Health assessment method for long-distance underground culvert structure

Abstract

The invention discloses a health assessment method for a long-distance underground culvert structure, and the method comprises the steps: carrying out the fusion processing based on acoustic and optical detection data, so as to recognize a structure defect; based on underwater rebound and coring test data, carrying out underwater-dry conversion and correction on a strength value, and determining corrected structural strength and material degradation parameters; in combination with defect, degradation and historical monitoring data, carrying out structure evolution modeling and mutation risk identification to obtain evolution characteristics and mutation risks of the structure; and integrating the corrected structural strength, evolution characteristics and abrupt change risks, performing full-field state assessment, and generating a structural health assessment and risk early warning result. According to the method, accurate, dynamic and full-field evaluation of the health state of the culvert can be realized, and the reliability and early warning capability of an evaluation result are improved.

Description

Technical Field

[0001] The present invention belongs to the field of civil engineering and structural health monitoring, and in particular to a method for assessing the structural health of a long-distance underground culvert. Background Art

[0002] Long-distance underground culvert structures are usually buried deep underground, crossing complex geological and environmental units such as rivers and highways, and serve for long periods of time under harsh conditions of flooded, high-velocity, and sand-entrained water. Factors such as the accumulation of hydraulic erosion, material aging, and geological changes over time inevitably lead to defects such as concrete cracking, material degradation, failure of joint sealing, soil leakage around the structure, and even the formation of piping channels. Once a structure suffers sudden damage, it will not only cause the project to be suspended, but may also cause secondary risks to the surrounding environment. Therefore, conducting systematic, accurate, and forward-looking health status assessments of long-distance underground culverts to promptly identify potential risks and issue early warnings has theoretical significance and engineering application value for ensuring their safe and efficient operation throughout their life cycle.

[0003] Currently, the inspection and assessment technology for long-distance underground culverts has made considerable progress. Regarding inspection methods, traditional manual diving exploration is gradually being supplemented or replaced by underwater robotic vehicles (ROVs) equipped with a variety of sensors. ROVs are typically equipped with high-definition optical cameras, forward-looking sonar, or multi-beam sounding systems. They can conduct preliminary surveys of siltation and surface concrete defects (such as cracks and spalling) without draining the culvert. For structural performance assessment, in-situ nondestructive testing (NDT) techniques are being applied. For example, in areas with suitable working conditions, rebound testing is used to sample concrete strength. Furthermore, fixed monitoring sensors such as piezometers, strain gauges, and displacement gauges are deployed at key locations within the culvert to obtain long-term monitoring data on key structural quantities. Furthermore, numerical simulation techniques based on the finite element method (FEM) are widely used to analyze the seepage and stress field distributions in culverts under design loads and boundary conditions, providing a theoretical basis for understanding the fundamental mechanical behavior of the structure. The application of these technologies and methods forms the foundation of current culvert health assessment work.

[0004] However, although existing technologies have achieved the detection of culverts to a certain extent, they still have significant limitations in the deep fusion of multi-source heterogeneous data and the assessment of the spatiotemporal evolution of structural health status. Specifically, at the data level, the detection data from different sources (such as sound, light, and mechanics) lack a unified physical model for spatiotemporal alignment, underwater environmental interference correction, and defect impact quantification, resulting in low accuracy and unclear physical meaning of the fusion results, forming information islands and inaccurate assessments. At the assessment framework level, existing methods usually stop at a static and discrete description of the current state. They can neither reveal the dynamic evolution of structural health with the coupling of multiple fields through a unified evolution function, nor reliably reconstruct sparse monitoring data into a full-field continuous state based on physical constraints. This dual limitation of data and framework makes it difficult for existing technologies to provide accurate and timely warnings of potential risks of long-distance culverts. Summary of the Invention

[0005] The purpose of the invention is to provide a long-distance underground culvert structure health assessment method to solve the above-mentioned problems existing in the prior art.

[0006] Technical solution, long-distance underground culvert structural health assessment method, including:

[0007] Based on acoustic and optical detection data, fusion processing is performed to identify and output structural defect data including soil type, defect type and geometric parameters;

[0008] Based on pre-stored underwater rebound and coring test data, combined with structural defect data, underwater-to-dry-land conversion and correction of strength values ​​are performed to determine the corrected structural strength and material degradation parameters;

[0009] Combining structural defect data, material degradation parameters, and pre-stored historical monitoring data, structural evolution modeling and mutation risk identification are performed to obtain the structural evolution characteristics and mutation risks;

[0010] The overall status assessment is conducted based on the comprehensively corrected structural strength, evolutionary characteristics and mutation risks to generate structural health assessment and risk warning results.

[0011] Beneficial effect: the present invention can realize accurate, dynamic and full-field assessment of the health status of culverts, and improve the reliability and early warning capability of the assessment results. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A flowchart of the steps of a long-distance underground culvert structure health assessment method provided in an embodiment of the present application.

[0013] Figure 2 A flowchart of the steps for obtaining the evolutionary characteristics and mutation risks of a structure provided in an embodiment of the present application.

[0014] Figure 3 A flowchart of the steps for generating an evolution degree time series provided in an embodiment of the present application.

[0015] Figure 4 A flowchart of the steps for dynamically calibrating the evolution weight coefficient provided in an embodiment of the present application. DETAILED DESCRIPTION

[0016] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0017] It should be noted that the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or are inherent to these processes, methods, products or apparatus.

[0018] The research revealed that data silos and misalignments lead to insufficient accuracy and completeness in the assessment basis. The multi-source data obtained by existing methods is physically fragmented and uncalibrated. The physical foundation for optical and acoustic data fusion is weak. Due to the physical separation and platform motion of the optical and acoustic sensors onboard the ROV, their data exhibit inherent temporal and spatial deviations, making simple image overlay ineffective for true feature fusion. Furthermore, the materials and flow conditions of different structural compartments (such as the crown and sidewalls) exhibit distinct attenuation and scattering patterns for acoustic and optical signals. Fixed fusion strategies cannot adapt to this spatial heterogeneity, limiting defect identification accuracy. The impact of the underwater environment on in-situ testing results has not been precisely quantified. For example, when testing concrete strength using the underwater rebound method, the rebound value is affected by multiple physical effects, including water pressure, water temperature, water flow, and water-medium coupling. Directly applying terrestrial conversion formulas can lead to significant misjudgments of structural strength. Existing methods lack physical models to decouple and comprehensively compensate for these interfering factors. The geometric information of a defect is disconnected from its mechanical impact. The relationship between the location, size, and other geometric information of detected defects like cracks and spalling and their actual weakening effect on the surrounding material is ambiguous. Assessments often rely on qualitative descriptions or rough reductions, failing to quantitatively and continuously reflect the spatial distribution of the defect's impact, leading to inaccurate local risk assessments.

[0019] Furthermore, the static and discrete nature of the assessment framework limits the holistic and forward-looking nature of the assessment results. Existing assessment models are often static patchworks rather than integrated, integrated systems. A unified framework for spatiotemporal evolution is lacking. Assessments typically stop at generating a current defect or strength distribution map, failing to address the core question of how the overall health of the structure has evolved from the past to the present. Existing methods struggle to unify multi-physics processes such as seepage, stress, and material degradation into a single evolutionary function capable of quantifying cumulative damage, thus failing to reveal the dynamic evolution of the structural health state. Assessment results are spatially discrete and lack physical context. Whether through in-situ testing or sensor monitoring, the data obtained is essentially discrete point or line information. Scientifically reconstructing the full-field state distribution of the entire culvert from this sparse data is a key challenge. Traditional spatial interpolation methods only consider geometric distances, ignoring the physical laws governing the propagation of evolving states within the structural medium (such as diffusion and propagation). Consequently, the reconstructed full-field state map can deviate significantly from the actual physical processes, making it difficult to accurately identify concentrated areas of risk and propagation paths.

[0020] like Figure 1 As shown in the figure, a long-distance underground culvert structural health assessment method is proposed, which includes the following steps:

[0021] Based on the acoustic and optical detection data, fusion processing is performed to identify and output structural defect data including soil type, defect type and geometric parameters.

[0022] In this example, acoustic inspection data includes time difference, amplitude, and angle of incidence information for multibeam and sidescan echoes, ultrasonic pulse-echo data, acoustic tomography data, and backscatter or scattering cross-section data for assessing the scattering intensity of the substrate and defects at different frequencies. Optical inspection data includes high-resolution visible light images and video frames, lidar point clouds, multispectral or hyperspectral imagery, and infrared thermal imaging data. This step is the starting point for data-driven analysis, aiming to overcome the limitations of single data sources in complex underwater environments (such as high turbidity water bodies) by integrating the strengths of different sensors, thereby accurately identifying and locating surface defects in culverts, such as cracks, spalling, and leaks. Structural defect data provides critical input for subsequent strength correction and evolution analysis.

