains the timeliness of risk control measures. All threshold changes are recorded on the blockchain evidence storage platform to ensure that data cannot be tampered with and operations are traceable.

[0037] Example 3: See Figure 4, focusing on the collaborative working mechanism between the in-situ remediation execution unit and the pollution source control unit, it focuses on the technical implementation path of the optimization of passivation agent addition and the regulation of vertical barrier walls. The three-dimensional diffusion equation of the in-situ remediation execution unit is constructed based on the solute transport theory in heterogeneous porous media, and the finite volume method is used for spatial discretization. The soil redox potential monitoring network consists of 48 platinum electrodes arranged in a 20m x 20m grid. Three depth layers (0.5m, 1.5m and 3.0m) are set at each measuring point, and the measurement value is recorded every 15 minutes by a multi-channel data acquisition instrument. The organic matter content is determined by the potassium dichromate oxidation method. The sample pretreatment process includes grinding and screening and hydrochloric acid to remove inorganic carbon. The measurement results are used to generate a spatial distribution field through an interpolation algorithm.

[0038] The construction process of the passivation agent optimization library includes three main steps: material characterization, spectral archiving and performance evaluation. 23 types of repair materials were tested by X-ray diffraction analysis and Fourier transform infrared spectroscopy to establish a database of crystal structure and functional group characteristics. An independent spectral response characteristic record is set for each material, including the reflectivity curve, characteristic absorption peak position and relative intensity parameters in the 400-2500nm band. The material performance data comes from standard batch experiments, recording the immobilization efficiency of heavy metals such as Cd, Pb, and As under different environmental conditions. The database adopts a hierarchical storage structure, with the top layer classified by material type, the middle layer indexed by pollutant type, and the bottom layer storing specific experimental conditions and treatment effect data sets.

[0039] The control logic of the intelligent grouting system is based on the concentration gradient field output by the diffusion equation, dividing the repair area into multiple control units. The grouting pressure is regulated using a graded control strategy, with the base pressure set at 0.3 MPa. This pressure is increased in steps of 0.1 MPa according to the degree of contamination, with the maximum operating pressure limited to 1.2 MPa. The flow control system uses an electromagnetic flowmeter for real-time monitoring and adjusts the speed of the variable frequency pump using a PID algorithm to maintain the set flow value within an error range of ±5%. The grouting point layout follows the spatial distribution characteristics of the pollutants, with dense grouting holes preferentially set in high-concentration areas. The hole spacing is dynamically adjusted based on the width of the pollution plume, with a minimum spacing of 5 meters.

[0040] The heavy metal leakage flux calculation model for pollution source control units integrates Darcy's law and solute flux equation:

[0041] in represents the total leakage flux (kg / d), is the soil dry density at the i-th monitoring point (g / cm³), is the pore water velocity (m / d), represents the controlled cross-sectional area (m²), is the heavy metal concentration (mg / kg). The calculation integrates monitoring well data and soil pore water pressure sensor readings in real time, updating the flux estimate every 30 minutes. The vertical barrier wall depth control algorithm is based on the limit equilibrium principle, considering the interaction between earth pressure, water pressure, and wall flexural strength. It outputs a recommended insertion depth (m) and construction parameters.

[0042] The hardware interface of the blocking device control submodule utilizes an industrial Ethernet protocol and connects to the control system of the HDPE membrane insertion device. Depth adjustment instructions, including three operating parameters: target depth, insertion speed, and vibration frequency, are transmitted to the hydraulic drive unit via a fieldbus. The system incorporates multi-level safety interlocks, automatically suspending construction if a sudden change in the groundwater level or abnormal soil displacement is detected. A feedback loop for control effectiveness transmits post-construction permeability test results back to the seepage flux calculation model, forming a closed-loop optimization mechanism.

[0043] The laser-induced breakdown spectroscopy system of the passivation agent reaction monitoring device uses a Nd:YAG pulsed laser with an output wavelength of 1064nm, a pulse energy of 100mJ, and a repetition rate of 10Hz. The optical collection system consists of a lens group with a focal length of 75mm and a fiber optic spectrometer, and the spectral range covers 200-900nm. The sample chamber is designed as a flow-type reaction tank, which receives the mixed liquid flow output by the grouting system in real time. The partial least squares regression algorithm is used for spectral data processing to establish a quantitative relationship model between the intensity of the characteristic spectral lines of heavy metals and the chelation rate. The monitoring results are sent to the central control system via a wireless transmission module, and the data update interval is set to 5 minutes.

