Underwater dam defect self-adaptive repairing method

Abstract

The application discloses an underwater dam defect self-adaptive repair method, which solves the problems of existing repair methods, such as solidification of a repair method process, dependence on experience for parameter setting, and lack of intelligent closed-loop feedback. The method comprises the following steps: acquiring multi-source detection data of a target defect area of an underwater dam, and extracting defect characteristic parameters; generating a customized repair scheme containing a repair process sequence and process parameters based on the defect characteristic parameters; performing a repair operation according to the customized repair scheme, and synchronously collecting process monitoring data; assessing the repair quality in real time based on the process monitoring data, and obtaining a repair quality conclusion; when the repair quality conclusion is unqualified, updating the defect characteristic parameters and the process parameters according to the unqualified degree and dimension, and re-performing the repair operation and quality assessment based on the updated parameters. The application improves the intelligent level and resource efficiency of underwater repair operation by constructing an adaptive closed loop of perception-planning-execution-evaluation-feedback.

Description

Technical Field

[0001] This invention relates to the field of underwater robot operation technology, and in particular to an adaptive repair method for defects in underwater dams. Background Technology

[0002] As a type of water conservancy infrastructure, the structural safety of underwater dams is crucial to flood control, water supply, and energy security. Due to prolonged exposure to high hydrostatic pressure and complex flow environments, defects such as cracks and erosion inevitably develop on the concrete surface and near-surface of the dam, potentially leading to seepage channels with safety hazards. Utilizing underwater rovers (ROVs) to perform online repairs of these defects, ensuring the long-term safe operation of the dam, is of significant engineering importance.

[0003] Currently, robotic repair operations for underwater dams have achieved a certain degree of automation. Existing technologies typically employ fixed, linear workflows, sequentially executing pre-set steps such as trenching, dredging, surface sealing, and drilling / grouting. Parameter settings largely rely on empirical formulas or fixed parameter tables derived from onshore engineering experience. For quality assessment, post-operation non-destructive testing methods such as ultrasonic echo are commonly used for inspection. This approach provides a fundamental solution for underwater repair operations and has been applied in some working conditions.

[0004] However, existing technologies still face several unresolved technical problems when dealing with complex and ever-changing underwater defect conditions. The core issue lies in the contradiction between the rigidity of the operational process and the individualized nature of the actual defects. For fully water-filled cracks underwater, grouting pressure calculation models based on land-based experience often fail to adequately consider the hindering effect of hydrostatic back pressure and the dilution effect caused by the mixing of the grout front with water, resulting in incomplete grouting or insufficient curing strength. Fixed repair procedures cannot adaptively adjust to the actual severity and type of defects, such as the presence of active leakage, potentially leading to overtreatment of minor defects or failure to repair urgent leakage defects due to a failure to prioritize water control. Furthermore, the open-loop quality assessment after completion is disconnected from the operational process. Once a substandard repair is detected, the system lacks an intelligent feedback mechanism to guide subsequent rework, often resulting in simple, repetitive operations that are inefficient and difficult to maintain stability. Summary of the Invention

[0005] The purpose of this invention is to provide an adaptive repair method for defects in underwater dams, in order to solve the aforementioned problems in the prior art.

[0006] Technical solution: An adaptive repair method for underwater dam defects, comprising:

[0007] Acquire multi-source detection data of the target defect area of ​​the underwater dam, and extract defect feature parameters based on the multi-source detection data;

[0008] Based on defect feature parameters, a customized repair scheme containing repair process sequence and process parameters is adaptively generated;

[0009] Repair work is carried out according to the customized repair plan, and process monitoring data of the repair work is collected simultaneously;

[0010] Based on process monitoring data, the repair quality is assessed in real time, and a repair quality conclusion is obtained;

[0011] Closed-loop feedback is implemented based on the repair quality conclusion: when the repair quality conclusion is unqualified, the defect characteristic parameters and process parameters are updated in a targeted manner according to the degree and dimension of unqualifiedness contained in the repair quality conclusion, and the repair operation and quality assessment are re-executed based on the updated parameters until the repair quality conclusion is qualified.

[0012] In an exemplary embodiment, the step of adaptively generating a customized repair scheme based on defect feature parameters specifically includes:

[0013] Based on the defect geometric parameters and activity status indicators included in the defect characteristic parameters, the comprehensive severity level of the defect is calculated.

[0014] Based on the comprehensive severity level and the defect type labels contained in the defect feature parameters, multiple preset repair process modules are adaptively trimmed and sorted to determine the repair process sequence; and based on the defect feature parameters and the requirements of each process module in the repair process sequence, the corresponding process parameters are calculated to form a customized repair solution.

[0015] In one exemplary embodiment, the step of calculating the overall severity level of the defect specifically includes:

[0016] The width, internal extension depth and surface extension length of the defect geometry parameters are normalized based on their respective reference values ​​and then weighted and summed to obtain the geometric severity index.

[0017] When the activity status index determines that the defect is in an active leakage state, an activity amplification factor greater than 1 is calculated based on the estimated leakage velocity in the index; otherwise, the activity amplification factor is set to 1.

[0018] The geometric severity index is multiplied by the activity amplification factor to obtain the comprehensive severity score, and the score is mapped to the comprehensive severity level according to the preset level threshold.

[0019] In one exemplary embodiment, the method further includes:

[0020] The process parameters are generated by using a pre-constructed coupled model of water-filled fracture slurry-water interface pressure difference driven permeation and frontal dilution.

[0021] The model couples the displacement permeation process of the slurry-water two-phase system under effective pressure difference, as well as the convection-diffusion dilution process between the slurry front and the water in the fracture.

[0022] The process parameters are calculated by using the geometric parameters in the defect feature parameters and the currently measured working water depth as inputs. The process parameters include grouting pressure or material injection volume.

[0023] In one exemplary embodiment, the step of assessing the repair quality further includes a post-repair quality assessment of the repaired area, specifically including:

[0024] A preset controlled micro-vibration disturbance linear frequency modulated Chirp signal is applied to the repaired area by an excitation source;

[0025] Vibration feedback from the repaired area is collected and converted into a spectral response.

[0026] The deep compactness of the repaired area is assessed based on the spectral response, and a conclusion on the repair quality is obtained.

[0027] In one exemplary embodiment, the deep compactness of the repaired region is assessed based on the spectral response, specifically by analyzing at least one of the following characteristic dimensions in the spectral response:

[0028] The deep compactness evaluation results of the repaired area are obtained by considering the main frequency offset, bandwidth ratio, higher harmonic energy, and attenuation slope.

[0029] In an exemplary embodiment, the step of providing closed-loop feedback based on the repair quality conclusion specifically includes:

[0030] The repair process ends when the repair quality is deemed satisfactory.

[0031] When the repair quality conclusion is that local reinforcement is required, the process parameters are fine-tuned for the unqualified dimensions, and a local reinforcement repair plan is generated using the fine-tuned process parameters, and the repair operation is performed again.

[0032] When the repair quality conclusion is that rework is required, the cause of non-conformity is fed back to the step of updating the defect feature parameters in a targeted manner, generating the updated defect feature parameters, and returning to the step of adaptively generating a customized repair plan.

[0033] In an exemplary embodiment, the step of updating the defect characteristic parameters when the repair quality conclusion is that local reinforcement or rework is required specifically includes:

[0034] The total amount of actual material injected as recorded in the process monitoring data is used as a material conservation constraint.

[0035] A pre-constructed coupling model of water-filled fracture slurry-water interface pressure difference driven by infiltration and frontal dilution was used as a forward model.

[0036] By using a reverse solution method and combining the constraint of material conservation, the residual spatial distribution of defects is derived, and updated defect characteristic parameters are generated based on the residual spatial distribution.

[0037] In one exemplary embodiment, the method further includes:

[0038] During the repair work, the working conditions data are monitored in real time by sensors mounted on the underwater work platform;

[0039] Based on operational data and a pre-set anomaly detection model, predict impending abnormal work interruptions.

[0040] When an abnormal interruption is anticipated, a blocking strategy is implemented for the non-compliant dimensions to safely interrupt the repair operation.

[0041] In one exemplary embodiment, a pre-constructed coupling model of water-filled fracture slurry-water interface pressure difference-driven permeation and frontal dilution was obtained offline as follows:

[0042] A finite element numerical model of water-filled cracks was established, and the slurry-water replacement process under different working conditions was simulated to obtain simulation data; physical model experiments were conducted in a simulated underwater environment to obtain experimental data.