[0023] Based on pre-stored underwater rebound and coring test data, combined with structural defect data, underwater-to-dry-land conversion and correction of strength values ​​are performed to determine the corrected structural strength and material degradation parameters.

[0024] In this embodiment, this step aims to obtain the true performance parameters of the structural material. Given that culverts are permanently underwater, conventional dry-ground testing methods are no longer applicable. Therefore, a physical model is constructed to quantitatively compensate for the impact of the underwater environment (such as water pressure and temperature) on rebound testing, achieving underwater-to-dry-ground strength conversion. Using structural defect data, the strength values ​​around the defect are locally corrected, ultimately generating a strength distribution map that reflects the true state of the structure. Material degradation parameters are then calculated based on this strength distribution map.

[0025] Combining structural defect data, material degradation parameters and pre-stored historical monitoring data, structural evolution modeling and mutation risk identification are carried out to obtain the evolutionary characteristics and mutation risks of the structure.

[0026] In this example, the purpose of this step is to reveal the dynamic patterns of structural health changes over time and potential mutation risks. Specifically, by integrating multi-source information, an evolutionary model is constructed that reflects the coupled effects of multiple physical fields, such as penetration and stress. By solving this model, we can quantify the evolutionary characteristics of the structural cumulative damage. By performing differential analysis on the evolutionary process, we can identify the risk transition moments that may lead to structural failure, namely, mutation risks.

[0027] The overall status assessment is conducted based on the comprehensively corrected structural strength, evolutionary characteristics and mutation risks to generate structural health assessment and risk warning results.

[0028] In this embodiment, the assessment results from multiple dimensions obtained in the preceding steps are integrated to form a global, unified health status assessment. Sparse data reconstruction techniques can be used to expand discrete monitoring point information into a continuous state field covering the entire culvert. Based on this, a nonlinear fusion model is employed to comprehensively calculate the structural strength representing current performance, the evolutionary characteristics representing historical evolution, and the mutation risk representing future risks. Ultimately, the assessment results are intuitively presented in the form of a comprehensive health index distribution chart, providing a graded early warning solution.

[0029] According to one aspect of the present application, before performing fusion processing to identify and output structural defect data, a step of dynamic spatiotemporal alignment of acoustic and optical data is included. The purpose is to address the differences between acoustic and optical sensors in terms of physical location, propagation medium speed, and dynamic changes in the motion platform, so that the data basis for subsequent fusion processing is consistent in time and space. Specifically, the alignment step includes:

[0030] The basic propagation time difference is calculated based on the difference in the propagation speed of sound and light in the water medium and the spatial distance between the sensors.

[0031] In this embodiment, the spatial distance d between the sensors can be obtained by reading the sensor layout file of the underwater robot (ROV) system, specifically the center coordinates of the acoustic sensor (x s ,y s , z s ) and the optical camera center coordinates (x c ,y c , z c ) between the Euclidean distance d = sqrt[(x s -x c ) 2 +(y s -y c ) 2 +(z s -z c ) 2 ]. The speed of sound in water medium v s and the speed of light v c It can be calculated by empirical formula based on real-time monitoring of water temperature, water depth and other environmental parameters. For example, the speed of sound v s Can be obtained by v s =1449.2+4.6T-0.055T 2 +0.00029T 3 +(1.34-0.01T)(S-35)+0.016h, where T is water temperature, S is salinity, and h is water depth. Basic propagation time difference Δt base Then Δt base =d(1 / v s -1 / v c ).

[0032] At least one disturbance effect introduced by the dynamic change of the detection equipment is taken into account to obtain a dynamic time difference correction term.

[0033] In this embodiment, the disturbance effect preferably includes the motion speed effect of the underwater robot or the beam spreading effect of the acoustic sensor. For example, the correction term Δt introduced by the motion speed effect is motion It can be expressed as Δt motion =d·v ROV / (v s ·v c ), where v ROV is the real-time velocity vector of the ROV. The correction term Δt introduced by the beam spreading effect spread It can be expressed as Δt spread =d·tan(θ s / 2) / v s , where θ s is the beam angle of the acoustic sensor.

[0034] The basic propagation time difference and the dynamic time difference correction term are superimposed to generate the comprehensive time difference compensation Δt total For example, Δt total =Δt base +Δt motion +Δt spread .

[0035] Based on the comprehensive time difference compensation amount, the acquisition time axes of the acoustic data and the optical detection data are shifted and aligned.

[0036] Specifically, the timestamp of the acoustic data can be shifted by t s_new =t s_old +Δt total , and resample the interval that overlaps with the optical data in time after translation to unify the time resolution. s_old is the timestamp of the original acquisition of acoustic data, t s_new This is the new timestamp after the comprehensive time difference compensation is completed on the acoustic data time axis.

[0037] This embodiment calculates dynamic time-lag compensation by comprehensively considering the difference in acoustic and optical propagation speeds, the ROV velocity vector, and the acoustic beam diffusion effect, achieving precise spatiotemporal alignment of multi-sensor data on a mobile inspection platform. This solves the problem of defect location deviation caused by ROV acoustic data lag. Dynamic compensation improves defect location accuracy, enabling accurate differentiation of parallel cracks and providing reliable spatial coordinate information for subsequent precision repair.

[0038] According to one aspect of the present application, after completing the spatiotemporal alignment, a fusion process is performed to identify and output structural defect data, specifically including:

[0039] Acoustic feature maps and visual feature maps are analyzed from acoustic and optical detection data respectively.

[0040] For example, short-time Fourier transform is performed on the time-aligned acoustic data to extract frequency domain features and form an acoustic feature map; edge detection and texture analysis are performed on the time-aligned optical data to obtain a visual feature map.

[0041] According to the structural drawings of the culvert, the culvert body is spatially divided into multiple structural partitions with different physical properties.

[0042] This division is intended to enable adaptive adjustment of subsequent fusion parameters, as the materials and flow environments of different zones have different impacts on acoustic and optical signals. For example, a culvert cross-section can be divided into top, sidewall, and bottom zones based on structural mechanics.

[0043] For each structural partition, partition-specific fusion parameters are adaptively calculated and generated based on its unique material and water flow environment.

[0044] This embodiment is the key to achieving intelligent fusion. Specifically, it includes: for each structural partition, extracting the partition state parameters that characterize its physical nature from the design data and environmental monitoring data corresponding to the structural partition. This group of parameters includes at least two of the following: the material properties represented by the concrete thickness or steel bar distribution of the partition; the hydrodynamic characteristics represented by the water flow velocity or Reynolds number of the partition. Substitute the partition state parameters into the preset physical model for solution, and generate a unique partition-specific fusion parameter for the partition to weight the acoustic feature map and the visual feature map respectively. For example, an acoustic attenuation coefficient model α can be established i =α0×exp(-β a ×h i )×(1+0.1×ρ i )×(1+0.05×Re i ) and the optical scattering coefficient model β i =β0×(1+γ s ×TSS i )×(1-0.2×h i / h max ), and normalized, where h i is the thickness of concrete, h max is the maximum thickness of concrete, ρ i is the reinforcement ratio, Re i is the Reynolds number, TSS i is the suspended matter concentration, α0, β a ,β0,γ s etc. are model coefficients.

[0045] In each structural partition, the specific fusion parameters corresponding to the structural partition are used to perform weighted fusion of the acoustic feature map and the visual feature map to generate a fused feature map. Defects are identified based on the fused feature map and the structural defect data is finally output. The fusion formula can be expressed as F i =α i ×A i +β i ×V i , where A i and V i are the normalized acoustic and visual features, respectively.

[0046] This embodiment addresses the significant differences in inspection conditions across culverts by adaptively calculating optical and acoustic fusion weights based on physical parameters such as concrete thickness, rebar distribution, and the Reynolds number of each zone. Optical weighting is increased in the top, low-velocity area to obtain clear crack images, while acoustic weighting is increased in the bottom, high-sediment flow area to penetrate turbid media. A balanced fusion strategy is employed in the sidewall transition zone. This improves the overall defect detection rate for long culverts and provides reliable data support for preventive maintenance.

[0047] Furthermore, after weighted fusion, the fusion result verification and correction steps are also included to improve the robustness of the fusion result. Specifically:

[0048] For a specific structural partition, the cross-modal mutual information between the acoustic feature map and the visual feature map corresponding to the specific structural partition is calculated, and the cross-modal mutual information is used as a quantitative indicator to measure the consistency of the two modal data. i , V i ) is larger, the higher the consistency of the information provided by the two data sources. i , V i ) = ∑p(a,v)log[p(a,v) / (p(a)p(v))], where p(a,v) is the joint probability that the acoustic feature value falls at quantization level a and the visual feature value falls at quantization level v, p(a) is the marginal probability that the acoustic feature falls at quantization level a, and p(v) is the marginal probability that the visual feature falls at quantization level v.

