Method for detecting leakage points of underground complex buried pipe multi-water supply system

Abstract

The invention discloses a method for detecting leakage points of an underground complex buried pipe multi-water supply system. The method is suitable for a scene in which a plurality of independent water supply subsystems share a drainage path and need to be diagnosed without interruption. A composite tracer agent with a unique multi-mode code is injected into each subsystem, and signals are synchronously detected at a water seepage point; screening candidate subsystems based on the pipe network topology model and the positions of the water seepage points, and identifying leakage sources by adopting a weighted Bayesian model; then slow-release fluorescent microspheres are injected, an unmanned aerial vehicle or a ground sensing network is used for collecting ground surface fluorescence distribution, and leakage point coordinates are inverted based on a Darcy law coupling convection-diffusion equation. According to the method, shutdown excavation is not needed, the positioning precision reaches the sub-meter level, the problems of signal aliasing, environment interference, positioning fuzziness and the like can be solved on the premise that operation of any water supply subsystem is not interrupted, and accurate identification of a leakage source and accurate positioning of a leakage point are achieved.

Description

Technical Field

[0001] This invention relates to the field of leakage detection technology, and in particular to a method for detecting leakage points in complex underground buried pipe multi-water supply systems. Background Technology

[0002] To maintain the power plant's production and domestic water supply, various water supply systems are configured, including: a water supply system for generating units, a water supply system for main transformers, a fire-fighting water system, and an air conditioning cooling water system for the underground powerhouse. Due to manufacturing processes, insufficient processing, inadequate weld precision, cavitation, and other reasons, leaks frequently occur in the water supply network at weak points such as pipe joints, flange connections, and welds under long-term water pressure. While leaks in surface pipes can be quickly identified based on design drawings, underground water supply and drainage networks are complex and interconnected, with multiple systems intersecting in the same location, making leak detection less effective at detecting hidden leaks. As leakage persists, seepage forms on the surface. If it's difficult to quickly determine whether the seepage is caused by a ruptured pipe or seepage from the mountainside, the water pressure in the corresponding water supply system drops, making it impossible to provide a stable and reliable water source for daily production. In particular, if leaks occur in the water supply pipelines of important equipment such as fire protection, generating units, and main transformers, and the source of the leak cannot be quickly identified, the temperature of the bearings in various parts will rise rapidly. In severe cases, the bearings will burn out, causing the generating unit to shed its load, which in turn will affect the frequency and voltage of the power grid.

[0003] Common methods for detecting leaks in underground pipe networks include: acoustic leak detection, tracer gas leak detection, and ground-penetrating radar (GPR). Acoustic leak detection requires highly experienced personnel but cannot detect leaks more than 1 meter underground. Tracer gas leak detection is only suitable for leak detection in a single pipe network system. While GPR offers high accuracy in locating leak points, its magnetic wave penetration is limited, resulting in a small detection area. Furthermore, given the complex network of water supply systems in power plants, covering the entire plant, this method is unsuitable for detecting leaks in multiple underground pipe systems. Considering the high environmental noise, multiple intersecting water supply systems, and wide water supply range in power plants, the ability to quickly locate the leaking water supply system and its leak point without interrupting water supply system operation remains a challenge. Therefore, this application proposes a method for detecting leaks in complex underground pipe systems with multiple water supply systems. Summary of the Invention

[0004] The purpose of this invention is to provide a method for detecting leaks in complex underground pipe systems with multiple water supply points. This method can overcome problems such as signal aliasing, environmental interference, and ambiguous positioning without interrupting the operation of any water supply subsystem, and can achieve accurate identification of the leak source and precise location of the leak point.

[0005] To achieve the above objectives, this application provides a method for detecting leaks in a complex underground multi-pipe water supply system. The multi-pipe water supply system includes multiple underground and independent water supply subsystems. The method for detecting leaks includes the following steps:

[0006] S1. For each water supply subsystem, inject a composite tracer with uniquely identifiable physicochemical characteristics; S2. Collect seepage samples from the seepage area and test the seepage samples using detection equipment; S3. Based on the pipeline network topology model and the location of the seepage point, a set of candidate water supply subsystems is selected, and the target water supply subsystem where the leak occurred is determined by combining the concentration signal of the composite tracer. S4. Inject a slow-release fluorescent microsphere contrast agent upstream of the target water supply subsystem; S5. Real-time monitoring of surface fluorescence intensity distribution using ground-based sensing devices, and inversion model based on coupled Darcy's law and convection-diffusion equation to determine the geographic coordinates of pipeline leakage points.

[0007] In S1, the composite tracer is a multimodal encoded tracer containing two or more orthogonal response signals. The orthogonal response signals are selected from: near-infrared fluorescent dyes with different center wavelengths, superparamagnetic nanoparticles with different saturation magnetization intensities, and pH-sensitive azo color-changing groups. The signal combinations of the composite tracers used in each water supply subsystem do not overlap in the multidimensional feature space.

[0008] In S2, the detection device includes an integrated microfluidic chip, which has a multi-channel optical detection area and a magnetoresistive sensing array. It can simultaneously complete fluorescence spectral scanning and magnetic signal intensity measurement in a single sample injection, and output the quantitative concentration value of each tracer by executing signal normalization and background subtraction algorithms through an embedded processor.