[0044] The parameter update mechanism for the three-dimensional diffusion equation is based on the forward difference approximation. Upon receiving feedback on the passivator's effectiveness, the retardation factor in the equation is adjusted as follows: if the chelation rate increases by more than 10% of the baseline value, the retardation factor is increased by 5%; if the chelation rate decreases by more than 15% of the baseline value, the adsorption kinetic parameters are recalibrated. The equation solver uses an algebraic multigrid method to accelerate convergence, and a full-scale calculation takes approximately 8 minutes on a 16-core computing node. The updated parameter set is validated through sensitivity analysis and stored in the model knowledge base for optimization of subsequent predictive calculations.

[0045] The system's operation monitoring interface integrates multiple functional views, including 3D visualization of the pollution plume, real-time grouting parameter curves, and barrier wall construction progress charts. Operators can adjust control parameters via the touchscreen, with dual authentication required for critical operations. Historical data is stored in a time-series database, recording all sensor readings, control commands, and system events, with a five-year data retention period. A remote diagnostic module allows technical support personnel to access the system's operating status through a secure channel for troubleshooting and parameter tuning.

[0046] The emergency response system includes three levels of contingency plans: Level 1 responds to local parameter excursions by automatically adjusting the operating parameters of individual grouting points; Level 2 addresses regional anomalies and coordinates adjustments across multiple grouting points; and Level 3 addresses system-wide failures by activating backup control units and notifying management. All response actions are logged in detail, including triggering conditions, execution time, and results. The system also conducts regular simulation drills to test the processing procedures and equipment status at each response level.

[0047] Example 4: Focusing on the multidimensional data fusion and hierarchical response mechanism of the three-dimensional dynamic early warning platform, the system workflow is illustrated through an actual operational case study of a soil remediation project in an industrial park. The park covers an area of ​​86 hectares, and historical monitoring data shows combined cadmium and lead pollution, with maximum concentrations reaching 28.7 mg / kg and 1560 mg / kg, respectively. The platform deployment phase first completes basic data collection, including core sample analysis from 12 boreholes, laboratory test results from 42 soil sampling points, and meteorological and hydrological observation records from the past five years.

[0048] The core data processed by the pollution heat map generation engine comes from the heavy metal spatial distribution map generated in Example 1, and the original data format is a GeoTIFF file with a 50-meter grid. Data enhancement is implemented in the engine preprocessing stage, and the grid is encrypted to a resolution of 0.5 meters using the inverse distance weighted method. The influence of spatial variability of soil types is considered in the interpolation process. The heat map rendering uses HSV color space conversion technology to define color bands corresponding to five pollution levels: blue (<1 times the standard value), green (1-2 times), yellow (2-3 times), orange (3-5 times), and red (>5 times). The three-dimensional visualization module is implemented based on WebGL technology, supports interactive operations such as perspective rotation, profile cutting, and transparency adjustment, and VRAM video memory usage is controlled within 4GB.

[0049] The historical case library of the restoration solution deduction module stores structured data for 327 projects, each of which contains 14 types of characteristic parameters. The following table compares the key parameters of five typical restoration projects in the case library:

[0050] The case matching algorithm uses an improved cosine similarity calculation method to define a seven-dimensional feature vector: pollution type weight, concentration level, soil pH, organic matter content, groundwater level, remediation objectives, and budget constraints. After calculating similarity between the current project's feature vector and the case library, the remediation solution combinations for the top 10 most matching cases are output. The solution optimizer, based on constraint satisfaction problem-solving techniques, considers the balance between construction schedule, cost, and technical feasibility, ultimately generating three recommended paths: radical (focusing on effectiveness), balanced (balancing effectiveness and cost), and conservative (focusing on economic efficiency).

[0051] The comprehensive pollution risk index is calculated by integrating four indicators: spatial distribution index (0-35 points), migration risk index (0-30 points), ecological sensitivity index (0-25 points), and exposure pathway index (0-10 points). The index is updated hourly and is calculated using a weighted summation model, with weight coefficients determined through an expert Delphi method. Risk levels are categorized into four levels: Level I (<30 points) is considered acceptable risk, Level II (30-50 points) is considered low risk, Level III (50-70 points) is considered medium risk, and Level IV (>70 points) is considered high risk.

