Geomembrane integrity intelligent detection system and method based on quality self-adaption and multi-dimensional feature analysis

Abstract

The application discloses a geomembrane integrity intelligent detection system and method based on quality self-adaption and multi-dimensional feature analysis, and is applied to quality monitoring of environmental protection anti-seepage projects such as waste landfill sites, tailing ponds, sewage treatment facilities and the like. The system comprises a field detection terminal, a cloud analysis platform and a closed-loop feedback link. The field detection terminal is configured with a positioning module and a multi-channel potential acquisition unit, and the acquisition strategy is dynamically adjusted according to the positioning quality parameters; the cloud analysis platform establishes a project feature database, recommends detection parameters through feature similarity matching, carries out spatial correlation analysis on the multi-channel potential data, and distinguishes real damage from interference signals by using an electric field fingerprint identification technology; and the closed-loop feedback link realizes real-time guidance and accurate re-measurement of suspected abnormal points. The application solves the problems of incomplete data, high false alarm rate and difficult parameter setting in the prior art, realizes intelligent and self-adaptive integrity detection of the environmental protection anti-seepage structure, effectively prevents groundwater pollution and soil pollution, and has remarkable environmental protection value.

Description

Technical Field

[0001] This invention relates to the fields of environmental protection technology and geophysical exploration technology, specifically to an intelligent detection system and method for geomembrane integrity based on quality adaptation and multidimensional feature analysis, which can be applied to quality monitoring and pollution control of environmental protection seepage prevention projects such as landfills, tailings ponds, sewage treatment facilities, and hazardous waste landfills. Background Technology

[0002] Geomembrane is a high-molecular polymer seepage-proof material widely used in seepage-proof structures in landfills, tailings ponds, sewage treatment plants, and water conservancy projects. The integrity of the geomembrane is directly related to environmental protection and engineering safety.

[0003] Significance of environmental protection:

[0004] Landfills: Prevent leachate from contaminating groundwater and soil, and protect drinking water sources.

[0005] Tailings ponds: prevent the leakage of heavy metals and toxic substances, and prevent damage to the ecological environment.

[0006] Wastewater treatment facilities: ensure that wastewater does not leak out and protect the surrounding environment.

[0007] Hazardous waste landfills: Preventing pollution from highly toxic substances and ensuring public safety.

[0008] According to the "Law of the People's Republic of China on the Prevention and Control of Environmental Pollution by Solid Waste" and the "Standard for Pollution Control of Municipal Solid Waste Landfills" (GB 16889-2008), geomembrane integrity testing is a mandatory requirement for the quality acceptance of environmental protection seepage prevention projects.

[0009] Electrical methods (such as the dipole method and Arc Tester) are the mainstream techniques for geomembrane integrity testing. The principle involves applying a DC or low-frequency current above the geomembrane and identifying damage points by measuring the surface potential distribution. International standards such as ASTM D6747 and ASTM D8265 have standardized this method.

[0010] However, existing technologies have the following main problems:

[0011] Problem 1: Fixed data collection strategy leads to insufficient data integrity.

[0012] Traditional geomembrane testing equipment typically employs a fixed data acquisition strategy: data is only recorded when the positioning accuracy reaches a preset threshold (such as centimeters), and data is discarded when the accuracy is insufficient.

[0013] This strategy works well under ideal conditions (open space, no obstructions), but in complex environments (tree obstruction, near buildings, at the bottom of slopes, etc.), the positioning signal is often weakened, resulting in a large number of breakpoints in the data trajectory and forming detection blind spots.

[0014] Existing technologies lack the ability to dynamically adjust acquisition strategies based on positioning quality, and fail to compensate for deficiencies in conditions by adjusting acquisition parameters under non-ideal circumstances.

[0015] Technical consequences:

[0016] The detection trajectory is discontinuous, with a large number of blind spots.

[0017] Repeated rework is required, resulting in low work efficiency.

[0018] Full coverage detection is difficult to achieve in certain terrains.

[0019] Question 2: High false alarm rate due to a single criterion

[0020] Current technology primarily relies on the amplitude of the potential difference signal to determine the presence of damage; points exceeding a threshold are marked as suspected damage points. However, this method cannot effectively distinguish between actual damage points and various types of interference.

[0021] Metallic foreign objects (such as steel bars or metal fragments) generate strong abnormal signals.

[0022] Linear interference from underground pipelines

[0023] Water accumulation and changes in soil conductivity can generate false signals.

[0024] Poor electrode contact leads to abrupt signal changes

[0025] Existing technologies lack the ability to analyze the spatial correlation of multi-channel data. Real damage points exhibit strict physical consistency constraints across multiple independent measurement channels (electrode pair potential difference, single electrode potential, contact impedance, etc.), while interference signals often only satisfy some of these characteristics. Existing technologies fail to fully utilize this combination of multi-dimensional features for identification.

[0026] Technical consequences:

[0027] The false positive rate is high, requiring extensive manual verification.

[0028] Excavation verification costs are high

[0029] Insufficient ability to distinguish between actual damage points and interference.

[0030] Question 3: Parameter settings rely on experience and are difficult to optimize.

[0031] The setting of key parameters such as detection threshold and sampling interval directly affects the detection results, but the reasonable values ​​of these parameters are related to various factors such as geomembrane type, membrane thickness, cover layer type, and cover layer thickness. Experienced experts can quickly set these parameters based on their experience, but novices often need to spend a long time trying and failing, which limits the promotion of the technology.

[0032] Current technologies lack intelligent parameter recommendation mechanisms based on historical project experience. A large amount of historical project data has not been effectively transformed into systematic knowledge, meaning that parameters still need to be explored from scratch for each new project.

[0033] Technical consequences:

[0034] Parameter setup takes a long time and has a high learning curve.

[0035] Different operators have different standards, resulting in inconsistent data quality.

[0036] Historical experience was not fully utilized

[0037] Question 4: Lack of a closed-loop verification mechanism hinders continuous optimization.

[0038] The typical workflow of existing technologies is: on-site data collection → data analysis → marking of suspected damage points → excavation verification. The verification results are usually archived in the form of paper reports and are not fed back into the system for parameter optimization.

[0039] Even if certain parameter settings are found to lead to a high false alarm rate or false negative rate, existing technologies lack an automatic adjustment mechanism and can only rely on manual parameter modification based on human experience, resulting in low optimization efficiency.

[0040] Technical consequences:

[0041] The system cannot learn from the verification results.

[0042] Parameter optimization relies on manual intervention, which is inefficient.

[0043] False alarm rate and false negative rate are difficult to reduce continuously.

[0044] Question 5: Data silos and difficulties in collaboration

[0045] Data from existing devices is typically stored locally and needs to be copied via USB drive. Data analysis is performed on a personal computer, and results are shared via email or document.

