Underground anti-seepage structure leakage monitoring method and system based on discrete time

Abstract

The invention discloses an underground anti-seepage structure leakage monitoring method and system based on discrete time, and relates to leakage detection. The method comprises the following steps: collecting transient electromagnetic response data generated by an excitation signal; according to the initial conductivity parameter and the electrode position parameter of the monitoring area, establishing a three-dimensional geophysical model of the monitoring area through a full waveform inversion algorithm; obtaining electromagnetic abnormal data by using a finite element forward modeling algorithm; according to the temperature data, a distributed temperature field analysis method is adopted to calculate the space gradient and the time change rate of the temperature, and temperature abnormal data are obtained; according to the pressure data and time sequence change characteristics of the pressure data, determining pressure abnormal data; fusing the electromagnetic abnormal data, the temperature abnormal data and the pressure abnormal data by using a Bayesian algorithm to obtain fused data; and inputting the fusion data into a pre-trained machine learning model to obtain a leakage detection result. Aiming at the low dynamic leakage monitoring precision of the underground anti-seepage structure, the dynamic monitoring precision is improved.

Description

Technical Field

[0001] The present application relates to the field of leakage detection, and in particular to a discrete-time-based underground anti-seepage structure leakage monitoring method and system. Background Art

[0002] With the acceleration of urbanization and the deepening development of underground space, underground anti-seepage structures have become a critical component of modern construction projects. Statistics show that in structures below the groundwater level, leakage-induced concrete carbonization increases by 300% and the risk of steel corrosion by 47%. For example, in a typical deep foundation pit project, when the underground diaphragm wall depth exceeds 30 meters, the probability of leakage at the waterproofing joints is as high as 18.6%, seriously threatening the safety and durability of the structure. Therefore, high-precision dynamic leakage monitoring of underground anti-seepage structures is of great significance.

[0003] However, the existing leakage monitoring technology for underground anti-seepage structures has serious deficiencies in dynamic monitoring accuracy. The traditional natural potential method requires the electrodes to be re-deployed each time a test is performed, making it impossible to achieve continuous dynamic monitoring. In addition, the detection sensitivity for small leaks is low, with a detection accuracy of only about 70%. Although the tracer method can locate the leakage path, it requires drilling holes on the outside of the anti-seepage structure to inject tracers, which can only detect large-scale leaks and cannot capture the dynamic development process of the leak. More importantly, the existing technology lacks an effective multi-source data fusion mechanism: (1) The single physical field detection method is easily affected by environmental interference, resulting in a false alarm rate of up to 15-20%; (2) It is impossible to achieve spatiotemporal synchronous acquisition and comprehensive analysis of multi-physical field data such as electromagnetic fields, temperature fields, and pressure fields; (3) There is a lack of a dynamic data processing framework based on discrete time windows, making it difficult to accurately extract abnormal features caused by leakage; (4) There is no effective background field separation mechanism, and it is impossible to distinguish the signal differences between normal and leakage states. These problems have led to the inability of existing technologies to meet the needs of high-precision, real-time dynamic monitoring of underground anti-seepage structures throughout their entire life cycle. Summary of the Invention

[0004] In view of the low accuracy of dynamic leakage monitoring of underground anti-seepage structures, the present application provides a method and system for underground anti-seepage structure leakage monitoring based on discrete time, which uses a transient electromagnetic excitation source to actively detect and collect response data, and extracts electromagnetic anomaly, pressure and temperature anomaly data based on discrete time windows, performs multi-source fusion, etc., thereby improving the accuracy of dynamic monitoring.

[0005] One aspect of the present application provides a discrete-time-based underground anti-seepage structure leakage monitoring method, comprising: S1, collecting underground anti-seepage structure data including current data, temperature data, and pressure data through a sensor array; S2, sending an excitation signal to a monitoring area through a signal excitation source, and collecting transient electromagnetic response data generated by the excitation signal through a sensor array; S3, setting a discrete time window, and constructing a spatiotemporal data matrix based on the spatial coordinates of each sensor in the sensor array and the data acquisition timestamp; performing spatiotemporal alignment of underground anti-seepage structure data and transient electromagnetic response data based on the spatiotemporal data matrix; S4, establishing a three-dimensional geophysical model of the monitoring area through a full waveform inversion algorithm based on the initial conductivity parameters and electrode position parameters of the monitoring area; and performing a finite element forward modeling algorithm on the transient electromagnetic response data after the spatiotemporal alignment of S3 based on the three-dimensional geophysical model. The background field modeling is performed based on the response data to obtain the background electromagnetic field distribution in the no-leakage state; a differential operation is performed on the transient electromagnetic response data and the background electromagnetic field distribution after the spatiotemporal alignment in step S3 to obtain electromagnetic anomaly data; S5, based on the temperature data in the underground anti-seepage structure data after the spatiotemporal alignment in S3, a distributed temperature field analysis method is used to calculate the spatial gradient and time change rate of the temperature to obtain temperature anomaly data; S6, the pressure data in the underground anti-seepage structure data after the spatiotemporal alignment in S3 is analyzed, and the pressure anomaly data is determined based on the time series change characteristics of the pressure data; S7, the electromagnetic anomaly data, the temperature anomaly data and the pressure anomaly data are fused using the Bayesian algorithm to obtain fused data; S8, the fused data is input into the pre-trained machine learning model to obtain the leakage detection result, which includes the leakage probability, leakage location and leakage level.

[0006] In this application, underground anti-seepage structures refer to structural systems used to prevent groundwater infiltration in underground projects, including underground continuous walls, waterproof concrete structures, water-stop curtains, and anti-seepage membranes. These structures are typically constructed with low-permeability materials, and their integrity is directly related to the safety of underground projects. This program monitors potential leakage defects in these anti-seepage structures.

[0007] Transient electromagnetic response data refers to the time-varying electromagnetic field response signal generated by the underground medium under the action of an excitation signal. When the bipolar square wave excitation signal is suddenly turned off, the induced eddy currents in the underground conductive medium do not disappear immediately, but instead decay exponentially. The secondary electromagnetic field generated by this decay process is the transient response.

