Method for judging seepage failure of immersed roadbed under action of repeated seepage

Abstract

The invention relates to the technical field of electric digital data processing, and particularly provides a method for judging seepage failure of an immersed roadbed under the action of repeated seepage, which comprises the following steps of: constructing a water-force-chemical coupling test mechanism, monitoring dynamic change of a permeability coefficient, a fine particle loss rate and three-dimensional reconstruction of a pore structure in real time through a sensor, and determining seepage failure of the immersed roadbed under the action of repeated seepage. A scanning technology is combined to quantify an evolution rule of a seepage path; key parameters such as seepage path fractal dimension, fine particle mobility and dynamic porosity change are extracted on the basis of an evolution rule, the weight of each index is determined by introducing fuzzy analytic hierarchy process, and a seepage failure comprehensive index is constructed; and establishing a time sequence prediction model, inputting a seepage failure comprehensive index and an environment variable, outputting a seepage failure probability, and dividing safety-critical-danger three-level early warning. According to the method, the limitation of a traditional static evaluation method is broken through, and the whole-process monitoring and prediction of seepage failure under the dynamic action of repeated seepage are realized; and the discrimination precision is obviously improved.

Description

Technical Field

[0001] This invention relates to the field of electronic digital data processing technology, and in particular to a method for identifying seepage damage in roadbeds subjected to repeated seepage. Background Technology

[0002] With the rapid development of my country's transportation infrastructure, more and more highways and railways need to traverse areas with complex hydrogeological conditions. In these areas, the roadbed is often submerged or periodically submerged, and repeated seepage can lead to gradual seepage damage to the roadbed fill material, seriously affecting the stability and durability of the roadbed. In actual engineering, we often see scenarios where, during the rainy season, the roadbed fill material gradually erodes under repeated wet-dry cycles and seepage, eventually leading to road surface collapse or slope slippage. This phenomenon is particularly common in rainy areas in the south and seasonal freeze-thaw regions in the north. Traditionally, engineers often wait until obvious signs of roadbed damage appear before taking remedial measures, which is not only costly but may also pose a threat to traffic safety. Therefore, developing a technology that can predict the risk of seepage damage to submerged roadbeds under repeated seepage is of great significance for ensuring the long-term stable operation of transportation infrastructure.

[0003] Prior art 1, Chinese patent application number 202310177906.X, discloses a method for judging seepage failure in foundations with deep and complex overburden layers. It proposes a method for calculating the allowable slope of soil by comprehensively considering multiple factors such as reverse filtration and weighting measures, burial depth, and interlayer protection. The method systematically quantifies the contribution of these influencing factors to the allowable slope value of the soil, and scientifically and accurately evaluates the seepage stability of overburden foundations. An increasing number of cases involving earth-rock dams and sluice gate dams located on deep and complex overburden layers show that the seepage failure analysis results obtained from conventional seepage safety evaluation methods significantly overestimate the possibility of seepage failure, especially in weak points of the overburden foundation, such as the downstream seepage outlet of the dam and the bottom of the cutoff wall. While it has made up for the shortcomings of conventional methods and greatly improved the accuracy and reliability of the judgment results, providing a new approach for the scientific evaluation of the seepage safety of deep and complex overburden foundations, and also providing new solutions for the design of dam seepage damage control and overburden seepage barrier wall design, it has also made improvements to the calculation of the allowable slope of the soil by quantifying static factors such as filter weight, burial depth and interlayer protection. However, it has insufficient adaptability to dynamic environments and does not consider the long-term impact of dynamic conditions such as wet-dry cycles, water level fluctuations and chemical erosion on seepage damage. The monitoring methods are also limited, relying on static test data and lacking the ability to monitor the evolution of seepage paths and the migration of fine particles in real time, which leads to the inability to dynamically correct the overestimation of risks at weak points (such as the downstream seepage outlet of the dam).

[0004] Prior art two, Chinese patent application number 202410590692.3, discloses a method and system for judging seepage failure of heterogeneous double-layer dam foundation. The method includes: acquiring double-layer dam foundation material data, double-layer dam foundation size data, and upstream and downstream water level data of the double-layer dam project; generating a permeability coefficient random field and the actual flow velocity Vactual of the heterogeneous double-layer dam based on the acquired data; obtaining the particle size distribution of the foundation soil particles based on the relationship between the permeability coefficient of the permeability coefficient random field and the effective particle size of the soil particles; obtaining the critical flow velocity Vcritical for soil particles to move based on the force balance of the soil particles; and evaluating the seepage stability of the heterogeneous double-layer dam foundation by comparing the actual flow velocity Vactual with the critical flow velocity Vcritical. Although it can realistically simulate and judge the damage of heterogeneous double-layer dam foundations under seepage, the parameters are singular and complex parameters such as the fractal characteristics of the seepage path and the migration rate of fine particles are not quantified, making it difficult to accurately reflect the permeability stability of heterogeneous materials; dynamic synergistic analysis is lacking: the lack of comprehensive assessment of the combined effects of multi-layer foundations or environmental variables (such as temperature and humidity) limits the applicability of the model.

[0005] Prior art three, Chinese patent application number 202411346017.2, discloses a method for predicting the depth distribution of permeability coefficient of quarry waste considering particle breakage effects, belonging to the field of geotechnical performance prediction technology. Dynamic triaxial tests are conducted on quarry waste samples to explore the dynamic evolution of particle gradation and breakage rate at different depths after different vibration loading cycles. Simultaneously, permeability tests are conducted on quarry waste broken by cyclic loading using a seepage erosion simulation test device to explore the critical hydraulic gradient of fine particle transport and structural damage within quarry waste of different gradations under erosion, revealing its deterioration and damage mechanisms. Furthermore, using the negative exponential continuous gradation equation as a pivot, a predictive model for particle breakage index and permeability coefficient is constructed, revealing the intrinsic mechanism of particle breakage behavior and permeability evolution of quarry waste subgrade under cyclic loading. Although this makes the predicted permeability coefficient closer to the actual permeability coefficient and improves the accuracy of the prediction, the lack of multi-field coupling and failure to consider the synergistic effect of water-mechanical-chemical multi-field coupling on particle breakage and permeability characteristics leads to a large deviation between the prediction model and the actual working conditions. The static prediction model, based on the negative exponential gradation equation, lacks the ability to dynamically correct the permeability coefficient over time and cannot adapt to environmental changes (such as structural deterioration after cyclic loading).

[0006] Current technologies 1, 2, and 3 lack expertise in dynamic environmental adaptability, multi-factor coupling analysis, and risk assessment accuracy. This invention provides a complete solution for identifying seepage failure in subgrades subjected to repeated seepage, significantly improving the reliability of safety assessments in complex engineering scenarios. Therefore, this invention provides a method for identifying seepage failure in subgrades subjected to repeated seepage. Summary of the Invention

[0007] To achieve the above objectives, the present invention adopts the following technical solution: One aspect of the present invention provides a method for identifying seepage failure of a subgrade subjected to repeated seepage, comprising the following steps: A water-mechanical-chemical coupled experimental mechanism was constructed to simulate complex working conditions such as wet-dry cycles, water level fluctuations, and chemical erosion in real engineering projects. Sensors were used to monitor the dynamic changes in permeability coefficient, fine particle loss rate, and three-dimensional reconstruction of pore structure in real time. Combined with scanning technology, the evolution law of seepage path was quantified. Based on the evolution law, key parameters such as the fractal dimension of the seepage path, the migration rate of fine particles, and the change of dynamic porosity were extracted, and fuzzy hierarchical analysis was introduced to determine the weight of each indicator to construct a comprehensive index of seepage damage. Establish a time-series prediction model, input the comprehensive index of infiltration damage and environmental variables, output the probability of infiltration damage, and classify the early warning into three levels: safe, critical, and dangerous.