[0046] Technical consequences:

[0047] The data collection and analysis cycle is long.

[0048] Team members cannot view on-site data in real time.

[0049] Data from multiple projects and multiple devices is difficult to manage centrally.

[0050] Historical data query difficulties

[0051] Current state of technology

[0052] Existing geomembrane testing technologies primarily employ electrical methods, identifying damage points by applying an electric field to the geomembrane and measuring the potential distribution. However, these existing technologies have limitations in the following aspects:

[0053] (1) Positioning technology

[0054] Existing testing equipment is typically equipped with a GPS positioning module, but it has the following problems:

[0055] Some devices use ordinary GPS, with positioning accuracy only at the meter level, which cannot meet the requirements for precise location of damage points.

[0056] Although some devices employ high-precision positioning technology (such as RTK), their data acquisition strategies are fixed. When positioning conditions are not ideal (such as satellite signal obstruction), they cannot adaptively adjust the acquisition parameters, resulting in data trajectory breakpoints or degraded data quality.

[0057] (2) Data collection and analysis

[0058] Existing technologies have the following shortcomings in data acquisition and analysis:

[0059] The limited number of acquisition channels results in insufficient information dimensions, making multidimensional feature analysis difficult.

[0060] The lack of spatial correlation analysis between multi-channel data, relying mainly on the amplitude judgment of a single channel, easily leads to misjudging interference signals such as metallic foreign objects and stray currents as damage points, resulting in a high false alarm rate.

[0061] The lack of intelligent recognition technology based on physical models makes it impossible to effectively distinguish between real damage and various interference signals.

[0062] (3) Parameter settings

[0063] The parameter settings in existing technologies mainly rely on human experience:

[0064] Key parameters such as detection threshold and sampling interval need to be adjusted according to factors such as geomembrane type, thickness, and overburden conditions.

[0065] Experienced experts can quickly set appropriate parameters, but beginners need a longer time to figure it out, which limits the widespread adoption of the technology.

[0066] The lack of an intelligent parameter recommendation mechanism based on historical project data has prevented a large amount of historical experience from being effectively transformed into systematic knowledge.

[0067] (4) Closed-loop optimization

[0068] Existing technologies lack effective closed-loop verification optimization mechanisms:

[0069] There is a time lag between on-site testing and excavation verification; verification results are usually archived in the form of paper reports.

[0070] The verification results were not promptly fed back into the system for parameter optimization.

[0071] The system cannot learn from the validation results, making it difficult to continuously reduce the false alarm rate and false negative rate.

[0072] (5) Collaborative work

[0073] The existing equipment lacks sufficient data management and collaboration capabilities:

[0074] Data is typically stored locally and requires physical copying, resulting in long data processing cycles.

[0075] Data from multiple devices and projects is difficult to manage and query centrally.

[0076] Team members have low collaboration efficiency and cannot share on-site data in real time.

[0077] Technological gap: Existing technologies lack systematic solutions that deeply integrate adaptive control of positioning quality, spatial correlation analysis of multi-channel data, intelligent recognition based on physical models, cloud-based intelligent parameter recommendation, and closed-loop verification optimization. Summary of the Invention

[0078] Key term definitions

[0079] The "electric field fingerprint" referred to in this invention refers to the topological feature vector of the spatial distribution of abnormal potential on a multi-channel electrode, which reflects the physical nature of the electric field source.

[0080] Specifically, electric field fingerprints include:

[0081] Spatial distribution characteristics: amplitude distribution, polarity distribution, and gradient direction distribution of potential values ​​at each measuring point.

[0082] Topological relationship characteristics: the relative positional relationships of feature points (peak points, zero points)

[0083] Physical model fit: The degree of fit between the measured distribution and the theoretical electric field model (such as a point source model or a line source model).

[0084] The physical basis of electric field fingerprinting:

[0085] The actual failure point forms a point source current leakage path, where the electric field satisfies Laplace's equation ∇²V=0, exhibiting spherically symmetric decay V∝1 / r.

[0086] Metallic foreign objects: generate an induced electric field with an asymmetrical distribution, which does not conform to the point source model.

[0087] Stray currents: form a linear electric field, exhibit a dipole distribution, and are significantly different from point source characteristics.

[0088] Electric field fingerprint recognition can distinguish between real damage and various types of interference, elevating the process from "signal detection" to "physical modeling and analysis".

[0089] Purpose of the invention

[0090] The purpose of this invention is to provide an intelligent detection system and method for geomembrane integrity based on quality adaptation and multidimensional feature analysis, thereby solving the aforementioned problems in the prior art.

[0091] Technical solution

[0092] The present invention adopts the following technical solution:

[0093] Core innovations:

[0094] The positioning quality assessment is deeply coupled with the data acquisition strategy. Based on positioning quality parameters (solution status, geometric accuracy factor, number of available satellites, etc.), the acquisition conditions are divided into multiple quality levels, with different levels corresponding to different combinations of acquisition parameters.

[0095] The core idea is to make decisions based on the quality level of the positioning, rather than a rigid judgment based on all conditions.

[0096] Under non-ideal conditions, the shortcomings of the conditions can be compensated by adjusting the acquisition parameters (such as increasing the number of repeated samplings, extending the sampling window, etc.), so as to maximize data integrity while ensuring data quality.

[0097] Technical effects: The continuity of data trajectories is significantly improved, the detection blind zone is reduced, and the detection task can still be completed in complex environments.

[0098] Multiple potential signals (electrode-to-ground potential signal and single-electrode-to-ground potential signal) are acquired simultaneously, and multi-dimensional spatial distribution features are extracted from them. By utilizing the different physical consistency constraints of actual damage and various interferences in the multi-dimensional feature space, accurate classification is achieved through pattern recognition of feature combinations.

[0099] For example, real damage has the following physical characteristics:

[0100] The potential difference signal between the electrodes reaches its peak near the point of failure.

[0101] The single-electrode potential signal exhibits a positive and negative polarity distribution.

[0102] The peak position of the potential difference between the electrode pairs coincides spatially with the position of the characteristic point of the single electrode potential.

[0103] Contact resistance is within normal range

[0104] There are strict physical consistency constraints among these characteristics. Interference signals often only satisfy some of these characteristics and can be identified.

[0105] Technical effects: The false positive rate is significantly reduced, the accuracy of identifying true damage points is improved, and the workload of excavation verification is reduced.

[0106] A project feature database is established in the cloud to digitally store complete information about historical projects. When a new project is launched, the cloud-based feature matching and recommendation engine calculates the multi-dimensional feature similarity between the new project and historical projects, extracts the historical projects with the highest similarity, and performs statistical analysis on their detection parameters (using the median as the recommendation value and the interquartile range to represent uncertainty), thus realizing the transformation from individual experience to collective wisdom.