[0008] The conductivity parameter is a physical parameter that characterizes the electrical conductivity of a material, and its unit is Siemens / meter (S / m). In this application, the initial conductivity parameter This is an initial estimate of the conductivity of each stratum in the monitoring area based on geological survey data, typically ranging from 0.01 to 0.1 S / m. Leakage can alter local conductivity, so accurate modeling of conductivity distribution is fundamental to leak detection.

[0009] Electrode position parameters refer to the three-dimensional spatial coordinates of the power supply electrodes and measurement electrodes arranged in the monitoring area. and electrode spacing information. In this application, electrodes are arranged in a matrix to stimulate current and collect potential responses. Accurate determination of electrode positions (with an accuracy better than 0.1 m) is a prerequisite for ensuring inversion accuracy.

[0010] The full waveform inversion algorithm is a geophysical inversion method that uses iterative optimization to achieve the best fit between the theoretical model response and the measured waveform data. In this application, the algorithm uses both DC and AC data to minimize the objective function , the three-dimensional conductivity distribution model is obtained by inversion , providing a background model for identifying leakage anomalies.

[0011] A three-dimensional geophysical model is a mathematical model that describes the spatial distribution of the physical properties of the underground medium in the monitoring area. In this application, the model includes conductivity The three-dimensional distribution information of dielectric constant ε and magnetic permeability μ is discretized through finite element meshing, which is the basis for numerical simulation of electromagnetic fields.

[0012] Finite element forward algorithm is a numerical calculation method based on the finite element method to solve Maxwell's equations and simulate the propagation of electromagnetic fields in complex media. In this application, the algorithm calculates the electromagnetic field distribution under the action of the excitation signal based on the three-dimensional geophysical model and obtains the background field in the non-leakage state. , providing a reference benchmark for anomaly identification.

[0013] Furthermore, S2 sends an excitation signal to the monitoring area through a signal excitation source, and collects transient electromagnetic response data generated by the excitation signal through a sensor array, including: generating a bipolar square wave excitation signal; a bipolar square wave excitation signal, a square wave current signal with alternating positive and negative polarity. In the present application, its amplitude range is 0-±1000V, the pulse width is adjustable from 0.1ms to 1000ms, and the frequency is set according to the detection depth requirements. The bipolar design can eliminate DC bias and electrode polarization effects, and the wide spectrum characteristics of the square wave are conducive to the simultaneous acquisition of geoelectric information at different depths. According to the geological conditions and monitoring depth requirements of the monitoring area, the frequency parameters of the bipolar square wave excitation signal are adjusted; a multi-channel parallel excitation mode is adopted to send bipolar square wave excitation signals to multiple monitoring areas;

[0014] The transient electromagnetic response data of each monitoring area is collected through the sensor array. The transient electromagnetic response data includes natural potential baseline DC data and transient response AC data. Among them, the natural potential baseline DC data is the steady-state natural potential distribution measured when no artificial excitation is applied, which reflects the static electric field characteristics of the underground medium. In this application, the DC data amplitude range is usually -100mV to +100mV, which is mainly generated by natural processes such as groundwater flow and redox reactions, and can be used to identify existing stable leakage channels. Transient response AC data is the secondary field signal that decays with time and is measured after the excitation signal is turned off. In this application, the effective time window of AC data is 0.1ms-100ms, and the peak value can reach 200mV. This data contains the electrical structure information of the underground medium, is particularly sensitive to new leakage and dynamically changing leakage processes, and is the core observation quantity of the transient electromagnetic method.

[0015] In particular, traditional passive electrical monitoring relies solely on subtle changes in the natural electric field to detect leaks, resulting in low signal strength, susceptibility to interference, and poor positioning accuracy. Active excitation technology, on the other hand, generates a primary field of sufficient intensity by artificially applying a controllable electromagnetic excitation signal. When this primary field propagates through the underground medium, it generates a secondary induction field due to differences in the medium's electrical properties. By analyzing the total field response characteristics, underground anomalies can be accurately detected. Using a bipolar square wave as an excitation signal offers unique advantages: alternating positive and negative pulses eliminates DC bias and polarization effects, and the square wave's wide frequency spectrum allows for simultaneous excitation of multiple frequency components, facilitating the acquisition of geoelectrical information at varying depths. This active detection mode improves the signal-to-noise ratio by one to two orders of magnitude, significantly enhancing detection sensitivity and spatial resolution.

[0016] In this application, on the one hand, precise control of the detection depth is achieved by flexibly adjusting the frequency parameters of the bipolar square wave: low-frequency signals (0.1-10 Hz) have stronger penetration ability and can detect deep leaks (>30 m), while high-frequency signals (10-1000 Hz) have fast attenuation but high resolution and are suitable for shallow fine detection (<10 m). This adaptive frequency adjustment based on geological conditions and monitoring needs ensures effective detection of leaks at different depths; on the other hand, the innovative use of a multi-channel parallel excitation mode breaks through the efficiency bottleneck of traditional single-point excitation: multiple excitation sources work simultaneously but use different coding modulation (such as orthogonal code division), and the receiving end can separate the responses of each channel through decoding, realizing multi-region synchronous monitoring without increasing the measurement time, and improving the monitoring efficiency by 4-8 times. What is particularly important is that the system simultaneously collects natural potential baseline DC data and transient response AC data. The DC data reflects the steady-state electric field distribution and can identify existing leakage channels. The AC data contains dynamic information about the transient process and is extremely sensitive to new leakage and leakage development. The combination of the two enables complete monitoring of the entire life cycle of leakage, providing rich raw data for subsequent full waveform inversion and anomaly identification.

[0017] Furthermore, the amplitude range of the bipolar square wave excitation signal is 0-±1000V, and the pulse width range is 0.1ms-1000ms.

[0018] Further, S3, constructs a spatiotemporal data matrix; performs spatiotemporal alignment of underground anti-seepage structure data and transient electromagnetic response data based on the spatiotemporal data matrix, including: setting a discrete time window The time interval is 10s to 24h. Through time slicing, the continuous data stream is divided into discrete time segments. The discrete time segments are used as the basic time units for data fusion. According to the three-dimensional spatial coordinates of each sensor, and the data acquisition timestamp t, constructing a four-dimensional spatiotemporal data matrix , used to store the measurement values of each sensor in each discrete time segment; for sensor data with different sampling frequencies, discrete time windows are used Based on the spatiotemporal data matrix, the sensor data are synchronized to the same time node through the interpolation algorithm; the time-synchronized sensor data are spatially aligned to obtain spatiotemporal aligned data. , the underground anti-seepage structure data and transient electromagnetic response data after time synchronization are converted into the corresponding spatial coordinates The time node t is mapped to the corresponding position of the spatiotemporal data matrix to achieve unified alignment of data in the time dimension and space dimension, and obtain spatiotemporal aligned data.