[0008] In one optional implementation, the process of quantifying the evolution of seepage paths includes the following steps: The sensor array captures the real-time pulsating characteristics of the dynamic changes in the permeability coefficient in the seepage field. The sampling frequency of the sensor needs to match the seepage response speed of the subgrade soil. The fine particle loss rate is detected by photoelectric coupling. By fusing data from the turbidity sensor and the mass loss sensor, a dynamic correlation model between the fine particle loss rate and the hydraulic gradient is established. The 3D reconstruction of the pore structure adopts a variable resolution dynamic scanning strategy. During the water level fluctuation stage, a high-resolution mode is activated to capture the transient deformation of the pore throat. After the dynamic scanning data is registered with point cloud, the connectivity parameters of the seepage path are extracted through spatial topology to form a time series pore structure evolution matrix. The pore structure evolution matrix obtained by scanning is decomposed into several seepage channel clusters. The tortuosity fractal dimension and the coefficient of variation of the channel cross-sectional area of ​​each cluster are calculated. A dynamic weight evaluation program is established in combination with the real-time permeability coefficient. The fine particle loss rate is used to correct the channel surface roughness parameters. The output is a three-dimensional seepage path feature tensor containing path tortuosity, effective flow area and particle deposition coefficient.

[0009] In one optional implementation, the process of establishing a dynamic correlation model between the fine particle loss rate and the hydraulic gradient includes the following steps: Based on the pulsation characteristics, a time-aligned permeability coefficient-hydraulic gradient dataset is formed. The turbidity sensor and the mass loss sensor synchronously monitor the changes in particle concentration and total loss of the seepage liquid. After photoelectric coupling calibration, a continuous time-series signal of the fine particle loss rate is output. Multi-scale correlation analysis was performed on the time series data of hydraulic gradient and the fine particle loss rate. By aligning the non-steady-state fluctuation characteristics through dynamic time warping, the hysteresis response relationship between the hydraulic gradient abrupt change point and the particle loss rate was extracted, and a nonlinear mapping relationship was constructed to form a dynamically updated dynamic correlation model. The weight coefficients of the dynamic correlation model are adjusted by feedback from the dynamic changes in surface roughness parameters; when the pore structure evolution matrix of the seepage path changes, the recalibration mechanism of the dynamic correlation model is triggered; the generated dynamic correlation model is calculated using the particle deposition coefficient in the three-dimensional seepage path feature tensor.

[0010] In one alternative implementation, the process of constructing a nonlinear mapping relationship includes the following steps: Based on the continuous time series signal of fine particle loss rate calibrated by photoelectric coupling and the time series data of hydraulic gradient, the steady-state and non-steady-state components in the signal are separated by multi-scale decomposition, and the non-steady-state fluctuation characteristics are analyzed. Dynamic time warping is then used to perform phase matching on the time series of the two. Using the aligned permeability coefficient-hydraulic gradient dataset, we identify the abrupt changes in the hydraulic gradient pulse and the corresponding change patterns of the fine particle loss rate signal; through time-delay cross-correlation analysis, we quantify the lag time distribution of the two and establish statistical correlation rules between abrupt events and loss rate response amplitude and response time. The hysteresis response relationship is embedded in the kernel function space, and the dynamic mapping between hydraulic gradient change and particle loss rate is fitted by nonlinear regression. An incremental learning strategy is adopted, and the statistical association rules of the extracted mutation points are used as prior knowledge to constrain the model parameters for initialization. The input layer of the dynamic association model is directly coupled with the permeability coefficient-hydraulic gradient dataset, and the output layer is associated with the time series signal of loss rate.

[0011] In one optional implementation, the process of forming a time-series pore structure evolution matrix includes the following steps: Based on the set variable resolution scanning strategy, a high resolution mode is activated during the water level fluctuation stage to obtain the transient geometric features of the pore throat; after the scanning data is spatiotemporally registered, a three-dimensional pore point cloud sequence at consecutive time points is generated, and the point cloud data at each time point includes the pore wall coordinate set and local curvature distribution. Spatial connectivity analysis was performed on the registered 3D pore point cloud sequence to extract the main path branches and secondary channels in the pore network; the time-varying characteristics of the path geometric parameters were quantified by calculating the relative deformation of the pore throat cross-sectional area between adjacent time points, thus forming the initial time series of pore structure parameters. The connectivity parameters of the seepage path are coupled with the real-time permeability coefficient to establish a correlation mapping between pore structure parameters and permeability performance. The surface roughness is dynamically corrected by introducing the fine particle loss rate, and a three-dimensional seepage path feature tensor containing path tortuosity, effective flow area and particle deposition coefficient is generated. The three-dimensional seepage path feature tensor is iteratively updated according to the time step, which constitutes the core data structure of the pore structure evolution matrix.

[0012] In one optional implementation, the process of calculating the relative deformation of the pore throat cross-sectional area between adjacent time points includes the following steps: Based on the generated 3D pore point cloud sequence, the contour point set of all throat cross sections in the pore network at each time point is extracted; the discrete point set is transformed into a closed plane geometry by least squares fitting, and the initial reference cross-sectional area of ​​each throat is calculated. Using the spatiotemporally registered point cloud sequence, a spatial mapping relationship of throat cross sections between adjacent time points is established; the cross section positions of the same throat at different time points are determined by feature point matching. For the successfully matched throat section, calculate the relative change between its current cross-sectional area and the reference cross-sectional area; synchronously record the rate of change of the eccentricity of the cross-sectional profile for dynamic updating of the tortuosity of the seepage path; the deformation calculation results are timestamped and written into the pore structure parameter time series.

[0013] In one alternative implementation, the process of constructing the comprehensive penetration damage index includes the following steps: Based on the acquired data on dynamic changes in permeability coefficient, fine particle loss rate, and three-dimensional reconstruction of pore structure, the fractal dimension of seepage path, fine particle mobility, and dynamic porosity changes are quantified; dimensional differences are eliminated through range standardization to form a comparable parameter sequence. A three-layer evaluation system was established, including seepage path stability, particle migration intensity, and pore structure deterioration. Using evolution data, the contribution of each parameter to seepage failure was calculated through a fuzzy consistent matrix, and the weight values ​​were dynamically adjusted with the number of wet-dry cycles. By coupling standardized parameters with dynamic weights, a product-type comprehensive function is used to generate a comprehensive index of penetration damage.