[0009] In S3, the pipeline topology model is integrated with a Geographic Information System (GIS) to construct a three-dimensional hydraulic proximity map. The selection criteria for the candidate water supply subsystem set are: the Euclidean distance from the pipeline centerline to the leakage point is ≤50 m, and the hydraulic gradient direction points towards the leakage area. A weighted Bayesian model is used to calculate the... i Post-leakage probability of each candidate system Its expression is: ; In the formula, In the system i Under the premise of a confirmed leak, evidence was observed. E The likelihood probability is obtained by estimating the intensity of the tracer signal output by the detection device using Gaussian kernel density. In the first jUnder the premise of a leak in a water supply subsystem, the current detection signal was observed. E The likelihood probability; Representation system i It is the prior probability of the leakage source, that is, the probability given based on historical experience or risk assessment before observed data is available; Representation system j It is the prior probability of the leakage source; Prior probability , For the system i The standard deviation of operating pressure fluctuation, This refers to the service life of the pipeline. For historical leakage frequency, the function ( f, g, h) Let be a monotonically increasing normalized function in the interval [0,1], and let the weight coefficients satisfy... .

[0010] In S4, the sustained-release fluorescent microsphere contrast agent is made of polylactic acid-glycolic acid copolymer encapsulating quantum dot fluorescent material with a particle size of 0.6~0.9μm and a polyethylene glycol anti-adsorption layer grafted on the surface. The sustained-release fluorescent microsphere contrast agent is released in water with zero-order kinetics for 1~5 hours, ensuring the formation of a continuous and stable fluorescent migration trajectory downstream of the leak point.

[0011] In S5, the ground sensing device is a fluorescence sensor network deployed in the suspected area or a drone platform equipped with an ultraviolet laser excitation source and a high-sensitivity sCMOS camera; the collected surface fluorescence intensity field The input is fed into the inversion model, and the coordinates of the leakage point are determined by solving the following optimization problem. With leakage flux : ; In the formula, Based on soil saturation permeability coefficient K Hydraulic gradient i and effective porosity n The constructed groundwater velocity field The simulated fluorescence intensity distribution; For Tikhonov regularization terms, For regularization parameters; The simulation process uses the finite element method to solve the convection-diffusion control equations: ; In the formula, The concentration of the contrast agent at a certain point in space. The time from the start of injection, This is a local coordinate system established with the leakage point as the origin. The dispersion coefficient; These are the coordinates of the leak point, i.e., the three-dimensional geographic coordinates of the actual leak point; The final output includes the geographic coordinates and error ellipse of the leak point with a confidence level of ≥90%.

[0012] Before executing S1, there is also an interference verification step: tracers with the same chemical composition but different time tags are injected into two adjacent water supply subsystems at a preset time interval Δt. If the time difference between the two signals in the seepage sample is less than Δt / 2, it is determined that there is groundwater backflow or pipeline cross-contamination interference, triggering the re-inspection mechanism.

[0013] When S2 detects a mixture of multiple composite tracer signals, a nonnegative matrix factorization algorithm is used to process the original signal matrix. Perform blind source separation and solve the following optimization problem: ; In the formula, V for m Various detection channels are available. n The signal matrix at each sampling time, Here are the characteristic response basis matrices for each tracer. Here are the activation intensity matrices for each system. r The number of candidate water supply subsystems. The sparsity regularization coefficient; F Represents the Frobenius norm; The reconstructed pure response intensities of each tracer are taken from the corresponding row of H and used for accurate discrimination of S3.

[0014] The steps between S3 and S5 also include a cross-validation step: drilling a micro-observation hole near the initially located leakage point, extracting groundwater samples for secondary tracer detection, and confirming the validity of the location result if the target tracer concentration is significantly higher than the background value.

[0015] The injection concentration of the composite tracer is dynamically adjusted according to the pipeline operating conditions to satisfy: ; In the formula, for t Injection concentration (mg / L) at time. D The detection limit (mg / L) is the lowest detection limit of the testing equipment. The instantaneous flow rate of the subsystem is (m³ / h). For real-time pipeline pressure, For safety reasons, This is the pressure disturbance compensation factor; This strategy ensures that the tracer signal in the seepage sample is always above the signal-to-noise ratio threshold, while suppressing misjudgments caused by pressure fluctuations.

[0016] Compared with the prior art, the above-conceptual technical solution conceived in this application has the following beneficial effects: This invention addresses the long-standing engineering challenge of leak detection in complex underground multi-pipe water supply systems. Focusing on the core deficiency of existing technologies—the inability to accurately identify and precisely locate leak sources in situations where multiple systems share drainage paths and the system operates without interruption—this invention proposes a complete, collaborative, and feasible technical solution. Based on this solution, this application offers the following significant advantages: 1. This invention constructs a unique chemical fingerprint by assigning a multimodal encoded composite tracer to each subsystem. Even if the signals are partially mixed, precise decoupling can still be achieved through microfluidic synchronous detection and NMF blind source separation technology.

[0017] 2. This invention introduces a slow-release fluorescent microsphere contrast agent, forming a stable and continuous fluorescence migration trajectory downstream of the seepage point. Combined with a physical inversion model coupling Darcy's law and the convection-diffusion equation, the surface fluorescence intensity field is mapped to the underground three-dimensional seepage coordinates. This method fully considers site hydrological parameters (permeability coefficient, flow velocity), controlling the positioning error within 0.5 meters with a confidence level ≥90%, providing a reliable basis for precise remediation and significantly reducing the excavation area and engineering costs.