[0019] Further, S4, based on the initial conductivity parameters and electrode position parameters, a three-dimensional geophysical model of the monitoring area is established by full waveform inversion algorithm; the initial conductivity parameters of the monitoring area are obtained. , initial conductivity parameters Based on geological survey data, the typical value range is 0.01S / m-0.1S / m; for multi-layer geological structures, the initial conductivity value of each layer is set layer by layer; the electrode position parameters are obtained, including the three-dimensional spatial coordinates of the electrode and electrode spacing d, the electrodes are arranged in an m×n matrix with a spacing of d=2m-10m.

[0020] Extraction of spontaneous potential baseline DC data from spatiotemporally aligned data and transient response AC data The DC data is low-pass filtered (cut-off frequency 1Hz) to eliminate high-frequency noise; the AC data is time-windowed to retain the effective attenuation segment of 0.1ms-100ms after the excitation signal is turned off; in particular, the conductivity distribution of the underground anti-seepage structure reflects the physical state of the medium. The full waveform inversion algorithm can reconstruct the true three-dimensional spatial distribution of the conductivity of the underground medium by simultaneously utilizing the natural potential baseline (DC) and transient response (AC) data. This method is based on the propagation laws of electromagnetic fields in conductive media and obtains medium parameters by solving an inverse problem. Unlike the traditional homogeneous medium assumption, this application takes into account the heterogeneity of the underground medium, providing a reliable reference benchmark for subsequent anomaly identification.

[0021] Establish the objective function: ,in, is the measured data, is the model response, λ is the regularization parameter (value ranges from 0.01 to 0.1); the conjugate gradient method is used for iterative optimization, and the iterative step size is adaptively adjusted: , where k is the number of iterations; when the relative change of the objective function of two adjacent iterations is less than Or stop when the number of iterations reaches 100; output the three-dimensional conductivity distribution model , the spatial resolution reaches 0.5m×0.5m×0.5m;

[0022] based on , the monitoring area is meshed using tetrahedron elements, and the mesh size is adaptively adjusted according to the conductivity gradient: ,in, represents the minimum mesh size allowed in mesh generation, h represents the local mesh size, represents the magnitude of the conductivity gradient vector; Represents the maximum value of the conductivity gradient modulus in the monitoring area; solve Maxwell's equations in the time domain: , , where the conductivity term is introduced , E represents the electric field intensity, B represents the magnetic induction intensity, H represents the magnetic field intensity, D represents the electric displacement, and J represents the current density; boundary conditions are set: the far field boundary adopts the absorbing boundary condition, and the surface adopts the air-ground interface condition; a complete three-dimensional geophysical model including the spatial distribution of conductivity σ, the distribution of dielectric constant ε, and the distribution of magnetic permeability μ is established ;

[0023] Input excitation signal parameters: bipolar square wave, amplitude A, pulse width τ, frequency f; using finite element forward algorithm, at each time step , solve the electromagnetic field propagation equation; consider the reflection and refraction effects of the stratum interface, use the interface continuity condition to deal with multilayer media; calculate the background electromagnetic field under the non-leakage state , including the three components of the electric field ; The measured transient electromagnetic response data and background electromagnetic fields Perform point-wise differencing: ;

[0024] In particular, the background electromagnetic field in the no-leakage state is calculated by the finite element forward algorithm. , the background field includes the influence of normal underground geological structure on electromagnetic field. When leakage occurs, the leakage channel changes the local conductivity and generates additional abnormal electromagnetic field. This application eliminates the "false anomalies" caused by geological structure, such as natural electrical differences such as stratum interfaces and faults; improves the recognition ability of weak leakage signals, even if the abnormal signal only accounts for 5%-10% of the total signal, it can be detected; realizes quantitative abnormality assessment, through Directly obtain the change in electromagnetic field caused by leakage.

[0025] Furthermore, the finite element method considers the complete physical process of the electromagnetic field when solving Maxwell's equations in the time domain: the propagation of the primary field generated by the excitation source in the medium; the secondary field response caused by differences in the medium's electrical conductivity; and the eddy current effects and diffusion processes during transients. This forward modeling, based on physical mechanisms, more accurately describes the distribution characteristics of the electromagnetic field in complex underground environments than empirical formulas or simplified models, providing a high-fidelity background field model.

[0026] Finally, this application captures the dynamic process of leakage development and distinguishes instantaneous interference from continuous anomalies by differentiating the time-space aligned data with the accurate background field model. Traditional methods can only detect static anomalies at a certain moment, while this application can detect the dynamic process of leakage development and distinguish instantaneous interference from continuous anomalies. Dynamic monitoring of four-dimensional space-time anomalies is achieved.

[0027] Further, S5, obtaining temperature anomaly data, including: temperature data in the spatiotemporal alignment data , calculate the temperature spatial gradient of each monitoring point ; Calculate the temperature time change rate of each monitoring point ; When the temperature spatial gradient modulus of the monitoring point When the gradient is greater than the preset threshold, the corresponding monitoring point is marked as a temperature spatial anomaly point; when the temperature time change rate of the monitoring point is greater than the preset gradient threshold, the corresponding monitoring point is marked as a temperature spatial anomaly point; When the temperature rise is greater than the preset threshold, the corresponding monitoring point is marked as a time-varying temperature anomaly point. Based on the physical mechanism of local heat transfer anomaly caused by groundwater leakage, the temperature field distribution is analyzed by solving the heat conduction equation to verify whether the temperature spatial anomaly point and the temperature identification anomaly point meet the temperature field change characteristics caused by leakage. Among them, by solving the heat conduction equation: ,in, is density, c is specific heat capacity, k is thermal conductivity, and q is the heat source term (introduced by leakage). Determine whether the detected anomaly meets the thermodynamic characteristics of leakage and distinguish leakage anomalies from other heat sources (such as geothermal, pipelines, etc.). For abnormal points that pass verification, determine them as temperature anomaly monitoring points; extract the temperature data corresponding to the temperature anomaly monitoring points. , temperature spatial gradient and temperature-time rate of change , as temperature anomaly data.