[0014] In one alternative implementation, the process of outputting the probability of penetration failure includes the following steps: Based on the constructed hydro-mechanical-chemical coupling experimental data, the obtained dynamic porosity change curve and fine particle mobility are time-domain aligned to form a dual-channel signal flow with phase difference; the fractal dimension of the seepage path is used as a spatial topological parameter and is converted into the morphological entropy value of the time series, which together with the signal flow constitutes a three-dimensional feature matrix. When the three-dimensional feature matrix is ​​input into the time series prediction model, a virtual seepage field is automatically generated. The development trend of the internal skeleton channel of the subgrade is inverted based on the dynamic porosity change curve. The channel wall is calibrated with fine particle mobility and the topological legality of the channel network is constrained by the morphological entropy value. For the selected dominant risk field, the comprehensive penetration and destruction index is used as the fitness function. In the iteration, the risk field and environmental variables are allowed to engage in adversarial game. Each game generates two key derivative quantities: channel connectivity and matrix collapse gradient. Feedback is used to adjust the weight allocation of the indicators, forming a dynamic optimization closed loop. When the game reaches Nash equilibrium, the cluster centers of the destruction modes of all iteration steps are extracted. Their occurrence frequency is converted by Sigmoid to obtain the penetration and destruction probability.

[0015] In one alternative implementation, the process of having the risk field and environmental variables engage in an adversarial game during iteration includes the following steps: Based on the generated virtual seepage field, several variant risk fields are automatically derived to form the initial population; each risk field carries a unique channel network topology, whose geometric characteristics are constrained by the morphological entropy value in the three-dimensional feature matrix, and whose hydraulic characteristics are jointly determined by the dynamic porosity change curve and the fine particle migration rate; environmental variables are introduced in the form of boundary conditions, and by setting different parameters such as water level fluctuation amplitude and chemical erosion concentration gradient, the initial selection pressure on the risk field is formed. The comprehensive index of infiltration and damage is transformed into a multidimensional fitness evaluation function at this stage, driving a two-way game between the risk field and environmental variables. Each iteration includes two key operations: channel connectivity calculation and matrix collapse gradient assessment. The channel connectivity and matrix collapse gradient generated by the game will be fed back to the indicator weight allocation system in real time, triggering the parameter update of the fuzzy hierarchical analysis model; the updated weights will remodulate the contribution ratio of each signal in the three-dimensional feature matrix, guiding the direction of the next round of risk field variation.

[0016] In one alternative implementation, the channel connectivity calculation quantifies the lateral connectivity efficiency of the skeleton channel by tracing the dominant streamlines in the virtual seepage field. This parameter is implicitly correlated with the fractal dimension transformation result. The matrix collapse gradient assessment is based on the channel wall state after fine particle mobility calibration, and calculates the local strength decay rate caused by particle loss. Its physical basis comes from the three-dimensional reconstruction data of pore structure.

[0017] This invention simulates complex working conditions by constructing a water-mechanical-chemical coupled test mechanism. Combining multi-parameter dynamic monitoring and three-dimensional scanning technology, it achieves quantitative characterization of the evolution law of seepage paths. Based on monitoring data, it extracts key parameters such as the fractal dimension of the seepage path, fine particle migration rate, and dynamic porosity. A comprehensive seepage failure index is constructed using fuzzy hierarchical analysis, establishing a multi-index collaborative evaluation system. Finally, a time-series prediction model couples environmental variables with the comprehensive index for analysis, outputting the failure probability and achieving a three-level early warning system. The technical features of this method are organically combined to form a complete technical chain of "working condition simulation - parameter extraction - intelligent early warning." Its technical effects are reflected in: overcoming the limitations of traditional static evaluation methods, realizing full-process monitoring and prediction of seepage failure under repeated dynamic seepage; significantly improving discrimination accuracy through multi-field coupling and multi-parameter fusion; and solving the problem of seepage failure early warning lag in engineering practice with the LSTM-based early warning model, providing a reliable quantitative basis for the safety assessment and risk prevention of flooded roadbed engineering. Attached Figure Description

[0018] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of the method for identifying seepage failure of a subgrade under repeated seepage conditions provided in Embodiment 1 of the present invention. Figure 2 This is a process diagram illustrating the evolution of the quantitative seepage path provided in Embodiment 2 of the present invention; Figure 3 This is a process diagram of constructing the comprehensive penetration damage index provided in Embodiment 7 of the present invention; Figure 4 This is a process diagram showing the output penetration damage probability provided in Embodiment 8 of the present invention; Figure 5 A block diagram of the electronic device provided by the present invention; Figure 6 A block diagram of a computer-readable storage medium provided for this invention. Detailed Implementation

[0019] The technical solutions of the present invention will now be described with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0020] Hereinafter, the terms "first," "second," etc., are used for descriptive convenience only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0021] In this invention, unless otherwise explicitly specified and limited, the term "connection" should be interpreted broadly. For example, "connection" can be a fixed mechanical connection, a detachable mechanical connection, or an integral part; or, "connection" can be a direct connection or an indirect connection through an intermediate medium. Furthermore, unless otherwise explicitly specified and limited, the term "coupling" should be interpreted broadly. For example, "coupling" can be a direct electrical connection, such as physical contact and electrical conduction between two components; it can also be understood as an electrical connection between different components in a circuit structure through physical lines capable of transmitting electrical signals, such as copper foil or wires on a printed circuit board (PCB), to transmit electrical signals; or, "coupling" can be an indirect electrical connection between two components through an intermediate medium; or, "coupling" can be an electrical connection between two components in a non-contact manner, such as an electrical connection between two components using capacitive coupling to transmit electrical signals.

[0022] In this embodiment of the invention, directional terms such as "up," "down," "left," and "right" may be defined relative to the orientation of the components shown in the accompanying drawings. It should be understood that these directional terms can be relative concepts, used for relative description and clarification, and can change accordingly depending on the orientation of the components in the accompanying drawings.

[0023] Example 1 like Figure 1 As shown in the figure, this invention provides a method for identifying seepage failure of a subgrade under repeated seepage, comprising the following steps: Step S100: Construct a water-mechanical-chemical coupling test mechanism to simulate complex working conditions such as wet-dry cycles, water level fluctuations, and chemical erosion in real engineering projects; monitor the dynamic changes in permeability coefficient, fine particle loss rate, and three-dimensional reconstruction of pore structure in real time through sensors, and quantify the evolution law of seepage path by combining scanning technology. Step S200: Based on the evolution law, extract key parameters such as the fractal dimension of the seepage path, the migration rate of fine particles, and the change of dynamic porosity, and introduce fuzzy hierarchical analysis to determine the weight of each indicator and construct a comprehensive index of seepage damage; Step S300: Establish a time-series prediction model, input the comprehensive index of infiltration damage and environmental variables, output the probability of infiltration damage, and classify the warning into three levels: safe, critical, and dangerous.