[0049] When the cross-modal mutual information is lower than a preset confidence threshold (e.g., MI < 0.3), it is determined that there is modal information conflict in the partition, which usually means that one of the sensors in the area is severely interfered with or fails.

[0050] For partitions with conflicting modal information, the fusion strategy automatically switches from weighted fusion to a preferential selection strategy. This strategy selects the acoustic and visual features with higher confidence (for example, the one with better signal quality or feature strength) as the final fusion result for that partition, and the fused feature map is corrected. This effectively prevents contamination of the final fusion result due to poor quality of a single modal data, thereby improving the accuracy of defect identification.

[0051] This embodiment ensures reliable fusion results. When turbulent water flow in culvert bends causes blurry optical images but clear acoustic signals, the system intelligently selects acoustic data. Furthermore, in areas of smooth flow along straight sections, the system leverages the complementary strengths of both modalities. This reduces the false alarm rate of fusion detection while maintaining high sensitivity, making it particularly suitable for the complex environments of long culverts where detection conditions frequently change.

[0052] According to one aspect of the present application, the steps of converting the strength value from underwater to dry land and correcting the strength value specifically include:

[0053] Based on the core test data, the underwater-dry land comprehensive conversion coefficient was constructed and calibrated;

[0054] The underwater-dry land comprehensive conversion coefficient in this embodiment is intended to compensate for the impact of the underwater environment and systematically eliminate the interference of the underwater environment on the rebound test results. Specifically, it includes:

[0055] For the physical effects of the underwater environment, independent environmental correction factors are quantified and determined. The set of environmental correction factors includes at least two of the following, and preferably all four:

[0056] The pressure correction coefficient k that characterizes the effect of hydrostatic pressure on rebound value p Due to the hydrostatic pressure p w It will produce pre-compressive stress on the concrete surface and increase the energy dissipation of rebound impact, so it needs to be corrected. w =ρ w ×g×h,ρ w is the density of water (for example, 1000 kg / m 3 ), g is the acceleration due to gravity (about 9.8m / s 2 ), h is the water depth at the detection point. The pressure correction coefficient can be obtained by model k p =1 / (1+λ p ×p w ) calculation, where λ p is the pressure influence coefficient calibrated by experiment (for example, λ p =2×10 -6 / Pa).

[0057] Temperature correction coefficient k that characterizes the effect of water temperature on the elastic modulus of concrete T The water temperature T will affect the elastic modulus of concrete, and thus affect the rebound value. The correction factor can be calculated by k T =[1+α T ×(T-20)]×k age Calculate, where α T is the temperature sensitivity coefficient of the elastic modulus (e.g., α T =-0.0004 / ℃), k age is the age correction term to consider the influence of early hydration reaction. For example, when the concrete age is t age When it is less than 90 days, k age =0.9+0.1×(t age / 90).

[0058] Scour correction coefficient k that characterizes the effect of water scour on concrete surface hardnesse Long-term water flow will wear away the concrete surface and reduce its hardness. This effect can be explained by the model k e =1-β e ×η e Quantify, where η e is the surface hardness loss rate, which can be obtained from the cumulative scouring depth Δ e and the average aggregate particle size d agg The ratio (η e =Δ e / d agg ) estimate, Δ e And the water velocity v and service time t service Correlation; β e is the impact factor of the washout (e.g., β e =0.15).

[0059] The medium coupling coefficient k characterizes the effect of water medium on the energy transfer of rebound impact c The presence of water increases the damping during rebound impact and reduces the energy transfer efficiency. This effect can be seen by k c Make corrections, such as k c It can be calculated by taking into account the influence of water depth based on the benchmark efficiency ratio, such as k c =1.31×(1-0.001×h).

[0060] By coupling and integrating all environmental correction factors, the underwater-dry land comprehensive conversion coefficient is generated.

[0061] Specifically, the integration step includes: multiplying multiple independent environmental correction factors in the group to obtain a preliminary comprehensive coefficient k=k p ×k T ×k e ×k c The second-order interaction correction term Δk is further introduced to characterize the coupling effect between different physical effects, and the preliminary comprehensive coefficient is finally corrected to generate the underwater-dry land comprehensive conversion coefficient k final Because the physical effects are not completely independent, for example, the coupled effect of high water pressure and low temperature may be greater than the product of their independent effects. The second-order interaction correction term can be expressed as Δk=0.05×(k p -1)×(k T -1), the final comprehensive conversion coefficient is k final =k×(1+Δk).

[0062] The underwater-dry land comprehensive conversion coefficient is applied to convert the underwater rebound into a preliminary dry land equivalent strength value.

[0063] For example, the formula f' can be used c =kfinal ×(A×R w -B) Set the underwater rebound original value R w Converted to preliminary dry land equivalent strength value f' c , where A and B are the calibration parameters of the rebound hammer.

[0064] According to the defect type and geometric parameters, the preliminary dry land equivalent strength value is locally corrected to obtain the corrected structural strength. Based on the comparison between the corrected structural strength and the preset design strength, the material degradation parameter is generated.

[0065] In this embodiment, the local correction step specifically includes:

[0066] For each defect in the structural defect data, a three-dimensional impact field model I(r) is constructed based on the geometric parameters of the structural defect data, in which the impact intensity decays with increasing spatial distance. For example, the model can be expressed as a Gaussian attenuation function I(r) = I0×exp(-r 2 / 2σ 2 ), where r is the distance from any point in space to the defect center, I0 is the influence intensity of the defect center (normalized to 1), and σ is the influence radius, whose value is related to the length l and width w of the defect, such as σ=sqrt(l 2 +w 2 ) / 4.

[0067] Based on the defect type, set the foundation strength reduction factor that represents the maximum impact of the defect type. For example, for cracks, the reduction factor k crack Can be related to the crack width w, k crack =1-0.15×(w / w cr ) 0.5 , where w cr is the critical width; for spalling, the reduction factor k spall Can be related to the spalling depth d.

[0068] By integrating the three-dimensional impact field model with the foundation strength reduction factor, the defect correction coefficient field k that varies continuously in space is calculated. d (x, y, z). Specifically, for a single defect, k d (x, y, z) = 1-(1-k type )×I(r) / I0, where k type is the basic strength reduction coefficient of the defect type. Optionally, when the three-dimensional influence field models of multiple defects overlap in space, the method further includes: performing a probability superposition operation on the defect correction coefficients corresponding to each independent defect in the spatial overlapping area to generate a final correction coefficient k that can reflect the coupling effect of multiple defects. total, and use the final correction coefficient to correct the intensity value in the overlapping area. This probability superposition can be expressed as k total =1-Π(1-k i ), where k i is the correction coefficient caused by the i-th independent defect at this point.

[0069] Apply the defect correction factor field to the preliminary dry land equivalent strength value to obtain the corrected structural strength. c =k d ×f' c , complete the strength correction of the defect affected area.

[0070] This embodiment establishes a high-precision underwater-to-dryland strength conversion model. This model accounts for the decrease in rebound value caused by every 10-meter increase in water depth, the change in elastic modulus when the water temperature deviates from 20°C, the attenuation of surface hardness caused by long-term water erosion, and the effect of water as a coupling medium on the efficiency of impact energy transfer, thereby reducing the error in underwater concrete strength assessment. This allows for accurate strength distribution throughout long culverts without requiring extensive coring, improving detection efficiency and reducing costs. By establishing a three-dimensional Gaussian attenuation field model for defect effects and using probabilistic superposition to process overlapping regions of multiple defects, a refined quantification of the impact of defects on concrete strength is achieved. By considering the elliptical impact zone along the crack strike, the hemispherical impact zone of spalling, and the synergistic degradation effect of multiple defects, the spatial resolution of strength assessment is improved compared to traditional two-dimensional fixed-range reduction. By systematically eliminating the combined effects of the underwater environment and local defects on strength testing, more accurate and reliable structural strength assessment results are obtained than with traditional methods, providing a solid data foundation for subsequent structural health assessments.

[0071] like Figure 2 As shown, according to one aspect of the present application, the steps of performing structural evolution modeling and mutation risk identification to obtain structural evolution characteristics and mutation risks include:

[0072] By integrating structural defect data, material degradation parameters and historical monitoring data, we construct and solve the evolution function of multi-physics field coupling, and generate an evolution degree time series that quantifies the cumulative evolution degree of the structure.

[0073] In this embodiment, the evolution degree time series Φ(t) is a dimensionless parameter with a value range between [0, 1], where 0 represents the intact state of the structure and 1 represents the complete failure of the structure or reaching the terminal stage of evolution. Its numerical value intuitively reflects the cumulative damage or health deterioration of the structure.