[0107] Technical benefits: Parameter setting time is shortened, allowing even beginners to quickly obtain reasonable parameters, and historical experience is fully utilized.

[0108] Establish a closed-loop feedback chain: After identifying high-confidence anomalies in the cloud, spatial guidance instructions are generated in real time and sent to the on-site terminal. The terminal guides the operator to move precisely, perform multi-directional scanning, determine centimeter-level accurate coordinates, and verify the results to feed back to the cloud for parameter optimization.

[0109] Technical results: The accuracy of anomaly point location has been improved to the centimeter level, the verification results feedback forms a positive loop, and the system is continuously optimized.

[0110] A four-electrode orthogonal arrangement is adopted: two pairs of electrodes are arranged along orthogonal directions, with their geometric centers close to or coinciding. This configuration enables the simultaneous acquisition of information from two orthogonal directions in a single measurement, improving spatial resolution and feature extraction capabilities.

[0111] Technical benefits: Increased information content per measurement, more accurate spatial feature extraction, and improved detection efficiency.

[0112] It employs predictive power switching technology based on the rate of voltage change. Instead of monitoring the voltage amplitude, it monitors the time derivative of the voltage (through a differentiating circuit), predicting anomalies and triggering switching in advance as soon as the voltage begins to change, ensuring uninterrupted power supply.

[0113] Technical benefits: The system operates stably under abnormal conditions such as hot-swapping of power supplies, improving reliability.

[0114] In the receiving channel, a series frequency selection unit (such as a surface acoustic wave filter) is used to maintain low insertion loss in the operating frequency band, achieve high suppression of interference frequency bands, realize frequency domain isolation, and avoid mutual interference between multiple communication modules.

[0115] Technical benefits: Multiple communication modules can work simultaneously, improving communication reliability.

[0116] The device employs a thermally conductive metal casing, with thermally conductive protrusions on the inner wall corresponding to the heat-generating elements. A thermally conductive interface material is filled between the protrusions and the chip, and the entire internal space is filled with potting material. This structure balances heat dissipation (shortest path heat transfer) and protection (complete environmental isolation).

[0117] Technical benefits: Significantly improved equipment reliability and lifespan in harsh environments.

[0118] Break down data silos and achieve end-to-end digitalization:

[0119] Data reception: Field devices upload data in real time or in batches.

[0120] Time-series storage: Employs a time-series database to optimize large-scale data queries.

[0121] Intelligent analytics: Executes advanced algorithms (feature matching, spatial correlation analysis, etc.) in the cloud.

[0122] Visualization: Map display, statistical charts, 3D visualization

[0123] Collaborative work: Centralized management of multiple users, devices, and projects.

[0124] Access control: user roles, data isolation, operation auditing

[0125] Technical benefits: Data processing cycles are significantly shortened, team collaboration efficiency is significantly improved, and historical data is readily available.

[0126] To address the high false alarm rate in geomembrane testing, an electric field fingerprinting technology is introduced. This technology utilizes the physical principle that the electric field distribution at the actual damage point conforms to the spherically symmetric attenuation characteristic described by the Laplace equation, while interference sources such as metallic foreign objects and stray currents do not conform to this characteristic. A mathematical model is then used to differentiate between actual damage and interference.

[0127] Technical details:

[0128] The measured potential distribution data around suspected anomaly points were extracted, and the least squares method was used to fit the measured data with the theoretical point source electric field model V(r) = V0 / r.

[0129] Calculate the goodness-of-fit coefficient R²

[0130] When R² > 0.85 (adjustable), it indicates that the measured data conforms to the characteristics of a point source electric field, and is marked as a true damage.

[0131] When R² < 0.85, it indicates a distortion in the electric field distribution, which is identified as environmental noise or non-point source interference.

[0132] Physical principle:

[0133] The actual damage point forms a current leakage channel beneath the geomembrane. This channel can be approximated as a point source, and the electric field it generates in the homogeneous medium satisfies the Laplace equation ∇²V = 0, with the potential exhibiting a spherically symmetric decay distribution V∝1 / r. However, interference sources such as the induced electric field generated by metallic foreign objects and the linear electric field generated by stray currents do not conform to this distribution law.

[0134] Innovation: It moves from "experience-based threshold judgment" to "mathematical model verification", using the essential differences in physical laws for identification, rather than relying solely on signal amplitude.

[0135] Technical effects:

[0136] The false alarm rate (false positive rate) was significantly reduced, thus decreasing the amount of ineffective excavation verification work.

[0137] The ability to distinguish common interferences such as metallic foreign objects and stray currents has been greatly improved.

[0138] It improves the reliability of automated judgment and reduces reliance on human experience.

[0139] Beneficial effects

[0140] Compared with the prior art, the present invention has the following beneficial effects:

[0141] Data can still be acquired under non-ideal positioning conditions through an adaptive acquisition strategy.

[0142] Reduce trajectory breakpoints and eliminate detection blind spots

[0143] Improved detection success rate in complex environments (tree obstruction, building interference)

[0144] Through multidimensional feature space correlation analysis, we can accurately distinguish between actual damage and interference.

[0145] Further reduce the false alarm rate by using electric field fingerprint recognition technology.

[0146] The false positive rate is reduced, thus decreasing the amount of ineffective excavation verification work.

[0147] Improved accuracy in identifying actual damage points

[0148] By using cloud-based feature matching recommendations, reliance on personal experience is reduced.

[0149] Even beginners can quickly obtain reasonable parameters.

[0150] Historical project experience is being systematically utilized.

[0151] Through closed-loop verification feedback, the system learns from the verification results.

[0152] Parameters are automatically optimized, resulting in a continuous decrease in both false alarm and false negative rates.

[0153] Forming a positive cycle of "data-driven, continuous learning".

[0154] Through transient response power management, the system can operate stably during power failures.

[0155] Inter-module interference is avoided by using multi-band isolated communication.

[0156] With thermally conductive bosses for potting protection, it can operate reliably in harsh environments.

[0157] Data processing cycle significantly shortened

[0158] Team members view on-site data in real time.

[0159] Centralized management of multiple projects, historical data readily available.

[0160] By improving detection accuracy, damage to the impermeable layer can be detected in a timely manner, preventing the leakage of pollutants.

[0161] By reducing false alarm rates, unnecessary excavation and repair can be minimized, thus avoiding secondary damage to the impermeable layer.

[0162] In landfill applications, it effectively prevents leachate from contaminating groundwater and protects drinking water sources.

