A method for detecting integrity of a dam diaphragm based on an artificial power supply

Abstract

This invention discloses a method for detecting the integrity of dam seepage barriers based on artificial power sources. The method includes the following steps: injecting a wideband excitation current signal into the dam seepage barrier using a multi-frequency programmable constant current source; acquiring voltage data on the surface of the dam seepage barrier in real time using a modular distributed electrode array; obtaining high signal-to-noise ratio voltage data using filtering and digital phase-locked amplification technology; performing three-dimensional inversion calculations using an improved electrical impedance tomography algorithm to reconstruct an image of the internal conductivity distribution of the dam seepage barrier; analyzing the data using a deep learning model to identify and locate defects; constructing an electro-seepage coupling inversion model to convert the image of the internal conductivity distribution of the dam seepage barrier into a permeability coefficient distribution, and generating a dam seepage barrier integrity evaluation report. Compared with traditional detection methods, this invention offers high detection accuracy, wide applicability, and high detection efficiency, providing a novel technical means for assessing the health status of dam seepage barriers.

Description

Technical Field

[0001] This invention belongs to the field of water conservancy project safety monitoring technology, specifically relating to a method for detecting the integrity of dam seepage prevention bodies based on artificial power sources. Background Technology

[0002] As the core seepage prevention structure of water conservancy projects, the integrity of the dam's seepage barrier is directly related to the safety of the project. However, existing detection technologies face several technical bottlenecks. Among traditional detection methods, ground-penetrating radar (GPR) is simple to operate, but its detection depth and resolution have inherent contradictions. In water-bearing media such as clay cores, electromagnetic waves are severely attenuated, and the defect detection rate in areas deeper than 5 meters drops sharply to below 60%, making it almost impossible to identify microcracks smaller than 3 centimeters. While borehole sampling can obtain direct rock core samples, it damages the integrity of the dam's seepage barrier, and the detection range of a single borehole is limited. For conventional dams, dozens of boreholes are needed to obtain reliable data, and the detection cycle can take as long as 3-5 working days.

[0003] In the field of electrical testing, conventional resistivity methods employ fixed-frequency excitation, making it difficult to simultaneously meet the requirements of high resolution in shallow layers and high penetration in deep layers. Furthermore, the measurement error is generally 15%-20% due to electrode polarization effects and contact impedance. While electrical impedance tomography (EIT) can theoretically achieve non-destructive testing, existing methods have significant limitations: spatial resolution is constrained by the number of electrodes and algorithm limitations, typically only able to identify defects larger than 10 cm; in complex electromagnetic environments in the field, power frequency interference and random noise result in a signal-to-noise ratio generally below 60 dB; more importantly, existing technologies lack effective multi-physics coupling models, resulting in an accuracy of less than 70% in tracking seepage channels and a permeability coefficient inversion error exceeding 30%. These technical deficiencies make current methods insufficient to meet the needs of major water conservancy projects for identifying and quantitatively evaluating millimeter-level defects in dam seepage barriers, especially in critical scenarios such as high dams, large reservoirs, and aging dams that have been in operation for many years. There is an urgent need to develop new testing technologies to overcome the performance limitations of existing methods. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention proposes a method for detecting the integrity of dam seepage barriers based on artificial power sources. The aim is to reconstruct the three-dimensional electrical characteristic parameter distribution of the dam seepage barrier using an improved electrical impedance tomography algorithm; and to achieve intelligent diagnosis of seepage channels and structural defects by combining a deep learning-assisted defect identification algorithm. To achieve the above objectives, this invention provides the following technical solution: a method for detecting the integrity of dam seepage barriers based on artificial power sources, with the following specific steps:

[0005] Step S1: Inject a wideband excitation current signal into the seepage prevention body of the dam using a multi-frequency programmable constant current source;

[0006] Step S2: Based on the wideband excitation current signal from step S1, the voltage data on the surface of the dam seepage barrier is acquired in real time using a modular distributed electrode array.

[0007] Step S3: Based on the voltage data of the dam seepage prevention body surface obtained in step S2, high signal-to-noise ratio voltage data is obtained by filtering and digital phase-locked amplification technology.

[0008] Step S4: Based on the high signal-to-noise ratio voltage data obtained in step S3, a three-dimensional inversion calculation is performed using an improved electrical impedance tomography algorithm to reconstruct the conductivity distribution image inside the dam seepage barrier.

[0009] Step S5: Based on the image of the electrical conductivity distribution inside the dam seepage barrier obtained in Step S4, analyze it using a deep learning model to identify and locate defects.

[0010] Step S6: Construct an electro-seepage coupled inversion model, convert the electrical conductivity distribution image inside the dam seepage barrier based on step S4 into a permeability coefficient distribution, and generate a dam seepage barrier integrity evaluation report.

[0011] Furthermore, in step S1, a wideband excitation current signal is injected into the dam's seepage prevention body through a multi-frequency programmable constant current source; the specific steps are as follows:

[0012] Step S11: Establish a database of frequency response for dam seepage prevention materials, and pre-store the optimal excitation frequency bands for different materials;

[0013] Step S12: The multi-frequency programmable constant current source automatically identifies the type of dam seepage prevention material by rapidly scanning the impedance spectrum and intelligently matches the optimal excitation frequency band.

[0014] Step S13: The proportional-integral-derivative closed-loop control algorithm is used to dynamically adjust the multi-frequency programmable constant current source to inject a wideband excitation current signal of 0.1-1000 Hz into the seepage prevention body of the dam.

[0015] Step S14: The multi-frequency programmable constant current source integrates an overload protection module. When the multi-frequency programmable constant current source detects a contact resistance > 10 kΩ, it automatically cuts off the output and alarms.

[0016] Furthermore, in step S2, based on the broadband excitation current signal from step S1, a modular distributed electrode array is used to acquire the voltage data on the surface of the dam's seepage prevention body in real time; the specific steps are as follows:

[0017] Step S21: Modular distributed electrode array, deploying 128 high-precision silver-silver chloride electrodes with a 1-meter spacing, integrating a preamplifier circuit with a gain of 100 times and an impedance detection module with a contact impedance of <3 kΩ from the high-precision silver-silver chloride electrodes;

[0018] Step S22: Based on the modular distributed electrode array in step S21, multi-channel time synchronization is achieved through synchronous triggering via the Global Positioning System. A 24-bit sigma-delta analog-to-digital converter is used for synchronous sampling, and voltage data of the dam seepage prevention body surface is obtained in conjunction with 0.1-1.5 kHz adaptive bandpass filtering and 40-70 Hz adjustable dynamic power frequency notch filtering.

