Water-retaining sand hydrophobic performance detection method and system based on impedance tomography

Abstract

The application discloses a water-retaining sand hydrophobic performance detection method and system based on impedance tomography, and belongs to the technical field of material performance electrical detection. The method comprises the following steps: filling a water-retaining sand sample into an insulating annular container and compacting, injecting liquid simulation water into the center to simulate dynamic seepage of water, collecting boundary voltage data to generate a global impedance data set, constructing a conductivity distribution inversion model to solve internal conductivity distribution, outputting a global conductivity matrix of the water-retaining sand, generating a time-series conductivity evolution image sequence, establishing a calibration mapping relationship between conductivity and water content, converting the water content pixel by pixel to generate a time-series distribution cloud image, extracting water content statistical features, calculating water seepage rate, radial diffusion coefficient, global circumferential segregation degree and hydrophobic retardation factor to form a hydrophobic performance quantitative feature set, grading the water-retaining sand hydrophobic performance and identifying abnormal areas to form a hydrophobic performance detection report, and realizing integrated nondestructive characterization of internal water distribution, dynamic seepage and hydrophobic uniformity of the water-retaining sand.

Description

Technical Field

[0001] This invention belongs to the field of electrical testing technology for material properties, and relates to a method and system for testing the hydrophobic properties of water-retaining sand based on impedance tomography. Background Technology

[0002] Water-retaining sand, as a novel geotechnical engineering material with water-retaining, hydrophobic, and seepage-proof properties, can be applied to roadbed seepage prevention, ecological restoration, and water conservancy projects. Currently, the hydrophobic performance testing of water-retaining sand mainly employs traditional methods such as weighing, contact sensor measurement, surface observation, or slice sampling. However, these traditional methods are mostly contact-based and destructive, failing to achieve in-situ, visual observation of internal moisture content and the wetting interface. Furthermore, the test results are easily affected by human intervention and environmental interference, making it difficult to quantitatively characterize hydrophobic uniformity, seepage paths, and dynamic wetting processes. Simultaneously, existing testing methods cannot achieve full-domain imaging of the three-dimensional hydrophobic characteristics within water-retaining sand, resulting in low testing efficiency and large data dispersion, which fails to meet the needs of rapid testing for the research and development of new water-retaining sand materials and engineering quality.

[0003] Chinese patent CN115901524A discloses a method for testing the water retention of building mortar. This method uses specialized water retention testing equipment, which includes a feeding unit, a settling unit, a drying unit, and a mold. The technical solution involves placing a substrate slice directly on top of a mold filled with mortar mixture, allowing the lower surface of the substrate slice to directly contact the mortar mixture for water absorption measurement. This solves the problem that substrate slices, lacking deformability, cannot absorb water from the mortar mixture if a metal mesh is used as a barrier. It allows for testing the water retention of mortar on different substrates, providing a basis for optimizing the dosage of water-retaining agents for different substrates. Furthermore, by weighing the substrate slices dried to constant weight before water absorption and after water absorption, the influence of sand particles adhering to the surface of the substrate slices on the experimental data is effectively eliminated, improving the accuracy of the experimental data.

[0004] Although there is an existing method for testing the water retention of building mortar, which involves directly contacting a base layer slice with the mortar mixture to absorb water, using multi-stage drying and weighing, and correcting for errors caused by sand particle adhesion and environmental moisture absorption, thus achieving adaptability to different absorbent base layers, accurate calculation of mortar water retention rate, and providing a reliable basis for optimizing the dosage of water-retaining agent, it still has the problem of not being able to non-destructively characterize the internal moisture distribution, dynamic seepage, and hydrophobic uniformity of the water-retaining sand. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention aims to provide a method and system for detecting the hydrophobic properties of water-retaining sand based on impedance tomography. This method enables visualization of the dynamic seepage process of water within the sand and high-precision quantitative detection of hydrophobic uniformity, solving the problems of traditional detection methods failing to capture the microscopic seepage characteristics of water-retaining sand, lacking unified quantitative evaluation standards, and struggling to locate abnormal hydrophobic areas. By constructing a multi-frequency fusion conductivity distribution inversion model, establishing a calibration correction relationship between conductivity and water content, extracting core quantitative indicators to form a set of hydrophobic performance quantitative features, and designing a dual-indicator grading system, this invention achieves scientific grading of the hydrophobic properties of water-retaining sand and accurate identification of abnormal areas, providing reliable detection methods and system support for the performance evaluation, quality control, and engineering applications of water-retaining sand materials.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for detecting the hydrophobic properties of water-retaining sand based on impedance tomography includes:

[0008] The water-retaining sand sample was filled into an insulating ring container and compacted. Liquid was injected into the center of the sample to simulate the dynamic seepage process of water. Boundary voltage data was collected by cyclic scanning in adjacent excitation mode, and preprocessed to generate a global impedance dataset.

[0009] Based on the global impedance dataset, a conductivity distribution inversion model is constructed. The boundary voltage amplitude and phase under multi-frequency excitation are used as observation constraints to solve the internal conductivity distribution. The global conductivity matrix of water-retaining sand is output and mapped with spatial coordinates to generate a time-series conductivity evolution image sequence.

[0010] Establish a calibration mapping relationship between electrical conductivity and water content, convert the time-series electrical conductivity evolution image sequence pixel by pixel into water content, generate a time-series distribution cloud map, extract statistical features of water content, calculate water seepage rate, radial diffusion coefficient, global circumferential segregation and hydrophobic hindrance factor, and form a set of quantitative features of hydrophobic performance.

[0011] Based on the quantitative feature set of hydrophobic performance, a judgment threshold is set to classify and judge the hydrophobic performance of water-retaining sand. Abnormal areas are identified according to the global circumferential segregation and hydrophobic barrier factor, and a hydrophobic performance test report is generated.

[0012] Specifically, the steps for generating a time-series conductivity evolution image sequence include:

[0013] Based on the global impedance dataset, an inversion model of conductivity distribution is constructed according to the dielectric properties of water-retaining sand porous media.

[0014] Using multiple sets of global impedance datasets with different water contents and different seepage times as the training set, and the root mean square error of voltage residuals as the loss function, the conductivity distribution inversion model is trained by combining the Adam optimizer.

[0015] After training, the real-time collected global impedance dataset is input, and the global conductivity matrix of water-retaining sand at different seepage times is output through input layer parsing, impedance feature fusion layer multi-frequency feature weighted fusion, and inversion optimization layer iterative inversion optimization.

[0016] By using pseudo-color normalized encoding, the global conductivity matrix of water-retaining sand at different seepage times is visualized, generating a time-series conductivity evolution image sequence.

[0017] Specifically, the steps for constructing the conductivity distribution inversion model include:

[0018] A conductivity distribution inversion model is constructed, including an input layer, an impedance feature fusion layer, an inversion optimization layer, and an output layer;

[0019] The input layer receives and parses the global impedance dataset;

[0020] Impedance characteristic fusion layer: The water content of water-retaining sand is calculated based on the calibration formula of impedance amplitude and water content under different frequency excitations. ;

[0021] Set moisture content threshold , ;

[0022] like If so, the corresponding area is designated as a high moisture content area; if Then the corresponding area is divided into a dry-wet transition zone; if If so, the corresponding area is designated as a low moisture content area;

[0023] Extract the amplitude and phase characteristics of multi-channel complex impedance at low, medium, and high frequencies at a single moment from the global impedance dataset;

[0024] By setting low-frequency, mid-frequency, and high-frequency weights using a weighted fusion algorithm, a multi-frequency fusion feature vector is generated.

[0025] Specifically, the steps for constructing the conductivity distribution inversion model also include:

[0026] The inversion optimization layer uses the multi-frequency fusion feature vector as the observation constraint to construct the conductivity inversion objective function. The conductivity distribution to be solved is discretized into multiple units and substituted into the electric field control equation. The finite difference method is used for discretization, and the theoretical response value of the boundary voltage of each unit is obtained by iterative solution.

[0027] The difference between the theoretical response value and the measured boundary voltage value is calculated. After systematically collecting and organizing the boundary voltage measurement channels point by point, a voltage residual sequence is generated.

[0028] With the goal of minimizing the voltage residual sequence, the conjugate gradient method is used to solve for the minimum value of the conductivity inversion objective function. In each iteration, the gradient vector and the conjugate gradient direction are solved to update and correct the internal conductivity distribution, and the optimal conductivity distribution result in the global domain is output.

[0029] The output layer outputs the global conductivity matrix of water-retaining sand at the corresponding seepage time based on the optimal global conductivity distribution results, and maps the global conductivity matrix of water-retaining sand to spatial coordinates one by one.