[0018] 3. All steps of this invention are completed under normal system operation: online injection of tracer, non-invasive sampling of water seepage, and ground-based UAV scanning. The entire process takes 6-8 hours and does not affect production.

[0019] 4. This invention employs a dynamic concentration adjustment mechanism to optimize the injection dosage in real time based on instantaneous flow rate and pressure change rate, ensuring the signal always remains above the signal-to-noise ratio threshold. Simultaneously, it incorporates dual safeguards: interference verification (time-stamped tracing) and cross-validation (micro-well retesting) to proactively identify interferences such as groundwater backflow and pipeline collusion, and to verify the authenticity of the inversion results. This design enables the method to maintain high reliability even in high-noise, high-disturbance environments, significantly enhancing its engineering applicability.

[0020] 5. The method of the present invention does not require stopping the excavation, and the positioning accuracy reaches the sub-meter level. It can overcome the problems of signal aliasing, environmental interference and positioning ambiguity without interrupting the operation of any water supply subsystem, and achieve accurate identification of the leakage source and precise location of the leakage point. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0022] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0023] To more clearly illustrate the purpose, technical solution, and beneficial effects of this application, a further detailed description of this application is provided below in conjunction with illustrations and specific embodiments. It should be specifically noted that the specific embodiments described below are only for illustrating the technical content of this application and do not constitute a limitation on the scope of protection of this application.

[0024] Regarding the explanation of terminology: In this application, "and / or" is used to describe the relationship between related objects, covering three possible situations: taking "A and / or B" as an example, it can indicate the situation where only A exists, A and B exist simultaneously, or only B exists; the symbol " / " indicates the "or" relationship between related objects, such as "A / B" which refers to A or B.

[0025] Regarding the description of the embodiments: The terms "exemplary" and "for example" appearing in this application are only used to illustrate the technical solutions through specific examples. It should be particularly emphasized that any implementation method or design scheme marked as "exemplary" or "for example" should not be construed as having an advantage over other solutions. Such expressions are only used to present the technical concepts more intuitively.

[0026] Example 1: See Figure 1 This invention provides a method for detecting leaks in complex underground multi-pipe water supply systems. The multi-pipe water supply system includes multiple underground and independent water supply subsystems. The method for detecting leaks includes the following steps: S1. For each water supply subsystem, inject a composite tracer with uniquely identifiable physicochemical characteristics; S2. Collect seepage samples from the seepage area and test the seepage samples using detection equipment; S3. Based on the pipeline network topology model and the location of the seepage point, a set of candidate water supply subsystems is selected, and the target water supply subsystem where the leak occurred is determined by combining the concentration signal of the composite tracer. S4. Inject a slow-release fluorescent microsphere contrast agent upstream of the target water supply subsystem; S5. Real-time monitoring of surface fluorescence intensity distribution using ground-based sensing devices, and inversion model based on coupled Darcy's law and convection-diffusion equation to determine the geographic coordinates of pipeline leakage points.

[0027] This invention ensures that even with simultaneous, minute leaks from multiple systems, the source can be distinguished from the leak sample by assigning a unique composite tracer to each subsystem. It utilizes a single sample injection of the detection equipment to achieve simultaneous quantification of multiple signals, avoiding errors introduced by multiple sampling. A pipeline topology model is introduced to screen the candidate set, significantly narrowing the identification range and improving computational efficiency and accuracy. Slow-release fluorescent microspheres form stable migration trajectories, providing a reliable signal source for high-precision location. An inversion model based on coupled Darcy's law and convection-diffusion equations maps the surface fluorescence distribution to the underground leak point, achieving sub-meter-level location accuracy. The entire process requires no downtime, isolation, or excavation, making it suitable for critical facilities such as power plants, chemical plants, and data centers. Leak detection efficiency is reduced from several days to several hours, and the location accuracy error is less than 0.5 m. This invention utilizes a complete technical chain of multimodal tracer injection, synchronous detection, topological constraint identification, slow-release imaging, and physical inversion positioning to achieve leakage source identification and precise leakage point location without interruption of operation or excavation. This method completely overcomes the fundamental bottleneck of traditional tracer methods failing due to signal aliasing in multi-system collinear scenarios.

[0028] In S1, the composite tracer is a multimodal encoded tracer containing two or more orthogonal response signals, wherein the orthogonal response signals are selected from: near-infrared fluorescent dyes with different center wavelengths and superparamagnetic dyes with different saturation magnetization intensities. Nanoparticles, and pH-sensitive azo color-changing groups; the signal combinations of the composite tracers used in each water supply subsystem do not overlap in the multidimensional feature space.

[0029] The composite tracer employs a multimodal coding design, contains near-infrared fluorescent dyes with wavelength intervals ≥30 nm, and exhibits superparamagnetism. The design employs three orthogonal response signals: nanoparticles, a saturation magnetization Ms difference ≥15 emu / g, and a pH-sensitive chromogenic group. These signals endow each water supply subsystem with a unique and identifiable chemical fingerprint. This design addresses the problem of single tracers being susceptible to interference from soil adsorption, dilution, or color mixing. Because the three signals—optical, magnetic, and chemical—are physically independent, robust identification can still be achieved through a multidimensional feature space, even under strong background noise or partial signal attenuation.