[0019] Furthermore, in step S3, the voltage data on the surface of the dam seepage barrier obtained in step S2 is processed using filtering and digital phase-locked loop amplification techniques to obtain high signal-to-noise ratio voltage data; the specific steps are as follows:

[0020] Step S31: Perform primary noise suppression on the voltage data of the dam seepage prevention body surface using a hardware bandpass filter;

[0021] Step S32: An adaptive notch filter based on the least mean square algorithm is used to track and eliminate the 50 Hz power frequency and 50 Hz power frequency harmonic interference in real time, with a convergence time of <100 milliseconds.

[0022] Step S33: Combining digital lock-in amplification technology, coherent detection is performed using the wideband excitation current signal from step S1 as the reference frequency. In-phase and quadrature components are extracted through orthogonal vector decomposition. The output signal-to-noise ratio is monitored in real time. The passband width and stopband attenuation parameters of the 16th-order finite-length unit impulse response filter are dynamically adjusted. The signal-to-noise ratio can be stably maintained at >90 dB under complex working conditions, resulting in high signal-to-noise ratio voltage data.

[0023] Furthermore, in step S4, based on the high signal-to-noise ratio voltage data obtained in step S3, a three-dimensional inversion calculation is performed using an improved electrical impedance tomography algorithm to reconstruct the conductivity distribution image inside the dam's seepage barrier. The improved electrical impedance tomography algorithm consists of four steps: accurate forward problem modeling, inverse problem construction and improvement, efficient iterative solution, and post-image optimization. The specific steps are as follows:

[0024] Step S41: Accurate forward problem modeling, establishing a complete electrode model that includes contact resistance effects;

[0025] Step S42, Inverse problem construction and improvement: The sensitivity matrix is ​​calculated by admittance method and the objective function is constructed by combining improved Tikhonov regularization with a regularization matrix based on discrete Laplace operator to constrain the smoothness of the solution.

[0026] Step S43, Efficient Iterative Solution: The iterative process of solving the conductivity correction using the conjugate gradient method includes iterative solution updating, conjugate direction solving, and calculation of the objective function gradient.

[0027] Step S44, post-image optimization: A pre-trained deep residual network is used to optimize the image to restore details and suppress noise, as shown in the formula:

[0028] ;

[0029] in, Image showing the electrical conductivity distribution inside the seepage barrier of the dam. Let θ represent a convolutional neural network, and let θ represent the set of parameters of the convolutional neural network.

[0030] Further, in step S41, precise forward problem modeling is performed to establish a complete electrode model that includes the contact resistance effect; the specific steps are as follows:

[0031] Step S411: The governing equations for the complete electrode model are described using the Poisson equation to depict the potential distribution under the steady-state electric field, as shown in the formula:

[0032] ;

[0033] in, Denotes divergence, Indicates electrical conductivity. Represents the gradient. Represents electric potential;

[0034] Step S412, Boundary conditions for the complete electrode model, considering the contact impedance at the electrode-dielectric interface, as shown in the formula:

[0035] ;

[0036] in, This represents the voltage measured on the l-th electrode. Indicates the first One electrode, Represents electric potential The directional derivative in the direction n normal to the boundary. Indicates injection of the first Current of each electrode Indicates the first The contact resistance between an electrode and the dielectric surface.

[0037] Further, in step S42, the inverse problem is constructed and improved by calculating the sensitivity matrix using the admittance method and constructing the objective function using the improved Tikhonov regularization, along with a regularization matrix based on the discrete Laplace operator, to constrain the smoothness of the solution; the specific steps are as follows:

[0038] Step S421: Calculate the sensitivity matrix using the admittance method. The elements of the sensitivity matrix are shown in the formula:

[0039] ;

[0040] in, Let J represent the element in the m-th row and p-th column of the sensitivity matrix J, and Ω represent the integration domain of the entire solution region. This represents the electric field distribution calculated under the i-th current excitation mode. dΩ represents the electric field distribution calculated under the j-th current excitation mode, where i ≠ j, and dΩ represents a volume element.

[0041] Step S422: Construct the objective function using improved Tikhonov regularization; as shown in the formula:

[0042] ;

[0043] Where min represents minimizing, i.e. finding the minimum solution of the objective function value, Δσ represents the conductivity correction amount, ΔV represents the voltage change measured on the boundary electrode, λ represents the regularization parameter, and L represents the regularization matrix;

[0044] Step S423: The regularization parameter is obtained by using an adaptive decay strategy that decreases with the number of iterations; as shown in the formula:

[0045] ;

[0046] Where α represents the proportionality coefficient, β represents the decay coefficient, k represents the number of iterations, and exp(-βk) represents the exponential decay function. Representation matrix The F-norm;

[0047] Step S424: The regularization matrix L is constructed based on the discrete Laplace operator to constrain the smoothness of the solution; the elements of the regularization matrix are shown in the formula:

[0048] ;

[0049] in, This represents the element in the r-th row and s-th column of the regularization matrix L. Let represent the element in the r-th row and t-th column of the regularization matrix L.