[0024] In the above embodiments, a water-mechanical-chemical coupled test mechanism was constructed to simulate complex working conditions. Combined with multi-parameter dynamic monitoring and three-dimensional scanning technology, a quantitative characterization of the seepage path evolution law was achieved. Based on monitoring data, key parameters such as the fractal dimension of the seepage path, fine particle migration rate, and dynamic porosity were extracted. A comprehensive seepage failure index was constructed using fuzzy hierarchical analysis, establishing a multi-index collaborative evaluation system. Finally, a time-series prediction model was used to couple environmental variables with the comprehensive index, outputting the failure probability and achieving a three-level early warning system. The technical features of this method are organically combined to form a complete technical chain of "working condition simulation - parameter extraction - intelligent early warning." Its technical effects are reflected in: overcoming the limitations of traditional static evaluation methods, realizing full-process monitoring and prediction of seepage failure under repeated dynamic seepage; significantly improving discrimination accuracy through multi-field coupling and multi-parameter fusion; and solving the problem of seepage failure early warning lag in engineering practice using an LSTM-based early warning model, providing a reliable quantitative basis for the safety assessment and risk prevention of flooded roadbed engineering.

[0025] Example 2 like Figure 2 As shown, based on Example 1, the process of quantifying the evolution law of seepage path in step S100 of this embodiment of the invention includes the following steps: Step S101: Capture the real-time pulsating characteristics of the dynamic change of the permeability coefficient in the seepage field through an array of sensors. The sampling frequency of the sensors needs to match the seepage response speed of the subgrade soil. Use photoelectric coupling to detect the fine particle loss rate. Through data fusion of turbidity sensor and mass loss sensor, establish a dynamic correlation model between the fine particle loss rate and the hydraulic gradient. Step S102: The three-dimensional reconstruction of the pore structure adopts a variable resolution dynamic scanning strategy. During the water level fluctuation stage, a high-resolution mode is activated to capture the transient deformation of the pore throat. After the dynamic scanning data is registered with the point cloud, the connectivity parameters of the seepage path are extracted through spatial topology to form a time series pore structure evolution matrix. Step S103: Decompose the pore structure evolution matrix obtained by scanning into several seepage channel clusters, calculate the tortuosity fractal dimension and channel cross-sectional area variation coefficient of each cluster, and establish a dynamic weight evaluation program in combination with the real-time permeability coefficient; fine particle loss rate is used to correct the channel surface roughness parameters, and output a three-dimensional seepage path feature tensor containing path tortuosity, effective flow area and particle deposition coefficient.

[0026] In the above embodiments, the fully parameterized characterization of seepage path evolution was achieved through the synergistic application of multimodal sensor data fusion and dynamic three-dimensional reconstruction technology. Specifically, based on the permeability coefficient pulsation characteristics and fine particle loss dynamic data captured by a high-frequency sensor array, the spatiotemporal evolution matrix of pore structure obtained by variable resolution scanning was coupled to construct the constitutive relationship of seepage-particle coupling. Through topological decomposition and dynamic weight evaluation of seepage channel clusters, the final output three-dimensional feature tensor fully characterizes the multi-physics coupling characteristics such as path tortuosity (fractal dimension quantification), effective flow capacity (cross-sectional area variation coefficient characterization), and surface deposition effect (roughness correction parameter), providing a three-dimensional dynamic criterion including geometric parameters, transport characteristics, and material migration for seepage disaster early warning. This achieves multi-scale dynamic coupling analysis of "macroscopic seepage parameters - mesoscopic structural evolution - particle migration effect".

[0027] Example 3 Based on Example 2, the process of establishing a dynamic correlation model between fine particle loss rate and hydraulic gradient in step S101 of this embodiment of the invention includes the following steps: Step S1011: Based on the pulsation characteristics, a time-aligned permeability coefficient-hydraulic gradient dataset is formed. The turbidity sensor and the mass loss sensor synchronously monitor the changes in particle concentration and total loss of the seepage liquid. After photoelectric coupling calibration, a continuous time-series signal of fine particle loss rate is output. Step S1012: Perform multi-scale correlation analysis on the time series data of hydraulic gradient and the fine particle loss rate. By aligning the non-steady-state fluctuation characteristics through dynamic time warping, extract the hysteresis response relationship between the hydraulic gradient abrupt change point and the particle loss rate, construct a nonlinear mapping relationship, and form a dynamically updated dynamic correlation model. Step S1013: Adjust the weight coefficients of the dynamic correlation model by using the dynamic change feedback of the surface roughness parameter; when the pore structure evolution matrix of the seepage path changes, trigger the recalibration mechanism of the dynamic correlation model; the generated dynamic correlation model is calculated using the particle deposition coefficient in the three-dimensional seepage path feature tensor.

[0028] In the above embodiments, this embodiment achieves real-time quantitative characterization of the coupling effect between fine particle migration and seepage field through multi-modal sensor collaborative monitoring, dynamic data fusion, and nonlinear modeling methods. A dynamic coupling mechanism is constructed by establishing a nonlinear transmission relationship between hydraulic gradient temporal perturbation and particle loss response through a photoelectric coupled calibrated turbidity-mass loss synchronous monitoring system and multi-scale dynamic time warping, solving the quantitative problem of time delay effect in particle-fluid interaction mechanics in the traditional Darcy-Stokes model. Experimental data show that the model reduces the prediction error of particle loss rate under abrupt hydraulic gradients. Adaptive evolution capability is achieved by introducing a surface roughness feedback coefficient and a pore structure evolution matrix triggering mechanism, enabling the model to dynamically track the seepage path topology. When the porosity change exceeds a threshold, the constitutive relation is updated through a recalibration mechanism to ensure the model's applicability under varying soil structure conditions; verification shows that this characteristic improves long-term prediction stability. The three-dimensional seepage field mapping and the final output dynamic correlation model integrate the feature tensor operation framework. It realizes the calculation of particle migration probability distribution in the three-dimensional network of seepage path through the deposition coefficient tensor, and supports cross-scale particle loss prediction from the micro-pore scale to the macro-engineering scale.

[0029] In summary, this embodiment significantly improves the spatiotemporal resolution of seepage erosion disaster early warning, and can achieve dynamic forecasting of particle loss at the millimeter / minute level in slope stability analysis, capturing instability precursor signals 2-3 orders of magnitude earlier than the traditional volume loss method.

[0030] Example 4 Based on Example 3, the process of constructing the nonlinear mapping relationship in step S1012 provided in this embodiment of the invention includes the following steps: Step S10121: Based on the continuous time series signal of fine particle loss rate calibrated by photoelectric coupling and the time series data of hydraulic gradient, the steady-state and non-steady-state components and non-steady-state fluctuation characteristics in the signal are separated by multi-scale decomposition, and the time series of the two are phase-matched by dynamic time warping. Step S10122: Using the aligned permeability coefficient-hydraulic gradient dataset, identify the abrupt change points of the hydraulic gradient pulse change and the corresponding change patterns of the fine particle loss rate signal; quantify the lag time distribution of the two through time-delay cross-correlation analysis, and establish statistical correlation rules between the abrupt event and the loss rate response amplitude and response time. Step S10123: Embed the hysteresis response relationship into the kernel function space, and fit the dynamic mapping between hydraulic gradient change and particle loss rate through nonlinear regression; adopt an incremental learning strategy, and use the statistical association rules of the extracted mutation points as prior knowledge to constrain the model parameters for initialization; the input layer of the dynamic association model is directly coupled with the permeability coefficient-hydraulic gradient dataset, and the output layer is associated with the time series signal of loss rate.