[0163] In tailings dam applications, heavy metal leakage is prevented, protecting soil and water bodies.

[0164] Complies with environmental protection regulations such as the "Solid Waste Pollution Prevention and Control Law".

[0165] Provide reliable technical means for the quality acceptance of environmental protection projects Attached Figure Description

[0166] Figure 1 This is a schematic diagram of a four-electrode orthogonal configuration, where 110a-110d are the four detection electrodes and 111 is the reference electrode. The four detection electrodes are orthogonally arranged and located in the east, west, south, and north directions, respectively, forming a potential measurement system together with the reference electrode. Detailed Implementation

[0167] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.

[0168] Example 1: Overall System Structure

[0169] System components:

[0170] The on-site testing terminal integrates six core modules and three auxiliary modules.

[0171] Core modules:

[0172] Positioning Module 101: Employs a dual-frequency GNSS receiver, supporting multi-constellation positioning including GPS L1 / L2, BDS B1 / B2, GLONASS L1 / L2, and Galileo E1 / E5a. It receives differential correction data (RTCM 3.x format) in real-time via 4G / 5G networks, achieving a fixed RTK solution with positioning accuracy <2cm horizontally and <5cm vertically. An integrated Kalman filter smooths the position coordinates. Output positioning quality parameters include: solution status (FIXED / FLOAT / SINGLE), geometric precision factor (PDOP), and number of available satellites.

[0173] Multi-channel potential acquisition unit 102: Employs an ADS8698 series 24-bit Σ-Δ ADC, with 8 channels of simultaneous sampling and a sampling rate of up to 100kSPS. Four probe electrodes (A / B / C / D) are arranged orthogonally with an adjustable spacing of 0.5 meters. One remote reference electrode is positioned 20 meters from the geometric center of the electrode array. Synchronous acquisition channels: V_AB (AB potential difference), V_CD (CD potential difference), V_A, V_B, V_C, V_D (potentials of each electrode relative to ground), R_A, R_B, R_C, R_D (contact impedances of each electrode). ADC input range ±10V, resolution 0.6μV (24-bit), dynamic range 144dB.

[0174] The adaptive acquisition and control module 103 receives quality parameters from the positioning module in real time, determines the quality level according to preset rules, and dynamically configures the ADC sampling parameters. At the optimal level, it samples 64 points at a frequency of 2kHz without repetition; at the suboptimal level, it increases the number of repetitions to 10; at the normal level, it extends the sampling window to 512 points and reduces the frequency to 0.5kHz; and at the poor level, it refuses to acquire data.

[0175] Transient Response Power Management Module 104: The main power supply consists of three series-connected lithium batteries (11.1V 5000mAh), and the backup power supply is a supercapacitor (5.5V 10F). An RC differentiating circuit is used to detect the time derivative of the main power supply voltage, with a time constant τ = 1μs and a threshold set to -5V / μs. When dV / dt < -5V / μs is detected, a fast switching unit is triggered, and a high-speed MOSFET completes the main / backup power supply switching within 100μs, ensuring uninterrupted system power supply.

[0176] Communication unit 105 integrates a 4G / 5G module (operating frequency band 700MHz-2.6GHz) and a WiFi module (operating frequency band 2.4 / 5GHz). A surface acoustic wave (SAW) filter is connected in series in the 4G module's receiving channel, achieving an insertion loss of <2dB for the 4G band and a suppression of >35dB for the WiFi band, thus preventing mutual interference between the two modules. Data upload uses the MQTT protocol, and local access uses the HTTP protocol.

[0177] Human-Machine Interface Module 108: Features a 4.3-inch TFT color LCD touchscreen (480×272 resolution) that displays real-time positioning status, potential data, acquisition quality, and system status. It includes four function buttons (F1 for recording, F2 for pause, F3 for menu, and F4 for navigation). Five status LED indicators are provided (USB / WiFi / LoRa / RTK / battery level). A buzzer is included for audible alerts.

[0178] Auxiliary modules:

[0179] IMU Module 106: Integrates a three-axis accelerometer and a three-axis gyroscope for assisted positioning and attitude detection.

[0180] Storage module 107: 128GB eMMC storage, capable of storing over 10 million data records.

[0181] Protective Structure 109: Aluminum alloy 6061 shell, with thermally conductive protrusions on the inner wall corresponding to the main chip position. The space between the protrusions and the chip is filled with thermally conductive silicone grease (thermal conductivity >3W / m·K). The entire internal space is filled with epoxy resin potting material, with an IP67 protection rating.

[0182] The cloud platform is deployed on Alibaba Cloud / Tencent Cloud and adopts a microservice architecture.

[0183] Project Feature Database: A hybrid storage system using PostgreSQL relational database and InfluxDB time-series database. The relational database stores basic project information (project name, site type, geomembrane type, membrane thickness, overburden type, overburden thickness, etc.), detection parameters (current intensity, sampling interval, threshold settings, etc.), and verification results (excavation location, whether damage was confirmed, damage type, etc.). The time-series database stores raw data collected on-site (timestamp, location, potential, impedance, etc.), using 1-minute granularity aggregation to support efficient querying.

[0184] Feature-based matching recommendation engine: When a new project is created, its feature vector (site type, material type, thickness, overlay, etc.) is extracted, and the Euclidean distance or cosine similarity with historical projects is calculated. The 10 historical projects with the highest similarity are extracted, and their detection parameter distribution is statistically analyzed. The median is output as the recommendation value, and the interquartile range (IQR) is output as an uncertainty indicator. Credibility score: The more similar projects and the smaller the IQR, the higher the credibility.

[0185] Spatial correlation analysis module: Receives uploaded multi-channel data and performs the following processing:

[0186] 1. Data preprocessing: filtering and noise reduction, outlier removal

[0187] 2. Feature Extraction: Extract peak amplitude and peak position from V_AB and V_CD; extract polarity distribution (positive and negative regions) and feature point positions (zero potential crossover points) from V_A, V_B, V_C, and V_D.

[0188] 3. Spatial Correlation Analysis: Compare the spatial deviations between the peak position of V_AB and the characteristic point positions of V_A-V_B; compare the spatial deviations between the peak position of V_CD and the characteristic point positions of V_C-V_D; check whether the contact impedance is normal.

[0189] 4. Electric Field Fingerprint Recognition (Innovation 10): Extract potential distribution data within a ±3-meter range around suspected anomalies, fit a point source electric field model V(r) = V0 / r, and calculate the R² goodness of fit.