[0050] Further, step S43 involves efficient iterative solution, achieved through an iterative process of solving for the conductivity correction using the conjugate gradient method. This process specifically includes updating the iterative solution, determining the conjugate direction, and calculating the gradient of the objective function. The specific steps are as follows:

[0051] Step S431: The conductivity correction is obtained using the conjugate gradient method, as shown in the formula:

[0052] ;

[0053] in, This represents the conductivity correction amount in the (k+1)th iteration. This represents the conductivity correction amount in the k-th iteration. This represents the step size of the k-th iteration. This indicates the conjugate direction of the k-th iteration;

[0054] Step S432, solve for the conjugate direction, as shown in the formula:

[0055] ;

[0056] in, This represents the gradient of the objective function in the k-th iteration. This represents the transpose of the gradient of the objective function in the k-th iteration. This represents the transpose of the gradient of the objective function in the (k-1)th iteration. This indicates the conjugate direction of the (k-1)th iteration;

[0057] Step S433: Calculate the gradient of the objective function to understand its descent trend, as shown in the formula:

[0058] ;

[0059] in, This represents the transpose of the sensitivity matrix J. This represents the transpose of the regularization matrix L;

[0060] Step S434: During the iteration process of steps S431 to S433, a convergence condition is set: the change in the objective function value is less than a threshold or the maximum number of iterations is reached; when the iteration process terminates after satisfying the convergence condition, the final conductivity correction amount Δσ is obtained. final Adding this to the initial conductivity distribution σ0 yields the preliminary conductivity distribution image σ reconstructed by the improved electrical impedance tomography algorithm. EIT .

[0061] Furthermore, in step S5, based on the conductivity distribution image inside the dam seepage barrier obtained in step S4, a deep learning model is used to analyze and identify defects; the specific steps are as follows:

[0062] Step S51: Construct a simulation training database; establish a parametric geometric model library containing various dam structures, dam seepage prevention materials, and foundation types; embed typical defects with adjustable parameters, including seepage channels, cracks, and cavities, into the model base of the geometric model library; for each parametric model, perform forward problem calculations of electrical impedance tomography to simulate the excitation and measurement processes of steps S1 to S3, generating simulated voltage data; subsequently, input the simulated voltage data into the improved electrical impedance tomography algorithm in step S4 for inversion, generating the corresponding three-dimensional conductivity distribution image as input data for training samples; finally, generate a simulation training database containing more than 100,000 sets of samples.

[0063] Step S52: Design and construct a 3D defect recognition model; adopt a U-shaped network as the basic architecture; replace the encoder of the U-shaped network with a residual network-50 as the backbone feature extraction network; embed a spatial attention module at the jump connection between the encoder and the decoder; use a normalized exponential function at the end of the decoder to achieve pixel-level defect classification.

[0064] Step S53: Train the end-to-end 3D defect recognition model from step S52; use a hybrid loss function of Dess loss and cross-entropy loss to train the 3D defect recognition model and obtain the trained 3D defect recognition model.

[0065] In step S54, the conductivity distribution image inside the dam seepage prevention body obtained in step S4 is input into the trained three-dimensional defect recognition model. The encoder is used to downsample the image to 1 / 32 resolution and extract feature maps at different levels. Subsequently, the feature maps are upsampled by the decoder and fused with the same-scale encoder features weighted by skip connections and spatial attention modules. Finally, a defect classification probability map of the same size as the input image is output.

[0066] Step S55: For the defect classification probability map output in step S54, a three-dimensional connected component analysis algorithm is used to identify each independent defect; a sub-pixel localization algorithm is used to calculate the three-dimensional coordinates of the geometric center of each independent defect; finally, the category, three-dimensional coordinates of the geometric center, and equivalent size information of all defects are integrated to generate and output a three-dimensional defect distribution map with spatial coordinates.

[0067] Further, in step S6, an electro-seepage coupled inversion model is constructed to convert the conductivity distribution image inside the dam seepage barrier obtained in step S4 into a permeability coefficient distribution, and a dam seepage barrier integrity evaluation report is generated; the specific steps are as follows:

[0068] Step S601: Based on the inherent physical relationship between electrical conductivity and permeability in soil and rock media, and using the traditional Alzer's law as a foundation, an electro-permeability coupled inversion model is established, as shown in the formula:

[0069] ;

[0070] Where K represents the permeability coefficient, with units of centimeters per second or meters per day. Volume conductivity is expressed in Siemens per meter (S / m). The empirical coefficients of the model are represented by u, which depend on the mineral composition, pore structure and cementation of the dam seepage prevention material, and u = 1, 2, 3, 4.

[0071] Step S602: Accurately calibrate the empirical coefficients of the model through indoor geotechnical tests; prepare a series of dam seepage prevention material samples with different mix ratios, including different moisture contents, densities, and clay contents; for each dam seepage prevention material sample, use a permeameter to measure the permeability coefficient and a resistivity meter to measure the volumetric conductivity.

[0072] Step S603: Substitute the measured volumetric conductivity and permeability coefficient into the electro-seepage coupling inversion model, and use the nonlinear least squares method to perform regression analysis to solve for the specific coefficients of the dam seepage prevention material, i.e., the empirical coefficients of the model.

[0073] Step S604: Using the conductivity distribution image inside the dam seepage barrier from step S4 as input, and using the empirical coefficients of the model from step S603, the initial permeability coefficient distribution of the entire dam seepage barrier region is initially calculated through the electro-seepage coupling inversion model.

[0074] Step S605: Based on the initial permeability coefficient distribution of the entire dam seepage prevention area, construct the governing equations for the steady-state seepage field, as shown in the formula:

[0075] ;

[0076] Where K represents the permeability coefficient distribution and h represents the total head, which drives seepage.

[0077] Step S606: Discretize the three-dimensional structure of the dam seepage barrier into a finite element mesh; apply the corresponding head boundary or flow boundary to the finite element mesh according to the actual operating conditions of the dam seepage barrier; solve the control equation of the steady-state seepage field using the finite element method to obtain the head distribution within the entire dam seepage barrier area.

[0078] Step S607: Define the objective function, which is to minimize the difference between the head distribution and the measured head distribution; starting from the initial permeability coefficient distribution, use the gradient descent method or Gauss-Newton method to repeatedly adjust the permeability coefficient distribution, and return to step S606 to solve the control equation of the steady-state seepage field until the objective function value is less than the preset threshold, and finally obtain the high-precision permeability coefficient distribution.

[0079] Step S608: Perform spatial analysis on the obtained high-precision permeability coefficient distribution; set a critical permeability coefficient value K. crit Automatically identify and mark all permeability coefficient distribution values ​​K that are greater than or equal to the critical permeability coefficient value K. crit The spatial area is marked as the weak zone in seepage prevention, and its three-dimensional spatial coordinates and volume are recorded.