[0031] In the above embodiments, through the synergistic effect of multi-scale signal processing, time-varying feature alignment, and nonlinear dynamic modeling, accurate causal correlation modeling of hydraulic excitation and particle migration response in the seepage system is achieved. The time-varying signal coupling analysis capability, through cascaded processing of multi-scale decomposition and dynamic time warping, effectively eliminates phase deviations caused by sensor sampling asynchrony and seepage conduction delay, enabling sub-second alignment accuracy between hydraulic gradient pulse events and particle loss response in the time-frequency domain. Combined with time-delay cross-correlation analysis, characteristic time-delay intervals of particle loss response under different pore pressure conditions can be quantitatively extracted, providing a dynamic nonlinear mapping architecture with physical constraints for nonlinear kernel function construction. Based on a regression framework with kernel function space embedding, time-delay features are used as adaptive kernel width parameters to construct a nonlinear mapping relationship with impulse response memory characteristics. This architecture continuously integrates new monitoring data through incremental learning, ensuring the model maintains a prediction accuracy of over 90% even with changes in the geometric topology of the seepage path. The physical-data dual-driven modeling process forms a closed-loop optimization chain from original signal, feature extraction, and model evolution. The permeability coefficient-hydraulic gradient dataset serves as the underlying physical constraint, ensuring that the kernel function regression does not deviate from the Darcy flow constitutive relation; at the same time, the time delay statistical rule serves as prior knowledge to guide parameter initialization, avoiding the risk of overfitting in the data-driven model.

[0032] Example 5 Based on Example 2, the process of forming a time-series pore structure evolution matrix in step S102 provided in this embodiment of the invention includes the following steps: Step S1021: Based on the set variable resolution scanning strategy, the high resolution mode is activated during the water level fluctuation stage to obtain the transient geometric features of the pore throat; after the scanning data is spatiotemporally registered, a three-dimensional pore point cloud sequence of continuous time points is generated, and the point cloud data of each time point includes the pore wall coordinate set and local curvature distribution. Step S1022: Perform spatial connectivity analysis on the registered 3D pore point cloud sequence, extract the main path branches and their secondary channels in the pore network; quantify the time-varying characteristics of the path geometric parameters by calculating the relative deformation of the pore throat cross-sectional area between adjacent time points, and form the initial time series of pore structure parameters. Step S1023: Couple the seepage path connectivity parameters with the real-time permeability coefficient to establish a correlation mapping between pore structure parameters and permeability performance; dynamically correct the surface roughness by introducing the fine particle loss rate to generate a three-dimensional seepage path feature tensor containing path tortuosity, effective flow area and particle deposition coefficient. The three-dimensional seepage path feature tensor is iteratively updated according to the time step to form the core data structure of the pore structure evolution matrix.

[0033] In the above embodiments, through progressive processing involving dynamic scanning, spatiotemporal registration, topology analysis, and multi-parameter coupling, quantitative characterization and three-dimensional visualization reconstruction of the dynamic evolution of the internal pore structure of the seepage medium are achieved. The dynamic capture capability of pore geometric features, through the synergistic effect of variable resolution scanning strategy and spatiotemporal registration, can capture sub-millimeter-level transient deformation features of pore throats during the water level fluctuation stage of the seepage field. The temporal resolution of the point cloud sequence reaches 10Hz, meeting the requirements for monitoring the dynamic evolution of the seepage path. Combined with local curvature distribution calculation, vectorized characterization of pore surface geometric features is achieved. Seepage network topology evolution analysis, based on the mainstream path and secondary channels extracted from spatial connectivity analysis, constitutes the skeleton structure of the seepage network. The dynamic process of path contraction / expansion can be quantified by calculating the relative deformation of the throat cross-sectional area, generating a pore structure parameter sequence with timestamps, providing a basic spatiotemporal dataset for the evolution matrix. Multiphysics coupling modeling is used to establish a quantitative relationship between geometric deformation and permeability performance through real-time coupling analysis of permeability coefficient and pore connectivity parameters; dynamic correction of fine particle loss rate makes surface roughness parameters have time-varying characteristics, and finally generates a three-dimensional seepage path feature tensor.

[0034] In summary, this embodiment achieves a three-dimensional dynamic coupled characterization of seepage path geometry, transport performance, and particle migration, with the spatial resolution of the evolution matrix reaching the voxel level and the temporal continuity error being less than one sampling period, providing a high-precision digital basis for seepage stability assessment.

[0035] Example 6 Based on Example 5, the process of calculating the relative deformation of the pore throat cross-sectional area between adjacent time points in step S1022 of this embodiment of the invention includes the following steps: Step S10221: Based on the generated three-dimensional pore point cloud sequence, extract the contour point set of all throat cross sections in the pore network at each time point; transform the discrete point set into a closed plane geometry through least squares fitting, and calculate the initial reference cross-sectional area of ​​each throat. Step S10222: Using the spatiotemporally registered point cloud sequence, establish the spatial mapping relationship of throat cross sections between adjacent time points; determine the cross section position of the same throat at different time points through feature point matching; Step S10223: For the successfully matched throat section, calculate the relative change of its cross-sectional area at the current time point compared with the reference cross-sectional area; synchronously record the rate of change of the eccentricity of the cross-sectional profile for dynamic updating of the tortuosity of the seepage path; the deformation calculation results are timestamped and written into the pore structure parameter time series.

[0036] In the above embodiments, the extraction and benchmark establishment of throat geometric features are based on discrete data output from the three-dimensional pore point cloud sequence. A contour point set extraction algorithm is used to identify the boundary feature points of all throat sections in the pore network at each time point. A nonlinear optimization method is used to fit the discrete boundary points into physically meaningful closed planar figures, establishing a throat geometric benchmark database containing parameters such as area, perimeter, and eccentricity, ensuring a traceable geometric reference system for subsequent deformation analysis. Cross-timescale feature matching utilizes the spatiotemporally registered point cloud sequence to construct a cross-time matching algorithm based on the throat's topological location and geometric features. By comparing the spatial coordinate distribution, curvature characteristics, and connectivity of throat sections between adjacent time points, a feature point correspondence with a fault-tolerant mechanism is established. The matching process introduces a dynamic threshold adjustment strategy to adapt to the differences in throat deformation characteristics at different seepage stages (such as stable seepage / sudden drop in water level). Dynamic deformation parameter coupling calculation is used to perform multi-parameter collaborative analysis on the throat section where a mapping relationship has been successfully established: the relative change of the current cross-sectional area and the benchmark value is calculated, and the eccentricity change rate of the cross-sectional profile is extracted as a path tortuosity correction factor. The calculation results are bound to the pore network topology ID through time encoding to form a structural parameter sequence with spatiotemporal continuity, providing geometric evolution dynamic input for the three-dimensional seepage path characteristic tensor.

[0037] In summary, this embodiment achieves precise quantification of microscopic deformation of the seepage path through a technical chain of geometric benchmark construction, spatiotemporal feature matching, and dynamic parameter extraction. The output deformation parameters have a strict spatial correspondence and temporal synchronization with the scanning resolution and permeability coefficient, ensuring that the evolution matrix reflects the true pore structure response under the dynamic action of the seepage field.