[0030] Specifically, the steps for generating a time-series distribution cloud map include:

[0031] Five characteristic points were selected from typical moisture content ranges: dry state, low moisture content state, critically wet state, high moisture content state, and near-saturation state.

[0032] Prepare standard samples with corresponding moisture content, and prepare at least 3 parallel samples for each standard sample;

[0033] Standard samples were sequentially placed into an insulating ring container, and global impedance datasets at low, medium, and high frequencies were collected. The global conductivity matrix of water-retaining sand for each standard sample was output through the conductivity distribution inversion model. The average conductivity value corresponding to each moisture content was obtained by averaging the parallel samples.

[0034] Using the measured moisture content as the abscissa and the average electrical conductivity as the ordinate, a five-point calibration raw data pair was formed. A scatter plot was drawn and a baseline mapping curve was generated using cubic spline interpolation.

[0035] The local coefficient of variation was calculated based on the global conductivity matrix of water-retaining sand for each standard sample.

[0036] The correction coefficient fitting factor is determined by the least squares method, and the local correction coefficient obtained by the correction coefficient formula is used to generate the correction mapping curve, so as to minimize the deviation between the corrected moisture content and the measured moisture content.

[0037] Specifically, the steps for generating a time-series distribution cloud map also include:

[0038] Select validation samples that were not calibrated, measure the water-retaining sand global conductivity matrix of each validation sample, calculate the moisture content of each validation sample, and calculate the maximum absolute error by combining the measured moisture content;

[0039] Set the maximum error threshold;

[0040] If the maximum absolute error is greater than the maximum error threshold, adjust the node positions of the cubic spline interpolation or increase the number of calibration points, and recalculate the correction coefficients until the maximum absolute error meets the standard.

[0041] Align all frames to the spatial coordinate system of the first frame, read the conductivity value of each pixel in the time-series conductivity evolution image sequence, calculate the local coefficient of variation and correction coefficient, and use the correction mapping curve to calculate the corrected moisture content.

[0042] The time series of each pixel is smoothed by Kalman filtering, and the time series water content value is output to generate a time series water content image sequence.

[0043] A time-series moisture content image sequence was visualized using a moisture content grading pseudo-color mapping technique to generate a time-series distribution cloud map.

[0044] Specifically, the steps for forming a quantitative feature set of hydrophobic properties include:

[0045] The time-series moisture content image sequence was converted into polar coordinates; the insulating annular container was divided into equal-width circular bands along the radial direction and equal-angle sectors along the circumference.

[0046] The regional average moisture content of the overlapping areas of each equal-width circular zone and each equal-angle sector is statistically analyzed, and the radial average moisture content curve is obtained by calculating the circumferential average moisture content.

[0047] The location of the wetting front is extracted frame by frame, and the water seepage rate is calculated after filtering and smoothing. The water seepage rate time series is obtained by point-by-point difference.

[0048] A radial water content profile was fitted during the relatively stable middle and late stages of seepage to determine the radial extension characteristic length and calculate the radial diffusion coefficient.

[0049] Calculate the coefficient of variation of circumferential moisture content based on the regional average moisture content, define circumferential segregation and calculate global circumferential segregation;

[0050] Determine the theoretical seepage rate and reference diffusion coefficient, calculate the initial hydrophobic retardation factor by combining the water seepage rate and radial diffusion coefficient, and take the arithmetic mean of the seepage steady stage as the hydrophobic retardation factor.

[0051] All time-series indicators are aligned along the time axis, and statistical features of water content are extracted from the time-series distribution cloud map to form a set of quantitative features for hydrophobic performance.

[0052] Specifically, the steps for determining the classification include:

[0053] Extract the hydrophobic barrier factor and global circumferential segregation from the set of quantitative features of hydrophobic properties;

[0054] Set judgment thresholds, including the first judgment threshold. , , Second judgment threshold , , The water-retaining and water-draining properties of sand are classified and determined.

[0055] If hydrophobic blocking factor And global circumferential segregation If it is, it is judged to be a Class 1 hydrophobic property; if And global circumferential segregation If so, it is determined to be a level 2 hydrophobic property;

[0056] like And global circumferential segregation If it is positive, it is determined to be a Class III hydrophobic property; otherwise, it is determined to be a Class IV hydrophobic property.

[0057] Specifically, the steps for generating a hydrophobic performance test report include:

[0058] Based on the hierarchical judgment results, combined with the hydrophobic barrier factor and global circumferential segregation, abnormal regions are identified.

[0059] If hydrophobic blocking factor If the global circumferential segregation is greater than that of the surrounding area, it is determined to be a local hydrophobic failure zone;

[0060] like But global circumferential segregation If so, it is determined to be a water accumulation area;

[0061] If the hydrophobic barrier factor and the global circumferential segregation of the corresponding area both fluctuate drastically, it is determined to be an area with abnormal seepage.

[0062] Add hydrophobicity ratings for different regions to the time-series distribution cloud map;

[0063] Based on global circumferential parseness combined with a second judgment threshold Calculate the hydrophobicity uniformity index;

[0064] The results of the graded judgment and the results of the abnormal area identification are integrated to form structured report data, and a hydrophobic performance test report is generated using a visual template.

[0065] The water-retaining sand hydrophobic performance testing system based on impedance tomography includes: a global impedance acquisition module, an inversion imaging module, a hydrophobic performance quantification module, and a performance grading and determination module.

[0066] The global impedance acquisition module is used to fill and compact the water-retaining sand sample into an insulating annular container, inject liquid at the center of the sample to simulate the dynamic seepage process of water; acquire boundary voltage data, and preprocess to generate a global impedance dataset.

[0067] The inversion imaging module is used to construct a conductivity distribution inversion model. Using the boundary voltage amplitude and phase under multi-frequency excitation as observation constraints, it solves the internal conductivity distribution, outputs the water-retaining sand global conductivity matrix and maps it with spatial coordinates to generate a time-series conductivity evolution image sequence.

[0068] The hydrophobic performance quantification module is used to establish a calibration mapping relationship between electrical conductivity and water content, convert the time-series electrical conductivity evolution image sequence pixel by pixel into water content, generate a time-series distribution cloud map, extract statistical features of water content, calculate water seepage rate, radial diffusion coefficient, global circumferential segregation and hydrophobic hindrance factor, and form a set of hydrophobic performance quantification features.

[0069] The performance grading and judgment module, based on the hydrophobic performance quantitative feature set, sets a judgment threshold to grade and judge the hydrophobic performance of water-retaining sand, identifies abnormal areas based on global circumferential segregation and hydrophobic barrier factor, and generates a hydrophobic performance test report.

[0070] The beneficial effects of this invention are:

[0071] 1. This invention, based on impedance tomography (ETT) technology, constructs a conductivity distribution inversion model comprising an input layer, an impedance feature fusion layer, an inversion optimization layer, and an output layer. Combining the dielectric properties of porous water-retaining sand, and using amplitude and phase under multi-frequency excitation as observation constraints, it achieves efficient conductivity distribution inversion. Simultaneously, a five-point calibration and correction mechanism for conductivity and water content is established. Through sample calibration and Kalman filtering smoothing, a high-precision conversion from conductivity to water content is achieved. Pseudo-color normalized encoding is used to visualize the time-series water content in a cloud map. This invention overcomes the limitations of traditional detection methods that cannot capture the dynamic seepage process of water within water-retaining sand in real time. It achieves a full-link characterization from conductivity inversion to water content distribution and seepage characteristics, transforming the detection of the hydrophobic properties of water-retaining sand from macroscopic qualitative to microscopic quantitative, significantly improving the accuracy and visualization of the detection results.

[0072] 2. This invention constructs a quantitative feature set of hydrophobic performance based on hydrophobic barrier factor and global circumferential segregation, and designs a four-level classification judgment system with dual-index differential thresholds. Based on the numerical characteristics and fluctuation patterns of the two core indicators, it accurately identifies local hydrophobic failure zones, moisture accumulation zones, and abnormal seepage areas, and calculates the hydrophobic uniformity index. The classification results and abnormal area information are integrated to form a structured and visualized test report. This invention solves the problems of traditional water-retaining sand hydrophobic performance testing lacking unified quantitative standards, having ambiguous classifications, and being unable to locate abnormal areas. It not only provides a scientific and quantifiable evaluation basis for the hydrophobic performance of water-retaining sand, but also accurately points to the defective areas of the material's hydrophobic performance. This provides reliable practical guidance for the research and development optimization, quality control, and engineering application selection of water-retaining sand materials, effectively improving the practicality and engineering value of the testing system. Attached Figure Description

[0073] Figure 1 A schematic diagram of a method for detecting the hydrophobic properties of water-retaining sand based on impedance tomography;

[0074] Figure 2 This is a flowchart illustrating the generation of a time-series conductivity evolution image sequence in this invention;

[0075] Figure 3 This is a flowchart of the process for generating a time-series distribution cloud map in this invention;

[0076] Figure 4 This is a flowchart illustrating the process of forming a quantitative feature set of hydrophobic properties in this invention.