[0028] In particular, on the one hand, Reflects the rate of change of temperature in three-dimensional space. An abnormal temperature gradient will form at the leakage channel: horizontal gradient : Indicates the lateral extension of the leakage path, vertical gradient Reflects the vertical migration characteristics of leakage. The spatial gradient is sensitive to the boundaries of the leakage channel. The peak position of |∇T| corresponds to the leakage boundary. The gradient direction indicates the direction of heat flow and indirectly reflects the direction of water flow. The spatial continuity of the gradient anomaly can be used to track the leakage path.

[0029] on the other hand, It reflects the rate of temperature change over time. A positive value indicates warming, which may be caused by the infiltration of groundwater with a higher temperature; a negative value indicates cooling, which may be caused by the infiltration of surface water with a lower temperature. The magnitude of the rate of change reflects the leakage intensity and water flow velocity.

[0030] The present application utilizes temperature field changes as an indirect indicator of leakage, provides a detection method independent of the electromagnetic field, and enhances the reliability of the monitoring system.

[0031] Furthermore, the sensor array includes a potential sensor, a temperature sensor and a pressure sensor; the pressure sensor adopts a fiber grating sensor.

[0032] Further, S6, obtaining pressure anomaly data includes: extracting reflection wavelength data of the fiber Bragg grating sensor ; According to the wavelength-pressure relationship Calculate pressure change ,in, is the initial wavelength, is the wavelength change, K is the stress sensitivity coefficient; for pressure change Perform timing analysis, when the pressure change rate When the pressure change rate is greater than the preset threshold, the corresponding monitoring point is regarded as the pressure mutation point; when the pressure change value is greater than the preset threshold, the corresponding monitoring point is regarded as the pressure mutation point; When the pressure is greater than the threshold range, the corresponding monitoring point is regarded as an abnormal fluctuation point; the pressure data corresponding to the monitoring point containing the pressure mutation point and the abnormal fluctuation point is regarded as pressure abnormal data.

[0033] In particular, fiber Bragg grating pressure sensing technology has shown unique technical advantages in the leakage monitoring of underground anti-seepage structures. Its working principle is based on the optical properties of Bragg gratings: fiber Bragg gratings form Bragg reflections through periodic refractive index modulation. When external pressure acts on the optical fiber, it causes the grating period to and effective refractive index changes in the reflected wavelength This wavelength-encoded measurement method is inherently safe, requiring no power supply and fundamentally avoiding electrical safety hazards. Furthermore, wavelength information is unaffected by light source power fluctuations and transmission losses, ensuring long-term measurement stability. In practical applications, wavelength resolution can reach 1pm, corresponding to a pressure resolution of 0.1kPa, fully meeting the requirements for detecting minute leaks. Furthermore, optical signal transmission is completely immune to electromagnetic interference, making it particularly suitable for the complex electromagnetic environments found in underground anti-seepage structures.

[0034] In this application, on the one hand, by establishing a linear relationship between wavelength and pressure The precise pressure measurement model realizes the quantitative detection of pore water pressure changes caused by leakage. The calibration of stress sensitivity coefficient K ensures the accuracy and repeatability of measurement. On the other hand, the comprehensive identification of different leakage modes is achieved by setting dual abnormality criteria: pressure mutation point ( Greater than the threshold) is used to identify rapidly developing dangerous leaks, such as sudden damage to the anti-seepage structure causing a sharp change in pressure; abnormal fluctuation points ( A time derivative criterion (a value greater than a threshold) is used to detect steady but abnormal pressure states, such as persistent pressure deviations caused by slowly developing leak channels. This detection strategy, which combines the time derivative criterion with the amplitude criterion, can capture transient pressure pulses and identify long-term pressure excursions.

[0035] Further, S7, obtains fused data, including: calculating the conditional probability of electromagnetic anomalies and leakage events based on historical monitoring data , the conditional probability of temperature anomaly and leakage event , and the conditional probability of pressure anomaly and leakage events ; Based on the Bayesian algorithm, calculate the posterior probability P of multi-source data fusion; set the conditional probability according to the signal-to-noise ratio of each sensor 、 and The weight coefficient ; , , ,in, They are the signal-to-noise ratios of the potential sensor, temperature sensor, and pressure sensor respectively; the electromagnetic abnormality data, temperature abnormality data, and pressure abnormality data are calculated based on the weight coefficients Perform weighted calculation to obtain comprehensive anomaly indicators; obtain fused data based on the comprehensive anomaly indicators and the posterior probability P.

[0036] In particular, the Bayesian algorithm treats the detection results of different sensors as independent evidence and, by combining conditional probability with prior knowledge, calculates the posterior probability of a leakage event given multiple pieces of evidence. This probabilistic reasoning mechanism not only handles the uncertainty of various sensor data but also quantitatively assesses the contribution of each abnormal feature to leakage judgment. Compared to simple threshold judgment or linear superposition, Bayesian fusion can fully utilize the statistical laws inherent in historical data. This is particularly true in complex environments such as underground anti-seepage structure monitoring, where single physical field detection is susceptible to various interferences. Bayesian fusion significantly improves the accuracy and robustness of leakage detection through cross-validation of multi-source information.

[0037] In this application, on the one hand, by deeply mining historical monitoring data, a conditional probability model based on statistical learning is established: the conditional probability of electromagnetic anomalies and leakage events Reflects the reliability of electromagnetic field changes as leakage indicators, and the conditional probability of temperature anomalies and leakage events Quantified the correlation between temperature field disturbance and leakage, and the conditional probability between pressure anomaly and leakage event The probability that pressure change indicates leakage is represented. The accurate estimation of these three conditional probabilities lays a solid foundation for subsequent probability reasoning. On the other hand, the dynamic weight allocation mechanism based on signal-to-noise ratio is innovatively introduced. Formulas such as

[15] adaptively adjust the weight of each sensor in the fusion decision according to its real-time signal quality. Sensors with high signal-to-noise ratios are given greater weights, effectively suppressing the impact of data with severe noise pollution on the final judgment.