[0080] Step S609: Based on the results of steps S607 and S608, calculate the structural integrity score S of the dam's seepage prevention structure; as shown in the formula:

[0081] ;

[0082] Among them, V total V represents the total volume of the dam's seepage prevention structure. weak This indicates the total volume of the identified weak points in the seepage prevention system. This represents the average permeability coefficient across all weak zones.

[0083] Step S610: Based on the results of steps S608 and S609, generate a dam seepage prevention integrity evaluation report that includes a permeability coefficient distribution map, a weak zone location map, and a structural integrity score.

[0084] Beneficial effects of this invention:

[0085] This invention innovatively proposes a signal enhancement technique that combines multi-frequency excitation with adaptive filtering, effectively solving the signal interference problem in complex environments.

[0086] This invention achieves non-destructive evaluation of the permeability coefficient of impermeable bodies by constructing an electro-permeation coupled inversion model. Attached Figure Description

[0087] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0088] This invention provides a technical solution: a method for detecting the integrity of dam seepage prevention structures based on artificial power sources, such as... Figure 1 As shown, it includes the following steps:

[0089] Step S1: Inject a wideband excitation current signal into the seepage prevention body of the dam using a multi-frequency programmable constant current source;

[0090] Step S2: Based on the wideband excitation current signal from step S1, the voltage data on the surface of the dam seepage barrier is acquired in real time using a modular distributed electrode array.

[0091] Step S3: Based on the voltage data of the dam seepage prevention body surface obtained in step S2, high signal-to-noise ratio voltage data is obtained by filtering and digital phase-locked amplification technology.

[0092] Step S4: Based on the high signal-to-noise ratio voltage data obtained in step S3, a three-dimensional inversion calculation is performed using an improved electrical impedance tomography algorithm to reconstruct the conductivity distribution image inside the dam seepage barrier.

[0093] Step S5: Based on the image of the electrical conductivity distribution inside the dam seepage barrier obtained in Step S4, analyze it using a deep learning model to identify and locate defects.

[0094] Step S6: Construct an electro-seepage coupled inversion model, convert the electrical conductivity distribution image inside the dam seepage barrier based on step S4 into a permeability coefficient distribution, and generate a dam seepage barrier integrity evaluation report.

[0095] Furthermore, in step S1, a wideband excitation current signal is injected into the dam's seepage prevention body through a multi-frequency programmable constant current source; the specific steps are as follows:

[0096] Step S11: Establish a database of frequency response for dam seepage prevention materials, and pre-store the optimal excitation frequency bands for different materials;

[0097] Step S12: The multi-frequency programmable constant current source automatically identifies the type of dam seepage prevention material by rapidly scanning the impedance spectrum and intelligently matches the optimal excitation frequency band.

[0098] Step S13: The proportional-integral-derivative closed-loop control algorithm is used to dynamically adjust the multi-frequency programmable constant current source to inject a wideband excitation current signal of 0.1-1000 Hz into the seepage prevention body of the dam.

[0099] Step S14: The multi-frequency programmable constant current source integrates an overload protection module. When the multi-frequency programmable constant current source detects a contact resistance > 10 kΩ, it automatically cuts off the output and alarms.

[0100] Furthermore, in step S2, based on the broadband excitation current signal from step S1, a modular distributed electrode array is used to acquire the voltage data on the surface of the dam's seepage prevention body in real time; the specific steps are as follows:

[0101] Step S21: Modular distributed electrode array, deploying 128 high-precision silver-silver chloride electrodes with a 1-meter spacing, integrating a preamplifier circuit with a gain of 100 times and an impedance detection module with a contact impedance of <3 kΩ from the high-precision silver-silver chloride electrodes;

[0102] Step S22: Based on the modular distributed electrode array in step S21, multi-channel time synchronization is achieved through synchronous triggering via the Global Positioning System. A 24-bit sigma-delta analog-to-digital converter is used for synchronous sampling, and voltage data of the dam seepage prevention body surface is obtained in conjunction with 0.1-1.5 kHz adaptive bandpass filtering and 40-70 Hz adjustable dynamic power frequency notch filtering.

[0103] Furthermore, in step S3, the voltage data on the surface of the dam seepage barrier obtained in step S2 is processed using filtering and digital phase-locked loop amplification techniques to obtain high signal-to-noise ratio voltage data; the specific steps are as follows:

[0104] Step S31: Perform primary noise suppression on the voltage data of the dam seepage prevention body surface using a hardware bandpass filter;

[0105] Step S32: An adaptive notch filter based on the least mean square algorithm is used to track and eliminate the 50 Hz power frequency and 50 Hz power frequency harmonic interference in real time, with a convergence time of <100 milliseconds.

[0106] Step S33: Combining digital lock-in amplification technology, coherent detection is performed using the wideband excitation current signal from step S1 as the reference frequency. In-phase and quadrature components are extracted through orthogonal vector decomposition. The output signal-to-noise ratio is monitored in real time. The passband width and stopband attenuation parameters of the 16th-order finite-length unit impulse response filter are dynamically adjusted. The signal-to-noise ratio can be stably maintained at >90 dB under complex working conditions, resulting in high signal-to-noise ratio voltage data.

[0107] Furthermore, in step S4, based on the high signal-to-noise ratio voltage data obtained in step S3, a three-dimensional inversion calculation is performed using an improved electrical impedance tomography algorithm to reconstruct the conductivity distribution image inside the dam's seepage barrier. The improved electrical impedance tomography algorithm consists of four steps: accurate forward problem modeling, inverse problem construction and improvement, efficient iterative solution, and post-image optimization. The specific steps are as follows:

[0108] Step S41: Accurate forward problem modeling, establishing a complete electrode model that includes contact resistance effects;

[0109] Step S42, Inverse problem construction and improvement: The sensitivity matrix is ​​calculated by admittance method and the objective function is constructed by combining improved Tikhonov regularization with a regularization matrix based on discrete Laplace operator to constrain the smoothness of the solution.

[0110] Step S43, Efficient Iterative Solution: The iterative process of solving the conductivity correction using the conjugate gradient method includes iterative solution updating, conjugate direction solving, and calculation of the objective function gradient.