[0077] Figure 5 This is a structural diagram of a water-retaining sand hydrophobicity testing system based on impedance tomography. Detailed Implementation

[0078] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0079] Example 1

[0080] refer to Figures 1 to 4 As shown in the figure, this embodiment introduces a method for detecting the hydrophobic properties of water-retaining sand based on impedance tomography, including the following steps:

[0081] The water-retaining sand sample was evenly filled to an inner diameter of 100 mm. Height is 50 In an insulated annular container, compact to dry density, ensuring a smooth surface; the particle size is 0.1~0.5 mm, based on particle size distribution characteristics. The required dry density after hydrophobic modification was determined to be 1.6 kJ / L. Sixteen annular electrodes are embedded at equal intervals along the circumference of the sidewall of an insulated annular container, ensuring close contact between the electrodes and the water-retaining sand sample. A multi-channel impedance measurement module is connected to the electrode leads. Because contact resistance can easily arise between the 16-electrode array and the water-retaining sand sample due to loose contact, interface gaps, and residual air bubbles, leading to boundary voltage measurement errors and distorted moisture content detection data, online contact resistance compensation and contact defect correction of the 16-electrode array are required. This includes: acquiring the baseline impedance of each electrode channel under no-load conditions, eliminating the electrode's own contact resistance and lead impedance; and performing four-electrode comparison calibration on adjacent excitation-measurement channels to eliminate... To address contact impedance shifts caused by poor contact between the electrode and the water-retaining sand interface, neighbor mean interpolation was used to repair abnormal electrode channels with contact impedance exceeding the normal range, suppressing boundary voltage errors. The water-retaining sand sample was left to stand for 24 hours to eliminate stress unevenness during the filling process. Deionized water was then injected at a constant flow rate into the center of the sample using a micro-injection pump to simulate the dynamic seepage process of water. A multi-frequency impedance analyzer was used to cyclically scan in adjacent excitation mode, acquiring a set of boundary voltage data every 30 seconds within the 0-60 minute seepage time. Each set of data included complex impedance values ​​from 16 excitation locations and 13 measurement channels, with the frequency point selected being a low frequency of 10. 50 medium frequency High frequency 200 To enhance sensitivity to regions with varying moisture content, a standardized global impedance dataset was generated through noise reduction, drift removal, and normalization preprocessing; the electrode width was 5. The constant flow rate is 0.5 The complex impedance value includes both amplitude and phase. The mean baseline impedance for the normal range is determined by the statistical distribution of historical unloaded baseline impedance and the 95% confidence interval. 20%;

[0082] Based on the global impedance dataset, an inversion model of conductivity distribution of water-retaining sand adapted to multi-frequency excitation is constructed. The boundary voltage amplitude and phase under multi-frequency excitation are used as observation constraints. The conjugate gradient method is used to solve the internal conductivity distribution and output the global conductivity matrix of water-retaining sand at different seepage times. The global conductivity matrix of water-retaining sand is mapped one-to-one with the spatial coordinates. Through pseudo-color normalization encoding, a time-series conductivity evolution image sequence of one frame every 30 seconds within 0-60 minutes is generated.

[0083] Based on the time-series conductivity evolution image sequence, a five-point calibration method is used to establish the calibration mapping relationship between conductivity and water content. The time-series conductivity evolution image sequence is converted into water content pixel by pixel to generate a time-series distribution cloud map of water content inside water-retaining sand. Statistical features of water content are extracted from the time-series distribution cloud map along the radial and circumferential directions. Water seepage rate, radial diffusion coefficient, circumferential segregation, and hydrophobic retardation factor are calculated. The circumferential segregation characterizes the hydrophobic uniformity, and the hydrophobic retardation factor characterizes the hydrophobic effectiveness. The dynamic seepage process and internal hydrophobic uniformity are quantitatively characterized, forming a quantitative feature set of hydrophobic performance.

[0084] Based on the quantitative feature set of hydrophobic performance, the judgment threshold is set differently to classify and judge the hydrophobic performance of water-retaining sand. Abnormal areas are identified according to circumferential segregation and hydrophobic barrier factor. Local hydrophobic failure areas, water accumulation areas and seepage abnormal areas are identified non-destructively, and a hydrophobic performance test report is generated, including but not limited to time-series distribution cloud map, hydrophobic uniformity index, and hydrophobic performance level, so as to realize the integrated non-destructive characterization of water distribution, dynamic seepage and hydrophobic uniformity inside water-retaining sand.

[0085] Specifically, the steps for generating a time-series conductivity evolution image sequence include:

[0086] Based on a global impedance dataset, and according to the dielectric properties of porous water-retaining sand, an inversion model for the conductivity distribution of water-retaining sand adapted to multi-frequency excitation is constructed. This model includes an input layer, an impedance feature fusion layer, an inversion optimization layer, and an output layer. The dielectric properties of porous water-retaining sand are characterized by low dielectric constant and hydrophobic insulation in the solid phase sand particles, while the dielectric constants of the gas phase and the liquid phase deionized water inside the pores differ. These dielectric properties are the core basis for multi-frequency impedance feature extraction and conductivity inversion, and they align with the correlation between conductivity and water content during the seepage process of water-retaining sand.

[0087] The input layer is used to receive and parse the global impedance dataset;

[0088] An impedance characteristic fusion layer is used to calculate the water content of water-retaining sand based on the calibration formula of impedance amplitude and water content under different frequency excitations. The calibration expression is as follows:

[0089]

[0090] in, To calibrate the slope coefficient, an indoor calibration experiment was conducted, based on the linear fitting results of the impedance amplitude and moisture content data of multiple sets of standard samples with known moisture content. For the first The natural logarithmic value of the complex impedance amplitude acquired by the global electrode array under group frequency excitation; The calibration intercept coefficient under multi-frequency excitation was obtained by fitting an indoor calibration experiment conducted simultaneously with the calibration slope coefficient. The excitation frequency grouping number corresponds to the three excitation frequency bands: low frequency, medium frequency, and high frequency, and is assigned a value sequentially. , , Different frequency operating conditions are distinguished; calibration is performed through calibration formulas. The calibration process provides a basis for water content correlation for multi-frequency feature fusion, which indirectly supports the accuracy of conductivity inversion.

[0091] Based on the critical saturation of water-retaining sand seepage (22% of the critical saturation for liquid interconnection) and the engineering measured moisture content boundary standard of 15%~20%, a moisture content threshold is set. , ,like If so, the corresponding area will be classified as a high moisture content area; if Then the corresponding area will be divided into a dry-wet transition zone; if If the corresponding region is divided into low moisture content regions, then the region division is used to match the sensitivity differences of different frequencies, providing theoretical support for the weight allocation of the weighted fusion algorithm.

[0092] To address the varying sensitivity of different frequencies to water-retaining sand regions with high and low water content, amplitude and phase features of 16×13 channel complex impedances at single-moment low, mid, and high frequencies were extracted from the global impedance dataset. Redundant features caused by inter-channel crosstalk and environmental electromagnetic interference were removed. A weighted fusion algorithm was used to differentiate the weights of low, mid, and high frequencies to generate a multi-frequency fusion feature vector.

[0093] The weighted fusion algorithm calculation process includes using the min-max normalization method to map the amplitude and phase characteristics of the 16×13 channel complex impedance at low, medium, and high frequencies to the [0,1] interval to eliminate dimensional differences. Combining the frequency sensitivity characteristics—low frequencies have large penetration depth and strong sensitivity to low moisture content areas; medium frequencies are balanced and stable with high identification of dry-wet transition zones; and high frequencies have fast response and high sensitivity to high moisture content areas—the amplitude and phase characteristics sensitive to high moisture content areas are assigned a high-frequency weight of 0.5, those sensitive to dry-wet transition zones are assigned a medium-frequency weight of 0.3, and those sensitive to low moisture content areas are assigned a low-frequency weight of 0.2. The normalized amplitude and phase characteristics at a single moment are multiplied by the corresponding low-frequency, medium-frequency, and high-frequency weights according to the channel dimension, and accumulated channel by channel to obtain the initial value of the fusion feature. Through centralized noise reduction and dimensional compression processing, redundant fusion information is eliminated, the feature vector dimension is unified, and a multi-frequency fusion feature vector is generated.