[0038] S8, building a machine learning model, using historical leakage monitoring data, and simulating the electromagnetic field distribution characteristics, temperature field distribution characteristics, and pressure field distribution characteristics under different leakage states and different leakage scales to generate a training sample set to train the machine learning model;

[0039] The fused data is input into the trained machine learning model. The fused data includes electromagnetic anomaly data, temperature anomaly data, and pressure anomaly data weighted and fused according to the weight coefficient. The leakage probability is obtained through the output layer of the machine learning model. ;

[0040] Data points exceeding the preset abnormal threshold are extracted from the fused data as abnormal data points, and the abnormal data points contain their corresponding spatial coordinates. ;

[0041] According to the spatial coordinate distribution of abnormal data points, a three-dimensional coordinate clustering algorithm is used to determine the clustering center of the abnormal data points, and the clustering center is used as the leakage center. ;

[0042] Based on the numerical changes of abnormal data points in a continuous time window, the time change rate of the abnormal degree is calculated as the leakage rate;

[0043] Count the spatial distribution range of abnormal data points and calculate the minimum three-dimensional bounding box volume containing all abnormal data points as the leakage impact range;

[0044] Based on the leakage rate and leakage impact range, when the leakage rate is less than the first rate threshold and the leakage impact range is less than the first range threshold, it is judged as a minor leakage; when the leakage rate or the leakage impact range exceeds the second rate threshold or the second range threshold, it is judged as a serious leakage; all other cases are judged as moderate leakage.

[0045] Another aspect of the present application also provides an underground anti-seepage structure leakage monitoring system based on a discrete time dynamic grid, comprising: a sensor array, including distributed potential sensors, temperature sensors and pressure sensors, for collecting current data, temperature data and pressure data in the monitoring area; the pressure sensor adopts a fiber grating sensor; a signal excitation source, for generating a bipolar square wave excitation signal, and adjusting the frequency parameters according to the geological conditions and monitoring depth requirements of the monitoring area, and using a multi-channel parallel excitation mode to send excitation signals to multiple monitoring areas; a data acquisition module, for collecting underground anti-seepage structure data and transient electromagnetic response data generated by the excitation signal, the transient electromagnetic response data including natural potential baseline DC data and transient response AC data; a time-space alignment module, for setting a discrete time window , according to the spatial coordinates of each sensor in the sensor array and the data acquisition time stamp t to construct a four-dimensional spatiotemporal data matrix The full waveform inversion module uses the interpolation algorithm to align the sensor data of different sampling frequencies in time and space; according to the initial conductivity parameters and electrode position parameters, and minimize the fitting error between the model response and the measured data through iterative optimization to construct a three-dimensional conductivity distribution model ;

[0046] The background field modeling module uses the finite element method to solve the time-domain Maxwell equations based on the three-dimensional conductivity distribution model, establishes a three-dimensional geophysical model, and uses the finite element forward algorithm to calculate the background electromagnetic field under the non-leakage state. ; Anomaly detection module, which converts the measured transient electromagnetic response data and background electromagnetic fields Perform differential calculations to obtain electromagnetic anomaly data ; Calculate the spatial temperature gradient and time rate of change ,when Greater than the gradient threshold and When the temperature rise threshold is greater than the threshold, the corresponding temperature anomaly data is extracted; according to the wavelength-pressure relationship of the fiber Bragg grating Calculate the pressure change value, when the pressure change rate Greater than the rate of change threshold or pressure change value When the threshold range is exceeded, the corresponding pressure anomaly data is extracted; the Bayesian fusion module calculates the conditional probability of various anomalies and leakage events based on historical monitoring data 、 and , set the weight coefficient based on the signal-to-noise ratio of each sensor , generates fused data through the Bayesian algorithm; the leakage detection module has a built-in pre-trained machine learning model to receive fused data and output leakage detection results. The leakage detection results include leakage probability, leakage location and leakage level.

[0047] Compared with the existing technology, the advantages of this application are:

[0048] Through the transient electromagnetic excitation source active detection technology, a bipolar square wave excitation signal with an amplitude of 0-±1000V and a pulse width of 0.1ms-1000ms is generated. A three-dimensional geophysical model is established by combining the full waveform inversion algorithm, and the background field differential operation is performed. Accurately extract electromagnetic anomaly signatures caused by leaks. Compared to traditional passive detection methods, this active detection method can detect tiny leaks within a 5cm² range, with a detection sensitivity as high as 0.1mV, enabling accurate identification of early-stage tiny leaks.

[0049] Based on discrete time window Constructing a four-dimensional space-time data matrix , through time slicing processing and interpolation algorithm, the spatiotemporal alignment of sensor data with different sampling frequencies is achieved. Under the unified spatiotemporal framework, electromagnetic anomaly data are extracted separately. 、Temperature abnormality data (when Greater than 0.5℃ / h and exceeds the threshold) and abnormal pressure data (when This solution can continuously monitor the dynamic evolution of leakage, capturing changes in leakage rate and providing data support for leakage trend prediction.

[0050] Use Bayesian algorithm to calculate the conditional probability of various anomalies and leakage events 、 and , dynamically assign weight coefficients based on the sensor signal-to-noise ratio , achieving probabilistic fusion of three types of abnormal data: electromagnetic, temperature, and pressure. Using a pre-trained machine learning model, it intelligently identifies leaks and outputs leak probability, location (with 0.5m accuracy), and level. This multi-source data fusion mechanism effectively eliminates the impact of environmental interference from single-field detection, significantly improving the reliability of the monitoring system. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The present application will be further described in the form of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting, and in these embodiments, the same numbers represent the same structures, wherein:

[0052] Figure 1 is an exemplary flow chart of a discrete-time-based underground anti-seepage structure leakage monitoring method according to some embodiments of the present application;

[0053] Figure 2 is an exemplary flow chart for generating electromagnetic anomaly data according to some embodiments of the present application;

[0054] Figure 3 is an exemplary flow chart for generating temperature anomaly data according to some embodiments of the present application;

[0055] Figure 4 This is an exemplary flow chart of generating pressure anomaly data according to some embodiments of the present application. DETAILED DESCRIPTION