[0043] Based on the obtained simulation or experimental data, the key physical parameters of the coupled model are calibrated.

[0044] Beneficial effects: This invention improves the intelligence level and resource efficiency of underwater remediation operations by constructing an adaptive closed loop of perception-planning-execution-evaluation-feedback. To address the failure of grouting models under underwater conditions, this invention proposes a coupled model based on physical principles. This model no longer relies on onshore experience but models the grout-water two-phase replacement process in underwater water-filled cracks, coupling the hindering effect of hydrostatic back pressure and the dilution effect of the grout front as variables in the calculation. In this way, the grouting pressure and time required to ensure grout penetration depth and solidification strength can be determined, thus improving the efficiency of underwater grouting operations. Attached Figure Description

[0045] Figure 1 This is the overall framework and workflow diagram of the adaptive repair method for underwater dam defects in the embodiments of this application.

[0046] Figure 2This is a flowchart illustrating the steps involved in generating a customized repair scheme that includes a repair process sequence and process parameters in this embodiment of the application.

[0047] Figure 3 This is a flowchart illustrating the steps for assessing repair quality in an embodiment of this application. Detailed Implementation

[0048] Example 1: This example details the overall framework and workflow of an adaptive repair method for underwater dam defects provided by the present invention, such as... Figure 1 As shown.

[0049] It is understood that the method of this embodiment of the invention can be executed by an underwater robot operation system. This system preferably includes an underwater robot platform with high-precision navigation and positioning and autonomous adsorption and anchoring functions. The platform is equipped with a multi-functional robotic arm and an integrated multi-source detection module. Specifically, the multi-source detection module may include a high-definition camera for acquiring optical images of the dam surface, a scanning sonar for acquiring three-dimensional geometric shapes, an ultrasonic probe for detecting the internal structural state, and a seepage pressure sensor for monitoring the hydraulic state.

[0050] For ease of understanding, this embodiment and subsequent embodiments will be described using a specific repair scenario. For example, the repair target is a structural crack 30m underwater in a hydropower station dam. Preliminary detection shows that the crack has a surface width of about 5mm and extends to a depth of about 80mm into the dam body, indicating continuous active leakage.

[0051] Optionally, multi-source detection data of the target defect area of ​​the underwater dam can be obtained, and defect feature parameters can be extracted based on the multi-source detection data.

[0052] In this embodiment, the underwater robot platform navigates to the target area where the aforementioned crack is located and completes adsorption and anchoring. The platform's multi-source detection module scans the crack and its surrounding area to acquire multi-source detection data. Specifically, the multi-source detection data may include: a sequence of optical images of the dam surface captured by a high-definition camera, three-dimensional point cloud data acquired by scanning sonar, reflected echo signals collected by ultrasonic probes at multiple measuring points, and seepage pressure time-series data recorded by seepage pressure sensors. Based on this, the system's central control unit performs fusion analysis on the acquired multi-source detection data to extract defect feature parameters. These defect feature parameters are structured data used to quantitatively describe the state of the defect; for example, they may include a defect ID, defect type label (e.g., crack), geometric parameters (e.g., width, depth, and extension length), and activity status indicators (e.g., the presence of active seepage and its calculated estimated velocity).

[0053] Optionally, based on defect characteristic parameters, a customized repair scheme containing a repair process sequence and process parameters can be adaptively generated.

[0054] In this embodiment, the central control unit receives defect characteristic parameters as input. Unlike methods using a fixed process, this system generates a customized repair plan for the current defect through a series of analyses and calculations. This generation process is adaptive; based on the severity and type of the defect reflected by the defect characteristic parameters, it selects modules from a pre-set repair process module library and determines the execution order, forming a repair process sequence. Simultaneously, the system calculates the control parameters required for each process module based on the specific geometric dimensions of the defect and the operating environment parameters, forming process parameters. The final output customized repair plan is a combination of the repair process sequence and process parameters.

[0055] Optionally, repair work can be performed according to a customized repair plan, and process monitoring data of the repair work can be collected simultaneously.

[0056] In this embodiment, the underwater robotic platform's robotic arm automatically changes and operates different tools, such as grooving tools, high-pressure water guns, grouting injection heads, and sealing material applicators, according to a customized repair plan, performing repair operations sequentially. Throughout the entire operation, sensors installed on the tools or platform synchronously collect process monitoring data. This process monitoring data reflects the physical execution status of the operation in real time, and may include, for example, the time-series curves of pressure and flow rate during the grouting stage, the total amount of material injected and back pressure response during the sealing stage, and real-time readings of pressure sensors around the repair area.

[0057] Optionally, the repair quality can be assessed in real time based on process monitoring data to obtain a repair quality conclusion.

[0058] In this embodiment, the system compares and analyzes the real-time collected process monitoring data with the preset target values ​​of process parameters in the customized repair plan to evaluate the quality of the repair work in real time. For example, the system will determine whether the actual grouting pressure curve is within the preset upper and lower limits, and whether the actual injected material volume has reached the theoretical calculation value. Through this comprehensive evaluation based on process data, the system can obtain a preliminary conclusion on the current repair effect, i.e., a repair quality conclusion. This conclusion is the evaluation result, which can be specifically divided into three levels: qualified, requiring local reinforcement, or requiring rework.

[0059] Optionally, a closed-loop feedback is implemented based on the repair quality conclusion: when the repair quality conclusion is unqualified, the defect characteristic parameters and process parameters are updated in a targeted manner according to the degree and dimension of unqualifiedness contained in the repair quality conclusion, and the repair operation and quality assessment are re-executed based on the updated parameters until the repair quality conclusion is qualified.

[0060] Accordingly, the repair process ends when the repair quality conclusion is satisfactory. When the conclusion is unsatisfactory, the system enters a closed-loop feedback path. Specifically, if the degree of non-compliance is low, such as a repair quality conclusion requiring local reinforcement, the system may only make minor adjustments to the process parameters, such as increasing the injection amount of sealing material in the next round, and re-execute the corresponding repair work based on the updated parameters. If the degree of non-compliance is high, such as a repair quality conclusion requiring rework, the system may focus on updating the defect characteristic parameters, typically triggering a complete repair scheme generation process, i.e., returning to execution. This iterative process will continue until the latest assessment yields a satisfactory repair quality conclusion, ensuring the final repair effect.

[0061] Closed-loop feedback also includes iterative convergence control: updating the closed-loop iteration count counter each time the repair operation and quality assessment are re-executed;

[0062] When the closed-loop iteration counter reaches the preset maximum allowed number of iterations and the repair quality conclusion is still unqualified, the closed-loop feedback is terminated, the current defect is marked as repair non-convergence state, and the repair quality index sequence of each iteration, customized repair plan and repair execution file are summarized and output as non-convergence defect report.

[0063] Example 2: This example details how a customized repair scheme, including a repair process sequence and process parameters, is adaptively generated based on defect feature parameters. Figure 2 As shown, the system will calculate and classify defects, and dynamically reorganize the repair process sequence based on the classification results.

[0064] Optionally, the overall severity level of the defect can be calculated based on the defect geometric parameters and activity status indicators included in the defect feature parameters.

[0065] This embodiment establishes a calculable and reproducible computational evaluation model to replace qualitative judgments relying on human experience. The model assesses the geometric shape of the defect, corrects it based on its hydraulic activity, and ultimately outputs a level that reflects its engineering hazard.

[0066] Specifically, the calculation process may include the following sub-steps:

[0067] The geometric severity index is calculated, which aims to assess the structural risk posed by the static physical dimensions of a defect.

[0068] When the activity status index determines that the defect is in an active leakage state, an activity amplification factor greater than 1 is calculated based on the estimated leakage velocity in the index; otherwise, the activity amplification factor is set to 1.

[0069] The geometric severity index is multiplied by the activity amplification factor to obtain the comprehensive severity score, and the score is mapped to the comprehensive severity level according to the preset level threshold.

[0070] Specifically, the system extracts the defect width W, internal extension depth D, and surface extension length L from the defect feature parameters, and uses a set of preset reference values ​​W ref D ref L ref Normalization is performed. Preferably, the reference benchmark value can be set according to the design specifications or historical safety thresholds of the dam body. Different weighting coefficients w are assigned based on the importance of different geometric dimensions to structural safety. w w d w l The geometric severity index I is obtained by weighted summation. geom The sum of the weighting coefficients is 1. This calculation process can be expressed by the following formula:

[0071] I geom =w w *(W / W ref )+w d *(D / D ref )+w l *(L / L ref );

[0072] Among them, I geom The geometric severity index is represented by W, D, and L, which are the defect width, internal extension depth, and surface extension length, respectively. ref D ref L ref For the corresponding reference base value, w w w d w l These are the corresponding weighting coefficients.