[0074] Further, such as Figure 3 As shown in FIG, the steps of constructing and solving the evolution function of the multi-physics field coupling and generating the evolution degree time series include:

[0075] A plurality of physical field indices including at least hydraulic gradient i, seepage rate q and soil strain ε are extracted from historical monitoring data, and respective time increment series of the plurality of physical field indices are calculated.

[0076] Specifically, historical monitoring data can be derived from various sensors installed on the culvert, such as pressure gauges, flow meters, strain gauges, etc. In order to perform evolutionary analysis, these original time series data need to be preprocessed. For example, the time resolution is unified to 1 hour, and the change in a time step Δt is calculated to obtain the hydraulic gradient increment sequence Δi(t)=i(t)-i(t-Δt), where i(t) is the hydraulic gradient at time t, and Δt is the time step between two adjacent monitoring data; the seepage rate change sequence Δq(t)=[q(t)-q(t-Δt)] / A seep , where q(t) is the seepage rate at time t, A seep is the seepage area; and the strain rate increment sequence Δε(t)=ε*(t)-ε*(t-Δt), where ε*(t) is the soil strain rate at time t, and ε*(t)=dε(t) / dt.

[0077] With reference to soil type and material degradation parameters, the evolution weight coefficients corresponding to various physical field indicators are dynamically calibrated.

[0078] In this embodiment, the contribution of different physical field processes to structural degradation is not constant, but is closely related to the specific environment (such as soil type) and the state of the structure itself (such as the degree of material degradation). Figure 4As shown, the dynamic calibration of evolutionary weight coefficients includes: assigning base weight coefficients to each physical field indicator based on soil type; extracting a structural degradation index from material degradation parameters and modifying at least one coefficient value in the base weight coefficients accordingly to obtain adjusted weights; and performing a normalization calculation on the adjusted weights to generate evolutionary weight coefficients. Specifically, base weight coefficients are assigned to each physical field indicator based on the soil type contained in the structural defect data. For example, based on soil mechanical properties, for sandy soils that are more sensitive to seepage damage, higher base weights (e.g., α = 0.5, β = 0.3) can be assigned to hydraulic gradient i and seepage flow q; whereas for clay soils that are more sensitive to deformation, a higher base weight (e.g., γ = 0.3) can be assigned to soil strain ε. A structural degradation index (e.g., strength loss rate η) is extracted from material degradation parameters, and at least one coefficient value in the base weight coefficients is modified based on the structural degradation index to obtain adjusted weights. For example, when the strength loss rate η exceeds a certain threshold (e.g., η > 0.2), it indicates that the structural capacity has decreased and it is more sensitive to strain. In this case, the weight of the strain term can be increased, such as the modified weight γ' = γ × (1 + 0.5η). The adjusted weights are normalized so that the sum of the coefficients is 1, generating the final evolution weight coefficients α', β', and γ'.

[0079] The time increment series is weightedly fused using the evolution weight coefficient to obtain the evolution increment series dΦ / dt that reflects the instantaneous evolution rate of the structure.

[0080] In this embodiment, the instantaneous evolution rate can be expressed as dΦ / dt=α'×Δi(t)+β'×Δq(t)+γ'×Δε(t). Optionally, to eliminate the influence of measurement noise on the calculation results, a Kalman filter can be applied to the evolution increment sequence after calculation to perform filtering processing.

[0081] The evolution increment series is numerically integrated along the time dimension to generate the evolution degree time series.

[0082] Specifically, a numerical integration method, such as the trapezoidal integration method, can be used to accumulate dΦ / dt time-by-time, starting from the initial state Φ0 (for example, Φ0 = 0), resulting in the evolutionary degree time series Φ(t) = Φ0 + ∫(dΦ / dt)dt. This procedure yields a curve that continuously reflects the evolution of the culvert structure from its past to its current health state, providing a foundation for subsequent evolutionary stage delineation and mutation risk identification.

[0083] This implementation achieves precise matching of the multi-physics coupling model with actual geological conditions by dynamically adjusting the evolution weight coefficient based on soil type (clay / silt / sand) and the degree of material degradation. When the material degradation rate exceeds 20%, the system automatically increases the strain weight to focus on the impact of structural deformation. In sandy areas, the hydraulic gradient weight is increased to focus on monitoring the risk of seepage damage. This allows the same evolution model to adapt to the complex and changing geological conditions along the culvert. Compared to fixed-weight methods, this improves prediction accuracy, particularly at critical locations such as stratigraphic interfaces.

[0084] Differential analysis is performed on the evolutionary degree time series to identify potential mutation moments that represent risk transitions.

[0085] The purpose of this example is to capture nonlinear and non-stationary key nodes that indicate qualitative changes in the structural state from the continuous evolution curve. Specifically, it identifies potential mutation moments that represent risk transitions, including:

[0086] Solve the first-order derivative sequence and second-order derivative sequence of the evolution time series; use high-precision numerical methods to solve the first-order derivative sequence ΨΦ / Ψt and the second-order derivative sequence Ψ of the evolution time series Φ(t) 2 Φ / Ψt 2 , respectively characterize the evolution rate and evolution acceleration of the structure, where Ψ is the partial derivative.

[0087] In this embodiment, to improve computational accuracy and avoid amplification of numerical errors, a five-point central difference method is preferably used for derivation. For example, the first-order derivative can be calculated as [-Φ(t+2Δt)+8Φ(t+Δt)-8Φ(t-Δt)+Φ(t-2Δt)] / (12Δt). In addition, to eliminate the influence of raw data noise on derivative calculations, the derivative sequence can be first smoothed using a Savitzky-Golay filter.

[0088] The mutation sensitivity time series is constructed by calculating the ratio of the absolute values ​​of the second-order derivative sequence to the first-order derivative sequence at each moment.

[0089] Specifically, the mutation sensitivity time series S m (t)=|Ψ 2 Φ / Ψt 2 | / |ΨΦ / Ψt|. Mutation sensitivity is a dimensionless metric that amplifies the change in evolutionary acceleration relative to the current evolutionary rate. Physically, it characterizes the degree to which the evolutionary trend will experience an inflection point or a sudden change. When the structure is in a stable evolutionary phase, this ratio is small. However, when the structure is about to undergo or is undergoing a sudden change, the evolutionary acceleration changes dramatically, causing the ratio to peak.

[0090] Each sensitivity value in the mutation sensitivity time series is compared with a preset threshold, and the moment exceeding the threshold is identified as a potential mutation moment.

[0091] Specifically, the threshold value can be a fixed value (for example, S cr =2.5), or, in a preferred embodiment, an adaptive threshold is used. For example, by calculating the mutation sensitivity time series S m (t) Moving average S* within a sliding time window m and standard deviation σ s , dynamically adjust the mutation threshold to S cr =S* m +n×σ s (where n can be 2.5), so that the threshold can adapt to the baseline fluctuations in different evolution stages and improve the accuracy of recognition.

[0092] This embodiment constructs a mutation sensitivity index by calculating the ratio of the second-order derivative to the first-order derivative of the evolution degree time series, accurately capturing nonlinear mutation points in the culvert's seepage evolution process. This index peaks when the evolution rate changes dramatically, effectively identifying the critical moments of transition from the seepage stage to the seepage deformation stage, and from the seepage damage stage to the channel expansion stage. Compared to traditional methods that rely solely on seepage flow or displacement monitoring, this method can predict system instability before significant changes in physical quantities occur, shortening the warning time for major accidents such as culvert collapse and improving the safety and security capabilities of underground water conservancy facilities.

[0093] Furthermore, to verify the authenticity of the identified potential mutation moments and explore their physical causes, the following also applies:

[0094] For each potential mutation moment identified, extract the acoustic data segments within the predetermined time window before and after the potential mutation moment from the acoustic detection data. m When potential mutations are identified at any time, [t m -300s, t m +300s] interval.

[0095] Perform spectral analysis on the acoustic data segments to decode whether there are acoustic anomaly patterns associated with specific pre-determined failure mechanisms.

[0096] This embodiment aims to use acoustic emission signals that are directly related to physical processes such as structural deformation and cracking to cross-validate mutations inferred from macroscopic physical field data. Optionally, the identified acoustic anomaly pattern includes at least one of the following features: in the high-frequency band, the spectrum energy shows a sudden increase with an amplitude exceeding a predetermined decibel value. For example, in the frequency band greater than 20kHz, the energy shows a sudden increase of more than 10dB above the background noise. This pattern is usually associated with brittle failure events such as microcrack expansion of materials and tearing of water-stopping materials. In the low-frequency band, the vibration intensity shows a sustained jump significantly higher than the background noise level. For example, in the frequency band below 100Hz, the vibration energy is continuously more than 5 times higher than the normal level. This pattern may correspond to the resonance of the entire structure, particle scouring, or abnormal changes in water flow pulsation pressure.