[0190] 5. Physical consistency constraint determination:

[0191] -True damage: V_AB or V_CD exceeds the threshold + symmetrical polarity distribution + spatial deviation <0.5m + normal contact impedance + R²>0.85

[0192] - Metallic foreign object: V_AB and V_CD are both extremely high + polarity distribution distortion + R² < 0.85

[0193] - Poor contact: V_AB or V_CD sudden change + abnormal contact impedance

[0194] - Ambient noise: R² < 0.85 and no other obvious characteristics.

[0195] Anomaly detection engine: Calculates a confidence score (0-1) for the classification results. If the confidence score is >0.8, it is marked as "high confidence suspected damage" and a closed-loop feedback instruction is issued.

[0196] Visualization module: The web front-end uses React + MapboxGL for map display, ECharts for statistical charts, and Three.js for 3D visualization. The mobile app is developed using Flutter for cross-platform compatibility.

[0197] Report generation module: Automatically generates PDF test reports based on templates, including project overview, test scope, parameter settings, list of anomalies, potential distribution diagram, verification results, etc.

[0198] Users can access the cloud platform via web browser (computer browser) or mobile app (phone / tablet) to view real-time on-site data, analysis results, visualizations, and PDF reports.

[0199] Example 2: Positioning Quality Driven Adaptive Acquisition Control

[0200] The positioning module outputs the following quality parameters in real time:

[0201] Solution status: FIXED (fixed solution) / FLOAT (floating-point solution) / SINGLE (single-point solution)

[0202] Geometric precision factor: PDOP (Position precision factor)

[0203] Number of available satellites

[0204] The adaptive acquisition and control module determines the quality level based on the parameter combination:

[0205] Dynamically adjust ADC acquisition parameters according to quality level:

[0206] Key idea:

[0207] When positioning conditions are not ideal, instead of simply discarding data, we compensate for the deficiencies by increasing the number of repeated samplings or extending the sampling window, ensuring data quality while maximizing data integrity.

[0208] The system assesses the positioning quality 10 times per second, and immediately adjusts the acquisition strategy when the quality level changes.

[0209] Degrading from optimal to suboptimal: Automatically increasing the number of repetitions.

[0210] Degrading from suboptimal to average: Extend the sampling window, reduce the sampling frequency, and instruct operators to reduce their movement speed.

[0211] Deteriorated from fair to poor: Data collection paused, prompting a wait for conditions to improve.

[0212] Recovery from poor to average / suboptimal / optimal: Automatic data recovery

[0213] Technical effects:

[0214] In complex environments (tree obstruction, near buildings), the continuity of data trajectories is significantly improved, and the detection blind spots are greatly reduced.

[0215] Example 3: Multidimensional Feature Space Correlation Analysis and Electric Field Fingerprint Recognition

[0216] Four-electrode orthogonal configuration (electrodes A / B / C / D + reference electrode R), simultaneously acquiring data from 8 channels:

[0217] Differential channels: V_AB = V_A - V_B, V_CD = V_C - V_D

[0218] Grounding channels: V_A, V_B, V_C, V_D (relative to reference electrode R)

[0219] Impedance paths: R_A, R_B, R_C, R_D (contact impedance)

[0220] Peak amplitude: max(|V_AB|), max(|V_CD|)

[0221] Peak position: argmax(|V_AB|), argmax(|V_CD|)

[0222] Signal-to-noise ratio: SNR = 20·log 10 (Peak / Noise Standard Deviation)

[0223] Extracted from the ground access channel:

[0224] Polarity distribution: positive and negative distribution patterns of V_A, V_B, V_C, and V_D

[0225] Feature point locations: V_A-V_B zero-crossing point, V_C-V_D zero-crossing point

[0226] Gradient direction: the principal direction of potential change

[0227] Extracted from the impedance path:

[0228] Contact resistance: Are R_A, R_B, R_C, and R_D within the normal range (10kΩ-1MΩ)?

[0229] For each suspected anomaly, execute the electric field fingerprinting process:

[0230] Extract the potential distribution data V(x, y) within a range of ±3 meters around the anomaly point.

[0231] The actual failure point can be approximated as a point source, whose electric field satisfies the Laplace equation ∇²V = 0, and its potential decays spherically symmetrically in a homogeneous medium.

[0232] V(r) = V0 / r + V bg

[0233] Where V0 is the source intensity, r is the distance, and V bg This is the background potential.

[0234] The measured data were fitted using the least squares method:

[0235]

[0236] Where (x0, y0) are the coordinates of the outlier (to be estimated), V i Let be the measured potential at the i-th measuring point.

[0237] Calculate the coefficient of determination R²:

[0238]

[0239] in, These are the model's predicted values. This represents the average of the measured values.

[0240] The closer R² is to 1, the better the measured data matches the point source electric field model.

[0241] R² > 0.85: The measured data matches the characteristics of a point source electric field and is marked as "true damage".

[0242] R² < 0.85: The measured data does not conform to the point source electric field and is marked as "non-point source interference".

[0243] The final classification is based on a combination of multidimensional features:

[0244] Combination of true damage characteristics:

[0245] 1. The peak value of V_AB or V_CD exceeds the threshold (e.g., 50mV).

[0246] 2. The potential to ground exhibits a symmetrical distribution of positive and negative polarities (V_A>0, V_B<0, or V_C>0, V_D<0).

[0247] 3. The spatial deviation between the peak position of V_AB and the zero-crossing point of V_A-V_B is <0.5m.

[0248] 4. Contact impedances R_A, R_B, R_C, and R_D are all within the normal range (10kΩ-1MΩ).

[0249] 5. Electric field fingerprint recognition R² > 0.85

[0250] If all conditions are met, mark it as "High Confidence Suspected Damage".

[0251] Characteristics of interference from metallic foreign objects:

[0252] 1. Both V_AB and V_CD peak values ​​are extremely high (>200mV).

[0253] 2. Distortion (asymmetry) in the polarity distribution of ground potential.

[0254] 3. Electric field fingerprint recognition R² < 0.85 (because the metal induced electric field does not conform to the point source model)

[0255] Characteristics of poor contact:

[0256] 1. V_AB or V_CD mutation (jump amplitude > 500mV)

[0257] 2. Abnormal contact resistance of the corresponding electrode (>2MΩ or <1kΩ)

[0258] Environmental noise:

[0259] 1. V_AB or V_CD slightly exceeds the threshold (50-100mV)

[0260] 2. No obvious polarity distribution

[0261] 3. Electric field fingerprint recognition R² < 0.85

[0262] By combining multi-dimensional features and using electric field fingerprinting, we can accurately distinguish between genuine damage and various types of interference.

[0263] Reduced false alarm rate for metallic foreign objects: Electric field fingerprint recognition R² is typically <0.5, making it easy to distinguish.

[0264] Eliminating false alarms about poor contact: Direct identification through impedance detection

[0265] Reduced false alarms due to environmental noise: Electric field fingerprint recognition eliminates false positives due to weak signals.