[0073] Accordingly, an activity amplification factor is calculated. This factor is introduced because active seepage defects, even if small in size, pose a far greater long-term threat to dam safety than static defects of the same size. Therefore, their severity needs to be dynamically amplified. When the activity status index determines that the defect is inactive, the activity amplification factor A is... act The value is set to 1, and no correction is made for geometric severity. When an active leakage condition is determined, the system extracts the estimated leakage velocity v from the indicators and calculates an amplification factor greater than 1 based on this velocity. This calculation process can be expressed by the following formula:

[0074] A act =1+α*(v / v ref );

[0075] Among them, A act Here, α is the activity amplification factor, v is the estimated leakage velocity, and v is the activity amplification coefficient. ref This is a reference flow rate baseline value.

[0076] Further, a comprehensive severity score is calculated and mapped to a level. The system multiplies the geometric severity index obtained in the first two steps by the activity amplification factor to obtain the final comprehensive severity score S.

[0077] S=I geom *A act ;

[0078] Where S is the overall severity score, and I geom A is a geometric severity index. act This is the activity amplification factor.

[0079] Based on this, the system compares the calculated score S with a set of preset, incremental level thresholds Th1 and Th2, mapping the defect to a comprehensive severity level. For example, if S is less than or equal to Th1, the level is minor surface damage; if S is between Th1 and Th2, the level is moderate structural damage; and if S is greater than Th2, the level is severe structural damage.

[0080] As an example, continuing with the scenario of the above embodiment (W=5mm, D=80mm, L=1.5m, v=0.1L / s), and setting the parameters as follows: W ref =10mm,D ref =100mm,L ref =2m;w w =0.4,w d =0.5,w l =0.1; α=1.5, v ref =0.05L / s.

[0081] Then we can calculate:

[0082] I geom =0.4*(5 / 10)+0.5*(80 / 100)+0.1*(1.5 / 2)=0.2+0.4+0.075=0.675.

[0083] A act =1+1.5*(0.1 / 0.05)=1+3=4.

[0084] S = 0.675 * 4 = 2.7.

[0085] If the preset level thresholds Th1=1.0 and Th2=2.0, then since S=2.7 is greater than Th2, the defect is judged to be of the severe structural damage level.

[0086] Optionally, based on the comprehensive severity level and the defect type label contained in the defect feature parameters, multiple preset repair process modules are adaptively trimmed and sorted to determine the repair process sequence; and based on the defect feature parameters and the requirements of each process module in the repair process sequence, the corresponding process parameters are calculated to form a customized repair solution.

[0087] This embodiment uses a pre-defined rule base to convert the calculation level output by the previous embodiment into a specific, executable sequence of work steps. The pre-defined repair process modules are standardized, independently callable functional units, which may include, for example: a grooving and shaping module, a dredging and cleaning module, a surface sealing module, a drilling and grouting module, an emergency leak sealing module, and a surface protective coating module.

[0088] In this embodiment, since the calculated level is a severe structural damage level, the system will select and sort the above modules according to preset rules. A preferred implementation is that, for the severe structural damage level, the system will activate the grooving and shaping module, the dredging and cleaning module, the drilling and grouting module, and the surface sealing module.

[0089] The steps for adaptively tailoring and sorting multiple preset repair process modules based on the comprehensive severity level and defect type label include:

[0090] When the overall severity level is severe structural damage, the drilling and grouting module should be executed before the surface sealing module in the repair process sequence.

[0091] When the overall severity level is moderate structural damage, the surface sealing module should be executed before the borehole grouting module in the repair process sequence.

[0092] When the overall severity level is classified as minor on the surface, the borehole grouting module is removed from the repair process sequence.

[0093] In particular, the sequencing logic of this invention for this level demonstrates the advancement of its technical concept. For deep and severe defects, if surface sealing is performed first, the sealing layer will hinder the subsequent penetration of grouting material into the depth of the defect, which may result in unfilled cavities inside, creating only a false appearance of surface healing.

[0094] Preferably, the execution order of the drilling and grouting modules is reversed and placed before the surface sealing modules. Therefore, the final repair process sequence is: first, groove shaping; second, dredging and cleaning; third, drilling and grouting; fourth, surface sealing. This sequence ensures that the deep structure is reinforced first, and the surface sealing then protects the grouting material, improving the long-term reliability of the repair.

[0095] In contrast, in some alternative implementations, if the defect is determined to be of a moderate structural damage level, the system-generated repair process sequence will place the surface sealing module before the borehole grouting module, which is the conventional processing order. If the defect is determined to be of a minor surface level, such as a shallow microcrack, the system will exclude the grooving and shaping module and the borehole grouting module, activating only the dredging and surface sealing modules to avoid over-treatment of minor damage.

[0096] Furthermore, this invention also considers a special case. In the severity assessment, there exists a special active leakage urgency level. In addition to the overall severity score reaching a certain extremely high threshold, the determination criteria for this level also include a necessary condition: an activity amplification factor A. act The defect exceeds the preset emergency leakage threshold. When a defect is determined to be at this level, it poses the highest threat to dam safety, and the primary task at this time is water control. Therefore, the system will prioritize activating the emergency plugging module at the very beginning of the repair process sequence, using a quick-setting material to rapidly cut off the active water flow channel. After the emergency plugging operation is completed, since the activity state of the defect has been changed, i.e., the leakage velocity has decreased or reached zero, the system will use sensors to re-acquire the activity state index of the defect and re-execute the assessment process to obtain a new comprehensive severity level reflecting the post-plugging state. Based on the new level, the subsequent repair module sequence will continue to be executed. This dynamic reassessment mechanism ensures that subsequent repair operations are carried out under controlled, non-active leakage conditions, guaranteeing the curing effect of the repair material and the final repair quality.

[0097] When the overall severity level is classified as an emergency level for active leakage, the process for determining the repair procedure sequence also includes:

[0098] After the emergency leak-sealing module is completed, the completed leak-sealing surface is observed in real time to determine whether the active seepage has been effectively blocked. If it is determined that the active seepage has been effectively blocked, the comprehensive severity level of subsequent repairs is re-determined based on the geometric severity index of the defect. The geometric severity index is a pure geometric score after removing the activity amplification factor. The corresponding combination of subsequent process modules and execution order are selected based on the re-determined level. If it is determined that the active seepage has not been effectively blocked and there is still a reserve of leak-sealing material, the emergency leak-sealing module is executed again with the adjusted injection parameters, and real-time observation is performed again.

[0099] Example 3 describes the determination of process parameters in a customized repair solution. This example will describe two implementation methods for calculating process parameters: one is a basic implementation method based on empirical formulas, and the other is a preferred implementation method based on a physical model.

[0100] Understandably, different process modules in the repair process sequence require the calculation of different process parameters. For example, the grooving and shaping module needs to determine the depth and width of the grooving; the drilling and grouting module needs to determine the pressure range of the grouting; and the surface sealing module needs to determine the total amount of material to be injected.

[0101] As a basic implementation method, process parameters can be calculated using a set of preset formulas based on engineering experience. Specifically, this method can include:

[0102] When the repair process sequence includes a grooving and shaping module, the system calculates the geometry of the grooving based on the internal extension depth D and defect width W from the defect characteristic parameters. For example, the grooving depth D... cut The internal extension depth D can be determined by multiplying it by a preset depth extension factor (e.g., 1.2) and adding a basic cutting allowance (e.g., 10mm) to ensure a clean cut into the structural layer. The groove width W... cut Therefore, it can be determined by extending a lateral allowance to both sides based on the defect width W.

[0103] When the repair process sequence includes a borehole grouting module, the system will calculate the hydrostatic pressure P based on the current operating water depth H. static The lower limit P of grouting pressure g_low It can be set to static pressure P static The sum of the minimum effective injection pressure difference. The upper limit P of the grouting pressure. g_high The lower limit can be determined by adding the defect depth-to-width ratio, i.e., the additional margin related to D / W, to prevent excessive pressure from causing splitting damage to the surrounding structure.