[0094] The inversion optimization layer is used to construct the conductivity inversion objective function with multi-frequency fused feature vectors as observation constraints, as shown in the following expression:

[0095]

[0096] in, Let be the objective function for conductivity inversion. The conductivity distribution to be solved; This represents the theoretical response value of the boundary voltage. This is the measured boundary voltage value; The regularization coefficient is used to balance the fitting accuracy of the voltage residual with the smoothness of the conductivity distribution. The value of the regularization coefficient is determined by cross-validation and ranges from 0.01 to 0.1 to ensure that the fitting accuracy and distribution smoothness are optimally balanced. For regularization terms, a first-order smooth regularization constraint is applied. By applying L2 norm constraints to the gradient of conductivity distribution, inversion ill-conditioning is suppressed to ensure the spatial continuity of conductivity distribution and avoid local abrupt changes and distortion problems.

[0097] The conductivity distribution to be solved Discretize the data into multiple elements, each corresponding to a conductivity value, and substitute these values ​​into the electric field control equations. The electric field control equations are discretized using the finite difference method, and the calculated boundary voltage values ​​for each element are obtained through iterative solutions, which are the theoretical boundary voltage response values. ;in, The potential gradient represents the spatial rate of change of potential inside the water-retaining sand, reflecting the distribution of the electric field inside the water-retaining sand. The direction of the potential gradient is the direction of the fastest potential increase, and the magnitude of the potential gradient is the amount of potential change per unit distance.

[0098] Calculate the theoretical response value of the boundary voltage Compared with the measured boundary voltage value The difference is used to obtain the voltage deviation. By traversing the boundary voltage measurement channels point by point, the voltage deviation of each boundary voltage measurement channel is collected and processed in an orderly manner to generate a voltage residual sequence. The objective is to minimize the voltage residual sequence. The conjugate gradient method is used to solve for the minimum value of the conductivity inversion objective function. In each iteration, the gradient vector and conjugate gradient direction of the conductivity inversion objective function are calculated. The internal conductivity distribution is updated and gradually corrected using the optimal iteration step size until the residuals in the voltage residual sequence converge to the convergence threshold. The result of the globally optimal conductivity distribution is then output, achieving high-precision global conductivity inversion optimization. The accuracy requirement is based on the measured boundary voltage. The accuracy requirement for conductivity inversion is: The convergence threshold is determined as follows: Based on the inversion convergence characteristics of fast and stable convergence, no iterative oscillations, and a monotonically decreasing objective function for conductivity inversion, the optimal iteration step size for this embodiment is determined to be 0.01.

[0099] The output layer, based on the global optimal conductivity distribution results output by the inversion optimization layer, outputs the global conductivity matrix of water-retaining sand at the corresponding seepage time. A spatial coordinate mapping algorithm is used to complete the one-to-one mapping between the global conductivity matrix of water-retaining sand and spatial coordinates. The specific process of the spatial coordinate mapping algorithm includes: establishing a spatial coordinate system, with the center of the insulated annular container as the origin, and the radial direction as... Axial and circumferential directions are A two-dimensional polar coordinate system is constructed, and the inner diameter of the insulating annular container is defined as 100. Height 50 The corresponding coordinate range is , The global conductivity matrix of water-retaining sand is gridded, with each element in the matrix corresponding to a detection unit of the water-retaining sand sample. The center coordinates of each detection unit are determined as follows: ,in, , These are the row and column indices of the global conductivity matrix of water-retaining sand. A one-to-one correspondence is established between the detection unit and the elements of the global conductivity matrix of water-retaining sand. Each element is assigned the center coordinates of the corresponding detection unit. The coordinate error caused by electrode installation deviation and container geometric deviation is corrected by the calibration coordinate algorithm. This ensures that each conductivity value in the global conductivity matrix of water-retaining sand is matched with the actual spatial position inside the water-retaining sand sample with high precision. This achieves synchronous correspondence between the global conductivity distribution and spatial coordinates, providing reliable coordinate support for the visualization of subsequent time-series conductivity evolution images.

[0100] Using multiple sets of global impedance datasets with different water contents and different seepage times as the training set and the root mean square error of voltage residuals as the loss function, the conductivity distribution inversion model is trained by combining the Adam optimizer. After training, the real-time collected global impedance dataset is input, and the model is iteratively inverted and optimized through input layer analysis, impedance feature fusion layer multi-frequency feature weighted fusion, and inversion optimization layer to output the water-retaining sand global conductivity matrix at different seepage times.

[0101] By using pseudo-color normalized encoding, the global conductivity matrix of water-retaining sand at different seepage times is visualized, generating a time-series conductivity evolution image sequence with one frame every 30 seconds within 0-60 minutes. The pseudo-color normalized encoding process involves extracting the maximum and minimum conductivity values ​​from the global conductivity matrix of water-retaining sand at different seepage times as a benchmark, normalizing the conductivity at each seepage time to the 0~1 interval, and corresponding it one-to-one with the pseudo-color chromatogram.

[0102] For example, taking a water-retaining sand sample used in a certain project as the test object, an inner diameter of 100 mm was used. Height 50 An insulated annular container was fitted with 16 annular electrodes. Global impedance data was collected during the 0-60 minute seepage process. A complete inversion example at the 10-minute seepage time was provided by combining parameters from a conductivity distribution inversion model. The dielectric constant of the solid-phase sand particles in the water-retaining sand sample was also analyzed. pore gas phase permittivity Dielectric constant of deionized water in porous liquid phase The parameters to be set include multi-frequency excitation to low frequency 10. 50 medium frequency High frequency 200 ,correspond , , The calibration formula is: , , Moisture content threshold , The weighted fusion algorithm uses a low-frequency weight of 0.2, a mid-frequency weight of 0.3, and a high-frequency weight of 0.5; regularization coefficient. The convergence threshold is The optimal iteration step size is 0.01; the spatial coordinate system is a two-dimensional polar coordinate system with the center of the insulating annular container as the origin, and the coordinate range is... , The conductivity matrix of the entire water-retaining sand area is 64×64.

[0103] The input layer receives and parses the global impedance dataset at 10 minutes of seepage, obtaining complex impedance data across 3 frequency bands and 16×13 channels. The values ​​at the 8th excitation location and the 6th measurement channel are low-frequency values ​​(10). Complex impedance amplitude 1250 Phase 18.6 50 MHz Amplitude 820 Phase 12.3 High frequency 200 Amplitude 480 Phase 7.9 The impedance amplitude of the remaining channels ranges from 350 to 1400. The phase range is 5.2. ~21.7 All redundant data caused by channel crosstalk and electromagnetic interference have been removed;

[0104] The impedance characteristic fusion layer is substituted into the calibration formula to calculate the water content of the water-retaining sand in each frequency band, thus obtaining the water content at low frequencies. Moisture content of medium frequency High frequency moisture content ,because , , All are greater than the moisture content threshold. If the corresponding area belongs to the high moisture content area, then the amplitude and phase characteristics of each frequency band are subjected to minimum-maximum normalization processing. Taking the amplitude as an example, the normalization value of low frequency is 1.0, the normalization value of mid frequency is 0.44, and the normalization value of high frequency is 0.0. Then, combined with the low frequency weight, mid frequency weight and high frequency weight, the initial value of the fusion feature is obtained by accumulating channel by channel through 1.0×low frequency weight + 0.44×mid frequency weight + 0.0×high frequency weight. After centering noise reduction and dimensional compression, the multi-frequency fusion feature vector is [0.33, 0.29]. All detection units in the whole domain are calculated in the above way.

[0105] The inversion optimization layer uses multi-frequency fused feature vectors as observation constraints to construct the conductivity inversion objective function as follows: The water-retaining sand sample was discretized into 64×64 detection units with a size of 0.78. ×5.625 Substituting into the electric field control equation The potential gradient of the corresponding unit cell is obtained by discretization using the finite difference method. 0.023 Substitute the measured boundary voltage value Initial conductivity 0.015 The theoretical response value of the boundary voltage is obtained. Generate voltage residual sequence After 12 iterations using the conjugate gradient method with an optimal iteration step size of 0.01, the residuals in the voltage residual sequence converge to... The output of the global optimal conductivity distribution is as follows: The conductivity range of the full-domain detection unit is Among them, the electrical conductivity range corresponding to the high water content area is: The conductivity range corresponding to the dry-wet transition region is: The electrical conductivity range corresponding to low moisture content areas is: ;

[0106] Based on the global optimal conductivity distribution results output by the inversion optimization layer, the output layer outputs a 64×64 global conductivity matrix of water-retaining sand at the 10-minute seepage time. The corresponding cell center coordinates, after calibration, are (32... 180 The coordinate error is less than or equal to 0.05. ;

[0107] Fifty sets of global impedance datasets with different water contents and seepage times were selected as the training set. The root mean square error of the voltage residual was used as the loss function, and an Adam optimizer with a learning rate of 0.001 and a momentum of 0.9 was used to train the conductivity distribution inversion model for 200 rounds. The loss function converged to... The fitting accuracy was 99.7%.