[0052] The multi-level risk response mechanism is implemented through an industrial IoT platform, with execution terminals including PLC controllers, audible and visual alarms, and access control systems. When the index reaches the primary alert threshold (70 points), the system automatically sends a Modbus TCP command to the plant's DCS system, reducing the production load in the pollution source area to 50% of the design value and notifying the responsible engineer via SMS. When the intermediate alert threshold (85 points) is triggered, the on-site warning light (rotating red) and buzzer (85dB) activate, and emergency repair procedures are simultaneously initiated: the passivation agent dosage is increased by 30%, the grouting frequency is switched to continuous mode, and the emergency team is placed on standby. The response to the highest alert threshold (95 points) includes a site closure signal sent to the access control controller, which automatically locks all entrances and exits, activates the emergency lighting system, and transfers control to the higher-level monitoring platform via the OPCUA protocol.

[0053] The 3D dynamic early warning platform's human-computer interaction interface (HMI) design adheres to standards. The main console features a 55-inch 4K touchscreen display divided into six functional areas: a real-time monitoring area displays a 3D heat map and risk index dashboard; a scenario simulation area presents a radar chart comparing repair paths; an equipment status area displays grouting system operating parameters; an alarm log area scrolls through abnormal events; an environmental parameter area updates meteorological and hydrological data; and a system log area records all operational traces. The interface supports multi-user collaboration, with differentiated functional visibility corresponding to different permission levels. Iris authentication is required for critical operations.

[0054] The data management subsystem utilizes a hybrid storage architecture, storing real-time data in a time-series database (with a sampling interval of 1 minute), case library data in a relational database, and 3D model data in a graph database. The data backup strategy includes daily incremental backups and weekly full backups, with off-site disaster recovery nodes located in a backup data center 30 kilometers away. System security includes a network-layer firewall, application-layer permission control, and data-layer encrypted transmission, achieving Level 3 security certification.

[0055] The platform's maintenance mechanism encompasses both preventive and predictive maintenance. Preventive maintenance involves replacing key components on a regular basis: cleaning optical sensor lenses every three months, replacing the equipment room's precision air conditioning filters every six months, and annually calibrating all monitoring probes. Predictive maintenance relies on device status monitoring data, using an LSTM neural network to predict potential failures and generate maintenance work orders in advance. The system utilizes a dual-server hot standby architecture, with automatic switchover between primary and backup servers using heartbeat detection. Maximum recovery time is limited to 90 seconds.

[0056] The system, implemented in this industrial park, handled 12 risk warning events over six months of continuous operation, including three incidents of increased migration risk caused by heavy rain, four incidents of increased pollution due to production anomalies, and five repair interruptions caused by equipment failures. The platform-generated repair plans matched the actual project implementation results to the expected degree, with a median risk warning response delay of 2 minutes and 48 seconds. During the system's operation, a complete digital twin data chain was established, providing a comprehensive data foundation for the subsequent full lifecycle management of contaminated sites.

[0057] Example 5: This paper describes the complete process flow from parameter analysis to performance feedback, focusing on the coordinated operation mechanism of the intelligent grouting system and the passivation agent reaction monitoring device. The intelligent grouting system's hardware configuration utilizes a modular design, with core components including a Mitsubishi FX5U programmable controller, a plunger metering pump unit, a pressure buffer tank, and a distributed pipeline network. The controller receives the optimal passivation agent dosage signal from the contamination migration model via an Ethernet interface. This signal contains four parameters: target dosage concentration (g / m³), total grouting volume (m³), upper pressure limit (MPa), and flow rate range (L / min). The signal analysis module converts digital commands into control variables. The plunger pump stroke frequency is continuously adjustable within a range of 10-100 Hz based on flow requirements. The parallel design of the pump units ensures a maximum delivery capacity of 15 m³ / h.

[0058] The optical detection unit of the passivation agent reaction monitoring device adopts a closed flow path design and consists of three functional parts: a sample pretreatment chamber, a laser action chamber, and a spectral analysis chamber. The sample pretreatment chamber implements online filtration to remove suspended particles through a 5μm pore size ceramic filter element, and the constant temperature system maintains the reaction liquid temperature at 25±0.5°C. The laser action chamber is equipped with beam shaping optics to control the laser spot diameter at 0.8mm, and the pulse energy stability is better than ±2%. The spectral analysis chamber adopts a medium-step grating spectrometer system, coupled with a back-illuminated CCD detector, to achieve full spectrum coverage of the 200-900nm band, and the optical resolution reaches 0.1nm. The device has a built-in automatic calibration program, which performs wavelength calibration and intensity correction every 24 hours, and uses NIST standard reference materials to verify the accuracy of the system.