[0104] However, the aforementioned empirical formula-based methods have limited accuracy and reliability when facing complex underwater environments, especially deep water and turbulent water flow conditions. Therefore, a calculation method based on a physical model is preferred, particularly when calculating the process parameters of the borehole grouting module.

[0105] Optionally, the step of generating process parameters is to use a pre-constructed coupled model of water-filled crack grout-water interface pressure difference-driven infiltration and front-dilution. The model couples the displacement infiltration process of the grout-water two phases under effective pressure difference, as well as the convection-diffusion dilution process of the grout front and the water in the crack. The process parameters are calculated using the geometric parameters in the defect characteristic parameters and the currently measured working water depth as inputs. The process parameters include grouting pressure or material injection volume.

[0106] This embodiment models the grouting process in the specific physical scenario of underwater water-filled cracks using first-principles calculations. Many grouting models assume the crack interior is hollow or contains only air, which does not reflect the physical reality of underwater cracks filled with water and subject to hydrostatic back pressure. The coupled model constructed in this embodiment solves two problems: first, the grout penetrates into the narrow crack under the resistance of water back pressure, i.e., the displacement infiltration process; second, the grout front mixes with water inside the crack during its advancement, leading to a decrease in concentration and affecting the final curing strength, i.e., the front dilution process.

[0107] The establishment and calculation process of the coupling model, specifically, simplifies the defect into a flat channel defined by two parallel walls, uses the defect width in the defect geometry as the channel spacing, and uses the internal extension depth as the channel depth direction scale.

[0108] The effective driving pressure difference is defined as the difference between the grouting pressure and the hydrostatic back pressure of the water body in the crack, where the hydrostatic back pressure is calculated from the working water depth, water density, and gravitational acceleration.

[0109] Based on the modified Poiseuille flow relationship of the slurry section in the flat plate channel, a slurry front propulsion rate equation is established, which makes the propulsion rate proportional to the effective driving pressure difference and the square of the channel spacing, and inversely proportional to the equivalent apparent viscosity of the slurry and the distance already advanced.

[0110] By integrating the propulsion rate equation over time, the evolution of the slurry front position over time is obtained. The square of the front position is directly proportional to the effective driving pressure difference and time, and inversely proportional to the equivalent apparent viscosity.

[0111] At the slurry front position, based on the steady-state solution of the one-dimensional convection-diffusion equation, the concentration distribution of the dilution transition zone, in which the slurry volume concentration continuously decreases from the pure slurry end to the pure water end, is described by a complementary error function. The characteristic length of the dilution transition zone is proportional to the square root of the ratio of the effective diffusion-dispersion coefficient to the front advancement rate.

[0112] Accordingly, the process of calculating the target grouting pressure and the required injection time specifically includes:

[0113] Set a minimum effective concentration threshold, which is the minimum volume concentration of grout that meets the repair requirements after the grout body has cured.

[0114] The effective grouting depth is defined as the position of the grout front minus the effective depth reduction caused by the front dilution. The effective depth reduction is determined by the spatial offset distance corresponding to the concentration distribution in the dilution transition zone at the point where the concentration equals the lowest effective concentration threshold.

[0115] Using the internal extension depth in the defect geometry parameters as the target effective grouting depth, and combining the time evolution relationship of the vane position with the calculation relationship of the effective depth reduction, a set of constraint equations regarding grouting pressure and injection duration is established.

[0116] Solve the constraint equations within the feasible range of grouting pressure. The lower bound of the feasible range is the hydrostatic back pressure, and the upper bound is the upper limit of the splitting pressure determined based on the strength of the structural materials.

[0117] The lowest feasible grouting pressure and its corresponding shortest injection duration that satisfy the target effective grouting depth are obtained by solving the problem and are output as the target grouting pressure and the required injection time.

[0118] Optionally, in the coupled model, the square of the grout front penetration depth is directly proportional to the effective driving pressure difference formed by the grouting pressure and hydrostatic back pressure, and inversely proportional to the equivalent apparent viscosity of the injected grout; the model is based on a one-dimensional convection-diffusion equation to calculate the concentration distribution at the grout front and evaluate the front dilution effect.

[0119] Specifically, the first part of the coupled model, namely the displacement permeation process, can be described by a relationship derived from hydrodynamics. It is understandable that, under the assumption of laminar flow between smooth parallel plates, the propagation depth of the slurry front is related to several physical quantities. The calculation formula is as follows:

[0120] L p (t) 2 =(k*ΔP eff *t*W 2 ) / μ eq ;

[0121] Among them, L p (t) represents the grout front penetration depth as a function of time, k is the permeability coefficient, which is related to fluid properties and fracture wall roughness, ΔP eff For effective driving pressure difference, t is the injection time, W is the defect width, and μ eq The equivalent apparent viscosity of the injected slurry. Effective driving pressure difference ΔP eff That is, the externally applied grouting pressure P g The hydrostatic pressure and back pressure P generated by the operating water depth H static The difference. This formula clarifies the effective driving pressure difference required to achieve the target penetration depth, i.e., the internal extension depth D of the defect, within a given time period. The target grouting pressure P that needs to be applied can be calculated from this. g .

[0122] The second part of the coupled model, the front dilution process, addresses the issue of slurry solidification strength. The model simplifies the slurry front into a one-dimensional problem and utilizes the classic one-dimensional convection-diffusion equation to describe the concentration transition zone between the slurry and the water body. By solving this equation, for example using a complementary error function as its analytical solution, the concentration distribution curve at the slurry front's advance to any depth at any given time can be calculated.

[0123] By coupling these two processes, the system can solve for the parameters. Continuing with the scenario from the above embodiment (D=80mm, W=5mm, operating water depth H=30m), and assuming the pre-configured slurry parameters μ... eq Given that the system's target constraints are: firstly, within the set injection time, the slurry front penetration depth L... p (t) must be greater than or equal to the internal extension depth D, i.e., 80mm; secondly, in L p When (t) reaches 80mm, the grout concentration at its very tip must be higher than the preset minimum concentration threshold (e.g., 85%) to ensure normal material curing. The system iteratively solves the problem using the aforementioned coupled model, ultimately obtaining a set of optimal process parameter combinations, such as the target grouting pressure P. g The required injection time is 35 seconds, with a maximum pressure of 1.2 MPa.

[0124] Optionally, the pre-constructed coupled model of water-filled fracture slurry-water interface pressure difference driven permeation and frontal dilution is obtained offline by: establishing a finite element numerical model of the water-filled fracture and simulating the slurry-water replacement process under different working conditions to obtain simulation data; conducting physical model experiments in a simulated underwater environment to obtain experimental data; and calibrating the key physical parameters of the coupled model based on the obtained simulation data or experimental data.

[0125] Understandably, some key parameters in the aforementioned coupled model, such as the permeability coefficient k, are difficult to obtain through pure theoretical derivation and require calibration through experiments or simulations. In a preferred embodiment, before the project begins, a three-dimensional finite element numerical model incorporating the water body and the cracked dam structure is established. Computational fluid dynamics (CFD) software is then used to numerically simulate the grout-water replacement process under different pressures and crack widths, obtaining simulation data. Alternatively, a physical model can be built in a high-pressure water environment in a laboratory to conduct actual grouting experiments and obtain experimental data. By fitting the aforementioned relationships with this type of simulation or experimental data, the key physical parameters in the model can be calibrated, completing the pre-construction of the coupled model. The completed model will be stored in the system's control unit for online real-time retrieval.

[0126] Example 4: This example details how to assess repair quality in real time based on process monitoring data and arrive at a repair quality conclusion. It should be noted that the assessment based on process monitoring data is a preliminary, process-oriented quality judgment. After the repair work is completed, the system will perform a more comprehensive quality assessment. This example will describe a comprehensive assessment method with two parallel dimensions, and for deep compactness assessment, will introduce a basic implementation method and a preferred implementation method.

[0127] In one embodiment of the present invention, the assessment of repair quality is multi-dimensional, preferably including at least two dimensions: surface integrity assessment and deep compactness assessment.

[0128] The surface integrity assessment aims to calculate the surface smoothness of the repaired area and its connection with the surrounding intact dam surface. Specifically, after the repair work is completed, the underwater robot platform can use its onboard 3D scanning sonar to perform a high-precision scan of the repaired area, obtaining the repaired 3D point cloud data. The system then registers and compares this point cloud data with pre-stored original point cloud data of the area before repair or point cloud data of adjacent intact dam surfaces. A preferred calculation index is the root mean square elevation deviation of the point cloud in the repaired area. The smaller the deviation value, the smoother the repaired surface, the higher the fit with the original dam surface, and the better the surface integrity. The deep compactness assessment detects whether there are invisible defects such as voids, debonding, and material non-compactness inside the repair body.