[0108] After training, the global conductivity matrix of water-retaining sand at each seepage moment was pseudo-color normalized and encoded to generate a time-series conductivity evolution image sequence of 121 frames, one frame every 30 seconds within 0-60 minutes, which clearly presents the dynamic change process of conductivity and water content inside the water-retaining sand.

[0109] Specifically, the steps for generating a time-series distribution cloud map include:

[0110] Based on the typical moisture content range during the seepage process of water-retaining sand, with a saturated moisture content of 0% to 30%, five characteristic points were selected from this range, including dry state, low moisture content state, critically wetted state, high moisture content state, and near-saturated state. Five standard samples with different moisture contents were prepared, with at least three parallel samples prepared for each standard sample to ensure repeatability. The dry state had a moisture content of 0%, and after 105... Dry to constant weight; the moisture content of the low-moisture state is 5%, which is achieved by adding a measured amount of water and sealing and letting it stand for 24 hours to make the moisture content uniform; the moisture content of the critical wet state is 15%, corresponding to the lower limit of the dry-wet transition zone, i.e., the moisture content threshold. The water content in the high-water-content state is 22%, corresponding to the critical saturation of liquid connection; the water content in the near-saturated state is 28%, close to the maximum water-holding capacity. According to the water-holding test in the water-retaining sand chamber, the maximum water-holding capacity is determined to be 30%.

[0111] Five standard samples with different moisture contents were sequentially placed into an insulated ring container, and low-frequency 10 Hz data were collected. 50 medium frequency High frequency 200 The global impedance dataset is used to output the global conductivity matrix of water-retaining sand for each standard sample through a conductivity distribution inversion model. The average conductivity value corresponding to each water content is obtained by averaging the parallel samples. ,in, Based on measured moisture content The x-axis represents the average conductivity value corresponding to each moisture content. Using the vertical axis as the ordinate, five points are used to calibrate the original data pairs as follows: ), ( ), ( ), ( ), ( The measured moisture content was obtained by drying and weighing.

[0112] A scatter plot was plotted using the original data from the five-point calibration. Cubic spline interpolation was then used to generate a smooth curve that maps electrical conductivity to water content, serving as a baseline mapping curve. , Let the moisture content be the solution. The baseline mapping curve function from electrical conductivity to water content is defined; a monotonically increasing constraint is added to the interpolation process to ensure that the water content increases monotonically with increasing electrical conductivity; at the same time, boundary conditions are set including a slope of zero in the dry state and an asymptotic slope of zero in the near-saturated state, in order to conform to the physical laws of the pore water filling process.

[0113] Because the conductivity distribution within actual water-retaining sand exhibits local heterogeneity, directly using the baseline mapping curve would introduce errors. Therefore, based on the global conductivity matrix of five standard samples, the local variation coefficient for each pixel is calculated. The expression is as follows:

[0114]

[0115] in, The standard deviation of conductivity within the neighborhood of the current pixel. This represents the average conductivity within the neighborhood of the current pixel. It is the set of local conductivity formed by the current pixel and its neighboring pixels;

[0116] For each standard sample, the optimal parameters in the correction coefficient formula are determined using the least squares method. The specific process of the least squares method includes: aiming to minimize the sum of squared residuals between the corrected moisture content and the measured moisture content, fitting the solution based on the moisture content of the standard sample and the measured moisture content to obtain the optimal parameters that minimize the sum of squared residuals. The formula for the correction factor is as follows:

[0117]

[0118] in, This is a local correction factor for moisture content, used to compensate for the influence of microstructure differences on the relationship between conductivity and moisture content; parameters The fitting factor is the correction coefficient. This is the moisture content correction amount introduced by local conductivity non-uniformity;

[0119] The local correction coefficients obtained by combining the correction coefficient formula generate the corrected conductivity-to-water content correction mapping curve. , This is the corrected moisture content. The corrected conductivity-to-moisture content reference mapping curve function minimizes the deviation between the corrected moisture content in the corrected mapping curve and the measured moisture content; wherein, the measured moisture content is obtained by drying and weighing method;

[0120] Validation samples not included in the calibration were selected, such as those with moisture contents of 10%, 18%, and 25%. The overall conductivity matrix of the water-retaining sand for each sample was measured. The moisture content of each sample was calculated using the aforementioned baseline mapping and correction coefficient formula, and compared point-by-point with the measured moisture content. The maximum absolute error was obtained by calculating the absolute difference between the moisture content of each sample and the measured moisture content. Based on the accuracy requirements for moisture content monitoring during the seepage process... Set the maximum error threshold to 3%. If the maximum absolute error of the verification sample is greater than the maximum error threshold, adjust the node position of the cubic spline interpolation or increase the number of calibration points, and recalculate the correction coefficient until the maximum absolute error meets the accuracy requirements.

[0121] Since the seepage process may cause sample deformation, rigid registration based on mutual information is used to align all frames to the first frame. A spatial coordinate system (time) is used to ensure that the same spatial location corresponds to the same physical point at different times; for each pixel of each frame in the time-series conductivity evolution image sequence (time), Read the conductivity value. The local coefficient of variation is calculated using the expression for the local coefficient of variation. Calculate the correction factor according to the correction factor formula. The corrected moisture content was calculated using the corrected mapping curve. ;like If the moisture content exceeds the physical range [0%, 28%], the moisture content will be corrected. The mandatory constraint is the endpoint value of the physical range, i.e. When Set to 0%, When The moisture content is set at 28%; where 0% is the moisture content in the dry state and 28% is the moisture content in the near-saturated state.

[0122] To prevent abrupt noise on the timeline, the time series of each pixel is smoothed by Kalman filtering, and the time series water content value is output. The time series water content image sequence is generated by frame-by-frame reconstruction and spatiotemporal consistency verification to ensure the physical continuity of the evolution process.

[0123] A pseudo-color mapping technique based on moisture content grading was used to visualize the time-series moisture content image sequence. The color mapping rule was set as follows: low moisture content areas are blue, dry-wet transition areas are yellow, and high moisture content areas are red. The color depth corresponds to the moisture content level, with the color becoming darker as the moisture content increases. To enhance the practicality of the cloud map, the corresponding seepage time, maximum, minimum, and average moisture content were marked on each frame of the cloud map. At the same time, spatial coordinate scales were added, which corresponded to the radial and circumferential coordinates of the insulated annular container, generating a time-series distribution cloud map of moisture content for a single frame. The 121 frames of time-series distribution cloud maps of moisture content were integrated in the order of seepage time, and a time-series playback control was added to realize the dynamic playback of moisture content distribution, clearly presenting the dynamic seepage evolution process of moisture content inside the water-retaining sand within 0-60 minutes.

[0124] Specifically, the steps for forming a quantitative feature set of hydrophobic properties include:

[0125] Convert the spatial coordinates of the time-series moisture content image sequence to polar coordinates. ),in, Radial distance, For circumferential angles; at each moment The insulating annular container is divided into sections along the radial direction. The insulating annular container is divided into several equal-width circular bands along the circumferential angle. One equiangular sector;

[0126] For each equal-width circular band With each isoangular sector In the overlapping region, the average moisture content of all pixels within the overlapping region is calculated to obtain the region's average moisture content. ;in, , ;

[0127] For each equal-width circular band Based on the regional average moisture content, through Calculate the circumferential average moisture content. For the first At time, the equal-width circular bands The circumferential average moisture content, by measuring the radial direction The circumferential average moisture content was calculated sequentially for each of the three equal-width circular annular zones. A curve showing the radial average moisture content as a function of radial distance and time was plotted with radial distance as the abscissa and circumferential average moisture content as the ordinate. ;in, For the first Within a ring of equal width, The average water content of each equiangular sector The arithmetic mean of the values ​​is used to eliminate the influence of circumferential local fluctuations on radial moisture content distribution;

[0128] Using the critical wet state moisture content of 15% as the discrimination threshold, the definition of each time step is... moist front position To meet The minimum radial distance, i.e. ;like Then set ;

[0129] Extracting each moment frame by frame Corresponding moist front position The sequence of wet front locations was obtained. , For the first At each sampling time step, a 5-frame sliding window mid-range filter is applied to the wet front position sequence to obtain a smoothed wet front position sequence, suppressing local noise and anomalous jumps. The instantaneous velocity of the wet front is calculated using the central difference method to obtain the water infiltration rate. The time series of water seepage rates was obtained by point-by-point differencing the smoothed sequence of wetting front locations; where... This represents the difference in position between the moist front at the current moment and the previous moment. This represents the time difference between the current moment and the migration moment.