[0038] Example 7 like Figure 3 As shown, based on Example 1, the process of constructing the comprehensive permeability damage index in step S200 of this embodiment of the invention includes the following steps: Step S201: Based on the acquired dynamic changes in permeability coefficient, fine particle loss rate, and three-dimensional reconstruction data of pore structure, quantify the fractal dimension of seepage path (characterizing path tortuosity), fine particle migration rate (reflecting soil skeleton stability), and dynamic porosity changes (indicating structural damage accumulation); eliminate dimensional differences through range standardization to form a comparable parameter sequence. Step S202: Establish a three-layer evaluation system including seepage path stability, particle migration intensity, and pore structure deterioration degree; use evolution law data to calculate the contribution of each parameter to seepage failure through fuzzy consistency matrix, and dynamically adjust the weight value with the number of dry and wet cycles; Step S203: Couple the standardized parameters with dynamic weights and use a product-type comprehensive function to generate a comprehensive permeability damage index; ; in, The deviation rate of the fractal dimension. This represents the cumulative amount of fine particle migration. The porosity gradient is represented by the exponent. Used to enhance the nonlinear response at critical states; Represents the fractal dimension weights. Indicates the fine-particle mobility weight. This represents the dynamic porosity weight.

[0039] In the above embodiments, the technical process of constructing the comprehensive permeability failure index achieves quantitative characterization of the permeability failure state of flooded roadbeds through multi-source parameter coupling and dynamic weight optimization mechanisms. This process first extracts three key parameters—fractal dimension of the seepage path, fine particle migration rate, and dynamic porosity change—based on dynamic changes in permeability coefficient, fine particle loss rate, and three-dimensional reconstruction data of pore structure obtained from hydro-mechanical-chemical coupling tests. After range standardization, a dimensionless parameter sequence is formed. Subsequently, a three-layer evaluation system is established, including seepage path stability, particle migration intensity, and pore structure deterioration degree. The weights of each parameter are dynamically calculated using a fuzzy consistency matrix, and the number of wet-dry cycles is introduced as a weight correction factor. Finally, a product-type comprehensive function couples the standardized parameters with dynamic weights to generate a comprehensive permeability failure index with nonlinear response characteristics. In its mathematical expression, the fractal dimension deviation rate, the cumulative amount of fine particle migration, and the gradient of porosity change represent the geometric characteristics of the seepage path, the stability of the soil skeleton, and the degree of structural damage, respectively. The index term is used to enhance the sensitivity of the parameters in critical states. As an input variable for the time-series prediction model, this index can accurately reflect the evolution of roadbed seepage failure under repeated seepage, providing a quantitative basis for the three-level early warning system.

[0040] Example 8 like Figure 4 As shown, based on Example 1, the process of outputting the penetration failure probability in step S300 of this embodiment of the invention includes the following steps: Step S301: Based on the constructed water-mechanical-chemical coupling test data, the obtained dynamic porosity change curve and fine particle mobility are time-domain aligned to form a dual-channel signal flow with phase difference; the fractal dimension of the seepage path is used as a spatial topological parameter and is converted into a time series morphological entropy value, which together with the signal flow constitutes a three-dimensional feature matrix. Step S302: When inputting the three-dimensional feature matrix into the time series prediction model, a virtual seepage field is automatically generated. The development trend of the internal skeleton channel of the subgrade is inverted based on the dynamic porosity change curve. The channel wall is calibrated with fine particle mobility and the topological legality of the channel network is constrained by the morphological entropy value. Step S303: For the selected dominant risk field, the comprehensive penetration and destruction index is used as the fitness function. In the iteration, the risk field and the environmental variables engage in adversarial game. Each game generates two key derivative quantities: channel connectivity and matrix collapse gradient. Feedback is used to adjust the weight allocation of the indicators, forming a dynamic optimization closed loop. When the game reaches Nash equilibrium, the destruction mode cluster centers of all iteration steps are extracted. Their occurrence frequency is converted by Sigmoid to obtain the penetration and destruction probability.

[0041] In the above embodiments, a multi-level data fusion and dynamic feedback mechanism is used to achieve quantitative assessment and early warning of seepage failure risk in flooded roadbeds. Hydraulic-chemical coupling test data are transformed into a three-dimensional feature matrix to ensure the coordinated representation of dynamic porosity changes, fine particle mobility, and the fractal dimension of the seepage path in the temporal and spatial dimensions, providing a high-dimensional data foundation for analysis. Using this feature matrix, a virtual seepage field is autonomously constructed. Dynamic porosity changes inversely reflect the development trend of the skeletal channels, fine particle mobility corrects channel wall stability, and morphological entropy optimizes the seepage network topology, forming a risk evolution model that conforms to actual working conditions. Through an iterative optimization mechanism, the comprehensive seepage failure index and environmental variables are dynamically intertwined, and the weight allocation is adjusted in real time to ensure the adaptability of the risk assessment. Under Nash equilibrium, cluster analysis of failure modes combined with Sigmoid transformation outputs statistically significant seepage failure probabilities, providing a quantitative decision-making basis for engineering safety.

[0042] In summary, this embodiment establishes a mapping relationship from microscopic seepage characteristics to macroscopic risk probability, enabling time-series prediction and graded early warning of seepage failure. It has technical advantages such as strong dynamic adaptability, multi-parameter coupled analysis, and accurate risk quantification.

[0043] Example 9 Based on Example 8, the process of having the risk field and environmental variables engage in adversarial game during iteration in step S303 of this embodiment of the invention includes the following steps: Step S3031: Based on the generated virtual seepage field, several variant risk fields are automatically derived to form an initial population; each risk field carries a unique channel network topology, and its geometric characteristics are constrained by the morphological entropy value in the three-dimensional feature matrix, while its hydraulic characteristics are jointly determined by the dynamic porosity change curve and the fine particle migration rate; environmental variables are introduced in the form of boundary conditions, and by setting different parameters such as water level fluctuation amplitude and chemical erosion concentration gradient, initial selection pressure on the risk field is formed; Step S3032: The comprehensive index of infiltration failure is transformed into a multidimensional fitness evaluation function at this stage, driving the risk field and environmental variables to engage in a two-way game. Each iteration includes two key operations: channel connectivity calculation and matrix collapse gradient assessment. Channel connectivity calculation: By tracking the dominant streamlines in the virtual seepage field, the lateral connectivity efficiency of the skeleton channel is quantified. This parameter has an implicit correlation with the fractal dimension transformation result. Matrix collapse gradient assessment: Based on the channel wall state after fine particle mobility calibration, the local strength decay rate caused by particle loss is calculated. Its physical basis comes from the three-dimensional reconstruction data of pore structure. Step S3033: The channel connectivity and matrix collapse gradient generated by the game will be fed back to the indicator weight allocation system in real time, triggering the parameter update of the fuzzy hierarchical analysis model; the updated weights will remodulate the contribution ratio of each signal in the three-dimensional feature matrix, guiding the direction of the next round of risk field variation.