[0129] As a fundamental approach, deep compactness assessment can be performed using the conventional ultrasonic pulse-echo method. Specifically, the robotic arm of an underwater robot platform manipulates an ultrasonic probe to perform point-by-point detection along a pre-defined grid path on the surface of the repair area. At each measurement point, the probe emits a high-frequency ultrasonic pulse into the repair and receives the echo signals reflected back from different internal interfaces. By analyzing the arrival time of the echo signals, the system can infer the thickness of the repair layer; by analyzing the peak amplitude attenuation of the echo signals, it can assess the compactness of the internal medium. For example, the system can compare the peak echo value after repair with the original peak echo value before repair and calculate the attenuation ratio. A larger attenuation ratio usually indicates good internal material filling and effective absorption of acoustic energy; conversely, a smaller attenuation ratio or the appearance of abnormally strong reflection signals may indicate the presence of cavities or debonding interfaces within the repair.

[0130] However, conventional ultrasonic pulse-echo methods typically suffer from low signal-to-noise ratios in underwater environments, especially in the presence of noise interference such as water flow and suspended matter, which can affect detection accuracy and reliability. This embodiment preferably employs an active evaluation method based on controlled micro-vibration response.

[0131] Optionally, the steps for assessing the quality of the repair also include a post-repair quality assessment of the repaired area, such as... Figure 3 As shown, the specific steps include: applying a preset controlled micro-vibration disturbance linear frequency modulated Chirp signal to the area where the repair work has been completed through an excitation source; collecting the vibration feedback of the repair area and converting it into a spectral response.

[0132] In one optional implementation, a mechanical actuator mounted on an underwater work platform is used as the excitation source to press the surface of the repaired area in a contact manner. Based on the designed grouting depth of the grouting module and the designed longitudinal wave velocity of the grouting material in the customized repair scheme, the theoretical value of the first-order thickness resonance frequency of the repair layer is calculated. Using the theoretical value of the first-order thickness resonance frequency as a benchmark, the frequency coverage range of the Chirp signal is set so that the range covers at least the first three thickness resonance frequencies. After acquiring the vibration response time-domain signal of the repaired area, the response time-domain signal is cross-correlated with the known Chirp excitation signal to complete pulse compression and obtain the equivalent impact response signal with enhanced signal-to-noise ratio. The equivalent impact response signal is then subjected to spectral transformation to obtain the spectral response.

[0133] In this embodiment, the excitation source can preferably utilize the actuators built into the underwater robot platform, such as drilling tools held by the robotic arm or a specialized excitation head, without requiring additional large, specialized equipment. The controlled micro-vibration excitation chirp signal is a frequency-sweeping signal whose frequency varies linearly with time. Compared to impact pulse signals where energy is concentrated instantaneously, chirp signals disperse energy over a wider time and frequency range, exhibiting noise immunity. For example, the system can control the excitation source to apply a chirp signal with a small amplitude, linearly sweeping from 1 kHz to 10 kHz for a duration of 1 second, to the surface of the repair area. Simultaneously, vibration sensors (such as accelerometers) deployed near the platform or excitation source will synchronously collect vibration feedback signals from the repair area.

[0134] After obtaining the vibration feedback signal, the system performs signal processing, specifically pulse compression. Specifically, the system performs correlation operations between the acquired vibration feedback signal and a pre-stored copy of the chirp signal matched to the transmitted signal. According to signal processing principles, this operation can compress the wide-pulse chirp signal energy into a narrow, high peak value, suppressing random noise introduced during acquisition and improving the signal-to-noise ratio. Performing a Fourier transform on the time-domain signal after pulse compression processing yields a spectral response with peak values ​​and structural characteristics.

[0135] Furthermore, the measured peak position of the first-order thickness resonant frequency is located from the spectral response, and the relative deviation between the measured peak position and the theoretical value is calculated as the main resonant frequency offset. The offset reflects the change in equivalent stiffness caused by large-sized voids inside the grout body. The ratio of the bandwidth of the first-order resonant peak at the half-power point to the measured peak position is calculated as the resonant peak half-power bandwidth ratio. The bandwidth ratio reflects the change in vibration energy dissipation caused by debonding at the grout-substrate interface. The ratio of the signal energy in the high-frequency band above twice the first-order resonant frequency to the total signal energy of the entire frequency band is calculated as the higher-order harmonic energy ratio. The energy ratio reflects the selective excitation of higher-order modes by distributed micro-voids or dilution zones inside the grout body. The spectral amplitude is logarithmically linearly fitted in the high-frequency band, and the fitting slope is taken as the high-frequency attenuation slope. The attenuation slope reflects the scattering attenuation caused by the gradual change in acoustic impedance in the grout dilution zone. The main resonant frequency offset, resonant peak half-power bandwidth ratio, higher-order harmonic energy ratio, and high-frequency attenuation slope are normalized and then weighted and combined to obtain the deep compactness index.

[0136] Optionally, the deep compactness of the repaired area can be assessed based on the spectral response to obtain a conclusion on the repair quality.

[0137] The specific rules for determining whether a repair requires local reinforcement or rework in the repair quality assessment are as follows:

[0138] The quality indicators of each dimension involved in the repair quality assessment are compared with their respective compliance thresholds one by one. The number of dimensions that fail to meet the standards is counted, and the relative difference between the indicator value of each dimension that fails to meet the standards and its corresponding threshold is calculated. When only a single dimension fails to meet the standards and the relative difference of that dimension does not exceed the preset local reinforcement judgment margin, and no debonding or residual leakage abnormalities are detected, the repair quality conclusion is that local reinforcement is required. When two or more dimensions fail to meet the standards, or although only a single dimension fails to meet the standards, but the relative difference of that dimension exceeds the local reinforcement judgment margin, or debonding or residual leakage abnormalities are detected, the repair quality conclusion is that rework is required.

[0139] The closed-loop reentrancy paths that require local reinforcement and those that require rework differ in the scope of their steps:

[0140] When the repair quality conclusion is that local reinforcement is required, parameter correction is only performed on the single process module associated with the non-compliant dimension. The defect severity classification and repair process module trimming and sorting steps are not re-executed. The corrected parameters generate a local supplementary repair plan containing only a single process module and are then executed. When the repair quality conclusion is that rework is required, the updated residual defect feature parameters are re-executed from the defect severity classification to generate a complete adaptive repair plan, generating a new customized repair plan containing a complete repair process sequence and then executing it.

[0141] Optionally, the deep compactness of the repaired area can be evaluated based on the spectral response. Specifically, this is achieved by analyzing at least one of the following characteristic dimensions in the spectral response: dominant frequency offset, bandwidth ratio, higher-order harmonic energy, and attenuation slope, to obtain the evaluation result of the deep compactness of the repaired area.

[0142] In this embodiment, the system compares the spectral response with a pre-stored theoretical frequency response benchmark representing the intact structural state of the region, and infers internal defects by analyzing the differences. The theoretical frequency response benchmark can be pre-calculated by performing finite element modal analysis on the region, or obtained by actual measurement on adjacent, confirmed intact dam areas.

[0143] A preferred analytical dimension is the thickness resonance frequency shift. It is understood that a plate-like structure of a specific thickness, when subjected to vertical excitation, will resonate along its thickness at a specific frequency, which is related to its thickness and the sound velocity of the material. If the actual effective thickness of the repair layer matches the designed thickness, its resonant frequency should basically match the theoretical benchmark. If there is extensive debonding within the repair layer, resulting in a thinner actual load-bearing thickness, its resonant frequency will shift towards higher frequencies.

[0144] Another preferred analytical dimension is the higher harmonic energy ratio. When small-scale voids or cracks exist within the restoration, these defects undergo nonlinear opening and closing and collisions when subjected to vibration, exciting a series of higher harmonic components that are integer multiples of the fundamental frequency in the spectral response. By calculating the ratio of the total energy of the higher harmonics to the fundamental frequency energy, the presence of nonlinear defects can be identified. In a well-formed and dense structure, the higher harmonic energy ratio is typically very low.

[0145] In addition, characteristics such as the bandwidth ratio or attenuation slope of the spectrum can be analyzed to evaluate the damping properties of the material and infer its density.