[0130] Select the middle and late stages of seepage development where it is relatively stable ,in, The time (e.g., 30 minutes) corresponding to the first time the wetting front exceeds 30% of the radius of the insulating annular container. The total observation duration (e.g., 60 minutes) is based on the current position of the moist front. As the boundary, take This is a radial moisture content profile; for each time point... The radial moisture content profile is fitted to a simplified form of the Richards equation, as shown below:

[0131]

[0132] in, The water content is close to saturation. The moisture content is in the dry state. Let be the length of the radially extended feature to be fitted. Radial distance With radially extended characteristic length The ratio represents the normalized result of radial distance relative to diffusion range; The normalized distribution term of moisture content conforms to the physical law that the moisture content is high near the center and gradually tends to saturate with increasing radial distance during water infiltration.

[0133] Each time step is obtained by nonlinear least squares fitting. Radial expansion feature length to be fitted ,use Calculate the radial diffusion coefficient The radial diffusion coefficient reflects the ability of water to diffuse in the radial direction; among which, Radial extension feature length The ratio of the square of the value to the current time is used to convert the radial expansion feature length into a time-dependent diffusion coefficient, quantifying the rate and extent of moisture diffusion.

[0134] At every moment For each equal-width circular band According to each equal-width circular band Average regional moisture content of each sector Calculate the regional average moisture content of each sector. The standard deviation and arithmetic mean are used to calculate the coefficient of variation of the circumferential moisture content. Dividing the standard deviation by the arithmetic mean yields the coefficient of variation. Circumferential segregation Defined as the coefficient of variation over all equal-width circular bands. Calculate the maximum value of the coefficient of variation on all equal-width circular annular bands. The weighted average is used to obtain the global circumferential segregation. The weight is the area of ​​each equally wide circular ring; the calculation expression is as follows:

[0135]

[0136] in, For the first Area of ​​a circular ring of equal width for The sum of the products of the areas of the equal-width circular bands and their corresponding coefficients of variation is used to characterize the contribution weight of different radial positions to global uniformity. for The sum of the areas of the equal-width circular rings; the larger the circumferential segregation value, the more uneven the distribution of water in the circumferential direction, and the worse the hydrophobic uniformity.

[0137] The theoretical seepage rate was determined based on a hydrophilic control sample (without hydrophobic modification) that has the same particle size and porosity as the water-retaining sand to be tested, under the same seepage driving force, using the measured seepage rate. The reference diffusion coefficient is determined based on the average radial diffusion coefficient of the hydrophilic control sample during the same seepage period. Combined with water infiltration rate With radial diffusion coefficient Calculate the initial hydrophobic retention factor The expression is as follows:

[0138]

[0139] in, It is a tiny constant with a value of 0.01; The difference between the theoretical seepage rate and the water seepage rate characterizes the amount of inhibition of seepage by hydrophobic interaction. This is the theoretical seepage rate after incorporating a small constant, used to avoid calculation anomalies; The normalized seepage resistance rate characterizes the degree to which hydrophobic interactions inhibit the seepage rate. The ratio of the radial diffusion coefficient to the reference diffusion coefficient represents the degree to which hydrophobic interactions inhibit water diffusion. This is a diffusion inhibition correction term used to balance the contribution weights of seepage inhibition and diffusion inhibition.

[0140] To eliminate initial stage fluctuations, the seepage stabilization stage is selected ( Initial hydrophobic blocking factor (minutes) The arithmetic mean is the final hydrophobic blocking factor. ;

[0141] All time-series indicators are aligned along the time axis, and statistical features of water content, including seepage rate features, diffusion capacity features, uniformity features, and impedimentation effect features, are extracted from the time-series distribution cloud map along the radial and circumferential directions to form a quantitative feature set of hydrophobic performance; among which, seepage rate features include but are not limited to water seepage rate. Diffusion capability characteristics include, but are not limited to, radial diffusion coefficient. Uniformity characteristics include, but are not limited to, global circumferential segregation. The blocking effect characteristics include, but are not limited to, hydrophobic blocking factors. .

[0142] Specifically, the steps for generating a hydrophobic performance test report include:

[0143] Extracting hydrophobic blocking factors from the set of hydrophobic properties quantitative features Global circumferential segregation As a dual-core judgment indicator; among them, the hydrophobic barrier factor Characterization effectiveness, global circumferential segregation Characterizes uniformity;

[0144] The judgment thresholds are set differently according to the requirements of engineering applications, including the first judgment threshold. , , Second judgment threshold , , The water-retaining sand hydrophobic performance is graded and judged. The first and second judgment thresholds work together to take into account both the hydrophobic effectiveness and uniformity, so as to achieve accurate grading of the water-retaining sand hydrophobic performance. This meets the core requirements of water-retaining sand hydrophobic effect in different scenarios in engineering practice and avoids evaluation bias caused by a single index.

[0145] like and If it is, it is judged to be a Class 1 hydrophobic property; if and If it is, it is determined to be a level 2 hydrophobic property; if and If the hydrophobicity is positive, it is determined to be level three hydrophobicity; otherwise, it is determined to be level four hydrophobicity. The hydrophobicity increases progressively from level four to level one.

[0146] Each level of hydrophobic performance corresponds to a specific engineering application scenario, performance standard, and treatment requirement, including: Level 1 hydrophobic performance is the best, with the strongest overall hydrophobic effectiveness and uniform distribution, and no risk of local failure. It is suitable for high-requirement seepage prevention and water retention scenarios, such as precision hydraulic engineering. Level 2 hydrophobic performance is good, with slightly lower overall hydrophobic effectiveness and uniform distribution. It is suitable for conventional seepage prevention scenarios, such as seepage prevention in small-scale hydraulic facilities. Level 3 hydrophobic performance is qualified, possessing basic hydrophobic functions, but with substandard uniformity and local segregation. It is suitable for low-requirement water retention scenarios, such as temporary water retention. Level 4 hydrophobic performance is substandard, with insufficient hydrophobic effectiveness or uneven distribution, posing a risk of large-area hydrophobic failure. Rework is required, such as recoating the hydrophobic coating or adjusting the water-retaining sand ratio to ensure that the engineering application requirements are met.

[0147] Based on the classification results of the water-retaining and hydrophobic properties of sand, combined with the hydrophobic barrier factor Global circumferential segregation Perform abnormal region identification; if and If the area is larger than the surrounding area, it indicates that the corresponding area has completely lost its hydrophobic function and is judged as a local hydrophobic failure zone; if but This indicates that the corresponding area has insufficient hydrophobicity and is prone to water accumulation, thus being identified as a water accumulation zone; if the corresponding area has a hydrophobic barrier factor... Global circumferential segregation All three frames exhibited drastic fluctuations, indicating unstable seepage and a risk of sudden changes in hydrophobic properties, thus classifying them as areas of abnormal seepage. In this embodiment, the seepage pattern was defined as occurring within three consecutive frames. Fluctuation greater than or equal to 40% and A fluctuation of 30% or more is considered a violent fluctuation.

[0148] Adding hydrophobicity levels for different regions to the time-series distribution cloud map clearly reflects the dynamic evolution of hydrophobicity; based on global circumferential segregation... pass A hydrophobicity uniformity index was obtained to quantitatively characterize the overall hydrophobicity uniformity, with a value ranging from 0 to 1; among which, For the global circumferential segregation and the second judgment threshold The ratio is used to normalize the global circumferential segregation, eliminate the influence of dimensions, and make the hydrophobicity uniformity index in the range of 0 to 1, which is convenient for intuitive quantitative evaluation.

[0149] The grading results of water-retaining sand hydrophobic performance and the results of abnormal area identification are integrated to form structured report data. A visual template is used to output a hydrophobic performance test report. The hydrophobic performance test report includes, but is not limited to, time-series distribution cloud map, hydrophobic uniformity index, and hydrophobic performance level.