[0030] For example, when the fluorescence signal weakens due to soil shading, the magnetic signal can still be used as a criterion; when pH changes cause discoloration failure, the former two can still work together. Experiments show that in scenarios where multiple water supply subsystems are injected in parallel, this multimodal coding scheme improves the accuracy of leak source identification to over 96%, far exceeding the less than 70% of the single dye method. Furthermore, the selected materials, such as Cy7 and Fe3O4@SiO2, are low-toxicity and biodegradable substances, meeting industrial environmental protection requirements. This scheme not only improves detection reliability but also expands the applicability of chemical tracer methods in complex underground environments.

[0031] In S2, the detection device includes an integrated microfluidic chip, which has a multi-channel optical detection area and a magnetoresistive sensing array. It can simultaneously complete fluorescence spectral scanning and magnetic signal intensity measurement in a single sample injection, and output the quantitative concentration value of each tracer by executing signal normalization and background subtraction algorithms through an embedded processor.

[0032] This invention utilizes a multimodal detection device integrating a microfluidic chip to achieve simultaneous and quantitative detection of multiple tracer signals in a single water seepage sample. Traditional methods require separate testing using multiple devices such as fluorometers, magnetometers, and pH meters, which is cumbersome and prone to cross-contamination or human error. In contrast, this invention integrates a multi-channel optical detection area and a magnetoresistive sensing array into a single microfluidic chip, requiring only a 5-10 mL sample injection, enabling full signal scanning within minutes. The embedded processor simultaneously executes signal normalization, background subtraction, and temperature compensation algorithms, outputting precise concentration values ​​for each tracer, improving the signal-to-noise ratio by more than three times. This design significantly enhances on-site detection efficiency, allowing the entire S2 step to be completed within 30 minutes, providing data support for rapid subsequent decision-making.

[0033] More importantly, the microfluidic structure effectively suppresses sample diffusion and cross-interference, ensuring that low concentrations (<0.1 mg / L) of tracers can still be reliably detected. In a field test at a power plant, the repeatability RSD of this device for detecting five composite tracers was <5%, far superior to the >15% of discrete instruments. Therefore, this technology not only solves the engineering challenge of simultaneous detection of multiple signals but also lays the hardware foundation for the overall efficiency and accuracy of the method.

[0034] In S3, the pipeline topology model is integrated with a Geographic Information System (GIS), importing pipeline burial depth, diameter, direction, and hydraulic parameters to construct a three-dimensional hydraulic proximity map. The selection criteria for the candidate water supply subsystem set are: the Euclidean distance from the pipeline centerline to the leakage point is ≤50 m, and the hydraulic gradient direction points towards the leakage area. A weighted Bayesian model is used to calculate the... i Post-leakage probability of each candidate system Its expression is: ; In the formula, In the system i Under the premise of a confirmed leak, evidence was observed. E The likelihood probability is obtained by estimating the intensity of the tracer signal output by the detection device using Gaussian kernel density. In the first j Under the premise of a leak in a water supply subsystem, the current detection signal was observed. E The likelihood probability; Representation system iIt is the prior probability of the leakage source, that is, the probability given based on historical experience or risk assessment before observed data is available; Representation system j It is the prior probability of the leakage source; Prior probability , For the system i The standard deviation of operating pressure fluctuation, This refers to the service life of the pipeline. For historical leakage frequency, the function ( f, g, h) Let be a monotonically increasing normalized function in the interval [0,1], and let the weight coefficients satisfy... .

[0035] By introducing BIM-GIS fusion and a weighted Bayesian model, candidate systems are screened and the posterior probability of leakage is calculated, effectively overcoming the challenge of identifying multiple systems that are adjacent but only one is leaking. Traditional methods often rely on manual experience or simple threshold judgments, easily misidentifying nearby non-leaking systems as sources. This application constructs a three-dimensional hydraulic proximity map, using a distance ≤50 m and a hydraulic gradient pointing towards the leakage zone as hard screening conditions, reducing the candidate set from dozens to 2-3. By fusing three prior factors—operating pressure fluctuation (reflecting pipe wall fatigue), service life (reflecting aging risk), and historical leakage frequency (reflecting weak points)—through a Bayesian model, higher-risk systems are given higher weights. Subjective experience is transformed into objective probability, making the identification results interpretable and traceable.

[0036] In the embodiment, although both subsystems A and C detected signals, subsystem A had a high posterior probability of over 70% due to its large pressure fluctuations and long service life, successfully pinpointing the true leak source. This application can adapt to the historical data distribution of different plant areas.

[0037] In S4, the sustained-release fluorescent microsphere contrast agent is made of polylactic acid-glycolic acid copolymer encapsulating quantum dot fluorescent material with a particle size of 0.6~0.9μm and a polyethylene glycol anti-adsorption layer grafted on the surface. The sustained-release fluorescent microsphere contrast agent is released in water with zero-order kinetics for 1~5 hours, ensuring the formation of a continuous and stable fluorescent migration trajectory downstream of the leak point.