[0056] The method and system provided in the embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0057] like Figure 1As shown in the figure, underground anti-seepage structure data including current data, temperature data and pressure data are collected through the sensor array; an excitation signal is sent to the monitoring area through a signal excitation source, and the transient electromagnetic response data generated by the excitation signal is collected through the sensor array; a discrete time window is set, and a spatiotemporal data matrix is constructed according to the spatial coordinates of each sensor in the sensor array and the data acquisition timestamp; the underground anti-seepage structure data and the transient electromagnetic response data are spatiotemporally aligned based on the spatiotemporal data matrix; a three-dimensional geophysical model of the monitoring area is established through the full waveform inversion algorithm according to the initial conductivity parameters and electrode position parameters of the monitoring area; and a finite element forward algorithm is used to calculate the transient electromagnetic response data based on the three-dimensional geophysical model. The electromagnetic response data is used to model the background field and obtain the background electromagnetic field distribution in the no-leakage state; a differential operation is performed on the transient electromagnetic response data and the background electromagnetic field distribution to obtain electromagnetic anomaly data; based on the temperature data, the distributed temperature field analysis method is used to calculate the spatial gradient and time change rate of the temperature to obtain temperature anomaly data; the pressure data is analyzed and the pressure anomaly data is determined based on the temporal variation characteristics of the pressure data; the Bayesian algorithm is used to fuse the electromagnetic anomaly data, temperature anomaly data and pressure anomaly data to obtain fused data; the fused data is input into a pre-trained machine learning model to obtain leakage detection results, which include leakage probability, leakage location and leakage level.

[0058] S1, in a certain underground station anti-seepage structure monitoring project, a sensor array including potential sensors, temperature sensors and fiber Bragg grating pressure sensors was constructed.

[0059] The potentiometric sensors, made of composite electrode materials, achieve a sensitivity of 0.1mV and are capable of detecting micro-leakage current signals within a 5cm² range. They are deployed in a distributed matrix at a density of 4 sensors per square meter, focusing on high-risk areas such as underground diaphragm wall joints and weak points in anti-seepage structures. The sensors incorporate active noise cancellation circuitry to effectively suppress 50Hz power frequency interference, increasing the signal-to-noise ratio to over 60dB.

[0060] The temperature sensor uses a MEMS temperature chip with a temperature resolution of 0.1°C and a response time of less than 1 second. The sensor housing meets IP68 waterproof rating and is equipped with a waterproof and breathable membrane, allowing it to operate for extended periods at depths of 200 meters below the groundwater level.

[0061] Based on Fiber Bragg Grating technology, the device leverages the wavelength-stress sensitivity of FBGs to achieve precise pressure measurement with a range of 0-5 MPa and a resolution of 0.1 kPa. A tandem topology design allows for 64 measurement points to be connected in series on a single fiber. Materials and structures are optimized for corrosion resistance, with the fiber wrapped in a special acid- and alkali-resistant coating and sealed packaging. The device can operate stably for over 10 years in leaky water environments with a pH range of 3-11.

[0062] S2, the signal excitation source, generates a bipolar square wave excitation signal with an adjustable amplitude within the range of 0 ± 1000 V and a programmable pulse width between 0.1 ms and 1000 ms. Based on the geological conditions of the monitored area (e.g., soil resistivity of 50 Ω·m), the excitation signal amplitude is set to 500 V, the pulse width to 10 ms, and the frequency to 25 Hz.

[0063] Using a multi-channel parallel excitation mode, the monitoring area is divided into four sub-areas, and excitation signals with different parameters are sent simultaneously. The transient electromagnetic response data of each sub-area is synchronously collected using a potential sensor array, including natural potential baseline DC data (amplitude range -100mV to +100mV) and transient response AC data (decay time 0.1ms to 100ms).

[0064] S3, set the discrete time window Δt = 60s, and divide the 24-hour continuous monitoring data into 1440 discrete time segments through time slicing. Construct a four-dimensional spatiotemporal data matrix , where the spatial coordinate accuracy is 0.1m and the time accuracy is 1s.

[0065] For sensors with different sampling frequencies (potential sensor 1kHz, temperature sensor 1Hz, pressure sensor 10Hz), a cubic spline interpolation algorithm is used to synchronize all data to the 60s time node. Finally, a spatiotemporal aligned dataset is obtained, and each data point contains position information. , timestamp t and the corresponding measurement value.

[0066] like Figure 2 As described above, S4, input the initial conductivity parameters of the monitoring area Based on geological survey data, the electrode placement parameters were 32 electrodes distributed in an 8 × 4 matrix. DC data (average -50 mV) and AC data (peak 200 mV) were extracted from the spatiotemporally aligned data.

[0067] Through the full waveform inversion algorithm, after 50 iterations of optimization, a three-dimensional conductivity distribution model was constructed. The model grid size is 0.5m×0.5m×0.5m. Based on this model, the background electromagnetic field is obtained by finite element forward calculation. .

[0068] The measured transient electromagnetic response data With background field Perform differential calculations, and when ΔE>10mV, it is determined to be an electromagnetic anomaly. In this embodiment, three electromagnetic anomaly areas are detected, with abnormal values of 15mV, 22mV, and 18mV respectively.

[0069] like Figure 3 As shown in S5, temperature data is extracted from the spatiotemporal alignment data Calculate the spatial temperature gradient and time rate of change of each monitoring point. Set the temperature rise threshold to 0.5℃ / h and the gradient threshold to 0.2℃ / m.

[0070] During the monitoring process, the temperature time change rates at two monitoring points were found to be 0.8°C / h and 1.2°C / h, respectively, exceeding the temperature rise threshold. Simultaneously, the spatial temperature gradient moduli at these two points were 0.3°C / m and 0.5°C / m, respectively, exceeding the gradient threshold. The data from these two monitoring points were considered temperature anomaly data.

[0071] like Figure 4 As shown, S6, extract the reflection wavelength data of the fiber Bragg grating sensor, the initial wavelength , stress sensitivity coefficient K = 1.2pm / kPa. When the wavelength change is detected When , the pressure change value ΔP=2kPa is calculated.

[0072] The pressure change rate threshold is set to 0.5 kPa / min and the pressure variation threshold is set to 1.5 kPa. Monitoring found that the pressure change rate at one measuring point reached 0.8 kPa / min, and the pressure change value was 2 kPa. The data at this point was considered as pressure anomaly data.