[0150] Example 2

[0151] Please see Figure 5 Another embodiment of the present invention provides: a water-retaining sand hydrophobic performance detection system based on impedance tomography, comprising: a global impedance acquisition module, an inversion imaging module, a hydrophobic performance quantification module, and a performance grading determination module;

[0152] The global impedance acquisition module is used to uniformly fill the water-retaining sand sample into an insulating ring container and compact it to dry density; after the water-retaining sand sample is left to stand for 24 hours, deionized water is injected into the center of the sample at a constant flow rate using a micro-injection pump to simulate the dynamic seepage process of water; a set of boundary voltage data is collected by cyclic scanning in adjacent excitation mode, and the global impedance dataset is generated after preprocessing.

[0153] The inversion imaging module, based on the global impedance dataset, constructs an inversion model of conductivity distribution. Using the boundary voltage amplitude and phase under multi-frequency excitation as observation constraints, it uses the conjugate gradient method to solve the internal conductivity distribution, outputs the global conductivity matrix of water-retaining sand, and maps it one-to-one with spatial coordinates. Through pseudo-color normalization encoding, it generates a time-series conductivity evolution image sequence.

[0154] The hydrophobic performance quantification module, based on the time-series conductivity evolution image sequence, uses a five-point calibration method to establish the calibration mapping relationship between conductivity and water content. It converts the time-series conductivity evolution image sequence pixel by pixel into water content, generates a time-series distribution cloud map, extracts statistical features of water content, and calculates water seepage rate, radial diffusion coefficient, global circumferential segregation, and hydrophobic resistance factor to form a set of hydrophobic performance quantification features.

[0155] The performance grading and judgment module, based on the hydrophobic performance quantitative feature set, sets different judgment thresholds to grade and judge the hydrophobic performance of water-retaining sand, identifies abnormal areas based on global circumferential segregation and hydrophobic barrier factor, and generates a hydrophobic performance test report.

[0156] Working principle and effects:

[0157] Water-retaining sand samples were uniformly filled into an insulated annular container and compacted to a set dry density, then left to stand for 24 hours to allow internal stress equilibrium. Deionized water was injected into the center of the sample at a constant flow rate using a micro-injection pump to simulate the dynamic seepage process of water-retaining sand under actual working conditions. The annular electrode array was cyclically scanned using an adjacent excitation mode to collect boundary voltage data under multi-frequency excitation. After preprocessing, a global impedance dataset containing amplitude and phase information was generated. This process standardized the entire process of water-retaining sand sample preparation, seepage simulation, and global impedance data acquisition, eliminating interference from uneven sample filling, seepage disturbances, and measurement noise. This provided high-quality observational data for conductivity inversion and ensured the basic reliability of subsequent hydrophobic performance analysis.

[0158] Based on a global impedance dataset, an objective function for conductivity distribution inversion is constructed. The amplitude and phase of the boundary voltage under multi-frequency excitation are used as observation constraints, and a first-order smoothing regularization term is introduced to suppress ill-conditioned inversion. The conjugate gradient method is used to iteratively solve for the minimum value of the conductivity distribution inversion objective function, updating the internal conductivity distribution and outputting the global conductivity matrix of water-retaining sand. The global conductivity matrix of water-retaining sand is mapped one-to-one with the spatial coordinates of the insulated annular container. Through pseudo-color normalization encoding, a time-series conductivity evolution image sequence reflecting the internal conductivity distribution at different times is generated. This clearly presents the spatiotemporal evolution law of conductivity during water seepage, providing core data support for water content conversion and hydrophobic performance analysis, and solving the problem that traditional conductivity detection cannot achieve global, dynamic visualization.

[0159] Based on the time-series conductivity evolution image sequence, a nonlinear mapping relationship between conductivity and water content was established using a five-point calibration method. The time-series conductivity image was converted pixel by pixel into water content, generating a time-series distribution cloud map of water content within the water-retaining sand. Statistical features of water content were extracted along the radial and circumferential directions. The position of the wetting front was calculated to obtain the water seepage rate. The radial diffusion coefficient was obtained by fitting the radial water content profile. The global circumferential segregation was obtained by statistically analyzing the circumferential water content variation. Combined with seepage and diffusion features, a hydrophobic barrier factor was constructed, ultimately forming a quantitative feature set of hydrophobic performance. The time-series conductivity evolution image sequence was transformed into a quantifiable hydrophobic performance index, realizing a leap from "image observation" to "numerical quantification." This comprehensively characterizes the dynamic seepage process, radial diffusion capacity, circumferential hydrophobic uniformity, and hydrophobic barrier effectiveness of water-retaining sand, providing a multi-dimensional quantitative basis for subsequent classification and determination.

[0160] Based on a quantitative feature set of hydrophobic performance, and combined with differentiated thresholds for hydrophobic barrier factors and global circumferential segregation according to engineering application needs, the hydrophobic performance of water-retaining sand is classified into four levels, and matched with corresponding engineering scenarios and treatment requirements. According to the joint distribution of global circumferential segregation and hydrophobic barrier factors, local hydrophobic failure areas, water accumulation areas, and seepage anomaly areas are non-destructively identified, and their spatial location and temporal characteristics are marked. The classification results, anomaly identification information, temporal distribution cloud map, and quantitative indicators are integrated to form a hydrophobic performance test report. This achieves integrated non-destructive characterization of internal water distribution, dynamic seepage, and hydrophobic uniformity of water-retaining sand, and realizes automatic intelligent evaluation of the hydrophobic performance of water-retaining sand. It provides an efficient and reliable technical basis for the engineering application and graded maintenance of water-retaining sand, and improves the efficiency of resource allocation and the scientific nature of maintenance decisions.

[0161] Overall, a complete intelligent detection and evaluation system for the hydrophobic properties of water-retaining sand has been formed through a four-layer modular architecture consisting of full-domain impedance acquisition, conductivity inversion imaging, hydrophobicity quantification, and graded judgment. This system deeply integrates sample preparation, data acquisition, inversion imaging, quantitative analysis, and engineering evaluation, achieving high-precision, non-destructive, and dynamic characterization of internal water seepage and hydrophobic properties in water-retaining sand. It effectively solves the technical bottlenecks of traditional detection methods, such as the inability to achieve full-domain visualization, single quantitative dimensions, and lagging detection and evaluation, providing reliable technical support for the research, optimization, and engineering application of water-retaining sand materials.

[0162] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for detecting the hydrophobic properties of water-retaining sand based on impedance tomography, characterized in that, include: The water-retaining sand sample was filled into an insulated ring container and compacted. Liquid was injected into the center of the sample to simulate the dynamic seepage process of water. Boundary voltage data is collected by cyclic scanning in adjacent excitation modes, and preprocessed to generate a global impedance dataset. Based on the global impedance dataset, a conductivity distribution inversion model is constructed. The boundary voltage amplitude and phase under multi-frequency excitation are used as observation constraints to solve the internal conductivity distribution. The global conductivity matrix of water-retaining sand is output and mapped with spatial coordinates to generate a time-series conductivity evolution image sequence. Establish a calibration mapping relationship between electrical conductivity and water content, convert the time-series electrical conductivity evolution image sequence pixel by pixel into water content, generate a time-series distribution cloud map, extract statistical features of water content, calculate water seepage rate, radial diffusion coefficient, global circumferential segregation and hydrophobic hindrance factor, and form a set of quantitative features of hydrophobic performance. Based on the quantitative feature set of hydrophobic performance, a judgment threshold is set to classify and judge the hydrophobic performance of water-retaining sand. Abnormal areas are identified according to the global circumferential segregation and hydrophobic barrier factor, and a hydrophobic performance test report is generated.

2. The method for detecting the hydrophobic properties of water-retaining sand based on impedance tomography according to claim 1, characterized in that, The specific steps for generating a time-series conductivity evolution image sequence include: Based on the global impedance dataset, an inversion model of conductivity distribution is constructed according to the dielectric properties of water-retaining sand porous media. Using multiple sets of global impedance datasets with different water contents and different seepage times as the training set, and the root mean square error of voltage residuals as the loss function, the conductivity distribution inversion model is trained by combining the Adam optimizer. After training, the real-time collected global impedance dataset is input, and the global conductivity matrix of water-retaining sand at different seepage times is output through input layer parsing, impedance feature fusion layer multi-frequency feature weighted fusion, and inversion optimization layer iterative inversion optimization. By using pseudo-color normalized encoding, the global conductivity matrix of water-retaining sand at different seepage times is visualized, generating a time-series conductivity evolution image sequence.