[0097] If there is an acoustic anomaly pattern, the potential mutation moment is upgraded to a verified mutation risk, and the cause of the mutation risk is classified according to the acoustic anomaly pattern.

[0098] For example, if a sudden increase in high-frequency energy is identified, the risk can be preliminarily classified as structural cracking; if a sudden increase in low-frequency vibration is identified, it can be classified as structural instability or scour risk. This not only improves the reliability of risk warnings but also provides a direct physical basis for subsequent treatment decisions.

[0099] This embodiment implements a dual confirmation mechanism for mutation risk by performing temporal correlation verification between the potential mutation moment calculated by mathematical mutation theory and the abnormal acoustic spectrum pattern before and after that moment. When an abnormal peak appears in the second-order derivative of the evolution function, the system automatically extracts the acoustic data within the corresponding time window for spectrum analysis, identifying the sudden increase in high-frequency noise (>20kHz, amplitude increase >10dB) caused by the tearing of the waterstop or the low-frequency resonance (<100Hz) caused by structural vibration, and converting the abstract mathematical warning into a specific type of physical damage. This reduces the false alarm rate of mutations in culvert detection and can identify early signs of seepage damage in advance, providing accurate time windows and failure mode information for maintenance decisions.

[0100] Based on the evolution degree time series and the spatial location of the structural defect data, the evolution rate distribution map is derived and generated;

[0101] In this embodiment, the evolution rate of the time dimension is associated with the spatial position of the structural defect to generate a spatial dynamic evolution field, which intuitively presents the difference in degradation trends in different regions of the structure. Specifically, for each defect point x i , extract the corresponding local evolution rate Φ*(x i , t) (i.e. the first-order derivative ΨΦ / Ψt at position x i The value of Φ*(x i, t) is mapped to a continuous domain: Φ*(x, t)=∑ i=1 n λ i Φ*(x i , t); where the weight λ i Determined by optimizing the spatial covariance function, the interpolation result is ensured to conform to the spatial correlation of the structural material. Output the spatiotemporal evolution rate field Φ*(x, t) (three-dimensional space + time tensor); select the key time slice t k (such as the current moment, historical mutation point), extract the slice field Φ*(x, t k ); generate evolution rate distribution maps through contour plots or heatmap visualization.

[0102] The evolutionary degree time series, potential mutation moments and evolutionary rate distribution diagram together constitute the evolutionary characteristics and mutation risks of the structure.

[0103] In this embodiment, the temporal evolution, mutation nodes and spatial rate field are integrated to form a multi-dimensional risk criterion to support the structural safety classification decision. Specifically, based on the evolution degree time series, potential mutation moment and evolution rate distribution diagram, an evolution feature report is generated, including the cumulative evolution Φ(t now ), indicating the overall degradation degree of the structure; the current rate maxΦ*(x, t now ), represents the evolution speed of the most dangerous area; mutation risk S m (t now ) (sensitivity), which indicates the probability of a sudden change in the adjacent state. Build a dynamic risk grading model and define the risk level R as a weighted function: R=w1·Φ(t) / Φ crit +w2·maxΦ* / Φ* crit +w3·S m (t) / S cr ; The weight distribution (example) is: w1=0.3, w2=0.4, w3=0.3; Φ crit is the maximum allowable evolution degree (such as the concrete expansion limit of 0.5%); Φ*crit is the critical rate (such as the rock creep rate>1 mm / day); S cr is the sensitivity threshold.

[0104] According to one aspect of this application, since monitoring points are often sparse in actual engineering, it is impossible to directly obtain the health status of every point in the structure, and simple spatial interpolation methods ignore the physical laws of state propagation in the structural medium. Therefore, when performing full-field state assessment, a full-field reconstruction stage is included to generate a continuous state distribution. Specifically, this full-field reconstruction stage includes:

[0105] The evolutionary parameters contained in the evolutionary characteristics and mutation risks are arranged as sparse state monitoring points in the three-dimensional space of the culvert.

[0106] In this embodiment, the evolution parameters refer to the evolution degree Φ, evolution rate ΨΦ / Ψt, etc. of each key position. The selection of these key positions can be further optimized, for example, by giving priority to locations where the evolution rate gradient changes sharply or where mutation risks have been identified.

[0107] Construct physical constraint equations that describe the spatiotemporal evolution of the propagation law of the evolving state in the structured medium.

[0108] This embodiment incorporates engineering experience and physical mechanisms into the spatial interpolation process. In a specific embodiment, the spatiotemporal evolution physical constraint equation is a partial differential equation. The equation contains at least a time derivative term to describe the evolution state over time, and a space derivative term to describe the propagation and diffusion of the evolution state in space. For example, the following form of evolution propagation equation can be constructed: 2 Φ-(1 / c 2 )Ψ 2 Φ / Ψt 2 =S(x, y, z); where Φ is the degree of evolution of the spatial position (x, y, z) at time t, which is the unknown quantity to be solved; ▽ 2 is the Laplace operator, representing the spatial second-order derivative term (Ψ 2 / Ψx 2 +Ψ 2 / Ψy 2 +Ψ 2 / Ψz 2 ), which describes the diffusion or propagation trend of the evolution state in space; c is the evolution propagation speed, which is a physical parameter that characterizes the speed at which the degradation or damage effect propagates in the structural medium. Its value can be calibrated according to different media types (such as sand, silt, and concrete). For example, c can be taken as 2 m / day in sandy areas and 0.5 m / day in clay areas; Ψ 2 Φ / Ψt 2 is the time second-order derivative term, which describes the inertia of the evolution state over time; S(x, y, z) is the source term, which represents the endogenous growth rate of the evolution state at the spatial position (x, y, z). For example, it can be calculated by the evolution degree Φ of the point itself. local Determine, S(x, y, z) = κ × Φ local ×(1-Φ local ), where κ is the growth rate coefficient.

[0109] The data of sparse state monitoring points are used as solution boundaries or initial conditions to numerically solve the physical constraint equations and generate a full-field evolution degree distribution map that continuously covers the entire culvert.

[0110] Specifically, the partial differential equation can be solved using the finite element method or the finite difference method. A three-dimensional mesh is then generated for the culvert, preferably with adaptive mesh refinement in areas with large evolution gradients. The evolution degree values ​​of sparse monitoring points are applied to the model as known conditions (e.g., Dirichlet boundary conditions). Through iterative solution, the evolution degree Φ(x, y, z, t) of all mesh nodes within the entire solution domain (i.e., the entire culvert) can be obtained, forming a spatiotemporally continuous full-field evolution degree distribution map.

[0111] This embodiment realizes the physical constraint reconstruction from sparse monitoring points to full-field continuous distribution by establishing an evolution propagation partial differential equation containing time derivative terms and space derivative terms. The evolution propagation speed is determined according to the permeability characteristics of different soil bodies, so that the reconstruction results not only match the measured values ​​at the monitoring points, but also follow the physical laws of permeability damage in the entire spatial domain. Compared with pure statistical methods such as traditional Kriging interpolation, this physical information constraint enables the accurate identification of high-risk areas with drastic changes in evolution degree gradients under sparse arrangements, improves spatial resolution, and avoids non-physical oscillation phenomena that may be caused by statistical interpolation. The sparse and discrete monitoring data are reasonably interpolated and extrapolated to the entire structural space through partial differential equations containing physical mechanisms. The resulting full-field state distribution map can more realistically reflect the spatial continuity and evolution laws of the structural health state than traditional interpolation methods.

[0112] Furthermore, the assessment information from different dimensions (structural strength, evolutionary history, and mutation risk) is scientifically integrated into a single, intuitive comprehensive health index, thereby achieving a final assessment of the overall health of the structure. Specifically, the full-field condition assessment also includes a comprehensive index fusion stage. This fusion stage includes:

[0113] The full-field evolution degree distribution map, the corrected structural strength, and the mutation risk index in the evolutionary characteristics and mutation risk are normalized to form a standardized multi-source data set.

[0114] This embodiment is the data preparation work before data fusion, which aims to eliminate the dimension and value range differences between different data sources. Specifically, all data (for example, the full-field evolution degree distribution map Φ (x, y, z, t), the corrected intensity distribution map f c (x, y, z), and the mutation risk index P risk ) are resampled to a uniform 3D spatial grid. Each data source is normalized, for example, the intensity value f c Convert to a normalized intensity f in the range [0, 1] norm =(f c -f min ) / (f max -f min ), where fmin is the minimum intensity value, f max is the maximum intensity value. The evolution degree and mutation risk index have a value range between [0, 1] and can be used directly. Finally, the standardized data set {1-Φ norm , f norm , 1-P risk}, where Φ norm is the normalized full-field evolution degree, 1-Φ norm and 1-P risk They represent health contribution respectively.