[0038] The slow-release fluorescent microsphere contrast agent, coated with polylactic-co-glycolic acid and polyethylene glycol (PLGA-PEG), solves the problems of rapid diffusion and short-lived signals in soil associated with traditional transient tracers. Ordinary fluorescent dyes are rapidly dispersed by water flow after injection, making it difficult to form a continuous trajectory and resulting in ambiguous positioning. In contrast, the microspheres of this invention have a particle size controlled at 0.6–0.9 μm, allowing them to migrate with water flow without being easily trapped by soil pores. The PLGA shell enables zero-order kinetic release, lasting 3–5 hours, ensuring a stable and high-intensity fluorescent migration band downstream of the seepage point. The surface PEG layer effectively inhibits non-specific adsorption of the microspheres on the pipe wall or soil particles, ensuring signal integrity.

[0039] In actual measurements, the microspheres migrated up to 8 meters in the clay layer with a fluorescence intensity decay of less than 20%, while ordinary dyes fell below the detection limit at 3 meters. This provides a clear and persistent signal source for ground-based sensing devices, a prerequisite for achieving sub-meter level inversion positioning. Furthermore, both PLGA and PEG are FDA-approved biomaterials, leaving no harmful residues and meeting industrial safety standards.

[0040] In S5, the ground sensing device is a fluorescence sensor network deployed in the suspected area or a drone platform equipped with an ultraviolet laser excitation source and a high-sensitivity sCMOS camera; the collected surface fluorescence intensity field The input is fed into the inversion model, and the coordinates of the leakage point are determined by solving the following optimization problem. With leakage flux : ; In the formula, Based on soil saturation permeability coefficient K Hydraulic gradient i and effective porosity n The constructed groundwater velocity field The simulated fluorescence intensity distribution; For Tikhonov regularization terms, For regularization parameters; The simulation process uses the finite element method to solve the convection-diffusion control equations: ; In the formula, The concentration of the contrast agent at a certain point in space. The time from the start of injection, This is a local coordinate system established with the leakage point as the origin. The dispersion coefficient; The coordinates of the pipeline leak point are the three-dimensional geographic coordinates of the actual leak point; the final output is the geographic coordinates of the leak point and the error ellipse with a confidence level of ≥90%.

[0041] By coupling Darcy's law with the inversion model of the convection-diffusion equation, the surface fluorescence intensity field is accurately mapped to the coordinates of underground seepage points, achieving sub-meter-level positioning accuracy. Traditional methods such as ground-penetrating radar or thermal imaging can only provide ambiguous anomaly areas, often with errors exceeding 2 meters. This invention, however, is based on physical mechanism modeling and acquires high-resolution fluorescence intensity fields through UAVs or sensor networks. Then, we can establish an optimization problem.

[0042] This model fully considers site parameters such as soil permeability coefficient K and groundwater flow velocity v, ensuring physical consistency in the inversion results. In a power plant case study, the actual leakage point (12.3, 8.7) deviated from the inverted result (12.1, 8.9) by only 0.28 m, with a confidence level of 92%. Compared to purely data-driven models, such as neural networks, this method does not require a large number of training samples, and the results are interpretable and verifiable. Therefore, this technology elevates leakage location from empirical speculation to physical inversion, enhancing the scientific rigor, accuracy, and reliability of the method.

[0043] Before executing S1, there is also an interference verification step: tracers with the same chemical composition but different time tags are injected into two adjacent water supply subsystems at a preset time interval Δt. If the time difference between the two signals in the seepage sample is less than Δt / 2, it is determined that there is groundwater backflow or pipeline cross-contamination interference, triggering the re-inspection mechanism.

[0044] By introducing an interference verification step and employing a time-stamped tracer injection strategy, the robustness of the method is significantly improved by actively identifying interference factors such as groundwater backflow or pipeline cross-contamination. In complex underground environments, seepage may originate from rainwater backflow, leakage from adjacent systems, or water conduction through geological faults, rather than leakage from the target system. Failure to distinguish these factors can lead to misjudgments. Before formal testing, this application injects tracers of the same composition but with different time stamps into adjacent systems at intervals of Δt = 10~30 minutes. If the time difference between two signals in the seepage sample is less than Δt / 2, it indicates the existence of a rapid connection path, such as a crack or cross-contamination, triggering a re-inspection mechanism. This design is equivalent to setting up an interference detector, eliminating false positives at the source. In a power plant test, this step successfully identified cross-contamination between systems caused by internal leakage in aging valves, avoiding incorrect localization.

[0045] When S2 detects a mixture of multiple composite tracer signals, a nonnegative matrix factorization algorithm is used to process the original signal matrix. Perform blind source separation and solve the following optimization problem: ; In the formula, V for m Various detection channels are available. n The signal matrix at each sampling time, Here are the characteristic response basis matrices for each tracer. Here are the activation intensity matrices for each system. r The number of candidate water supply subsystems. The sparsity regularization coefficient; F Represents the Frobenius norm; The reconstructed pure response intensity of each tracer is taken from the corresponding row of H and used to calculate the posterior probability of leakage of the candidate water supply subsystem set in S3, so as to achieve accurate identification of the target water supply subsystem.

[0046] Nonnegative matrix factorization (NMF) algorithm is used for blind source separation of mixed tracer signals, effectively solving the identification distortion problem caused by the overlap of signals from multiple systems. When multiple subsystems leak simultaneously, the tracer signals in the leaked water sample are superimposed on each other in the detection channel. Traditional linear unmixing methods, such as least squares, are prone to non-physical solutions such as negative concentrations. However, NMF ensures that the decomposition results conform to physical reality by constraining W≥0 and H≥0.