[0111] Step S44, post-image optimization: A pre-trained deep residual network is used to optimize the image to restore details and suppress noise, as shown in the formula:

[0112] ;

[0113] in, Image showing the electrical conductivity distribution inside the seepage barrier of the dam. Let θ represent a convolutional neural network, and let θ represent the set of parameters of the convolutional neural network.

[0114] Further, in step S41, precise forward problem modeling is performed to establish a complete electrode model that includes the contact resistance effect; the specific steps are as follows:

[0115] Step S411: The governing equations for the complete electrode model are described using the Poisson equation to depict the potential distribution under the steady-state electric field, as shown in the formula:

[0116] ;

[0117] in, Denotes divergence, Indicates electrical conductivity. Represents the gradient. Represents electric potential;

[0118] Step S412, Boundary conditions for the complete electrode model, considering the contact impedance at the electrode-dielectric interface, as shown in the formula:

[0119] ;

[0120] in, This represents the voltage measured on the l-th electrode. Indicates the first One electrode, Represents electric potential The directional derivative in the direction n normal to the boundary. Indicates injection of the first Current of each electrode Indicates the first The contact resistance between an electrode and the dielectric surface.

[0121] Further, in step S42, the inverse problem is constructed and improved by calculating the sensitivity matrix using the admittance method and constructing the objective function using the improved Tikhonov regularization, along with a regularization matrix based on the discrete Laplace operator, to constrain the smoothness of the solution; the specific steps are as follows:

[0122] Step S421: Calculate the sensitivity matrix using the admittance method. The elements of the sensitivity matrix are shown in the formula:

[0123] ;

[0124] in, Let J represent the element in the m-th row and p-th column of the sensitivity matrix J, and Ω represent the integration domain of the entire solution region. This represents the electric field distribution calculated under the i-th current excitation mode. dΩ represents the electric field distribution calculated under the j-th current excitation mode, where i ≠ j, and dΩ represents a volume element.

[0125] Step S422: Construct the objective function using improved Tikhonov regularization; as shown in the formula:

[0126] ;

[0127] Where min represents minimizing, i.e. finding the minimum solution of the objective function value, Δσ represents the conductivity correction amount, ΔV represents the voltage change measured on the boundary electrode, λ represents the regularization parameter, and L represents the regularization matrix;

[0128] Step S423: The regularization parameter is obtained by using an adaptive decay strategy that decreases with the number of iterations; as shown in the formula:

[0129] ;

[0130] Where α represents the proportionality coefficient, β represents the decay coefficient, k represents the number of iterations, and exp(-βk) represents the exponential decay function. Representation matrix The F-norm;

[0131] Step S424: The regularization matrix L is constructed based on the discrete Laplace operator to constrain the smoothness of the solution; the elements of the regularization matrix are shown in the formula:

[0132] ;

[0133] in, This represents the element in the r-th row and s-th column of the regularization matrix L. Let represent the element in the r-th row and t-th column of the regularization matrix L.

[0134] Further, step S43 involves efficient iterative solution, achieved through an iterative process of solving for the conductivity correction using the conjugate gradient method. This process specifically includes updating the iterative solution, determining the conjugate direction, and calculating the gradient of the objective function. The specific steps are as follows:

[0135] Step S431: The conductivity correction is obtained using the conjugate gradient method, as shown in the formula:

[0136] ;

[0137] in, This represents the conductivity correction amount in the (k+1)th iteration. This represents the conductivity correction amount in the k-th iteration. This represents the step size of the k-th iteration. This indicates the conjugate direction of the k-th iteration;

[0138] Step S432, solve for the conjugate direction, as shown in the formula:

[0139] ;

[0140] in, This represents the gradient of the objective function in the k-th iteration. This represents the transpose of the gradient of the objective function in the k-th iteration. This represents the transpose of the gradient of the objective function in the (k-1)th iteration. This indicates the conjugate direction of the (k-1)th iteration;

[0141] Step S433: Calculate the gradient of the objective function to understand its descent trend, as shown in the formula:

[0142] ;

[0143] in, This represents the transpose of the sensitivity matrix J. This represents the transpose of the regularization matrix L;

[0144] Step S434: During the iteration process of steps S431 to S433, a convergence condition is set: the change in the objective function value is less than a threshold or the maximum number of iterations is reached; when the iteration process terminates after satisfying the convergence condition, the final conductivity correction amount Δσ is obtained. final Adding this to the initial conductivity distribution σ0 yields the preliminary conductivity distribution image σ reconstructed by the improved electrical impedance tomography algorithm. EIT .

[0145] Furthermore, in step S5, based on the conductivity distribution image inside the dam seepage barrier obtained in step S4, a deep learning model is used to analyze and identify defects; the specific steps are as follows:

[0146] Step S51: Construct a simulation training database; establish a parametric geometric model library containing various dam structures, dam seepage prevention materials, and foundation types; embed typical defects with adjustable parameters, including seepage channels, cracks, and cavities, into the model base of the geometric model library; for each parametric model, perform forward problem calculations of electrical impedance tomography to simulate the excitation and measurement processes of steps S1 to S3, generating simulated voltage data; subsequently, input the simulated voltage data into the improved electrical impedance tomography algorithm in step S4 for inversion, generating the corresponding three-dimensional conductivity distribution image as input data for training samples; finally, generate a simulation training database containing more than 100,000 sets of samples.

[0147] Step S52: Design and construct a 3D defect recognition model; adopt a U-shaped network as the basic architecture; replace the encoder of the U-shaped network with a residual network-50 as the backbone feature extraction network; embed a spatial attention module at the jump connection between the encoder and the decoder; use a normalized exponential function at the end of the decoder to achieve pixel-level defect classification.

[0148] Step S53: Train the end-to-end 3D defect recognition model from step S52; use a hybrid loss function of Dess loss and cross-entropy loss to train the 3D defect recognition model and obtain the trained 3D defect recognition model.

[0149] In step S54, the conductivity distribution image inside the dam seepage prevention body obtained in step S4 is input into the trained three-dimensional defect recognition model. The encoder is used to downsample the image to 1 / 32 resolution and extract feature maps at different levels. Subsequently, the feature maps are upsampled by the decoder and fused with the same-scale encoder features weighted by skip connections and spatial attention modules. Finally, a defect classification probability map of the same size as the input image is output.