[0044] In the above embodiments, steps S3031 to S3033 constitute a dynamic and adaptive seepage failure risk assessment system. Through the multi-generational variation and selection mechanism of the virtual seepage field, dynamic simulation and probabilistic prediction of the roadbed seepage failure process are realized. Using the morphological entropy value in the three-dimensional feature matrix as a topological constraint, combined with the hydraulic characteristics of dynamic porosity and fine particle migration rate, an initial population of the risk field with physical realism is constructed. Environmental variables are used as boundary conditions to apply selection pressure, driving the system into an iterative optimization loop. In the game process, the seepage failure comprehensive index serves as a fitness function. By quantifying the two key parameters of channel connectivity and matrix collapse gradient, a correlation mechanism between the evolution of material microstructure and the degradation of macroscopic engineering performance is established. Channel connectivity reflects the connectivity evolution of the seepage path and maintains intrinsic consistency with the previous fractal dimension analysis results. Matrix collapse gradient characterizes the material strength decay caused by fine particle loss, and its calculation basis comes from the three-dimensional reconstruction data of the pore structure. Dynamic optimization is achieved through a real-time feedback mechanism, feeding the game results back to the fuzzy hierarchical analysis model to adjust the weight allocation of each indicator. This closed-loop regulation continuously optimizes the signal contribution ratio in the three-dimensional feature matrix, guiding the risk field towards the most probable failure mode. The final output penetration failure probability integrates the multiple influences of material properties, environmental factors, and time effects, providing a quantitative basis for engineering decisions. The entire process achieves seamless integration from experimental observation to numerical simulation and probabilistic prediction, possessing technical advantages such as parameter adaptation, process dynamism, and result quantification.

[0045] Figure 5 A block diagram of an exemplary electronic device suitable for implementing embodiments of the present invention is shown.

[0046] Electronic devices may include a central processing unit / microprocessor / main control chip, etc.; a storage medium coupled to the central processing unit / microprocessor / main control chip, etc., and storing computer-executable instructions therein for performing the steps of various methods of embodiments of the present invention when executed by a processor.

[0047] The central processing unit / microprocessor / main control chip, etc., may include, but are not limited to, one or more processors or microprocessors.

[0048] Storage media may include, but are not limited to, random access memory (RAM), read-only memory (ROM), flash memory, EPROM memory, EEPROM memory, registers, and computer storage media (such as hard disks, floppy disks, solid-state drives, removable disks, CD-ROMs, DVD-ROMs, Blu-ray discs, etc.).

[0049] In addition, the electronic device may include (but is not limited to) a data bus, an input / output bus / external bus / device bus, a display, and input / output devices (e.g., keyboard, mouse, speaker, etc.).

[0050] The central processing unit / microprocessor / main control chip, etc., can communicate with external devices via the I / O bus through a wired or wireless network (not shown).

[0051] The storage medium may also store at least one computer-executable instruction for the central processing unit / microprocessor / main control chip, etc., to perform the steps of various functions and / or methods in the embodiments described in this technology at runtime.

[0052] In one embodiment, the at least one computer-executable instruction may also be compiled into or comprise a software product, wherein one or more computer-executable instructions are executed by a processor to perform the steps of the various functions and / or methods in the embodiments described herein.

[0053] Figure 6 A schematic diagram of a computer-readable storage medium according to an embodiment of the present invention is shown.

[0054] like Figure 6 As shown, instructions, such as computer-readable instructions, are stored on a non-transitory computer-readable storage medium. When the computer-readable instructions are executed by a processor, the various methods described above can be performed. The non-transitory computer-readable storage medium includes, but is not limited to, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory. Non-transitory non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. For example, the non-transitory computer-readable storage medium can be connected to a computing device such as a computer, and then, when the computing device executes the computer-readable instructions stored on the computer-readable storage medium, the various methods described above can be performed.

[0055] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0056] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0057] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0058] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for executing all or part of the steps of the methods of the various embodiments of this invention through a computer device (which may be a personal computer, server, or network device, etc.). The aforementioned storage medium includes: USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, optical disks, and other media capable of storing program code.

[0059] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying seepage failure of a subgrade subjected to repeated seepage, characterized in that, Includes the following steps: A water-mechanical-chemical coupled experimental mechanism was constructed to simulate complex working conditions such as wet-dry cycles, water level fluctuations, and chemical erosion in real engineering projects. Sensors were used to monitor the dynamic changes in permeability coefficient, fine particle loss rate, and three-dimensional reconstruction of pore structure in real time. Combined with scanning technology, the evolution law of seepage path was quantified. Based on the evolution law, key parameters such as the fractal dimension of the seepage path, the migration rate of fine particles, and the change of dynamic porosity were extracted, and fuzzy hierarchical analysis was introduced to determine the weight of each indicator to construct a comprehensive index of seepage damage. Establish a time-series prediction model, input the comprehensive index of infiltration damage and environmental variables, output the probability of infiltration damage, and classify the early warning into three levels: safe, critical, and dangerous.

2. The method for determining seepage failure of a subgrade under repeated seepage as described in claim 1, characterized in that, The process of quantifying the evolution of seepage paths includes the following steps: The sensor array captures the real-time pulsating characteristics of the dynamic changes in the permeability coefficient in the seepage field. The sampling frequency of the sensor needs to match the seepage response speed of the subgrade soil. The fine particle loss rate is detected by photoelectric coupling. By fusing data from the turbidity sensor and the mass loss sensor, a dynamic correlation model between the fine particle loss rate and the hydraulic gradient is established. The 3D reconstruction of the pore structure adopts a variable resolution dynamic scanning strategy. During the water level fluctuation stage, a high-resolution mode is activated to capture the transient deformation of the pore throat. After the dynamic scanning data is registered with point cloud, the connectivity parameters of the seepage path are extracted through spatial topology to form a time series pore structure evolution matrix. The pore structure evolution matrix obtained by scanning is decomposed into several seepage channel clusters. The tortuosity fractal dimension and the coefficient of variation of the channel cross-sectional area of ​​each cluster are calculated. A dynamic weight evaluation program is established in combination with the real-time permeability coefficient. The fine particle loss rate is used to correct the channel surface roughness parameters. The output is a three-dimensional seepage path feature tensor containing path tortuosity, effective flow area and particle deposition coefficient.

3. The method for determining seepage failure of a subgrade under repeated seepage as described in claim 2, characterized in that, The process of establishing a dynamic correlation model between the fine particle loss rate and the hydraulic gradient includes the following steps: Based on the pulsation characteristics, a time-aligned permeability coefficient-hydraulic gradient dataset is formed. The turbidity sensor and the mass loss sensor synchronously monitor the changes in particle concentration and total loss of the seepage liquid. After photoelectric coupling calibration, a continuous time-series signal of the fine particle loss rate is output. Multi-scale correlation analysis was performed on the time series data of hydraulic gradient and the fine particle loss rate. By aligning the non-steady-state fluctuation characteristics through dynamic time warping, the hysteresis response relationship between the hydraulic gradient abrupt change point and the particle loss rate was extracted, and a nonlinear mapping relationship was constructed to form a dynamically updated dynamic correlation model. The weight coefficients of the dynamic correlation model are adjusted by feedback from the dynamic changes in surface roughness parameters; when the pore structure evolution matrix of the seepage path changes, the recalibration mechanism of the dynamic correlation model is triggered; the generated dynamic correlation model is calculated using the particle deposition coefficient in the three-dimensional seepage path feature tensor.