[0059] The dynamic adjustment process of grouting parameters is based on a feedforward-feedback composite control strategy. The feedforward control link pre-calculates the initial working parameters of each grouting hole based on the concentration gradient distribution of the heavy metal spatial distribution map. The feedback adjustment link collects the measurement values ​​of the pipeline pressure sensor, electromagnetic flowmeter and online density meter in real time, and fine-tunes the operating status of the pump group through the PID algorithm. When the pipeline pressure fluctuation is detected to exceed 15% of the set value, the pressure buffer tank automatically compensates for the system fluctuation, and the buffer volume is designed to be 3 times the normal flow rate. The grouting terminal is equipped with an intelligent distribution valve group, which dynamically adjusts the slurry distribution ratio of each grouting point according to the real-time update results of the pollution heat map, and the distribution accuracy is controlled within ±5%.

[0060] The data acquisition process for laser-induced breakdown spectroscopy (LIBS) technology includes four stages: pulse excitation, plasma formation, spectrum acquisition, and data processing. Each analysis cycle emits 20 laser pulses, and gated detection technology is used to collect the characteristic spectrum when the plasma cools to 50 μs. The heavy metal identification algorithm is based on spectral line database matching, focusing on monitoring the characteristic spectral lines of 228.8 nm for Cd, 405.8 nm for Pb, and 193.7 nm for As. The chelation rate calculation module compares the characteristic peak intensity ratio of free heavy metals and passivator-bound heavy metals, establishing a three-dimensional calibration curve to eliminate matrix effects. Each batch of analysis results generates a data packet containing a timestamp, coordinate position, and spectral fingerprint, which is uploaded to a central database via the industrial wireless network.

[0061] The transmission and processing of performance feedback signals utilizes a layered architecture. Field-layer devices send raw monitoring data to edge computing nodes via IO-Link interfaces, where data is cleaned and formatted in a standardized manner. The network layer utilizes the Time-Sensitive Networking (TSN) protocol to ensure real-time data transmission, with end-to-end latency less than 50ms in the worst case. The platform layer deploys a stream processing engine that performs sliding window analysis on continuously arriving performance data, with the window size set to the most recent 30 minutes of data. A moving average is calculated as the current performance indicator. When the chelation rate is detected to have decreased by more than 10% over three consecutive windows, the system generates a parameter update request and pushes it to the pollution migration model.

[0062] The collaborative working mechanism between the grouting system and the monitoring device is reflected in three key interactive links. The first is parameter pre-calibration before grouting begins. The optimal laser detection parameters are predicted based on the historical data of the monitoring device to avoid spectral saturation caused by excessively high energy or insufficient signal-to-noise ratio caused by too low energy. The second is adaptive adjustment during operation. When abnormal chelation efficiency is detected in a certain area, the sampling frequency of the grouting points in that area is automatically increased to three times the normal value. The third is the coordinated scheduling of maintenance cycles. Preventive maintenance of optical components is synchronized with the planned downtime of the grouting system to minimize system interruption time.

[0063] The emergency response mechanism has a two-level response process. The primary response, for localized performance degradation events, automatically triggers a flushing procedure for the grouting point, using a dilute nitric acid solution with a pH of 5 for two minutes to clean the pipes and mixer. The advanced response handles system-level failures, such as laser energy attenuation exceeding 40% or the spectrometer CCD temperature exceeding 50°C. It immediately activates the backup analysis module and notifies maintenance personnel. All emergency events are recorded in a detailed log, including the trigger time, disposal measures, and recovery status. The log files use WORM storage technology to prevent tampering.

[0064] The system calibration and verification system consists of four levels. Equipment-level calibration involves daily optical system energy testing and wavelength calibration, using neodymium glass standards to verify laser performance. Method-level calibration involves weekly standard sample testing, covering five concentration gradients of heavy metal solutions. System-level verification is performed monthly, injecting known concentrations of reference materials through the grouting system, with full tracking and analysis of test accuracy. External comparisons are conducted quarterly, correlating online monitoring results with laboratory ICP-MS analysis data to maintain an R² > 0.95 consistency standard.