[0266] Overall results: The false positive rate was reduced by more than 50%, and the amount of excavation verification work was significantly reduced.

[0267] Threshold source:

[0268] Recommended historical cases

[0269] On-site background field test (mean + 2-3 times standard deviation)

[0270] Dynamic optimization:

[0271] Based on the verification feedback statistics, the characteristic distribution of true positive and false positive cases was analyzed.

[0272] Optimize thresholds using ROC curves, machine learning, and other methods.

[0273] Example 4: Real-time Closed-Loop Feedback Link

[0274] The cloud generates guidance instructions for high-confidence suspected damage points, including:

[0275] Target point coordinates

[0276] Suspected type

[0277] Confidence

[0278] Feature description

[0279] Priority

[0280] The data is sent to the terminal via wireless network.

[0281] The terminal calculates the distance and direction between the current location and the target location in real time.

[0282] Distance calculation: using the Haversine formula (considering the Earth's curvature).

[0283] Direction calculation: Calculate the azimuth (0° is due north, increasing clockwise).

[0284] Calculation cycle: 5-10 updates per second

[0285] Long distance (>0.5m): Displays direction and distance.

[0286] Medium distance (0.05-0.5m): Add audio-visual cues (buzzer + LED)

[0287] Precise re-measurement (<0.05m): Switch to precise scan mode

[0288] Guide operators to perform cross-scan or multi-directional radiation scan:

[0289] East-west scanning: obtaining signal profile

[0290] North-South Scan: Obtaining Signal Profile

[0291] Peak location: Analyze two profiles to find the peak position

[0292] Precise coordinate calculation: Calculate the precise coordinates of the damage point based on the offset.

[0293] Site marking: Plant flags to mark coordinates

[0294] The repair team excavated and verified the findings, then provided feedback. The cloud platform then processed the feedback.

[0295] Statistical analysis of true positive rate and false positive rate

[0296] Analyze case characteristics

[0297] Dynamically optimize threshold

[0298] Push optimization parameters

[0299] The verification results are stored in the database.

[0300] A closed loop is formed, and the system is continuously optimized.

[0301] Example 5: System Integration Workflow

[0302] Phase 1: Project Initialization → Input Parameters → Cloud Recommendation → Parameter Distribution → Background Field Testing

[0303] Phase 2: On-site data acquisition → Adaptive acquisition → Data upload

[0304] Phase 3: Real-time cloud-based analysis → Spatial correlation analysis → Electric field fingerprint recognition → Generation of guidance instructions

[0305] Phase 4: Closed-loop precise retesting → Navigation to target → Precise scanning → Marking

[0306] Phase 5: Verification and Feedback → Excavation Verification → Feedback Results → Parameter Optimization

[0307] Phase 6: Case Data Storage → Complete Data Stored in Database

[0308] Phase 7: Report Generation → Automatic Generation of Test Reports

[0309] Example 6: Electric Field Fingerprint Recognition Verification Experiment

[0310] To verify the effectiveness of electric field fingerprint recognition technology, a special verification experiment was conducted.

[0311] To verify whether electric field fingerprint recognition technology can effectively distinguish between real damage points and interference sources such as metallic foreign objects and stray currents.

[0312] Building an experimental prototype system:

[0313] Geomembrane sample: HDPE geomembrane (1.5mm thick), dimensions 30m × 30m

[0314] Artificial damage point: 5mm diameter round hole

[0315] Interference source 1: Steel sheet (10cm×10cm), buried under the geomembrane.

[0316] Interference source 2: Simulated metal pipeline (copper conductor), parallel to the geomembrane.

[0317] Multi-channel acquisition unit: 4 electrodes orthogonally configured, electrode spacing 0.5m

[0318] Current source: constant current output, current intensity 100mA

[0319] Artificial holes with a diameter of 5 mm were set on the geomembrane, and potential distribution data of a surrounding 3m×3m area were collected after power was applied (sampling interval 1m, a total of 30×30=900 measuring points).

[0320] Data processing:

[0321] 1. Extract the potential distribution V(x, y)

[0322] 2. Fitting the point source model: V(r) = V0 / r + V bg

[0323] 3. Calculate the R² goodness of fit.

[0324] A 10cm×10cm steel sheet was buried under the geomembrane, and potential distribution data of the same area were collected after power was applied.

[0325] Data processing: Same as scenario 1

[0326] A copper conductor (3m long) was laid under the geomembrane, and potential distribution data of the same area were collected after energizing it.

[0327] Data processing: Same as scenario 1

[0328] Results analysis:

[0329] The potential distribution at the actual damage point conforms to a spherically symmetric decay model (1 / r law), with R² = 0.89, significantly higher than the threshold of 0.85.

[0330] The induced electric field generated by the metal sheet does not conform to the point source model, with R² = 0.31, which is far below the threshold.

[0331] The linear electric field generated by the metal pipeline is even less consistent with the point source model, R²=0.24

[0332] The potential decreases spherically with respect to the point of failure, and the equipotential lines are concentric circles. The potential decreases inversely with distance.

[0333] Potential distribution of interference from metal sheet:

[0334] The equipotential lines are distorted, not concentric circles, and the potential distribution is asymmetrical, resulting in a local "hot spot" effect.

[0335] Potential distribution of interference from metal pipelines:

[0336] The equipotential lines appear as parallel stripes (rather than concentric circles), with the potential changing slowly along the pipeline and decreasing rapidly perpendicular to the pipeline.

[0337] The experiment verified the effectiveness of the electric field fingerprint recognition technology:

[0338] It can accurately identify the actual damage point (R²>0.85).

[0339] It can effectively eliminate interference from metallic foreign objects (R²<0.5)

[0340] It can effectively eliminate interference from metal pipelines (R²<0.5)

[0341] This technology distinguishes between real damage and interference from a physical principle perspective, and has stronger robustness and accuracy compared to simply relying on signal amplitude.

[0342] Example 7: Verification Experiment of Other Key Technologies

[0343] To verify the effectiveness of other key technologies of the present invention, a principle verification experiment was conducted.

[0344] Verify the effectiveness of the following key technologies:

[0345] 1. Can the adaptive acquisition strategy maintain data acquisition when positioning conditions change?

[0346] 2. Can the transient response power supply maintain system stability during power switching?

[0347] Building an experimental prototype system:

[0348] Positioning module: Dual-frequency receiver

[0349] Multi-channel ADC: 8 channels, 24 bits

[0350] Electrodes: 4 probe electrodes + 1 reference electrode

[0351] Power Management: Differentiating Circuit + MOSFET Switching

[0352] Data processing: Embedded processor + host computer software

[0353] In a laboratory environment, a decrease in positioning quality was simulated using a shielded antenna.