[0150] Step S55: For the defect classification probability map output in step S54, a three-dimensional connected component analysis algorithm is used to identify each independent defect; a sub-pixel localization algorithm is used to calculate the three-dimensional coordinates of the geometric center of each independent defect; finally, the category, three-dimensional coordinates of the geometric center, and equivalent size information of all defects are integrated to generate and output a three-dimensional defect distribution map with spatial coordinates.

[0151] Further, in step S6, an electro-seepage coupled inversion model is constructed to convert the conductivity distribution image inside the dam seepage barrier obtained in step S4 into a permeability coefficient distribution, and a dam seepage barrier integrity evaluation report is generated; the specific steps are as follows:

[0152] Step S601: Based on the inherent physical relationship between electrical conductivity and permeability in soil and rock media, and using the traditional Alzer's law as a foundation, an electro-permeability coupled inversion model is established, as shown in the formula:

[0153] ;

[0154] Where K represents the permeability coefficient, with units of centimeters per second or meters per day. Volume conductivity is expressed in Siemens per meter (S / m). The empirical coefficients of the model are represented by u, which depend on the mineral composition, pore structure and cementation of the dam seepage prevention material, and u = 1, 2, 3, 4.

[0155] Step S602: Accurately calibrate the empirical coefficients of the model through indoor geotechnical tests; prepare a series of dam seepage prevention material samples with different mix ratios, including different moisture contents, densities, and clay contents; for each dam seepage prevention material sample, use a permeameter to measure the permeability coefficient and a resistivity meter to measure the volumetric conductivity.

[0156] Step S603: Substitute the measured volumetric conductivity and permeability coefficient into the electro-seepage coupling inversion model, and use the nonlinear least squares method to perform regression analysis to solve for the specific coefficients of the dam seepage prevention material, i.e., the empirical coefficients of the model.

[0157] Step S604: Using the conductivity distribution image inside the dam seepage barrier from step S4 as input, and using the empirical coefficients of the model from step S603, the initial permeability coefficient distribution of the entire dam seepage barrier region is initially calculated through the electro-seepage coupling inversion model.

[0158] Step S605: Based on the initial permeability coefficient distribution of the entire dam seepage prevention area, construct the governing equations for the steady-state seepage field, as shown in the formula:

[0159] ;

[0160] Where K represents the permeability coefficient distribution and h represents the total head, which drives seepage.

[0161] Step S606: Discretize the three-dimensional structure of the dam seepage barrier into a finite element mesh; apply the corresponding head boundary or flow boundary to the finite element mesh according to the actual operating conditions of the dam seepage barrier; solve the control equation of the steady-state seepage field using the finite element method to obtain the head distribution within the entire dam seepage barrier area.

[0162] Step S607: Define the objective function, which is to minimize the difference between the head distribution and the measured head distribution; starting from the initial permeability coefficient distribution, use the gradient descent method or Gauss-Newton method to repeatedly adjust the permeability coefficient distribution, and return to step S606 to solve the control equation of the steady-state seepage field until the objective function value is less than the preset threshold, and finally obtain the high-precision permeability coefficient distribution.

[0163] Step S608: Perform spatial analysis on the obtained high-precision permeability coefficient distribution; set a critical permeability coefficient value K. crit Automatically identify and mark all permeability coefficient distribution values ​​K that are greater than or equal to the critical permeability coefficient value K. crit The spatial area is marked as the weak zone in seepage prevention, and its three-dimensional spatial coordinates and volume are recorded.

[0164] Step S609: Based on the results of steps S607 and S608, calculate the structural integrity score S of the dam's seepage prevention structure; as shown in the formula:

[0165] ;

[0166] Among them, V total V represents the total volume of the dam's seepage prevention structure. weak This indicates the total volume of the identified weak points in the seepage prevention system. This represents the average permeability coefficient across all weak zones.

[0167] Step S610: Based on the results of steps S608 and S609, generate a dam seepage prevention integrity evaluation report that includes a permeability coefficient distribution map, a weak zone location map, and a structural integrity score.

[0168] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for detecting the integrity of dam seepage prevention structures based on artificial power sources, characterized in that, Includes the following steps: Step S1: Inject a wideband excitation current signal into the seepage prevention body of the dam using a multi-frequency programmable constant current source; Step S2: Based on the wideband excitation current signal from step S1, the voltage data on the surface of the dam seepage barrier is acquired in real time using a modular distributed electrode array. Step S3: Based on the voltage data of the dam seepage prevention body surface obtained in step S2, high signal-to-noise ratio voltage data is obtained by filtering and digital phase-locked amplification technology. Step S4: Based on the high signal-to-noise ratio voltage data obtained in step S3, a three-dimensional inversion calculation is performed using an improved electrical impedance tomography algorithm to reconstruct the conductivity distribution image inside the dam seepage barrier. Step S5: Based on the image of the electrical conductivity distribution inside the dam seepage barrier obtained in Step S4, analyze it using a deep learning model to identify and locate defects. Step S6: Construct an electro-seepage coupled inversion model, convert the electrical conductivity distribution image inside the dam seepage barrier obtained in Step S4 into a permeability coefficient distribution, and generate a dam seepage barrier integrity evaluation report; the specific steps are as follows: Step S601: Based on the inherent physical relationship between electrical conductivity and permeability in soil and rock media, and using the traditional Alzer's law as a foundation, an electro-permeability coupled inversion model is established, as shown in the formula: ； Where K represents the permeability coefficient, with units of centimeters per second or meters per day. Volume conductivity is expressed in Siemens per meter (S / m). , , , The empirical coefficients of the model are represented, and their values ​​depend on the mineral composition, pore structure, and cementation of the dam's seepage prevention material. Step S602: Accurately calibrate the empirical coefficients of the model through indoor geotechnical tests; prepare a series of dam seepage prevention material samples with different mix proportions, including different moisture contents, densities, and clay contents; for each dam seepage prevention material sample, use a permeameter to measure the permeability coefficient and a resistivity meter to measure the volumetric conductivity. Step S603: Substitute the measured volumetric conductivity and permeability coefficient into the electro-seepage coupling inversion model, and use the nonlinear least squares method to perform regression analysis to solve for the specific coefficients of the dam seepage prevention material, i.e., the empirical coefficients of the model. Step S604: Using the conductivity distribution image inside the dam seepage barrier from step S4 as input, and using the empirical coefficients of the model from step S603, the initial permeability coefficient distribution of the entire dam seepage barrier region is initially calculated through the electro-seepage coupling inversion model. Step S605: Based on the initial permeability coefficient distribution of the entire dam seepage prevention area, construct the governing equations for the steady-state seepage field, as shown in the formula: ； in, Indicates the permeability coefficient distribution. This represents the total head, which drives seepage. Denotes divergence, Represents the gradient; Step S606: Discretize the three-dimensional structure of the dam seepage barrier into a finite element mesh; apply the corresponding head boundary or flow boundary to the finite element mesh according to the actual operating conditions of the dam seepage barrier; solve the control equation of the steady-state seepage field using the finite element method to obtain the head distribution within the entire dam seepage barrier area. Step S607: Define the objective function, which is to minimize the difference between the head distribution and the measured head distribution; starting from the initial permeability coefficient distribution, use the gradient descent method or Gauss-Newton method to repeatedly adjust the permeability coefficient distribution, and return to step S606 to solve the control equation of the steady-state seepage field until the objective function value is less than the preset threshold, and finally obtain the high-precision permeability coefficient distribution. Step S608: Perform spatial analysis on the obtained high-precision permeability coefficient distribution; set a critical permeability coefficient value. Automatically identify and label all permeability coefficient distribution values. Greater than or equal to the critical permeability coefficient value The spatial area is marked as the weak zone in seepage prevention, and its three-dimensional spatial coordinates and volume are recorded. Step S609: Based on the results of steps S607 and S608, calculate the structural integrity score of the dam's seepage prevention structure. As shown in the formula: ； in, This indicates the total volume of the dam's seepage prevention structure. This indicates the total volume of the identified weak points in the seepage prevention system. This represents the average permeability coefficient across all weak zones. Step S610: Based on the results of steps S608 and S609, generate a dam seepage prevention integrity evaluation report that includes a permeability coefficient distribution map, a weak zone location map, and a structural integrity score.