[0065] The design of the human-computer interaction interface focuses on intuitive operation and information integrity. The main control screen is divided into real-time monitoring area, historical trend area, alarm information area and operation control area. The real-time monitoring area dynamically displays the grouting pressure curve, flow distribution diagram and chelation rate thermogram, with an update frequency of 1Hz. The historical trend area can call up the parameter change curve of any time period and supports multi-variable superposition display. The alarm information area displays the current alarm according to the severity level, with disposal suggestion guidelines. The operation control area implements hierarchical authority management. Routine parameter adjustment is authorized to on-site operators, and key parameter modification requires engineer-level authority. The interface color scheme adopts a blue-gray main color that conforms to ergonomics, and important alarms use orange flashing prompts.

[0066] The data management subsystem enables full lifecycle tracking. Raw spectral data retains complete waveform information and is compressed and stored in HDF5 format, with a single file containing data from 1,000 pulses. Process parameter data is stored in a time-series database, with a tagging system that includes a triple index of device ID, parameter type, and timestamp. All data is retained for five years after the project ends, supporting multi-dimensional searches by time range, spatial location, and heavy metal type. The data export function generates EPA-compliant report files containing test results, quality control data, and metadata descriptions.

[0067] The mechanical design of the intelligent grouting system takes into account its adaptability to harsh environments. The pump unit and piping are constructed of 316L stainless steel, offering corrosion resistance within the pH range of 2-11. The laser optical components are housed in a shockproof, constant-temperature chamber, maintaining an ambient temperature between 20°C and 30°C and a relative humidity between 30% and 60%. The electrical system meets IP65 protection standards, with redundant backups for critical circuits. The on-site equipment layout adheres to modular principles, with each functional unit maintaining an access path of at least 1.2 meters. Heavy components are equipped with lifting rings for easy maintenance and replacement.

Claims

1. A soil heavy metal pollution identification system, characterized in that: include: A multispectral remote sensing perception module is configured with a high-resolution imaging unit and a ground spectrum acquisition unit. The high-resolution imaging unit acquires surface reflectance data via a satellite payload, and the ground spectrum acquisition unit collects in-situ soil spectrum data using a vehicle-mounted mobile platform. The multispectral remote sensing perception module is integrated with a heterogeneous data fusion gateway, which receives surface reflectance data and in-situ soil spectrum data and generates a multi-band spectral response signal. The pollution risk assessment module includes a spatial distribution analysis unit, a migration risk prediction unit, and a pollution threshold definition unit. The spatial distribution analysis unit processes the multi-band spectral response signal and outputs a heavy metal spatial distribution map. The migration risk prediction unit combines the heavy metal spatial distribution map with meteorological and hydrological data to generate a pollution migration probability cloud map. The pollution threshold definition unit defines ecological safety constraints based on the multi-band spectral response signal and outputs a soil remediation safety threshold. The governance decision-making execution module includes an in-situ remediation execution unit, a pollution source control unit and a three-dimensional dynamic early warning platform. The in-situ remediation execution unit receives the pollution migration probability cloud map and adjusts the passivation agent injection parameters. The pollution source control unit regulates the operating status of the pollution source blocking equipment based on the soil remediation safety threshold and the multi-band spectral response signal, and simultaneously collects the real-time monitoring signal of soil pore water. The three-dimensional dynamic early warning platform integrates the heavy metal spatial distribution map and the pollution migration probability cloud map to generate a comprehensive pollution risk index.

2. The soil heavy metal pollution identification system according to claim 1 is characterized in that The multispectral remote sensing perception module also includes an unmanned aerial vehicle near-ground detection unit and a groundwater quality sensing network. The unmanned aerial vehicle near-ground detection unit obtains vegetation stress characteristic data through a multi-rotor flight platform equipped with a hyperspectral sensor. The groundwater quality sensing network is deployed in the monitoring wells of the contaminated site and collects groundwater ion concentration data. The heterogeneous data fusion gateway receives vegetation stress characteristic data and groundwater ion concentration data and performs spatiotemporal alignment processing to form a multi-band spectral response signal containing spectral characteristic signals, vegetation response signals and ion diffusion signals. The multi-band spectral response signal is specifically divided into spatial resolution spectral signals, migration prediction spectral signals and threshold definition spectral signals.

3. The soil heavy metal pollution identification system according to claim 2 is characterized in that The spatial distribution analysis unit uses a convolutional neural network to construct a geostatistical inversion model. The spatial analysis spectral signal includes visible light-near infrared band reflectance data, short-wave infrared absorption characteristic data and historical pollution census data. The geostatistical inversion model extracts spectral spatial features through a deep separable convolution layer, and inputs the visible light-near infrared band reflectance data, short-wave infrared absorption characteristic data and historical pollution census data into a three-dimensional convolution kernel to generate a heavy metal spatial distribution map. At the same time, the vegetation response signal and ion diffusion signal are fused through a feature pyramid network to output the pollution source tracing results.