[0115] Based on the spatial information entropy of each data source in the standardized multi-source dataset, the data fusion weight is dynamically calculated.

[0116] This embodiment aims to determine the importance of each evaluation dimension objectively and data-drivenly. Specifically, the spatial information entropy H of each data source i can be calculated. i =-Σp ij ×log(p ij ), where p ij is the probability that the value of the i-th data source falls in the j-th interval. The data source with greater information entropy indicates that its distribution changes in the entire structure space are richer and contains more information, so it should be given a higher weight. The initial weight can be set to w i 0 =H i / ΣH k , where H k is the spatial information entropy of the kth data source. Optionally, the initial weights can be modified according to the reliability of each data source (e.g., reconstruction uncertainty, detection signal-to-noise ratio, etc.) to obtain the final dynamic weight vector [w1, w2, w3].

[0117] The standardized multi-source data sets are calculated using data fusion weights through a nonlinear fusion model including interaction terms to generate a comprehensive health index distribution map, which constitutes the core part of the structural health assessment and risk warning results.

[0118] In this embodiment, the reason for using a nonlinear model is that there is often a coupling amplification effect between different risk factors, and simple linear weighting cannot capture this relationship. The nonlinear model can be expressed as H = H linear +H interact ; Among them, the linear part H linear =w1×(1-Φ norm )+w2×f norm +w3×(1-P risk ). Interaction term part H interact It contains the product of different factors, such as H interact =λ12 ×(1-Φ norm )×f norm +..., where λ 12 is the interaction coefficient between the i-th and j-th factors.

[0119] Furthermore, the nonlinear fusion model also includes a risk avoidance criterion, the implementation of which includes:

[0120] The data source values ​​in the standardized multi-source dataset are checked point by point to determine whether any value is below a preset risk threshold. For example, the risk threshold can be set to 0.4.

[0121] If the value is lower than the risk threshold (for example, the normalized intensity f norm is only 0.3), the weighted fusion calculation at that point is terminated, and the lowest value at that data point is directly adopted as the final comprehensive health index, thereby achieving risk amplification of the short-board effect.

[0122] In other words, in this case, the point's comprehensive health index, H, is directly determined to be 0.3, even though its evolutionary degree and mutation risk indicators are high. This risk avoidance principle follows the wooden barrel theory, ensuring that serious deficiencies in any single dimension are not masked by good performance in other dimensions, making the assessment more conservative and secure. The resulting comprehensive health index distribution map, H(x, y, z), has a value range between [0, 1] and can be color-coded for three-dimensional visualization. Health levels and risk warnings can be determined based on different index ranges (for example, H ≥ 0.85 is healthy, 0.70 ≤ H < 0.85 is basically healthy, etc.).

[0123] This implementation achieves intelligent fusion and risk amplification of multi-source heterogeneous data. Data sources with high information entropy (e.g., those with a dramatic change in evolution degree in areas of seepage damage) are automatically weighted. When critical indicators show signs of danger, the system proactively adopts the most conservative assessment strategy. This addresses the subjectivity inherent in traditional expert weighting methods. Furthermore, through risk avoidance criteria, it prevents severe local deterioration from being masked by normal indicators. This improves the reliability of culvert health assessments and avoids structural failures caused by weak links.

[0124] In a specific embodiment, assuming that detection is performed on a side wall of a long underground culvert, the specific environment and detection parameters of the measurement point are as follows:

[0125] Environmental parameters are: water depth h = 10m, water temperature T = 15°C, water velocity v = 1.5m / s (assuming long-term service), water density ρ w =1000kg / m 3The parameters of the rebound hammer are: impact energy E0=2.207J, calibration parameters A=10, B=100. The original value of underwater rebound measurement R measured at this measuring point w =17.5. 0.3 m near the measuring point, there is a vertical crack with a width of w = 2 mm. The thickness of the component is 0.5 m. Calculate the pressure correction coefficient k p : Hydrostatic pressure p w =ρ w ×g×h=1000×9.8×10=98000Pa. Assuming the pressure influence coefficient λ p =2×10 -6 / Pa, then k p =1 / (1+2×10 -6 ×98000)=1 / 1.196≈0.836. Calculate the temperature correction coefficient k T : Assuming that the concrete age is much greater than 90 days, k age =1. Assuming the temperature sensitivity coefficient α T =-0.0004 / °C, then k T =[1+(-0.0004)×(15-20)]×1=1+0.002=1.002. To simplify the description, the water flow correction coefficient k is tentatively set in this embodiment. e and medium coupling correction factor k c According to the calibration results, they are k e = 0.98 and k c =1.25. (In practical applications, these two items should also be calculated using the corresponding model) Calculate the preliminary comprehensive coefficient k: k=k p ×k T ×k e ×k c =0.836×1.002×0.98×1.25≈1.025. Calculate the second-order interaction correction term Δk and obtain k final :Δk=0.05×(k p -1)×(k T -1)=0.05×(0.836-1)×(1.002-1)≈-0.0000164. k final =k×(1+Δk)=1.025×(1-0.0000164)≈1.02498. For clarity, k is taken here as final =1.025. Calculate the preliminary dry land equivalent strength f' c :f' c =k final ×(A×R w -B)=1.025×(10×17.5-100)=1.025×75=76.875MPa. Calculate the foundation strength reduction factor kcrack :Assume that the critical crack width w cr =5mm. k crack =1-0.15×(w / w cr ) 0.5 =1-0.15×(2 / 5) 0.5 ≈1-0.15×0.632=1-0.0948=0.9052. Calculate the spatial variation defect correction coefficient k d :The defect influence radius σ is assumed to be σ=0.5m. The distance from the measuring point to the defect center is r=0.3m. The influence intensity I(r)=exp(-r 2 / 2σ 2 )=exp(-0.3 2 / (2×0.5 2 ))=exp(-0.09 / 0.5)=exp(-0.18)≈0.835. Defect correction coefficient k d =1-(1-k crack )×I(r)=1-(1-0.9052)×0.835=1-0.0948×0.835≈1-0.0792=0.9208. Calculate the final corrected structural strength f c :f c =k d ×f' c =0.9208×76.875≈70.78MPa.

[0126] This example demonstrates that the original underwater rebound reading of 17.5, without correction, might correspond to a higher strength value. However, through multi-factor environmental correction (correcting the strength from approximately 75 MPa to 76.875 MPa) and correction for key defect effects (further correcting the strength from 76.875 MPa to 70.78 MPa), a strength value more consistent with the actual health of the structure was obtained.

[0127] In another embodiment of the present application, a method for assessing the structural health of a long-distance underground culvert includes:

[0128] Step 1: Collect the acoustic raw data and optical image sequences carried by the ROV, perform spatiotemporal fusion through propagation time difference compensation and partition adaptive correction, identify and locate structural defects, and output spatiotemporal aligned fusion detection data, defect feature maps and defect location coordinate sets.

[0129] Specifically, the acoustic waveform data (sampling rate ≥ 100kHz) and optical image sequence (frame rate ≥ 30fps) of the ROV-mounted sensor are read, and data integrity check and time stamp synchronization are performed to obtain a time-stamped acoustic data stream and a time-stamped image data stream. The time-stamped acoustic data stream and the time-stamped image data stream are read, and the difference in the propagation speed of sound and light in water is calculated (the speed of sound is 1500m / s, the speed of light is 2.25×10 8 m / s), and time-axis translation is performed based on the sensor spacing d to obtain time-aligned acoustic data and time-aligned optical data. The culvert structure CAD drawing is read and the culvert is divided into the top area, side wall area, and bottom area. The acoustic attenuation coefficient α of each area is calculated based on the material properties (concrete thickness, steel distribution) and water flow conditions of each area. i and the optical scattering coefficient β i , get the partition fusion parameter table. Read the time-aligned acoustic data, perform short-time Fourier transform to extract frequency domain features (0.1-50kHz), and identify the defect characteristic frequency; read the time-aligned optical data, perform edge detection and texture analysis, and get the acoustic feature map and visual feature map. Read the acoustic feature map, visual feature map and partition fusion parameter table, and apply weighted fusion F to each partition. i =α i ×A i +β i ×V i (A i is the acoustic characteristic, V i The fused feature map is read and threshold segmentation and connected domain analysis are performed to identify cracks (width > 0.2 mm), leakage (wetted area > 100 cm 2 ), peeling (depth> 5mm) and other defects, and calculate the three-dimensional coordinates of the defects in combination with the ROV position data to obtain the defect position coordinate set, defect type label and defect geometric parameters. Read the fusion feature map and the original acoustic feature map and visual feature map, and calculate the fusion gain index G=SNR fusion / (SNR acoustic +SNR visual ), where SNR fusion is the signal-to-noise ratio of the fused acoustic and optical features, SNR acoustic is the signal-to-noise ratio of the original acoustic signal, SNR visual is the signal-to-noise ratio of the original optical image; verify the fusion effect (G>1.5 is effective fusion), and generate spatiotemporally aligned fusion detection data and defect feature maps.