[0146] Specifically, the system integrates quantitative indicators of surface integrity assessment with deep compactness evaluation indicators calculated based on one or more of the aforementioned spectral feature dimensions. Through a preset multi-parameter comprehensive judgment model, such as weighted scoring or logical decision tree, it outputs the final repair quality level conclusion, such as qualified, requiring local reinforcement, or requiring complete rework.

[0147] Example 5: This example details the steps of closed-loop feedback, takes the repair quality conclusion as an example, describes the three-level differentiated feedback path, and introduces a basic parameter update method and a preferred defect reconstruction method based on the reverse solution of the physical model for two unqualified situations.

[0148] Optionally, the closed-loop feedback step based on the repair quality conclusion specifically includes: when the repair quality conclusion is qualified, ending the repair process; when the repair quality conclusion is that local reinforcement is needed, making local fine-tuning of the process parameters for the unqualified dimensions, and using the fine-tuned process parameters to generate a local reinforcement repair plan, and executing the repair operation again; when the repair quality conclusion is that rework is needed, feeding back the reason for the unqualified to the step of targeted updating of the defect feature parameters, generating updated defect feature parameters, and returning to the step of adaptively generating a customized repair plan.

[0149] This embodiment adopts error correction strategies with different costs and complexities based on the degree and dimensions of failure determined by the evaluation.

[0150] When the output repair quality conclusion is qualified, the system determines that the repair operation is successful and the process terminates normally.

[0151] When the repair quality conclusion indicates the need for localized reinforcement, it typically corresponds to minor, localized repair defects. For example, a surface integrity assessment might reveal that the smoothness of the repair area's edge is slightly below the threshold, but the deep compactness meets the standard. In this case, the system does not need to perform a complete defect assessment and repair planning again. Furthermore, the system activates a simplified feedback path, making localized fine-tuning of the relevant process parameters only for the non-compliant dimension of surface integrity. For example, the system might generate a localized reinforcement repair plan that includes only a surface protection coating module and increases its coating thickness by 10%. The robot will then perform only the localized reinforcement work, completing the repair at the lowest possible cost.

[0152] When the repair quality conclusion indicates that a complete rework is required, it usually corresponds to some issues affecting the overall repair quality, such as deep unfilled cavities discovered through Chirp micro-vibration response. In this case, parameter fine-tuning is insufficient; a more accurate assessment of the defect's current state is necessary, followed by a complete re-planning of the repair, requiring the updating of defect characteristic parameters.

[0153] As a fundamental implementation method, especially in scenarios requiring local reinforcement, the step of updating defect feature parameters can be achieved through a scalar correction factor fine-tuning mechanism. Specifically, the system acquires pre-configured geometric attenuation factors for different defect types; for example, for crack repair, this factor can be set to 0.8. The system extracts the pre-repair geometric parameters, such as width W and depth D, from the original defect feature parameter set and performs a scalar multiplication operation on these parameters with the geometric attenuation factor to calculate the reduced surface residual size and internal residual depth. These reduced parameters will be used as updated defect feature parameters to guide the next round of local reinforcement operations.

[0154] In this embodiment, a defect reconstruction method based on reverse solving of a physical model is adopted.

[0155] Optionally, when the repair quality conclusion is that local reinforcement or rework is required, the step of updating the defect characteristic parameters specifically includes: using the actual total amount of material injected recorded in the process monitoring data as a material conservation constraint; using a pre-constructed water-filled crack slurry-water interface pressure difference-driven permeation and frontal dilution coupling model as a forward model; and using a reverse solution method, combined with the material conservation constraint, to inversely derive the residual spatial distribution of the defect, and generate updated defect characteristic parameters based on the residual spatial distribution.

[0156] Optionally, the residual spatial distribution of the defect can be inverted by solving in reverse. Specifically, during the solution process, the actual total amount of material injected recorded in the process monitoring data is used as the material conservation volume constraint, and the deep compactness spatial distribution data obtained based on the repair quality assessment is used as the physical state boundary condition.

[0157] The process of inverting the residual spatial distribution through reverse engineering specifically includes: extracting the actual grouting pressure, actual injection duration, and actual total material injection from the actual execution records of the repair operation; substituting the actual grouting pressure and actual injection duration into the water-filled crack grout-water interface pressure difference-driven infiltration and frontal dilution coupling model, running in forward calculation mode, to obtain the predicted grout spatial concentration distribution and predicted total injection under ideal conditions; comparing the actual total material injection with the predicted total injection, and simultaneously comparing the deep compaction evaluation results in the repair quality conclusion with the compliance threshold, determining the spatial distribution pattern of residual defects based on the combination of the injection volume deviation direction and quality deviation type, the distribution pattern including at least one of the deep void concentration pattern, the distributed insufficient concentration pattern, and the interface debonding pattern; generating a virtual residual defect model carrying a pattern label based on the determined spatial distribution pattern, where the geometric parameters in the virtual residual defect model replace the corresponding parameter items in the original defect feature parameters.

[0158] Furthermore, when regenerating customized repair schemes using updated defect feature parameters, differentiated process parameter adjustment strategies are selected based on the mode label of the virtual residual defect model: when the mode label is the deep void concentration mode, the target grouting pressure is increased or the injection duration is extended when recalculating the grouting module process parameters to increase the effective grouting penetration depth; when the mode label is the distributed insufficient concentration mode, the water-cement ratio of the injected grout is reduced when recalculating the grouting module process parameters to improve the grout's anti-dilution ability; when the mode label is the interface debonding mode, a high-pressure flushing step is added before the grouting module in the regenerated repair process sequence to improve the bonding conditions between the grouting material and the substrate.

[0159] This embodiment uses a forward physical model, i.e. a coupled model, to describe the grouting process. The actual measurement results are used as constraints to invert the model and calculate the unknown internal state that leads to the current result, i.e. the morphology of the residual defect.

[0160] Specifically, after concluding that a complete rework is required, the system will initiate a reverse engineering process.

[0161] The system will construct a volume constraint equation based on the principle of matter conservation: the total volume of the original defects, minus the volume of the actually injected and effectively solidified material, should equal the volume of the residual defects, i.e., the cavity. The equation is as follows:

[0162] V residual_cavity =V initial_defect -V injected_effective ;

[0163] Among them, V residual_cavity Let V be the residual cavity volume to be solved. initial_defect The initial volume of the defect before repair can be calculated using the original defect characteristic parameters, V. injected_effective This represents the actual effective volume of injected material, which can be obtained from the total actual material injected by extracting process monitoring data.

[0164] Accordingly, the system utilizes deep density spatial distribution data as physical state boundary conditions. For example, the density assessment results at different locations obtained through Chirp micro-vibration response analysis can define the possible spatial range of residual cavities.

[0165] The system uses volume constraint equations as the core constraints and physical state boundary conditions as the solution range. It uses a coupled model as the forward model and performs inverse calculations on the three-dimensional topological morphology of residual defects through iterative inverse root-finding operator operations (such as the Gauss-Newton method or the Levenberg-Marquardt algorithm).

[0166] Optionally, the output of this process is not scalar parameters, but a three-dimensional virtual geometric model describing the shape, size, and spatial location of the residual cavity. The system extracts equivalent geometric parameters (such as the maximum width, depth, and length of the residual cavity) from this model and uses these parameters as updated defect feature parameters. This class of parameters, containing physical inversion information, will serve as input to the next round of repair planning and be returned for execution, achieving intelligent closed-loop replanning based on physical diagnosis.

[0167] Example 6: This example details a parallel abnormal interruption handling mechanism designed to address potential emergencies in complex underwater environments, ensuring system robustness and operational safety.

[0168] Understandably, underwater robot operations face numerous uncertainties, such as communication interruptions, hydraulic system malfunctions, and power source fluctuations, all of which can passively interrupt ongoing repair work. For certain process modules that require time continuity, such as grouting or sealing, interruptions may cause the injected material to initially set within the pipeline or defect, affecting subsequent operations.

[0169] Optionally, the method further includes: during the repair operation, monitoring the operation status data in real time through sensors mounted on the underwater operation platform; predicting the impending abnormal interruption conditions based on the operation status data and a preset anomaly judgment model; and when an abnormal interruption condition is predicted, implementing a blocking strategy for the non-compliant dimensions to safely interrupt the repair operation.

[0170] In this embodiment, the work status data is a multi-dimensional data stream that the system control unit collects in real time from each subsystem of the robot platform when performing repair work.