[0073] S7, based on historical monitoring data statistics, calculates: (Conditional probability of electromagnetic anomaly and leakage), (conditional probability of temperature anomaly and leakage), (Conditional probability of pressure anomaly and leakage).

[0074] According to the sensor signal-to-noise ratio ( 、 、 ), calculate the weight coefficient: 、 、 .

[0075] The posterior probability P of multi-source data fusion is calculated using the Bayesian algorithm, which is 0.92. The electromagnetic anomaly data (22 mV), temperature anomaly data (1.2 °C / h), and pressure anomaly data (2 kPa) are weighted according to the weight coefficient to obtain a comprehensive anomaly index of 18.5. The fused data is generated based on the posterior probability.

[0076] S8, input the fused data into the pre-trained deep learning model (including 3 convolutional layers and 2 fully connected layers), and the model outputs the leakage probability According to the spatial distribution of abnormal data, the leakage center position was determined to be (x=125.3m, y=48.7m, z=-15.2m) through the three-dimensional coordinate clustering algorithm.

[0077] The calculated leakage rate is 0.15 L / min, and the leakage area is 2.5 m². Based on the preset classification standards (mild: <0.1 L / min and <1 m²; moderate: 0.1-0.5 L / min or 1-5 m²; severe: >0.5 L / min or >5 m²), this is considered a moderate leak.

[0078] In practical applications, this embodiment transmits monitoring data in real time via a 4G / Wi-Fi dual-mode network, achieving a data compression rate of 85%. After six months of operation, the system successfully detected five leaks with a detection accuracy of 98.2%, a false alarm rate of only 1.8%, and leak location accuracy better than 0.5 meters. Thanks to the intelligent push notification system, maintenance personnel's average response time has been reduced from four hours to 30 minutes, effectively ensuring the safe operation of underground anti-seepage structures.

[0079] The invention of the present application and its implementation methods are described schematically above. This description is not restrictive. Without departing from the spirit or basic features of the present application, the present application can be implemented in other specific forms. What is shown in the accompanying drawings is only one of the implementation methods of the invention of the present application, and the actual structure is not limited to this. Therefore, if a person of ordinary skill in the art is inspired by it, without departing from the purpose of the invention, a structural method and embodiment similar to the technical solution are designed without creativity, which should all fall within the scope of protection of the present application. In addition, the word "including" does not exclude other elements or steps, and the word "one" before an element does not exclude the inclusion of "multiple" elements. Words such as first and second are used to indicate names and do not indicate any specific order.

Claims

1. A method for monitoring leakage of underground anti-seepage structures based on discrete time, characterized in that: include: S1, collecting underground anti-seepage structure data including current data, temperature data and pressure data through a sensor array; S2, sending an excitation signal to the monitoring area through a signal excitation source, and collecting transient electromagnetic response data generated by the excitation signal through a sensor array; S3, set a discrete time window and construct a spatiotemporal data matrix based on the spatial coordinates and data acquisition timestamps of each sensor in the sensor array; Perform spatiotemporal alignment of underground anti-seepage structure data and transient electromagnetic response data based on spatiotemporal data matrix; S4, establishing a three-dimensional geophysical model of the monitoring area using a full waveform inversion algorithm based on the initial conductivity parameters and electrode position parameters of the monitoring area; Based on the three-dimensional geophysical model, the finite element forward algorithm is used to model the background field of the transient electromagnetic response data after S3 time and space alignment to obtain the background electromagnetic field distribution under the no leakage state; Performing a differential operation on the transient electromagnetic response data and the background electromagnetic field distribution after the time-space alignment in step S3 to obtain electromagnetic anomaly data; S5, based on the temperature data in the underground anti-seepage structure data after the time-space alignment in S3, the distributed temperature field analysis method is used to calculate the spatial gradient and time change rate of the temperature to obtain the temperature anomaly data; S6, analyzing the pressure data in the underground anti-seepage structure data after the spatiotemporal alignment in S3, and determining the pressure anomaly data according to the temporal variation characteristics of the pressure data; S7, using a Bayesian algorithm to fuse the electromagnetic anomaly data, the temperature anomaly data, and the pressure anomaly data to obtain fused data; S8, inputs the fused data into a pre-trained machine learning model to obtain leakage detection results, which include leakage probability, leakage location and leakage level.

2. The method for monitoring underground anti-seepage structure leakage based on discrete time according to claim 1, characterized in that: S2, collects transient electromagnetic response data generated by the excitation signal, including: Generate a bipolar square wave excitation signal; Adjust the frequency parameters of the bipolar square wave excitation signal according to the geological conditions of the monitoring area and the monitoring depth requirements; Adopting multi-channel parallel excitation mode, bipolar square wave excitation signals are sent to multiple monitoring areas; The transient electromagnetic response data of each monitoring area is collected through the sensor array. The transient electromagnetic response data includes natural potential baseline DC data and transient response AC data.

3. The method for monitoring leakage of underground anti-seepage structures based on discrete time according to claim 2, characterized in that: The amplitude range of the bipolar square wave excitation signal is 0-±1000V, and the pulse width range is 0.1ms-1000ms.

4. The method for monitoring underground anti-seepage structure leakage based on discrete time according to claim 1, characterized in that: S3, constructs a spatiotemporal data matrix, including: Set the discrete time window ,Through time slicing processing, the continuous data stream is divided into discrete time segments; According to the three-dimensional space coordinates of each sensor and the data acquisition timestamp t, constructing a four-dimensional spatiotemporal data matrix , used to store the measurement values of each sensor in each discrete time segment; For sensor data with different sampling frequencies, discrete time windows are used As a benchmark, the sensor data are synchronized to the same time node through the interpolation algorithm; The time-synchronized sensor data are spatially aligned to obtain spatiotemporally aligned data.