[0130] Step 2: Read the original rebound measurement value, core test data and defect location coordinate set, perform strength calculation through the underwater-dry land conversion model, and output the corrected concrete strength distribution map and material degradation parameters.

[0131] Specifically, read the rebound measurement original value R of the underwater rebound hammer w (16 readings per measuring point), simultaneously record water depth h, water temperature T and water velocity v, remove outliers (exceeding the mean ±3σ), and obtain a valid rebound value set and environmental parameter set. Read the concrete core sample collected by the core drill, perform laboratory compressive strength test, and record the core sample compressive strength value f c ; Compare the effective rebound value set at the same position, establish the rebound value-strength correspondence, and obtain the calibration data pair. Read the calibration data pair and the environmental parameter set, and calculate the pressure correction term k according to the water pressure influence p =1-0.002h, calculate the temperature correction term k based on the temperature influence T =1+0.01(T-20), the conversion coefficient k=k p ×k T ×k0 (k0 is the base coefficient 1.15) to obtain the underwater conversion coefficient. Read the defect position coordinate set and defect geometric parameters, apply strength reduction to the rebound measurement points within 50cm around the defect according to the defect type: 5-15% for cracks and 10-25% for spalling, and calculate the correction coefficient k d , get the defect correction coefficient distribution. Read the effective rebound value set, underwater conversion coefficient and defect correction coefficient distribution, and calculate the concrete strength f of each measuring point c =k×k d ×(10×R w -100), reconstruct the strength distribution between the measuring points, and obtain the corrected concrete strength distribution diagram. Read the corrected concrete strength distribution diagram and the design strength value, and calculate the strength loss rate η=(f design -f current ) / f design , analyze the intensity gradient ▽f to identify the degradation concentration area, and extract the degradation rate v d =Δf / Δt (based on historical data), to obtain the material degradation parameters. design is the design strength value, f current is the currently measured concrete strength value, and Δf is the strength change.

[0132] Step 3: Read the defect feature map, material degradation parameters and historical monitoring data, and output the evolution state parameters, mutation risk indicators and evolution rate distribution through the continuous evolution model and mutation detection algorithm.

[0133] Specifically, read the defect feature map, identify the water seepage traces and water flow path, and calculate the water seepage area A.w and wet perimeter P w ; Read material degradation parameters, extract porosity growth rate and permeability coefficient change, and obtain permeability characteristic parameters and initial evolution state. Read historical monitoring data (including water level, flow, strain, etc.), perform data cleaning and interpolation, and unify the time resolution to 1 hour; calculate hydraulic gradient i, seepage velocity v s and soil strain rate ε* to obtain multi-physics time series data. Read multi-physics time series data and permeability characteristic parameters, calculate the instantaneous evolution increment dΦ / dt=α×Δi+β×Δq+γ×Δε (weight coefficients α=0.4, β=0.35, γ=0.25 are determined according to the soil type), perform time integration Φ(t)=∫dΦ / dt×dt, and obtain the evolution degree time series. Read the evolution degree time series and determine the current evolution stage based on the threshold: Φ<0.2 is the seepage stage, 0.2≤Φ<0.4 is the seepage deformation stage, 0.4≤Φ<0.6 is the seepage destruction stage, 0.6≤Φ<0.8 is the channel expansion stage, and Φ≥0.8 is the riverbed collapse stage. The current evolution stage identifier and stage transition moment are obtained. Read the evolution degree time series and calculate the first-order derivative ΨΦ / Ψt and the second-order derivative Ψ 2 Φ / Ψt 2 , calculate the mutation sensitivity S m =|Ψ 2 Φ / Ψt 2 | / |ΨΦ / Ψt|, when S m When the sensitivity is >2.5, it is marked as a potential mutation point, and the mutation sensitivity curve and potential mutation moment set are obtained. Read the time series of the acoustic feature map and the potential mutation moment set, analyze the spectrum changes before and after the mutation moment, identify the characteristics of high-frequency noise surge (>20kHz, amplitude increase >10dB) and low-frequency vibration (<100Hz), confirm the mutation type, and obtain the mutation risk index and mutation type label. Optionally, obtaining the mutation risk index and mutation type label includes: reading the acoustic feature map and the potential mutation moment set, for each mutation moment t m , extract [t m -300s, t m Acoustic data within the [+300s] time window; perform short-time Fourier transform (window length 1024, overlap 50%), calculate power spectral density PSD(f, t); divide the frequency bands into low frequency (<100Hz), medium frequency (100Hz-10kHz), and high frequency (>10kHz), and calculate the energy E of each frequency band. low 、E mid 、E high , get the acoustic feature matrix at the mutation moment. Read the acoustic feature matrix at the mutation moment and calculate the statistics of the characteristic time series: high-frequency energy sudden increase rate r high =(E high_max -E high_mean) / E high_mean where E high_max is the maximum value of the high-frequency acoustic energy, and E high_mean is the average acoustic energy in the high-frequency band; the low-frequency vibration intensity I low =∫E low (f < 50 Hz) df; Identify abnormal patterns: waterstop tearing (r high > 3 and lasting > 10 s), structural vibration (I low > 5 times the mean value), particle erosion (energy increase in 2 kHz < f < 5 kHz > 10 dB); Obtain the acoustic abnormal pattern label and abnormal intensity index. Read the set of potential mutation times and acoustic abnormal pattern labels, and calculate the time difference Δt lag = t acoustic - t evolution , where t acoustic is the central time point when a certain acoustic abnormal pattern (such as a sudden increase in high-frequency noise or low-frequency vibration) occurs, and t evolution is the time of a certain evolutionary mutation point identified in the mutation sensitivity analysis; Statistically analyze the typical time delays of different abnormal patterns: waterstop tearing (-60 s ~ +30 s), structural vibration (-120 s + 180 s); Establish the correlation intensity index R ae = exp(-|Δt lag | / τ), where τ = 60 s is the characteristic time; Obtain the acoustic-mutation correlation matrix. Read the preliminary judgment of the mutation type, acoustic abnormal pattern label, and acoustic-mutation correlation matrix, and perform multi-source information fusion: When R ae > 0.7 and the abnormal pattern matches, confirm the mutation type; Calculate the comprehensive risk index R risk = w1×I m + w2×r high + w3×R ae , with weights w1 = 0.3, w2 = 0.4, w3 = 0.3, and I m being the acoustic abnormal intensity index of this mutation event; Risk classification: R risk > 0.8 is high risk, 0.5 - 0.8 is medium risk, < 0.5 is low risk; Output the mutation risk index and mutation type label. Read the evolutionary degree time series, defect location coordinate set, and current evolutionary stage identifier, calculate the local evolutionary rate v Φ = dΦ / dt of each defect point, perform spatial interpolation based on the defect spacing and hydraulic connectivity, and obtain the evolutionary rate distribution and evolutionary state parameters.

[0134] Step 4. Read the evolutionary state parameters, mutation risk index, evolutionary rate distribution, and corrected concrete strength distribution map, and output the full-field health status distribution map, structural health assessment report, and graded warning plan through sparse reconstruction and comprehensive evaluation model.