3. The method for detecting the hydrophobic properties of water-retaining sand based on impedance tomography according to claim 2, characterized in that, The specific steps for constructing an inversion model of conductivity distribution include: A conductivity distribution inversion model is constructed, including an input layer, an impedance feature fusion layer, an inversion optimization layer, and an output layer; The input layer receives and parses the global impedance dataset; Impedance characteristic fusion layer: The water content of water-retaining sand is calculated based on the calibration formula of impedance amplitude and water content under different frequency excitations. ; Set moisture content threshold , ; like If so, the corresponding area is designated as a high moisture content area; if Then the corresponding area is divided into a dry-wet transition zone; if If so, the corresponding area is designated as a low moisture content area; Extract the amplitude and phase characteristics of multi-channel complex impedance at low, medium, and high frequencies at a single moment from the global impedance dataset; By setting low-frequency, mid-frequency, and high-frequency weights using a weighted fusion algorithm, a multi-frequency fusion feature vector is generated.

4. The method for detecting the hydrophobic properties of water-retaining sand based on impedance tomography according to claim 3, characterized in that, The specific steps for constructing an inversion model of conductivity distribution also include: The inversion optimization layer uses the multi-frequency fusion feature vector as the observation constraint to construct the conductivity inversion objective function. The conductivity distribution to be solved is discretized into multiple units and substituted into the electric field control equation. The finite difference method is used for discretization, and the theoretical response value of the boundary voltage of each unit is obtained by iterative solution. The difference between the theoretical response value and the measured boundary voltage value is calculated. After systematically collecting and organizing the boundary voltage measurement channels point by point, a voltage residual sequence is generated. With the goal of minimizing the voltage residual sequence, the conjugate gradient method is used to solve for the minimum value of the conductivity inversion objective function. In each iteration, the gradient vector and the conjugate gradient direction are solved to update and correct the internal conductivity distribution, and the optimal conductivity distribution result in the global domain is output. The output layer outputs the global conductivity matrix of water-retaining sand at the corresponding seepage time based on the optimal global conductivity distribution results, and maps the global conductivity matrix of water-retaining sand to spatial coordinates one by one.

5. The method for detecting the hydrophobic properties of water-retaining sand based on impedance tomography according to claim 4, characterized in that, The specific steps for generating a time series distribution cloud map include: Five characteristic points were selected from typical moisture content ranges: dry state, low moisture content state, critically wet state, high moisture content state, and near-saturation state. Prepare standard samples with corresponding moisture content, and prepare at least 3 parallel samples for each standard sample; Standard samples were sequentially placed into an insulating ring container, and global impedance datasets at low, medium, and high frequencies were collected. The global conductivity matrix of water-retaining sand for each standard sample was output through the conductivity distribution inversion model. The average conductivity value corresponding to each moisture content was obtained by averaging the parallel samples. Using the measured moisture content as the abscissa and the average electrical conductivity as the ordinate, a five-point calibration raw data pair was formed. A scatter plot was drawn and a baseline mapping curve was generated using cubic spline interpolation. The local coefficient of variation was calculated based on the global conductivity matrix of water-retaining sand for each standard sample. The correction coefficient fitting factor is determined by the least squares method, and the local correction coefficient obtained by the correction coefficient formula is used to generate the correction mapping curve, so as to minimize the deviation between the corrected moisture content and the measured moisture content.

6. The method for detecting the hydrophobic properties of water-retaining sand based on impedance tomography according to claim 5, characterized in that, The specific steps for generating time series distribution cloud maps also include: Select validation samples that were not calibrated, measure the water-retaining sand global conductivity matrix of each validation sample, calculate the moisture content of each validation sample, and calculate the maximum absolute error by combining the measured moisture content; Set the maximum error threshold; If the maximum absolute error is greater than the maximum error threshold, adjust the node positions of the cubic spline interpolation or increase the number of calibration points, and recalculate the correction coefficients until the maximum absolute error meets the standard. Align all frames to the spatial coordinate system of the first frame, read the conductivity value of each pixel in the time-series conductivity evolution image sequence, calculate the local coefficient of variation and correction coefficient, and use the correction mapping curve to calculate the corrected moisture content. The time series of each pixel is smoothed by Kalman filtering, and the time series water content value is output to generate a time series water content image sequence. A time-series moisture content image sequence was visualized using a moisture content grading pseudo-color mapping technique to generate a time-series distribution cloud map.

7. The method for detecting the hydrophobic properties of water-retaining sand based on impedance tomography according to claim 6, characterized in that, The specific steps for forming a quantitative feature set of hydrophobic properties include: The time-series moisture content image sequence was converted into polar coordinates; the insulating annular container was divided into equal-width circular bands along the radial direction and equal-angle sectors along the circumference. The regional average moisture content of the overlapping areas of each equal-width circular zone and each equal-angle sector is statistically analyzed, and the radial average moisture content curve is obtained by calculating the circumferential average moisture content. The location of the wetting front is extracted frame by frame, and the water seepage rate is calculated after filtering and smoothing. The water seepage rate time series is obtained by point-by-point difference. A radial water content profile was fitted during the relatively stable middle and late stages of seepage to determine the radial extension characteristic length and calculate the radial diffusion coefficient. Calculate the coefficient of variation of circumferential moisture content based on the regional average moisture content, define circumferential segregation and calculate global circumferential segregation; Determine the theoretical seepage rate and reference diffusion coefficient, calculate the initial hydrophobic retardation factor by combining the water seepage rate and radial diffusion coefficient, and take the arithmetic mean of the seepage steady stage as the hydrophobic retardation factor. All time-series indicators are aligned along the time axis, and statistical features of water content are extracted from the time-series distribution cloud map to form a set of quantitative features for hydrophobic performance.

8. The method for detecting the hydrophobic properties of water-retaining sand based on impedance tomography according to claim 7, characterized in that, The specific steps for determining the classification include: Extract the hydrophobic barrier factor and global circumferential segregation from the set of quantitative features of hydrophobic properties; Set judgment thresholds, including the first judgment threshold. , , Second judgment threshold , , The water-retaining and water-draining properties of sand are classified and determined. If hydrophobic blocking factor And global circumferential segregation If it is, it is judged to be a Class 1 hydrophobic property; if And global circumferential segregation If so, it is determined to be a Class II hydrophobic property; like And global circumferential segregation If it is positive, it is determined to be a Class III hydrophobic property; otherwise, it is determined to be a Class IV hydrophobic property.

9. The method for detecting the hydrophobic properties of water-retaining sand based on impedance tomography according to claim 8, characterized in that, The specific steps for generating a hydrophobic performance test report include: Based on the hierarchical judgment results, combined with the hydrophobic barrier factor and global circumferential segregation, abnormal regions are identified. If hydrophobic blocking factor If the global circumferential segregation is greater than that of the surrounding area, it is determined to be a local hydrophobic failure zone; like But global circumferential segregation If so, it is determined to be a water accumulation area; If the hydrophobic barrier factor and the global circumferential segregation of the corresponding area both fluctuate drastically, it is determined to be an area with abnormal seepage. Add hydrophobicity ratings for different regions to the time-series distribution cloud map; Based on global circumferential parseness combined with a second judgment threshold Calculate the hydrophobicity uniformity index; The results of the graded judgment and the results of the abnormal area identification are integrated to form structured report data, and a hydrophobic performance test report is generated using a visual template.

10. A system for detecting the hydrophobic properties of water-retaining sand based on impedance tomography, used to implement the method for detecting the hydrophobic properties of water-retaining sand based on impedance tomography as described in any one of claims 1-9, characterized in that, include: Global impedance acquisition module, inversion imaging module, hydrophobicity quantification module, and performance grading determination module; The global impedance acquisition module is used to fill and compact the water-retaining sand sample into an insulating annular container, inject liquid at the center of the sample to simulate the dynamic seepage process of water; acquire boundary voltage data, and preprocess to generate a global impedance dataset. The inversion imaging module is used to construct a conductivity distribution inversion model. Using the boundary voltage amplitude and phase under multi-frequency excitation as observation constraints, it solves the internal conductivity distribution, outputs the water-retaining sand global conductivity matrix and maps it with spatial coordinates to generate a time-series conductivity evolution image sequence. The hydrophobic performance quantification module is used to establish a calibration mapping relationship between electrical conductivity and water content, convert the time-series electrical conductivity evolution image sequence pixel by pixel into water content, generate a time-series distribution cloud map, extract statistical features of water content, calculate water seepage rate, radial diffusion coefficient, global circumferential segregation and hydrophobic hindrance factor, and form a set of hydrophobic performance quantification features. The performance grading and judgment module, based on the hydrophobic performance quantitative feature set, sets a judgment threshold to grade and judge the hydrophobic performance of water-retaining sand, identifies abnormal areas based on global circumferential segregation and hydrophobic barrier factor, and generates a hydrophobic performance test report.

Images