[0047] This algorithm can run in real time on embedded devices, providing support for on-site decision-making. Therefore, this technology not only solves the core challenge of signal aliasing, but also provides the algorithmic foundation for the scalability and intelligence of the method.

[0048] The steps between S3 and S5 also include a cross-validation step: drilling a micro-observation hole near the initially located leakage point, extracting groundwater samples for secondary tracer detection, and confirming the validity of the location result if the target tracer concentration is significantly higher than the background value.

[0049] By drilling micro-observation wells for secondary testing after initial positioning, false positives caused by parameter errors or model simplification in the inversion model can be effectively prevented. Although physical inversion is highly accurate, it still depends on the accuracy of soil parameters, such as the soil permeability coefficient K and groundwater flow velocity v. If the on-site survey has a large deviation, it may output incorrect coordinates.

[0050] This application involves drilling micro-holes with a diameter of only 20-50 mm within ±1 m of the inversion results to extract groundwater samples for tracer retesting. If the target system concentration is significantly higher than the background, the location is confirmed as valid; otherwise, model parameter correction or expansion of the search range is initiated. This step elevates the reliability of the location from model confidence to engineering confirmation. Traditional methods lack such verification mechanisms, often leading to ineffective excavation. This approach enhances the engineering credibility and implementation success rate of the method.

[0051] The injection concentration of the composite tracer is dynamically adjusted according to the pipeline operating conditions to meet the following requirements: ; In the formula, for t Injection concentration (mg / L) at time. DThe detection limit (mg / L) is the lowest detection limit of the testing equipment. The instantaneous flow rate of the subsystem is (m³ / h). For real-time pipeline pressure, For safety reasons, This is the pressure disturbance compensation factor.

[0052] This strategy ensures that the tracer signal in the seepage sample is always above the signal-to-noise ratio threshold, while suppressing misjudgments caused by pressure fluctuations.

[0053] Traditional fixed-concentration injections are prone to signal weakness (high-flow dilution) or over-strength (low-flow saturation) during flow fluctuations, affecting quantitative accuracy. This application adjusts the baseline concentration inversely to the instantaneous flow rate Q(t) and introduces an absolute value term for the rate of pressure change to compensate for transient disturbances such as water hammer and pump start-up / shutdown. Because sudden pressure changes exacerbate tracer pulse diffusion, additional dosage is required to maintain signal strength. Parameter k provides a safety margin, and γ is experimentally calibrated.

[0054] Example 2: This embodiment is applied to the leakage detection task of the underground water supply network of a power plant. The plant has five independent underground water supply subsystems, labeled A, B, C, D, and E, with the following functions: System A is circulating cooling water, DN800, made of carbon steel, buried at a depth of 3.2 m; System B is boiler feedwater, DN200, made of stainless steel, buried at a depth of 2.8 m; System C is fire-fighting water, DN300, made of ductile iron, buried at a depth of 3.0 m; System D is desulfurization process water, DN250, made of UPVC, buried at a depth of 2.5 m; System E is domestic water supply, DN150, made of PE, buried at a depth of 2.6 m.

[0055] A composite tracer with a unique coding feature is configured for each water supply subsystem. The specific formulation is shown in Table 1.

[0056] Two hours after injection, a 10 mL sample was collected from the seepage point. Analysis was performed using a multimodal detection device, the core of which is a 5 cm × 3 cm polydimethylsiloxane (PDMS) microfluidic chip integrating four functional areas: a sample pretreatment area, a fluorescence detection area (containing five-channel filters of 780 / 820 / 750 / 800 / 810 nm), a magnetoresistive sensing array (based on the giant magnetoresistive (GMR) effect), and a waste collection area.

[0057] The test results are shown in Table 2.

[0058] The results showed that the tracer concentrations in subsystems A and C were significantly higher than the detection limit of 0.1 mg / L, and the leak source was preliminarily determined to be located in one of them.

[0059] Based on the BIM-GIS fusion model, the Euclidean distance from the centerline of each pipeline to the seepage point (coordinates X=1250.3m, Y=876.5m) was calculated. The results show that systems A (nearest point distance 42.1 m) and C (48.7 m) meet the condition of ≤50 m, and their hydraulic gradient directions both point towards the seepage area, so they are included in the candidate set; the distances of the remaining systems all exceed 60 m and are excluded.

[0060] Subsequently, the operating data of the two subsystems were retrieved: Subsystem A: Standard deviation of pressure fluctuations over the past 30 days = 0.08 MPa, service life = 18 years, number of historical leaks = 1 time; Subsystem C: = 0.03 MPa, = 5 years, = 0 times.

[0061] The normalization function is defined as follows: ; The weighting coefficients were determined through historical case regression as w1=0.5 (pressure fluctuation weight), w2=0.3 (year weight), and w3=0.2 (historical frequency weight).

[0062] Calculate the prior probability: , For the system i The standard deviation of operating pressure fluctuation, This refers to the service life of the pipeline. For historical leakage frequency, the function ( f, g, h) Let be a monotonically increasing normalized function in the interval [0,1], and let the weight coefficients satisfy... .

[0063] ; ; Likelihood function The probability density was obtained through Gaussian kernel density estimation: using the standard deviation of repeated measurements (0.05 mg / L) as the bandwidth, the probability density corresponding to the observed concentrations (A: 0.81, C: 0.77) was calculated. =0.91, =0.85.