[0171] Preferably, the data may include: the outlet pressure and flow rate of the hydraulic pump, the motor current of the robotic arm joint, the outlet pressure of the grouting pump, the sensor reading of the remaining material reserves, and the quality of the communication link with the ground control station.

[0172] The preset anomaly detection model can be a rule-based logical model or a machine learning prediction model trained on historical data. In a preferred embodiment, the model includes a series of safety thresholds for different operating conditions. For example, the model continuously compares the real-time monitored hydraulic pump outlet pressure with a preset minimum operating pressure threshold. When the system detects that the pressure value is continuously decreasing and approaching the threshold, it can predict that an operation interruption caused by a power system failure is about to occur.

[0173] Upon anticipating an impending anomaly, the system will execute a shutdown strategy. This strategy does not halt all operations, but rather interrupts work in an orderly and recoverable manner. Specifically, the system will stop injecting new material into the defect, control the robotic arm to safely withdraw the work tools from the dam surface to a preset standby position, and record all critical status parameters at the moment of interruption. These critical status parameters may include: which repair process module is currently being executed, the progress of that module's work (such as the volume of material injected or the length of the cut), and the precise interruption timestamp. This approach maximizes equipment safety and preserves complete on-site information for subsequent recovery or rework decisions.

[0174] In this embodiment, the post-interruption processing logic will be described using a specific scenario. Assume that during the calculated grouting operation, the system anticipates an abnormal hydraulic power and safely interrupts the operation. At the time of interruption, according to process monitoring data, of the planned 2 liters of grout, 1.2 liters have been injected, representing a progress of 60%.

[0175] Upon interruption, the system will start an interruption timer and retrieve the pre-configured initial setting time threshold for the currently used slurry from the material property database. The initial setting time threshold refers to the time required for the slurry to lose its fluidity from the start of mixing; for example, it can be set to 300 seconds. The system will make different recovery decisions based on the relationship between the actual duration of the interruption and this threshold.

[0176] In scenario A, the interruption can be resumed before the material initially sets:

[0177] Assume the robot's hydraulic system recovers within 120 seconds of the interruption. Since 120 seconds is less than the preset 300-second initial setting time threshold, the system determines that the injected 1.2 liters of grout and the remaining grout in the pipeline still have good fluidity, and the operation can continue. At this point, the system will generate a mark indicating that some work is yet to be done. Upon resumption of execution, the system will deduct the completed portion from the original planned process parameters. Specifically, the system will adjust the subsequent grouting task volume to the originally planned 2 liters minus the injected 1.2 liters, i.e., 0.8 liters. The robot will reposition itself to the work point and only perform the remaining 0.8 liters of grouting.

[0178] Scenario B: The interruption time exceeds the initial setting time of the material.

[0179] Suppose the hydraulic system malfunction is severe, and the interruption lasts for 400 seconds before being restored. Since 400 seconds exceeds the preset 300-second initial setting time threshold, the system determines that the injected 1.2 liters of grout, as well as the grout remaining in the injection pipeline, has already begun to set, lost its fluidity, and formed a physical blockage. At this point, forcibly continuing grouting will not only fail to inject further but may also damage the equipment.

[0180] In this situation, the system will determine that the original repair plan can no longer be executed and generate a flag indicating an abnormal interruption requiring rework. This flag will trigger a comprehensive rework process. Specifically, the system will consider that the current state of the defect has changed: it is no longer the original water-filled crack, but a more geometrically complex defect blocked by partially solidified material. Therefore, the system will trigger an update of the defect characteristic parameters, for example, adding attributes describing the location and size of the physical blockage interface to its geometric parameters. The passively changed new defect characteristic parameters will be used as input and returned to the planning step to generate a completely new repair plan for the current complex situation. For example, the new plan may require first removing the solidified blockage with a drill bit before performing secondary grouting.

[0181] Specifically, the differentiated processing logic based on material curing time constraints can robustly address common abnormal interruption issues in underwater operations, ensuring the safety of the operation and the final repair quality.

[0182] Example 7: This example presents a complete closed-loop feedback process, supplemented with exit and documentation mechanisms to handle extreme cases. While the adaptive closed-loop repair method can handle most repair scenarios, in some complex defects, or even defects exceeding the current model's cognitive boundaries, the system may fall into a cycle of repeated repairs with unsatisfactory results.

[0183] To avoid resource waste and potential structural damage caused by infinite loops, this embodiment describes a closed-loop exit and intelligent documentation mechanism based on a preset maximum iteration limit. It is understood that this mechanism is a top-level monitoring and safeguard strategy operating outside the overall closed-loop process.

[0184] In a preferred embodiment of the present invention, the implementation of this mechanism includes the following steps:

[0185] Before starting a repair operation on a new defect, the system retrieves the pre-set maximum allowed rework re-entry threshold from its global configuration. This threshold is a circuit breaker parameter set for the entire closed-loop repair process, and its value can be set based on experience or risk assessment; for example, it can be set to 3 times. Simultaneously, the system initializes the repair round counter for the current defect, with an initial value of 0.

[0186] Each time the system enters the rework path due to unsatisfactory repair quality, that is, when the process of fully updating the defect feature parameters and regenerating a complete repair plan is executed, the system will increment the repair round counter by 1 while performing the operation.

[0187] The system will continuously compare the current value of the repair round counter with the maximum allowed rework re-entry threshold.

[0188] When the repair quality level consistently fails to meet the standards, and the repair round counter reaches the maximum allowed rework re-entry threshold, the system will terminate the automatic repair process for the current defect. This termination operation may result in a repair loop that cannot automatically converge upon voluntary exit.

[0189] After the process is terminated, the system will automatically execute a data packaging and archiving procedure. This procedure will retrieve and summarize historical data related to the current non-convergence defect from the system's runtime database and log system.

[0190] Specifically, the contents of the data package may include:

[0191] Before the initial repair, the original set of defect feature parameters is extracted through initial inspection. During each rework and re-entry process, residual defect feature parameters for all versions are generated through reverse engineering and other methods. A repair execution file corresponds to each repair operation, recording the repair process sequence used for each operation, the planned process parameters, and the actual process monitoring data. After each repair operation is completed, a repair quality assessment report is obtained using a multi-dimensional evaluation method, including surface integrity indicators and deep compactness indicators.

[0192] The system integrates and correlates the above data in a structured manner to form an anomaly diagnostic report library with non-convergence markers. This report library records the results of repair failures and traces the entire process from the initial diagnosis to each attempt and each failed diagnosis, providing data support for subsequent steps.

[0193] Accordingly, the anomaly diagnosis report library will be output to the offline expert system. By analyzing the archives containing the iteration history, the reasons for the failure of the automated system can be identified. For example, it may be due to a complex interconnected structure inside the defect that was not considered in the model, or the selected repair material being incompatible with the chemical environment of the defect. Based on the conclusions of this type of offline analysis, the system's model, algorithm, or material library can be fine-tuned, or special repair schemes can be developed to solve such repair problems.

[0194] Example 8: According to one aspect of this application, the method further includes:

[0195] At the operational level, the manned submersible serves as a crucial underwater command platform for mission implementation, combining information acquisition, positioning guidance, and local decision-making functions. Through wide-field observation windows, sonar scanning, and underwater ranging, it detects and spatially locates defects such as leaks, cracks, and pits in dams or hydraulic structures. Its positioning data is transmitted back to the shore-based control system in real time, serving as the basis for subsequent deep-water platform deployment and pinpoint operations. Furthermore, the manned submersible can collaborate with ROVs to form a flexible multi-platform underwater detection and guidance network.

[0196] After receiving the coordinates of the defect location, onshore personnel use lifting equipment (such as a crane and underwater support system) to lower the deep-water inspection and repair platform to the designated water area below. The platform is equipped with multiple high-powered thrusters, giving it strong underwater maneuverability and enabling it to navigate in complex hydrodynamic environments. Guided by a manned submersible and an ROV, the platform gradually approaches the target area and ultimately uses microneedle vortex suction cups to achieve adhesion and anchoring on the dam wall. This method solves the problem of underwater platforms struggling to operate stably in high-velocity or deep-water conditions.

[0197] When the deep-water operation platform is stably attached to the dam surface by the underwater vortex suction cup, it selectively grabs underwater equipment under the operation of its own robotic arm. The actual dam surface crack treatment process generally involves grooving, dredging, crack sealing, and drilling grouting. This completes the integrated commissioning scheme for deep-water detection and disaster damage control technologies.