4. The soil heavy metal pollution identification system according to claim 3 is characterized in that The migration risk prediction unit includes a pollution migration model and a Monte Carlo simulator. The migration prediction spectral signal includes soil texture parameters, groundwater flow rate data and rainfall erosion intensity data. The pollution migration model processes the migration prediction spectral signal by coupling the convection-diffusion equation with the adsorption-desorption kinetic equation, and outputs a heavy metal migration flux prediction value. The Monte Carlo simulator performs thousands of random sampling operations based on the heavy metal migration flux prediction value to generate a pollution migration probability cloud map. The pollution migration probability cloud map is fed back to the geostatistical inversion model to update the pollution source tracing results.

5. The soil heavy metal pollution identification system according to claim 2 is characterized in that The pollution threshold definition unit processes the threshold definition spectral signal through the ecotoxicology constraint model. The ecotoxicology constraint model is integrated with a species sensitivity distribution analysis module. The species sensitivity distribution analysis module calls the biological half-lethal concentration data in the local biological database to define the ecological safety constraint conditions, outputs the soil remediation safety threshold, and simultaneously adopts the model predictive control algorithm to dynamically adjust the soil remediation safety threshold to not be lower than the statutory limit standard.

6. The soil heavy metal pollution identification system according to claim 5, characterized in that The in-situ remediation execution unit is configured with a pollution migration model and a passivator optimization library. The pollution migration model constructs a three-dimensional diffusion equation based on Fick's second law. The input parameters of the three-dimensional diffusion equation include soil redox potential data, organic matter content data and the pollution migration probability cloud map, and outputs the optimal passivator dosage signal. The passivator optimization library stores the spectral response characteristic data of historical remediation materials, and generates a passivator type selection instruction by matching the spectral response characteristic data with the optimal passivator dosage signal.

7. The soil heavy metal pollution identification system according to claim 6, characterized in that The pollution source control unit includes a pollution flux monitoring submodule and a blocking device regulation submodule. The pollution flux monitoring submodule receives the real-time monitoring signal of soil pore water and the spatial distribution map of heavy metals, and calculates the heavy metal leakage flux in the pollution source area. The blocking device regulation submodule generates a vertical barrier wall depth adjustment instruction based on the difference between the heavy metal leakage flux and the soil remediation safety threshold. The vertical barrier wall depth adjustment instruction is transmitted to the three-dimensional diffusion equation to update the optimal passivator dosage signal.

8. The soil heavy metal pollution identification system according to claim 1 is characterized in that The three-dimensional dynamic early warning platform is deployed with a pollution heat map generation engine and a remediation scheme deduction module. The pollution heat map generation engine uses the Kriging interpolation algorithm to render the heavy metal spatial distribution map into a three-dimensional pollution heat map. The remediation scheme deduction module calls the historical remediation case library based on the comprehensive pollution risk index to generate a multi-scheme remediation path. The multi-scheme remediation path is input into the in-situ remediation execution unit to trigger the dynamic adjustment of the passivation agent parameters.

9. The soil heavy metal pollution identification system according to claim 8, characterized in that The three-dimensional dynamic early warning platform implements a multi-level risk response mechanism. When the comprehensive pollution risk index reaches the primary warning threshold, the pollution source area production slowdown instruction is initiated. When the comprehensive pollution risk index reaches the intermediate warning threshold, the contaminated site isolation protection instruction is activated and the in-situ remediation execution unit is triggered to urgently add passivating agent. When the comprehensive pollution risk index reaches the highest warning threshold, an emergency site closure signal is sent to the governance decision execution module.

10. The soil heavy metal pollution identification system according to claim 6, characterized in that The in-situ repair execution unit specifically includes an intelligent grouting system and a passivator reaction monitoring device. The intelligent grouting system analyzes the optimal passivator dosage signal and dynamically adjusts the grouting pressure and grouting flow rate. The passivator reaction monitoring device uses laser induced breakdown spectroscopy technology to detect the heavy metal chelation rate of the passivator reaction product in real time to generate a passivator efficiency feedback signal. The passivator efficiency feedback signal is transmitted to the pollution migration model to update the three-dimensional diffusion equation parameters.

Images