[0064] According to the weighted Bayesian model, the general formula for calculating the posterior probability of leakage is: ; Where n represents the number of candidate water supply subsystems. In this embodiment, n = 2 (systems A and C), substituting, we get: ; ; because If the value is greater than 0.7, system A is determined to be the target of leakage.

[0065] The contrast agent was injected upstream of system A at a concentration of 2 mg / L, with a total volume of 3 liters, over an injection period of 20 minutes. The microspheres migrated with the water flow within the pipeline, entering the soil pores upon encountering a leak. Due to the PEG anti-adsorption layer, they were not easily captured by soil particles, continuously releasing fluorescent substances for up to 4 hours and forming a clear migration trajectory.

[0066] A drone platform was used for surface scanning. The drone was equipped with a 365 nm ultraviolet laser with a power of 500 mW as the excitation source and an sCMOS camera as the detector. The flight altitude was 5 m, and the scanning area was 10 m × 10 m, centered on the seepage point.

[0067] Obtaining the fluorescence intensity field Then, input the inversion model. The groundwater velocity field v is calculated using Darcy's law from the following site parameters: Soil saturated permeability coefficient K Through on-site double-ring water injection test, it was determined that... K =1.0×10 -5 m / s; Hydraulic gradient i Calculated from the digital elevation model (DEM) and groundwater level monitoring data, i = 0.005; Effective porosity n Based on the soil type table, n=0.35; Therefore, the groundwater flow velocity is: ; This groundwater flow velocity is used to construct a groundwater flow velocity field. v This provides core parameters for subsequent solutions to the convection-diffusion equations and simulations of fluorescence intensity distribution.

[0068] The following optimization problem is established to determine the parameters of the leakage point: ; Among them, simulated fluorescence intensity field The following was obtained by solving the convection-diffusion control equations using the finite element method:

[0069] D is the dispersion coefficient, taken as 1×10⁻⁶. -6m² / s; the regularization parameter λ=0.01 was determined by the L-curve method.

[0070] After iterative optimization, the optimal solution is output: =12.1 m, =8.9 m, =-3.2 m (relative to the coordinate system of the seepage point), meaning the actual coordinates of the seepage point in the pipeline are... =12.1 m, =8.9 m, depth is 3.2 m. = 0.8 L / min, confidence level 92%. Error ellipse major axis 0.4 m, minor axis 0.3 m.

[0071] To ensure the reliability of the results, cross-validation was performed: a 40mm diameter micro-observation borehole was drilled at the (12.1, 8.9) location to a depth of 3.5 m, and a 100 mL groundwater sample was extracted. Microfluidic chip analysis showed that the concentration of tracer A in system A reached 0.79 mg / L, far exceeding the background value, confirming the effective localization.

[0072] Before injecting the main tracer, tracers with the same composition but different time labels were injected into systems A and B at intervals of Δt = 20 minutes. The time difference between the two signals in the seepage sample was 18 minutes (>Δt / 2 = 10 minutes), indicating no groundwater backflow interference.

[0073] During the testing period, the flow rate Q(t) of system A fluctuated between 80 and 120 m³ / h, and the pressure change rate reached a maximum of 0.02 MPa / min.

[0074] Based on the dynamic injection concentration formula for composite tracers; ; In the formula, for t Injection concentration (mg / L) at time. D The detection limit (mg / L) is the lowest detection limit of the testing equipment. The instantaneous flow rate of the subsystem is (m³ / h). For real-time pipeline pressure, For safety reasons, This is the pressure disturbance compensation factor. The absolute value sign is introduced to reflect the intensity of the pressure change disturbance; whether the pressure increases or decreases, the tracer concentration needs to be increased to maintain the signal-to-noise ratio.

[0075] In this embodiment, k=1.8, D=0.1 mg / L, and γ=0.2 were selected. The injection concentration was calculated and adjusted in real time, ranging from 0.85 to 1.25 mg / L, to ensure that the signal-to-noise ratio of the seepage sample was always higher than 15dB.

[0076] Furthermore, when simulating a three-system aliasing scenario, tracers A, C, and E are artificially mixed, and the nonnegative matrix factorization (NMF) algorithm is used to perform blind source separation on the signal matrix V: ; With α=0.01 and β=0.005, the separation error is 8.3%, which is significantly better than the 14.2% of principal component analysis (PCA), verifying the superiority of the present invention in signal demixing.

[0077] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0078] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A method for detecting leakage points in a complex underground multi-pipe water supply system, characterized in that: The multiple water supply system includes multiple underground and independent water supply subsystems. The method for detecting leaks includes the following steps: S1. For each water supply subsystem, inject a composite tracer with uniquely identifiable physicochemical characteristics; S2. Collect seepage samples from the seepage area and test the seepage samples using detection equipment; S3. Based on the pipeline network topology model and the location of the seepage point, a set of candidate water supply subsystems is selected, and the target water supply subsystem where the leak occurred is determined by combining the concentration signal of the composite tracer. S4. Inject a slow-release fluorescent microsphere contrast agent upstream of the target water supply subsystem; S5. Real-time monitoring of surface fluorescence intensity distribution using ground-based sensing devices, and inversion model based on coupled Darcy's law and convection-diffusion equation to determine the geographic coordinates of pipeline leakage points.