[0354] Scenario A: Normal conditions (FIXED, PDOP < 2.0) → The system uses fast single sampling

[0355] Scenario B: Suboptimal conditions simulating partial antenna obstruction (FIXED, PDOP=2.5) → The system automatically increases the number of repeated samplings.

[0356] Scenario C: Severe occlusion simulation under general conditions (FLOAT, PDOP=4.0) → System extends sampling window

[0357] Scenario D: Completely masking analog difference conditions (SINGLE, PDOP>5.0) → System refuses to collect data.

[0358] Results: The system can automatically adjust the data acquisition strategy based on the positioning quality, and can acquire data in both scene A and scene D, while refusing to acquire data in scene D. This verifies the effectiveness of the adaptive mechanism.

[0359] Use an oscilloscope to monitor the power switching process.

[0360] Test A: Traditional amplitude monitoring scheme → Voltage drops to 8V during switching, system resets.

[0361] Test B: The rate of change monitoring scheme of this invention → voltage fluctuation during switching is <0.3V, and the system operates stably.

[0362] Results: The transient response power supply scheme of the present invention can maintain system stability during power switching, verifying the effectiveness of predictive switching.

[0363] The principle verification experiment shows that the core technical solution of the present invention is feasible and effective, and can solve the key problems existing in the prior art.

[0364] Example 8: Industrial Application Prospects and Environmental Value

[0365] The technical solution of this invention has been verified in principle and has good prospects for industrial application and significant environmental protection value.

[0366] Main application areas:

[0367] Application scenarios: Municipal solid waste landfills, construction waste landfills

[0368] Environmental value: Prevents leachate from polluting groundwater and soil, protecting drinking water source safety.

[0369] Regulatory requirements: The "Pollution Control Standard for Municipal Solid Waste Landfills" (GB 16889-2008) mandates...

[0370] Application scenarios: Tailings ponds in metal mines and non-metal mines

[0371] Environmental value: Prevents the leakage of heavy metals (lead, mercury, cadmium, arsenic, etc.) and toxic substances, protecting the ecological environment.

[0372] Regulatory requirements: "Safety Technical Regulations for Tailings Dams" (AQ 2006-2005)

[0373] Application scenarios: equalization tanks, oxidation tanks, and sludge tanks in wastewater treatment plants.

[0374] Environmental value: Ensures that wastewater does not leak out, protecting the surrounding environment and groundwater.

[0375] The regulations require compliance with the "Discharge Standard of Pollutants for Urban Wastewater Treatment Plants" (GB 18918-2002).

[0376] Application scenarios: industrial hazardous waste landfills, medical waste landfills

[0377] Environmental value: Preventing pollution from highly toxic substances and ensuring public safety.

[0378] Regulatory requirements: Standard for Pollution Control of Hazardous Waste Landfill (GB 18598-2019)

[0379] Market size: Continued growth, with increasingly stringent environmental regulations.

[0380] Environmental benefit analysis:

[0381] By improving detection accuracy and efficiency, and particularly by reducing false alarm rates through electric field fingerprinting technology, this invention can:

[0382] Timely detection of damage: Increase the damage detection rate from 75% to over 95% to prevent long-term leakage of contaminants.

[0383] Reduce secondary damage: False alarm rate reduced by 50%, reducing unnecessary excavation damage to the impermeable layer.

[0384] Reduce repair costs: Precisely locate the damaged point (within centimeters), reduce the excavation area, and minimize environmental disturbance.

[0385] Improve acceptance efficiency: shorten the testing cycle by 30-50% and accelerate the acceptance process of environmental protection projects.

[0386] This invention is of great significance for promoting technological progress in the environmental protection industry, enhancing environmental governance capabilities, and protecting the ecological environment.

Claims

1. An intelligent detection system for geomembrane integrity, characterized in that, include: On-site testing terminals include: The positioning module uses differential positioning technology, which wirelessly receives differential correction data to achieve centimeter-level positioning accuracy. It also integrates a position smoothing algorithm to output smoothed position coordinates and positioning quality evaluation parameters. A multi-channel synchronous potential acquisition unit includes at least four detection electrodes arranged in a preset spatial relationship and a remote reference electrode. The spatial arrangement of the detection electrodes is as follows: a first electrode pair is arranged along a first direction, and a second electrode pair is arranged along a second direction, with the angle between the first and second directions ranging from 60 to 120 degrees. The unit synchronously acquires multiple potential signals, including: the electrode pair potential difference signal between the first and second electrode pairs; the single-electrode potential signal of each detection electrode relative to the remote reference electrode; and the contact impedance signal of each detection electrode. The adaptive acquisition control module divides the acquisition conditions into multiple quality levels according to the positioning quality assessment parameters. Different quality levels correspond to different combinations of acquisition parameters. The acquisition parameter combinations include the number of sampling points per time, the sampling frequency, and the number of repeated samplings. The positioning quality assessment parameters include a combination of solution status, geometric precision factor, and the number of available satellites. The adaptive acquisition control module divides the acquisition conditions into four quality levels: optimal, suboptimal, average, and poor. The transient response power management module includes: a voltage change rate detection unit, which uses an RC differentiating circuit to detect the time derivative dV / dt of the main power supply voltage, wherein the time constant τ = RC = 0.1 to 10 μs; a threshold comparison unit, which compares the time derivative dV / dt with a preset change rate threshold, wherein the preset change rate threshold is set to -3 to -10 V / μs; and a fast switching unit, which, when dV / dt is detected to be less than the preset change rate threshold, completes the switching to the backup power supply before the main power supply voltage drops to the system operating threshold, thereby achieving predictive power switching, and the switching response time is less than a preset proportion of the voltage drop time. Cloud-based intelligent analytics platform, including: The project characteristic database stores site parameters, material parameters, testing parameters, and verification results of historical projects; The feature matching recommendation engine calculates the multi-dimensional feature similarity between new and historical items, and outputs recommendation detection parameters and credibility indicators. The spatial correlation analysis module extracts the spatial distribution characteristics and feature point locations of multi-channel potential data, and classifies anomalies based on the physical consistency constraints of the spatial relationships of feature points. The physical consistency constraints include combinations of true damage features and combinations of interference features. The true damage feature combination simultaneously satisfies: the amplitude of the electrode pair potential difference signal exceeds a first threshold; the single electrode potential signal exhibits a positive and negative polarity distribution; the spatial deviation between the peak position of the electrode pair potential difference signal and the feature point position of the single electrode potential signal is less than a second threshold; and the contact impedance is within the normal range. The interference feature combination is: the amplitude of the electrode pair potential difference signal exceeds a threshold, but the polarity distribution of the single electrode potential signal is abnormal or the contact impedance is abnormal. The spatial correlation analysis module also includes an electric field fingerprinting unit, used for: mathematically fitting the measured potential distribution data with the theoretical point source electric field model; calculating the goodness-of-fit coefficient R²; marking a suspected damage point when R² is greater than a preset goodness-of-fit threshold, otherwise marking it as environmental noise or non-point source interference; wherein: the theoretical point source electric field model is based on the spherically symmetric electric field attenuation characteristics described by the Laplace equation; the preset goodness-of-fit threshold is 0.80 to 0.90; the electric field fingerprinting unit reduces the false alarm rate caused by metallic foreign objects and stray current interference sources by distinguishing between point source electric field signals and non-point source interference signals; The real-time closed-loop feedback link generates spatial guidance instructions based on the high-confidence anomalies identified by the cloud-based intelligent analysis platform, guiding the on-site detection terminal to move to the target location to perform precise retesting, and the retesting results are fed back to optimize the detection parameters.