2. The method for detecting the integrity of dam seepage prevention bodies based on artificial power supply according to claim 1, characterized in that, In step S1, a wideband excitation current signal is injected into the dam's seepage prevention body using a multi-frequency programmable constant current source; the specific steps are as follows: Step S11: Establish a dam seepage prevention material-frequency response database and pre-store the optimal excitation frequency bands for different materials; Step S12: The multi-frequency programmable constant current source automatically identifies the type of dam seepage prevention material by rapidly scanning the impedance spectrum and intelligently matches the optimal excitation frequency band. Step S13: The proportional-integral-derivative closed-loop control algorithm is used to dynamically adjust the multi-frequency programmable constant current source to inject a wideband excitation current signal of 0.1-1000 Hz into the seepage prevention body of the dam. Step S14: The multi-frequency programmable constant current source integrates an overload protection module. When the multi-frequency programmable constant current source detects a contact resistance > 10 kΩ, it automatically cuts off the output and alarms.

3. The method for detecting the integrity of dam seepage prevention bodies based on artificial power supply according to claim 2, characterized in that, In step S2, based on the broadband excitation current signal from step S1, a modular distributed electrode array is used to acquire the voltage data on the surface of the dam's seepage prevention body in real time; the specific steps are as follows: Step S21: Modular distributed electrode array, deploying 128 high-precision silver-silver chloride electrodes with a 1-meter spacing, integrating a preamplifier circuit with a gain of 100 times and an impedance detection module with a contact impedance of <3 kΩ from the high-precision silver-silver chloride electrodes; Step S22: Based on the modular distributed electrode array in step S21, multi-channel time synchronization is achieved through synchronous triggering via the Global Positioning System. A 24-bit sigma-delta analog-to-digital converter is used for synchronous sampling, and voltage data of the dam seepage prevention body surface is obtained in conjunction with 0.1-1.5 kHz adaptive bandpass filtering and 40-70 Hz adjustable dynamic power frequency notch filtering.

4. The method for detecting the integrity of dam seepage prevention bodies based on artificial power supply according to claim 3, characterized in that, In step S3, based on the voltage data of the dam seepage barrier surface obtained in step S2, high signal-to-noise ratio voltage data is obtained by filtering and digital phase-locked amplification technology; the specific steps are as follows: Step S31: Perform primary noise suppression on the voltage data of the dam seepage prevention body surface using a hardware bandpass filter; Step S32: An adaptive notch filter based on the least mean square algorithm is used to track and eliminate the 50 Hz power frequency and 50 Hz power frequency harmonic interference in real time, with a convergence time of <100 milliseconds. Step S33: Combining digital lock-in amplification technology, coherent detection is performed using the wideband excitation current signal from step S1 as the reference frequency. In-phase and quadrature components are extracted through orthogonal vector decomposition. The output signal-to-noise ratio is monitored in real time. The passband width and stopband attenuation parameters of the 16th-order finite-length unit impulse response filter are dynamically adjusted. The signal-to-noise ratio can be stably maintained at >90 dB under complex working conditions, resulting in high signal-to-noise ratio voltage data.

5. The method for detecting the integrity of dam seepage prevention bodies based on artificial power supply according to claim 4, characterized in that, In step S4, based on the high signal-to-noise ratio voltage data obtained in step S3, a three-dimensional inversion calculation is performed using an improved electrical impedance tomography algorithm to reconstruct the conductivity distribution image inside the dam's seepage barrier. The improved electrical impedance tomography algorithm consists of four steps: accurate forward problem modeling, inverse problem construction and improvement, efficient iterative solution, and post-image optimization. The specific steps are as follows: Step S41: Accurate forward problem modeling, establishing a complete electrode model that includes contact resistance effects; Step S42, Inverse problem construction and improvement: The sensitivity matrix is ​​calculated by admittance method and the objective function is constructed by combining improved Tikhonov regularization with a regularization matrix based on discrete Laplace operator to constrain the smoothness of the solution. Step S43, Efficient Iterative Solution: The iterative process of solving the conductivity correction using the conjugate gradient method includes iterative solution updating, conjugate direction solving, and calculation of the objective function gradient. Step S44, post-image optimization, uses a pre-trained deep residual network to optimize the image, in order to restore details and suppress noise, as shown in the formula: ； in, Image showing the electrical conductivity distribution inside the seepage barrier of the dam. This represents a preliminary conductivity distribution image. Let θ represent a convolutional neural network, and let θ represent the set of parameters of the convolutional neural network.