5. The method for monitoring leakage of underground anti-seepage structures based on discrete time according to claim 1, characterized in that: S4, establish a three-dimensional geophysical model of the monitoring area using the full waveform inversion algorithm; Based on the three-dimensional geophysical model, the finite element forward algorithm is used to model the background field of the transient electromagnetic response data after S3 time and space alignment to obtain the background electromagnetic field distribution under the no leakage state; Perform differential calculation based on the transient electromagnetic response data collected in step S2 and the background electromagnetic field distribution to obtain electromagnetic anomaly data, including: Obtain the initial conductivity parameters of the monitoring area and electrode position parameters; initial conductivity parameters is the initial estimated value of the conductivity of each layer in the monitoring area; Extract the natural potential baseline DC data and transient response AC data from the time-space aligned data as input data for the full waveform inversion algorithm; The full waveform inversion algorithm is used to iteratively optimize the initial conductivity parameters. , minimize the fitting error between the model response and the natural potential baseline DC data and transient response AC data, and construct a three-dimensional conductivity distribution model ; Based on the three-dimensional conductivity distribution model and electrode position parameters, the monitoring area is meshed using the finite element method, and the time-domain Maxwell equations are solved to establish a three-dimensional geophysical model containing the spatial distribution information of conductivity; Using a three-dimensional geophysical model, the finite element forward algorithm simulates the electromagnetic field response after applying the excitation signal under the current conductivity distribution conditions, and calculates the background electromagnetic field under the non-leakage state. ; Align the measured transient electromagnetic response data in time and space Background electromagnetic field Perform a difference operation: , get the electromagnetic anomaly data caused by leakage .

6. The method for monitoring underground anti-seepage structure leakage based on discrete time according to claim 1, characterized in that: S5, obtain temperature anomaly data, including: Temperature data in spatiotemporal alignment data , calculate the temperature spatial gradient of each monitoring point ; Calculate the temperature time change rate of each monitoring point ; When the spatial gradient modulus of the temperature at the monitoring point When the gradient is greater than the preset threshold, the corresponding monitoring point is marked as a temperature spatial anomaly point; When the temperature change rate of the monitoring point When the temperature rise is greater than the preset threshold, the corresponding monitoring point will be marked as a temperature time-varying abnormal point; Based on the physical mechanism of local heat transfer anomalies caused by groundwater leakage, the temperature field distribution is analyzed by solving the heat conduction equation to verify whether the temperature spatial anomaly points and temperature identification anomaly points conform to the temperature field change characteristics caused by leakage; For abnormal points that have passed the verification, the corresponding monitoring points are determined as temperature abnormality monitoring points; Extract temperature data corresponding to temperature anomaly monitoring points , temperature spatial gradient and temperature-time rate of change , as temperature anomaly data.

7. The method for monitoring underground anti-seepage structure leakage based on discrete time according to claim 1, characterized in that: The sensor array includes a potential sensor, a temperature sensor, and a pressure sensor; The pressure sensor uses a fiber grating sensor.

8. The method for monitoring leakage of underground anti-seepage structures based on discrete time according to claim 7, characterized in that: S6, obtains abnormal pressure data, including: Extracting reflected wavelength data from fiber Bragg grating sensors ; According to the wavelength-pressure relationship Calculate pressure change ,in, is the initial wavelength, is the wavelength variation, K is the stress sensitivity coefficient; Pressure change value Perform timing analysis, when the pressure change rate When the change rate is greater than the preset threshold, the corresponding monitoring point is regarded as a pressure mutation point; When the pressure changes When it is greater than the threshold range, the corresponding monitoring point is regarded as an abnormal fluctuation point; The pressure data corresponding to the monitoring points containing pressure mutation points and abnormal fluctuation points are regarded as pressure abnormality data.

9. The method for monitoring leakage of underground anti-seepage structures based on discrete time according to claim 1, characterized in that: S7, obtain fused data, including: Calculate the conditional probability of electromagnetic anomalies and leakage events based on historical monitoring data , the conditional probability of temperature anomaly and leakage event , and the conditional probability of pressure anomaly and leakage events ; Based on the Bayesian algorithm, calculate the posterior probability P of multi-source data fusion; Set the conditional probability based on the signal-to-noise ratio of each sensor 、 and The weight coefficient ; The electromagnetic anomaly data, temperature anomaly data and pressure anomaly data are calculated based on the weight coefficients. Perform weighted calculation to obtain comprehensive abnormality indicators; According to the comprehensive abnormality index and posterior probability P, the fused data is obtained.

10. A leakage monitoring system for underground anti-seepage structures based on discrete-time dynamic grids, characterized in that: include: A sensor array, including distributed potential sensors, temperature sensors, and pressure sensors, is used to collect current data, temperature data, and pressure data in the monitoring area; The pressure sensor uses a fiber Bragg grating sensor; The signal excitation source is used to generate a bipolar square wave excitation signal, adjust the frequency parameters according to the geological conditions and monitoring depth requirements of the monitoring area, and use a multi-channel parallel excitation mode to send excitation signals to multiple monitoring areas; The data acquisition module collects underground anti-seepage structure data and transient electromagnetic response data generated by the excitation signal. The transient electromagnetic response data includes natural potential baseline DC data and transient response AC data. Spatiotemporal alignment module, setting discrete time windows , according to the spatial coordinates of each sensor in the sensor array and the data acquisition time stamp t to construct a four-dimensional spatiotemporal data matrix and align sensor data with different sampling frequencies in time and space through interpolation algorithms; Full waveform inversion module, based on the initial conductivity parameters and electrode position parameters, and minimize the fitting error between the model response and the measured data through iterative optimization to construct a three-dimensional conductivity distribution model ; The background field modeling module uses the finite element method to solve the time-domain Maxwell equations based on the three-dimensional conductivity distribution model, establishes a three-dimensional geophysical model, and uses the finite element forward algorithm to calculate the background electromagnetic field under the non-leakage state. ; The abnormality detection module converts the measured transient electromagnetic response data and background electromagnetic fields Perform differential calculations to obtain electromagnetic anomaly data ; Calculate the spatial temperature gradient and time rate of change ,when Greater than the gradient threshold and When the temperature rise threshold is greater than the threshold, the corresponding temperature anomaly data is extracted; according to the wavelength-pressure relationship of the fiber Bragg grating Calculate the pressure change value, when the pressure change rate Greater than the rate of change threshold or pressure change value When the threshold value is exceeded, the corresponding abnormal pressure data is extracted; Bayesian fusion module, which calculates the conditional probability of various anomalies and leakage events based on historical monitoring data 、 and , set the weight coefficient based on the signal-to-noise ratio of each sensor , generate fusion data through Bayesian algorithm; The leakage detection module has a built-in pre-trained machine learning model that receives fused data and outputs leakage detection results, which include leakage probability, leakage location, and leakage level.

Images