2. The system according to claim 1, characterized in that: At the optimal quality level, a high-efficiency and fast acquisition mode is adopted, with 64 sampling points per sample and a sampling frequency of 2kHz, without repeated sampling. At the sub-high quality level, the single sampling parameters remain unchanged, the number of repeated samplings is increased to 10, and the results from multiple samplings are statistically processed to improve the signal-to-noise ratio. At the normal quality level, the sampling frequency is reduced to 0.5kHz, the number of sampling points per sampling is increased to 512, the sampling time window is extended, and the operator is prompted to reduce the moving speed. At the poor quality level, sampling is refused and the operator is prompted to wait for the positioning conditions to improve.

3. The system according to claim 1, characterized in that, The field testing terminal also includes a multi-band isolated communication unit, comprising a first communication module operating in a first frequency band and a second communication module operating in a second frequency band. A frequency selection unit is set in the signal path of the first communication module. The insertion loss of the frequency selection unit for the first frequency band is less than a first preset value, and the suppression degree for the second frequency band is greater than a second preset value.

4. The system according to claim 1, characterized in that, The method for calculating multi-dimensional feature similarity in the feature matching recommendation engine includes: assigning weight coefficients to the project parameters of site type, seepage prevention material type, material thickness, cover layer type, and cover layer thickness, wherein the weight coefficients are determined according to the degree of influence of each parameter on the detection results; calculating the similarity scores of new projects and each historical project on the project parameters; performing a weighted summation of the similarity scores to obtain a comprehensive similarity score; extracting several historical projects with the highest scores; performing statistical analysis on the detection parameters of the historical projects, outputting the median as the recommendation value, and outputting the interquartile range as an uncertainty index.

5. The system according to claim 1, characterized in that, The spatial distribution features extracted by the spatial correlation analysis module include: peak amplitude, signal-to-noise ratio, and peak position extracted from the electrode pair potential difference signal; polarity distribution, amplitude distribution, gradient direction, and feature point position extracted from the single electrode potential signal; and feature point position deviation and signal value consistency extracted from the relationship between the electrode pair potential difference signal and the single electrode potential signal.

6. The system according to claim 1, characterized in that, The outer shell of the on-site testing terminal is made of thermally conductive metal material. Thermally conductive protrusions are set on the inner wall of the outer shell for the main heating element. The space between the thermally conductive protrusions and the heating element is filled with thermally conductive interface material. The entire internal space is filled with epoxy resin or silicone potting material.

7. The system according to claim 1, characterized in that, The real-time closed-loop feedback link includes: a distance and orientation calculation unit, which calculates the spherical distance and azimuth angle between the current position and the target anomaly point in real time; a phased guidance unit, which adopts different human-computer interaction methods according to the distance, displaying the direction and distance at a long distance, adding audio and visual prompts when approaching, and switching to a precise retest mode after arrival; and a multi-directional scanning unit, which guides the operator to perform multi-directional scanning around the target position and determines the precise coordinates of the anomaly point based on the peak position of the scanning data.

8. The system according to claim 1, characterized in that, When the positioning quality deteriorates, the adaptive acquisition and control module further supports dynamically adjusting the sampling weight coefficients of different electrode pairs, thereby offsetting directional electromagnetic interference by changing the differential combination method.

9. A method for intelligent detection of geomembrane integrity based on the system described in any one of claims 1 to 8, characterized in that, Includes the following steps: S1: Receives site parameters and material parameters of new projects, calculates multi-dimensional feature similarity in the cloud project feature database, extracts similar historical projects, and recommends detection parameters after statistical analysis and sends them to the on-site detection terminal; S2: The on-site detection terminal evaluates the positioning quality in real time, adaptively adjusts the acquisition parameters according to the positioning quality level, and simultaneously acquires multi-channel potential data and position data; S3: Upload the collected data to the cloud-based intelligent analysis platform, extract spatial distribution features, classify anomalies based on the physical consistency constraints of the spatial relationship of feature points, and use electric field fingerprint recognition technology to distinguish between real damage and interference; S4: Generate spatial guidance instructions for high-confidence anomalies, guide the on-site detection terminal to move to the target position in real time, perform multi-directional precise re-measurement, and determine centimeter-level precise coordinates; S5: Based on the retest results and verification feedback, dynamically optimize the detection parameter thresholds and store the verification results in the project feature database.

10. The method according to claim 9, characterized in that: The electric field fingerprinting in step S3 involves: extracting potential distribution data around suspected anomalies; fitting the measured data with the theoretical point source electric field model V(r)=V0 / r using the least squares method; calculating the goodness of fit R²; and determining the anomaly type based on the R² value, with R² greater than 0.85 marking it as a true damage and R² less than 0.85 marking it as interference. The adaptive adjustment in step S2 involves: real-time monitoring of the solution status, geometric precision factor, and number of available satellites output by the positioning module; determining the quality level based on the monitoring parameters; dynamically switching the combination of acquisition parameters when the quality level changes; and prompting the operator to adjust the position or wait for satellite conditions to improve when acquisition is refused. The precise retest in step S4 involves: calculating the distance and direction between the current position and the target position, and updating the display in real time; adding audio-visual prompts when approaching the target to guide the operator to move precisely; and performing east-west and north-south cross-scans after reaching the target area. Analyze the two scanning profiles to determine the signal peak location and calculate the precise coordinates; The parameter optimization in step S5 involves: statistically analyzing the excavation verification results and calculating the true positive rate and false positive rate; analyzing the differences in the characteristic distribution of true positive and false positive cases; optimizing the threshold using ROC curve analysis or machine learning methods; and pushing the optimized parameters to the on-site detection terminal to achieve continuous optimization.

Images