[0198] Once the platform is stably anchored to the dam surface, its underwater robotic arm begins operation. The robotic arm can selectively grab and operate different tools according to the operational needs. The platform is equipped with four types of core small underwater operating equipment: (1) Underwater cutting equipment: used to grooving or shaping cracks or defects to create good interface conditions for subsequent repair work; (2) Underwater drilling and grouting equipment: capable of completing local drilling and injecting repair materials to achieve deep structural reinforcement; (3) Underwater sealing equipment: using flexible filling materials to seal surface cracks and holes, restoring the dam's seepage prevention and integrity; (4) Underwater dredging equipment: using high-speed water flow to flush and clean the attachments on the dam surface.

[0199] In the specific operation process, the repair of cracks and defects on the dam surface generally follows the technical process of cutting grooves, dredging, sealing cracks, and drilling and grouting: First, the cutting equipment processes the grooves at the cracks or pits to remove irregular surfaces; second, the platform uses high-pressure water flow or a sludge suction device to dredge the grooves and surrounding sediments to ensure a clean interface; third, the sealing equipment performs preliminary defect sealing to restore the integrity of the dam surface; finally, in areas requiring deep reinforcement, the drilling and grouting equipment completes drilling and injects grouting material to achieve deep structural reinforcement and seepage control.

[0200] Throughout the process, the ROV primarily undertakes the tasks of close-range auxiliary guidance and monitoring, ensuring that the deep-water operation platform can reach the target and carry out operations safely under complex conditions. Once the platform completes anchoring and operation commencement, its operational processes largely operate autonomously, while the shore-based control system is responsible for real-time status monitoring and emergency intervention.

[0201] Through this multi-platform, multi-equipment collaborative operation system, deep-water inspection, repair, and reinforcement of reservoir and dam systems can be completed efficiently and safely. This not only meets the needs for rapid response and handling of structural defects in complex underwater engineering projects but also provides strong technical support for the long-term safe and stable operation of reservoir and dam systems. Ultimately, an integrated deep-water inspection and disaster damage repair solution has been formed, encompassing underwater inspection, defect location, platform deployment, platform adsorption, operation implementation, and repair completion.

[0202] This invention introduces a defect calculation and grading mechanism with adaptive process reconfiguration. This mechanism uses a comprehensive scoring model that includes geometric dimensions and activity indicators to objectively and reproducibly calculate and rate the severity of defects. Based on the rating results, preset repair modules are tailored and prioritized. For example, a optimized sequence of grouting followed by sealing is used for severe defects, while emergency plugging is prioritized for active leakage defects. This ensures that the repair plan matches the specific condition of each defect, avoiding over-treatment or repair failures due to improper processes, and solving the problem of rigid repair processes that cannot adapt to individual defects.

[0203] Furthermore, an intelligent closed-loop feedback system was constructed. When the repair quality is unsatisfactory, the system does not simply repeat the operation. Instead, it utilizes the prior physical model, combined with process data such as the actual injection volume, to reconstruct the spatial morphology of the internal residual defects in three dimensions through reverse engineering. This diagnostic feedback provides target input for the next round of replanning, transforming rework from repetitive attempts into targeted repair driven by physical diagnosis, thereby improving repair efficiency and final quality.

[0204] It should be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the present invention will not describe the various possible combinations separately.

Claims

1. An adaptive repair method for defects in underwater dams, characterized in that, include: Acquire multi-source detection data of the target defect area of ​​the underwater dam, and extract defect feature parameters based on the multi-source detection data; Based on defect feature parameters, a customized repair scheme containing repair process sequence and process parameters is adaptively generated; Repair work is carried out according to the customized repair plan, and process monitoring data of the repair work is collected simultaneously; Based on process monitoring data, the repair quality is assessed in real time, and a repair quality conclusion is obtained; Closed-loop feedback is implemented based on the repair quality conclusion: when the repair quality conclusion is unqualified, the defect characteristic parameters and process parameters are updated in a targeted manner according to the degree and dimension of unqualifiedness contained in the repair quality conclusion, and the repair operation and quality assessment are re-executed based on the updated parameters until the repair quality conclusion is qualified.

2. The method according to claim 1, characterized in that, The steps for adaptively generating customized repair solutions based on defect feature parameters specifically include: Based on the defect geometric parameters and activity status indicators included in the defect characteristic parameters, the comprehensive severity level of the defect is calculated. Based on the comprehensive severity level and the defect type labels contained in the defect feature parameters, multiple preset repair process modules are adaptively trimmed and sorted to determine the repair process sequence; and based on the defect feature parameters and the requirements of each process module in the repair process sequence, the corresponding process parameters are calculated to form a customized repair solution.

3. The method according to claim 2, characterized in that, The steps for calculating the overall severity level of a defect specifically include: The width, internal extension depth and surface extension length of the defect geometry parameters are normalized based on their respective reference values ​​and then weighted and summed to obtain the geometric severity index. When the activity status index determines that the defect is in an active leakage state, an activity amplification factor greater than 1 is calculated based on the estimated leakage velocity in the index; otherwise, the activity amplification factor is set to 1. The geometric severity index is multiplied by the activity amplification factor to obtain the comprehensive severity score, and the score is mapped to the comprehensive severity level according to the preset level threshold.

4. The method according to claim 1, characterized in that, The method also includes: The process parameters are generated by using a pre-constructed coupled model of water-filled fracture slurry-water interface pressure difference-driven permeation and frontal dilution. The model couples the displacement permeation process of the slurry-water two-phase system under effective pressure difference, as well as the convection-diffusion dilution process between the slurry front and the water in the crack. The process parameters are calculated by using the geometric parameters in the defect feature parameters and the currently measured working water depth as inputs. The process parameters include grouting pressure or material injection volume.

5. The method according to claim 1, characterized in that, The steps for assessing the quality of restoration also include a post-restoration quality assessment of the restored area, specifically including: A preset controlled micro-vibration disturbance linear frequency modulated Chirp signal is applied to the repaired area using an excitation source; Vibration feedback from the repaired area is collected and converted into a spectral response. The deep compactness of the repaired area is assessed based on the spectral response, and a conclusion on the repair quality is obtained.

6. The method according to claim 5, characterized in that, The assessment of the deep compactness of the repaired area based on the spectral response is specifically achieved by analyzing at least one of the following characteristic dimensions in the spectral response: The deep compactness evaluation results of the repaired area are obtained by considering the main frequency offset, bandwidth ratio, higher harmonic energy, and attenuation slope.

7. The method according to claim 1, characterized in that, The steps for closed-loop feedback based on repair quality conclusions specifically include: The repair process ends when the repair quality is deemed satisfactory. When the repair quality conclusion is that local reinforcement is required, the process parameters are fine-tuned for the unqualified dimensions, and a local reinforcement repair plan is generated using the fine-tuned process parameters, and the repair operation is performed again. When the repair quality conclusion is that rework is required, the cause of non-conformity is fed back to the step of updating the defect feature parameters in a targeted manner, generating the updated defect feature parameters, and returning to the step of adaptively generating a customized repair plan.

8. The method according to claim 7, characterized in that, When the repair quality conclusion indicates that local reinforcement or rework is required, the steps for updating the defect characteristic parameters specifically include: The total amount of actual material injected as recorded in the process monitoring data is used as a material conservation constraint. A pre-constructed coupling model of water-filled fracture slurry-water interface pressure difference driven by infiltration and frontal dilution was used as a forward model. By using a reverse solution method and combining the constraint of material conservation, the residual spatial distribution of defects is derived, and updated defect characteristic parameters are generated based on the residual spatial distribution.

9. The method according to claim 1, characterized in that, The method also includes: During the repair work, the working conditions data are monitored in real time by sensors mounted on the underwater work platform; Based on operational data and a pre-set anomaly detection model, predict impending abnormal work interruptions. When an abnormal interruption is anticipated, a blocking strategy is implemented for the non-compliant dimensions to safely interrupt the repair operation.

10. The method according to claim 4, characterized in that, The pre-constructed coupled model of water-filled fracture slurry-water interface pressure difference-driven infiltration and frontal dilution was obtained offline as follows: A finite element numerical model of water-filled cracks was established, and the slurry-water replacement process under different working conditions was simulated to obtain simulation data; physical model experiments were conducted in a simulated underwater environment to obtain experimental data. Based on the obtained simulation or experimental data, the key physical parameters of the coupled model are calibrated.

Images