[0135] Specifically, read the evolution rate distribution and defect location coordinate set, identify the evolution rate extreme points and areas with drastic gradient changes, determine the locations of key monitoring points, ensure that the distance between any two points is less than 50m and cover all high-risk areas, and obtain the optimized monitoring point set. Read the evolution state parameters of the optimized monitoring point set and establish the evolution propagation equation ▽ 2 Φ-(1 / c 2 )Ψ 2 Φ / Ψt 2 =S(x, y, z), where c is the evolution propagation velocity (0.5-2m / day) and S is the source term intensity. Finite element solution is performed to obtain the full-field evolution degree distribution Φ(x, y, z, t). Specifically, the process of obtaining the full-field evolution degree distribution is as follows: read the defect position coordinate set and evolution rate distribution, perform mesh encryption in high gradient areas (|▽Φ|>0.1 / m), with a minimum cell size of 0.5m; use a sparse grid in slow evolution areas (|▽Φ|<0.01 / m), with a maximum cell size of 5m; use Delaunay triangulation to generate an unstructured grid, so that the grid quality index is >0.7; obtain the adaptive finite element grid and grid node coordinates. Read the measured evolution degree value Φ of the optimized monitoring point set measured , perform data quality checks and remove outliers that deviate from the mean by 3σ; use anisotropic kriging interpolation with range parameters of 50m longitudinally (along the direction of water flow), 20m laterally, and 10m vertically; calculate interpolation uncertainty σ kriging , mark the sparse monitoring point area (nearest monitoring point >30m) as low confidence; get the initial evolution degree field Φ0(x, y, z) and confidence distribution map. Read the initial evolution degree field, evolution propagation equation coefficient field and adaptive finite element mesh, and construct the discretized equation group: [M]d 2 Φ / dt 2 +[C]dΦ / dt+[K]Φ=[F], where [M], [C], and [K] are the mass, damping, and stiffness matrices; [F] is the external load (source term) vector; implicit time integration (Newmark-β method, β=0.25, γ=0.5) is used; the convergence criterion is set to ‖Φ (n+1) -Φ n ‖ / ‖Φ n ‖<10 -4 , where Φ (n+1) is the evolution degree value vector after the n+1th iteration; the spatiotemporal evolution degree field Φ(x, y, z, t) is obtained by iterative solution. Read the measured values ​​of the spatiotemporal evolution degree field and the optimized monitoring point set, and calculate the reconstruction error e i =|Φ reconstructed -Φ measured | / Φ measured , where Φ reconstructedThe reconstructed evolution degree value is obtained by cross-validation, removing one monitoring point at a time and reconstructing, and calculating the RMSE (root mean square of the reconstruction error); analyzing the uncertainty propagation, superimposing a ±20% disturbance band in the low confidence area; outputting the full-field evolution degree distribution Φ(x, y, z, t) and the reconstruction uncertainty cloud map. Reading the full-field evolution degree distribution, the corrected concrete strength distribution map and the mutation risk index, calculate the comprehensive health index H=w1×(1-Φ)+w2×(f c / f design )+w3×(1-P risk ), with weights w1=0.4, w2=0.35, and w3=0.25, to obtain a comprehensive health index distribution. The comprehensive health index distribution is read and categorized according to a five-level standard: H ≥ 0.85 for health, 0.70 ≤ H < 0.85 for basic health, 0.55 ≤ H < 0.70 for sub-health, 0.40 ≤ H < 0.55 for disease, and H < 0.40 for severe disease. This generates a color-coded full-field health status distribution map. Evolution state parameters, evolution rate distribution, and mutation risk indicators are read to establish a time series prediction model. The evolution degree Φ(t+Δt) is extrapolated for the next 30 / 90 / 180 days, identifying time windows that may exceed critical thresholds and generating risk evolution prediction curves and warning time nodes. The system reads the site-wide health status distribution map, warning time nodes, and mutation type labels, generating differentiated warnings based on risk levels: blue warning (within 180 days) for routine monitoring, yellow warning (within 90 days) for intensive monitoring, orange warning (within 30 days) for preparation for maintenance, and red warning (within 7 days) for emergency response. It then outputs a graded warning plan. The site-wide health status distribution map, risk evolution prediction curve, graded warning plan, and all key parameters are integrated to generate a structural health assessment report that includes current status diagnosis, evolution trend analysis, risk level assessment, and recommended response measures.

[0136] The preferred embodiments of the present invention are described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all fall within the scope of protection of the present invention.

Claims

1. A method for assessing the structural health of long-distance underground culverts, characterized in that: include: Based on acoustic and optical detection data, fusion processing is performed to identify and output structural defect data including soil type, defect type and geometric parameters; Based on pre-stored underwater rebound and coring test data, combined with structural defect data, underwater-to-dry-land conversion and correction of strength values ​​are performed to determine the corrected structural strength and material degradation parameters; Combining structural defect data, material degradation parameters, and pre-stored historical monitoring data, structural evolution modeling and mutation risk identification are performed to obtain the structural evolution characteristics and mutation risks; The overall status assessment is conducted based on the comprehensively corrected structural strength, evolutionary characteristics and mutation risks to generate structural health assessment and risk warning results.

2. The method according to claim 1, characterized in that Obtain the evolutionary characteristics and mutation risks of the structure, including: Integrate structural defect data, material degradation parameters, and historical monitoring data to construct and solve the evolution function of multi-physics field coupling, and generate an evolution degree time series that quantifies the cumulative evolution degree of the structure; Perform differential analysis on the evolutionary degree time series to identify potential mutation moments that represent risk transitions; Based on the evolution degree time series and the spatial location of the structural defect data, the evolution rate distribution map is derived and generated; The evolutionary degree time series, potential mutation moments and evolutionary rate distribution diagram together constitute the evolutionary characteristics and mutation risks of the structure.

3. The method according to claim 2, characterized in that Generates evolutionary time series, including: Extract physical field indicators including at least hydraulic gradient, seepage rate and soil strain from historical monitoring data, and calculate their respective time increment series; Dynamically calibrate the evolution weight coefficients corresponding to various physical field indicators by referring to soil type and material degradation parameters; The time increment series is weighted and fused using the evolution weight coefficient to obtain the evolution increment series that reflects the instantaneous evolution rate of the structure. The evolution increment series is numerically integrated along the time dimension to generate the evolution degree time series.

4. The method according to claim 3, characterized in that Dynamically calibrate the evolution weight coefficient, including: According to the soil type, configure the basic weight coefficient for each physical field indicator; Extracting a structural degradation degree index from the material degradation parameter, and modifying at least one coefficient value in the basic weight coefficient based on the index to obtain an adjusted weight; Perform normalization calculation on the adjusted weights to generate evolution weight coefficients.

5. The method according to claim 2, characterized in that Identify potential moments of disruption that signal a risk transition, including: Solve the first-order derivative sequence and second-order derivative sequence of the evolution degree time series; By calculating the ratio of the absolute value of the second-order derivative sequence to the absolute value of the first-order derivative sequence at each moment, the mutation sensitivity time series is constructed. Each sensitivity value in the mutation sensitivity time series is compared with a preset threshold, and the moment exceeding the threshold is identified as a potential mutation moment.

6. The method according to claim 5, characterized in that Also includes: For each identified potential mutation moment, extract the acoustic data segment within a predetermined time window before and after the potential mutation moment from the acoustic detection data; Perform spectral analysis on the acoustic data segments to decode whether there are acoustic anomaly patterns associated with specific pre-defined failure mechanisms; If there is an acoustic anomaly pattern, the potential mutation moment is upgraded to a verified mutation risk, and the cause of the mutation risk is classified according to the acoustic anomaly pattern.

7. The method according to claim 1, characterized in that Determine corrected structural strength and material degradation parameters, including: Based on the core test data, the underwater-dry land comprehensive conversion coefficient was constructed and calibrated; Apply the underwater-dry land comprehensive conversion coefficient to convert the underwater rebound into a preliminary dry land equivalent strength value; Based on the defect type and geometric parameters, the preliminary dry land equivalent strength value is locally corrected to obtain the corrected structural strength. Based on the comparison with the preset design strength, the material degradation parameter is generated.

8. The method according to claim 7, characterized in that Construct and calibrate the underwater-dry land comprehensive conversion coefficient, including: Quantify and determine independent environmental correction factors for the physical effects of the underwater environment, including at least two of the following: The pressure correction factor that characterizes the effect of hydrostatic pressure on the rebound value; Temperature correction coefficient that characterizes the effect of water temperature on the elastic modulus of concrete; Scour correction factor that characterizes the effect of water scour on the surface hardness of concrete; The medium coupling coefficient characterizes the effect of water medium on the energy transfer of rebound impact; By coupling and integrating all environmental correction factors, the underwater-dry land comprehensive conversion coefficient is generated.

9. The method according to claim 7, characterized in that Obtain corrected structural strength, including: For each defect in the structural defect data, a three-dimensional impact field model is constructed based on its geometric parameters, in which the impact intensity decays with increasing spatial distance. Based on the defect type, set the foundation strength reduction factor that represents its maximum impact; The three-dimensional impact field model and foundation strength reduction factor are integrated to calculate the defect correction factor field that changes continuously in space. The defect correction factor field is applied to the preliminary dry-ground equivalent strength values ​​to obtain the corrected structural strength.

10. The method according to claim 1, characterized in that Perform full-field state evaluation, including a full-field reconstruction phase to generate a continuous state distribution, specifically: The evolutionary parameters contained in the evolutionary characteristics and mutation risks are arranged as sparse state monitoring points in the three-dimensional space of the culvert; Construct physical constraint equations that describe the spatiotemporal evolution of the propagation laws of the evolving state in the structured medium; The data of sparse state monitoring points are used as solution boundaries or initial conditions to numerically solve the physical constraint equations and generate a full-field evolution degree distribution map that continuously covers the entire culvert.

Images