6. The method for detecting the integrity of dam seepage prevention bodies based on artificial power supply according to claim 5, characterized in that, Step S41, accurate forward problem modeling, establishing a complete electrode model including contact resistance effects; specific steps are as follows: Step S411: The governing equations for the complete electrode model are described using the Poisson equation to depict the potential distribution under the steady-state electric field, as shown in the formula: ； in, Denotes divergence, Indicates electrical conductivity. Represents the gradient. Represents electric potential; Step S412: Boundary conditions for the complete electrode model, considering the contact impedance of the electrode-dielectric interface.

7. The method for detecting the integrity of dam seepage prevention bodies based on artificial power supply according to claim 6, characterized in that, Step S42, Inverse Problem Construction and Improvement: The sensitivity matrix is ​​calculated using the admittance method, and the objective function is constructed using the improved Tikhonov regularization, along with a regularization matrix based on the discrete Laplace operator, to constrain the smoothness of the solution; the specific steps are as follows: Step S421: Calculate the sensitivity matrix using the admittance method. The elements of the sensitivity matrix are shown in the formula: ； in, Let represent the element in the m-th row and p-th column of the sensitivity matrix J, and Ω represent the integration domain of the entire solution region. This represents the electric field distribution calculated under the i-th current excitation mode. This represents the electric field distribution calculated under the j-th current excitation mode. , Represents a volume element; Step S422: Construct the objective function using improved Tikhonov regularization; as shown in the formula: ； Where min represents minimizing, i.e. finding the minimum solution of the objective function value, Δσ represents the conductivity correction amount, ΔV represents the voltage change measured on the boundary electrode, λ represents the regularization parameter, and L represents the regularization matrix; Step S423: The regularization parameter is obtained by using an adaptive decay strategy that decreases with the number of iterations; as shown in the formula: ； Where α represents the proportionality coefficient and β represents the attenuation coefficient. Indicates the number of iterations. Represents the exponentially decaying function. Representation matrix The F-norm; Step S424, the regularization matrix L is constructed based on the discrete Laplace operator to constrain the smoothness of the solution.

8. The method for detecting the integrity of dam seepage prevention bodies based on artificial power supply according to claim 7, characterized in that, Step S43, efficient iterative solution, involves an iterative process to solve for the conductivity correction using the conjugate gradient method. This includes updating the iterative solution, determining the conjugate direction, and calculating the gradient of the objective function. The specific steps are as follows: Step S431: The conductivity correction is obtained using the conjugate gradient method, as shown in the formula: ； in, Indicates the first The conductivity correction amount in the next iteration. Indicates the first The conductivity correction amount in the next iteration. Indicates the first The step size of the next iteration. Indicates the first The conjugate direction of the next iteration; Step S432, solve for the conjugate direction, as shown in the formula: ； in, Indicates the first The gradient of the objective function in the next iteration Indicates the first Transpose of the gradient of the objective function in the next iteration Indicates the first Transpose of the gradient of the objective function in the next iteration Indicates the first The conjugate direction of the next iteration; Step S433: Calculate the gradient of the objective function to understand its descent trend, as shown in the formula: ； in, This represents the transpose of the sensitivity matrix J. Represents the regularization matrix Transpose of; Step S434: During the iteration process of steps S431 to S433, a convergence condition is set: the change in the objective function value is less than a threshold or the maximum number of iterations is reached; when the iteration process terminates after satisfying the convergence condition, the final conductivity correction amount Δσ is obtained. final Adding this to the initial conductivity distribution σ0 yields the preliminary conductivity distribution image reconstructed by the improved electrical impedance tomography algorithm. .

9. The method for detecting the integrity of a dam seepage barrier based on artificial power source according to claim 8, characterized in that, In step S5, based on the internal conductivity distribution image of the dam seepage barrier obtained in step S4, a deep learning model is used for analysis to identify and locate defects; the specific steps are as follows: Step S51: Construct a training database for simulation; establish a parametric geometric model library containing various dam structures, dam seepage prevention materials, and foundation types; embed typical defects with adjustable parameters, including seepage channels, cracks, and cavities, into the model base of the geometric model library. For each parameterized model, the forward problem calculation of electrical impedance tomography is performed to simulate the excitation and measurement process in steps S1 to S3 and generate simulated voltage data. Subsequently, the simulated voltage data is input into the improved electrical impedance tomography algorithm in step S4 for inversion to generate the corresponding three-dimensional conductivity distribution image as input data for training samples. Finally, a simulation training database containing more than 100,000 samples is generated. Step S52: Design and construct a three-dimensional defect recognition model; adopt a U-shaped network as the basic architecture; replace the encoder of the U-shaped network with a residual network-50 as the backbone feature extraction network; At the jump connection between the encoder and decoder, a spatial attention module is embedded; at the end of the decoder, a normalized exponential function is used to achieve pixel-level defect classification. Step S53: Train the end-to-end 3D defect recognition model from step S52; use a hybrid loss function of Dess loss and cross-entropy loss to train the 3D defect recognition model and obtain the trained 3D defect recognition model. Step S54: Input the conductivity distribution image inside the dam seepage prevention body from step S4 into the trained three-dimensional defect recognition model, and gradually downsample it to 1 / 32 resolution through the encoder to extract feature maps at different levels. Subsequently, the feature map is progressively upsampled by the decoder and fused with the same-scale encoder features weighted by skip connections and spatial attention modules, finally outputting a defect classification probability map of the same size as the input image; Step S55: For the defect classification probability map output in step S54, a three-dimensional connected component analysis algorithm is used to identify each independent defect; a sub-pixel localization algorithm is used to calculate the three-dimensional coordinates of the geometric center of each independent defect; finally, the category, three-dimensional coordinates of the geometric center, and equivalent size information of all defects are integrated to generate and output a three-dimensional defect distribution map with spatial coordinates.

Images