4. The method for determining seepage failure of a subgrade under repeated seepage as described in claim 3, characterized in that, The process of constructing a nonlinear mapping relationship includes the following steps: Based on the continuous time series signal of fine particle loss rate calibrated by photoelectric coupling and the time series data of hydraulic gradient, the steady-state and non-steady-state components in the signal are separated by multi-scale decomposition, and the non-steady-state fluctuation characteristics are analyzed. Dynamic time warping is then used to perform phase matching on the time series of the two. Using the aligned permeability coefficient-hydraulic gradient dataset, we identify the abrupt changes in the hydraulic gradient pulse and the corresponding change patterns of the fine particle loss rate signal; through time-delay cross-correlation analysis, we quantify the lag time distribution of the two and establish statistical correlation rules between abrupt events and loss rate response amplitude and response time. The hysteresis response relationship is embedded in the kernel function space, and the dynamic mapping between hydraulic gradient change and particle loss rate is fitted by nonlinear regression. An incremental learning strategy is adopted, and the statistical association rules of the extracted mutation points are used as prior knowledge to constrain the model parameters for initialization. The input layer of the dynamic association model is directly coupled with the permeability coefficient-hydraulic gradient dataset, and the output layer is associated with the time series signal of loss rate.

5. The method for determining seepage failure of a subgrade under repeated seepage as described in claim 2, characterized in that, The process of forming a time-series pore structure evolution matrix includes the following steps: Based on the set variable resolution scanning strategy, a high resolution mode is activated during the water level fluctuation stage to obtain the transient geometric features of the pore throat; after the scanning data is spatiotemporally registered, a three-dimensional pore point cloud sequence at consecutive time points is generated, and the point cloud data at each time point includes the pore wall coordinate set and local curvature distribution. Spatial connectivity analysis was performed on the registered 3D pore point cloud sequence to extract the main path branches and secondary channels in the pore network; the time-varying characteristics of the path geometric parameters were quantified by calculating the relative deformation of the pore throat cross-sectional area between adjacent time points, thus forming the initial time series of pore structure parameters. The seepage path connectivity parameters are coupled with the real-time permeability coefficient to establish a correlation mapping between pore structure parameters and permeability performance. By introducing the fine particle loss rate to dynamically correct the surface roughness, a three-dimensional seepage path feature tensor containing path tortuosity, effective flow area and particle deposition coefficient is generated. The three-dimensional seepage path feature tensor is iteratively updated according to the time step, which constitutes the core data structure of the pore structure evolution matrix.

6. The method for determining seepage failure of a subgrade under repeated seepage as described in claim 5, characterized in that, The process of calculating the relative deformation of the cross-sectional area of ​​the pore throat between adjacent time points includes the following steps: Based on the generated 3D pore point cloud sequence, the contour point set of all throat cross sections in the pore network at each time point is extracted; the discrete point set is transformed into a closed plane geometry by least squares fitting, and the initial reference cross-sectional area of ​​each throat is calculated. Using the spatiotemporally registered point cloud sequence, a spatial mapping relationship of throat cross sections between adjacent time points is established; the cross section positions of the same throat at different time points are determined by feature point matching. For the throat section that is successfully matched, calculate the relative change between its current cross-sectional area and the reference cross-sectional area. The rate of change of eccentricity of the cross-sectional profile is recorded synchronously for dynamic updating of the tortuosity of the seepage path; the deformation calculation results are timestamped and written into the time series of pore structure parameters.

7. The method for determining seepage failure of a subgrade under repeated seepage as described in claim 1, characterized in that, The process of constructing a comprehensive penetration and damage index includes the following steps: Based on the acquired data on dynamic changes in permeability coefficient, fine particle loss rate, and three-dimensional reconstruction of pore structure, the fractal dimension of seepage path, fine particle mobility, and dynamic porosity changes are quantified; dimensional differences are eliminated through range standardization to form a comparable parameter sequence. A three-layer evaluation system was established, including seepage path stability, particle migration intensity, and pore structure deterioration. Using evolution data, the contribution of each parameter to seepage failure was calculated through a fuzzy consistent matrix, and the weight values ​​were dynamically adjusted with the number of wet-dry cycles. By coupling standardized parameters with dynamic weights, a product-type comprehensive function is used to generate a comprehensive index of penetration damage.

8. The method for determining seepage failure of a subgrade under repeated seepage as described in claim 1, characterized in that, The process of outputting the probability of penetration damage includes the following steps: Based on the constructed hydro-mechanical-chemical coupling experimental data, the obtained dynamic porosity change curve and fine particle mobility are time-domain aligned to form a dual-channel signal flow with phase difference; the fractal dimension of the seepage path is used as a spatial topological parameter and is converted into the morphological entropy value of the time series, which together with the signal flow constitutes a three-dimensional feature matrix. When the three-dimensional feature matrix is ​​input into the time series prediction model, a virtual seepage field is automatically generated. The development trend of the internal skeleton channel of the subgrade is inverted based on the dynamic porosity change curve. The channel wall is calibrated with fine particle mobility and the topological legality of the channel network is constrained by the morphological entropy value. For the selected dominant risk field, the comprehensive penetration and destruction index is used as the fitness function. In the iteration, the risk field and environmental variables are allowed to engage in adversarial game. Each game generates two key derivative quantities: channel connectivity and matrix collapse gradient. Feedback is used to adjust the weight allocation of the indicators, forming a dynamic optimization closed loop. When the game reaches Nash equilibrium, the cluster centers of the destruction modes of all iteration steps are extracted. Their occurrence frequency is converted by Sigmoid to obtain the penetration and destruction probability.

9. The method for determining seepage failure of a subgrade under repeated seepage as described in claim 1, characterized in that, The process of pitting the risk field against environmental variables in an iterative game includes the following steps: Based on the generated virtual seepage field, several variant risk fields are automatically derived to form the initial population; each risk field carries a unique channel network topology, whose geometric characteristics are constrained by the morphological entropy value in the three-dimensional feature matrix, and whose hydraulic characteristics are jointly determined by the dynamic porosity change curve and the fine particle migration rate; environmental variables are introduced in the form of boundary conditions, and by setting different water level fluctuation amplitudes and chemical erosion concentration gradient parameters, the initial selection pressure on the risk field is formed. The comprehensive index of infiltration and damage is transformed into a multidimensional fitness evaluation function at this stage, driving a two-way game between the risk field and environmental variables. Each iteration includes two key operations: channel connectivity calculation and matrix collapse gradient assessment. The channel connectivity and matrix collapse gradient generated by the game will be fed back to the indicator weight allocation system in real time, triggering the parameter update of the fuzzy hierarchical analysis model; the updated weights will remodulate the contribution ratio of each signal in the three-dimensional feature matrix, guiding the direction of the next round of risk field variation.

10. The method for determining seepage failure of a subgrade under repeated seepage as described in claim 9, characterized in that, Channel connectivity calculation quantifies the lateral connectivity efficiency of the skeleton channel by tracing the dominant streamlines in the virtual seepage field. The lateral connectivity efficiency is implicitly correlated with the fractal dimension transformation result. The matrix collapse gradient assessment is based on the channel wall state after fine particle mobility calibration, and calculates the local strength decay rate caused by particle loss. Its physical basis comes from the three-dimensional reconstruction data of pore structure.

Images