2. The detection method according to claim 1, characterized in that, In S1, the composite tracer is a multimodal encoded tracer containing two or more orthogonal response signals. The orthogonal response signals are selected from: near-infrared fluorescent dyes with different center wavelengths, superparamagnetic nanoparticles with different saturation magnetization intensities, and pH-sensitive azo color-changing groups. The signal combinations of the composite tracers used in each water supply subsystem do not overlap in the multidimensional feature space.

3. The method of claim 1, wherein, In S2, the detection device includes an integrated microfluidic chip, which has a multi-channel optical detection area and a magnetoresistive sensing array. It can simultaneously complete fluorescence spectral scanning and magnetic signal intensity measurement in a single sample injection, and output the quantitative concentration value of each tracer by executing signal normalization and background subtraction algorithms through an embedded processor.

4. The detection method according to claim 1, characterized in that, In S3, the pipeline topology model is integrated with a Geographic Information System (GIS) to construct a three-dimensional hydraulic proximity map. The selection criteria for the candidate water supply subsystem set are: the Euclidean distance from the pipeline centerline to the leakage point is ≤50 m, and the hydraulic gradient direction points towards the leakage area. A weighted Bayesian model is used to calculate the... i Post-leakage probability of each candidate system Its expression is: ； wherein is the likelihood probability of observing the evidence i given that there is a leak, the tracer signal strength output by the detection device is estimated by a Gaussian kernel density estimator; E is the likelihood probability of observing the current detection signal given that there is a leak in the first j water supply subsystem; E is the likelihood probability of observing the current detection signal is the prior probability that the system i is the source of the leak, i.e. the probability given by historical experience or risk assessment before observing the data; is the prior probability that the system j is the source of the leak. Prior probability , For the system i The standard deviation of operating pressure fluctuation, This refers to the service life of the pipeline. For historical leakage frequency, the function ( f, g, h) Let be a monotonically increasing normalized function in the interval [0,1], and let the weight coefficients satisfy... .

5. The detection method according to claim 1, characterized in that, In S4, the sustained-release fluorescent microsphere contrast agent is made of polylactic acid-glycolic acid copolymer encapsulating quantum dot fluorescent material with a particle size of 0.6~0.9μm and a polyethylene glycol anti-adsorption layer grafted on the surface. The sustained-release fluorescent microsphere contrast agent is released in water with zero-order kinetics for 1~5 hours, ensuring the formation of a continuous and stable fluorescent migration trajectory downstream of the leak point.

6. The detection method according to claim 1, characterized in that, In S5, the ground sensing device is a fluorescence sensor network deployed in the suspected area or a drone platform equipped with an ultraviolet laser excitation source and a high-sensitivity sCMOS camera; the collected surface fluorescence intensity field The input is fed into the inversion model, and the coordinates of the leakage point are determined by solving the following optimization problem. With leakage flux : ； In the formula, Based on soil saturation permeability coefficient K Hydraulic gradient i and effective porosity n The constructed groundwater velocity field The simulated fluorescence intensity distribution; For Tikhonov regularization terms, For regularization parameters; The simulation process uses the finite element method to solve the convection-diffusion control equations: ； In the formula, The concentration of the contrast agent at a certain point in space. The time from the start of injection, This is a local coordinate system established with the leakage point as the origin. The dispersion coefficient; These are the coordinates of the leak point, i.e., the three-dimensional geographic coordinates of the actual leak point; The final output includes the geographic coordinates and error ellipse of the leak point with a confidence level of ≥90%.

7. The detection method according to claim 1, characterized in that, Before executing S1, there is also an interference verification step: tracers with the same chemical composition but different time tags are injected into two adjacent water supply subsystems at a preset time interval Δt. If the time difference between the two signals in the seepage sample is less than Δt / 2, it is determined that there is groundwater backflow or pipeline cross-contamination interference, triggering the re-inspection mechanism.

8. The detection method according to claim 1, characterized in that, When S2 detects a mixture of multiple composite tracer signals, a nonnegative matrix factorization algorithm is used to process the original signal matrix. Perform blind source separation and solve the following optimization problem: ； In the formula, V for m Various detection channels are available. n The signal matrix at each sampling time, Here are the characteristic response basis matrices for each tracer. Here are the activation intensity matrices for each system. r The number of candidate water supply subsystems. The sparsity regularization coefficient; F Represents the Frobenius norm; The reconstructed pure response intensities of each tracer are taken from the corresponding row of H and used for accurate discrimination of S3.

9. The detection method according to claim 1, characterized in that, The steps between S3 and S5 also include a cross-validation step: drilling a micro-observation hole near the initially located leakage point, extracting groundwater samples for secondary tracer detection, and confirming the validity of the location result if the target tracer concentration is significantly higher than the background value.

10. The detection method according to claim 1, characterized in that, The injection concentration of the composite tracer is dynamically adjusted according to the pipeline operating conditions to satisfy: ； In the formula, for t Injection concentration (mg / L) at time. D The detection limit (mg / L) is the lowest detection limit of the testing equipment. The instantaneous flow rate of the subsystem is (m³ / h). For real-time pipeline pressure, For safety reasons, This is the pressure disturbance compensation factor; This strategy ensures that the tracer signal in the seepage sample is always above the signal-to-noise ratio threshold, while suppressing misjudgments caused by pressure fluctuations.

Images