Method and system for analyzing durability of ship lock concrete structure under coupling of multiple factors of erosion

Abstract

The application discloses a ship lock concrete structure durability analysis method and system under the coupling action of multiple factors, which comprises the following steps: acquiring internal structure images of different erosion areas of a ship lock concrete test piece; identifying and quantitatively analyzing the internal structure images of different erosion areas by a deep learning image segmentation algorithm to generate an erosion structure factor set; calculating environmental characteristics corresponding to multiple environmental parameters by a multi-physical field coupling inversion algorithm; constructing a three-dimensional microstructure model of different erosion areas; dynamically simulating the three-dimensional microstructure model according to the environmental characteristics; performing spatio-temporal correlation analysis on simulation result data by a hypergraph neural network to generate a prediction result; and comparing an expert knowledge base and performing construction maintenance based on text information of the prediction result by an Euclidean distance similarity algorithm. The application introduces high-precision three-dimensional model multi-factor coupling simulation, and improves the ship lock concrete structure analysis precision through high-precision coupling prediction of the hypergraph neural network.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence analysis technology for the durability of concrete structures in ship locks, and more specifically, to a method and system for analyzing the durability of concrete structures in ship locks under the coupled effects of multiple erosion factors. Background Technology

[0002] Ship locks, as key facilities in waterway engineering, play a vital role in ensuring the safe and efficient navigation of vessels. The durability and stability of their concrete structures directly affect the lock's service life and operational safety. However, the environment in which ship locks are located is extremely complex, facing erosion from multiple factors that interact and have a serious negative impact on the concrete structure. Ship locks are typically located near rivers, lakes, or oceans, and their concrete structures are constantly exposed to the aquatic environment. Chemical substances in the water, such as salts (chloride ions, sulfate ions, etc.) and acidic substances, can penetrate into the concrete through infiltration and diffusion. Furthermore, the environment in which ship locks are located can also be affected by temperature changes.

[0003] The effects of multi-factor coupling on the concrete structure of ship locks are complex and severe. While some progress has been made in understanding the coupling effects on concrete structures, research on the effects of multi-factor coupling is still insufficient. Current shortcomings in research on coupling effects include the following: 1) A high-precision three-dimensional multi-factor coupled simulation model was not used for erosion simulation; 2) The system does not use artificial intelligence to accurately predict the coupling of multiple factors in the erosion of the lock concrete structure, and when the erosion of the concrete structure is predicted to be at risk, it is unable to take accurate and rapid maintenance measures for the lock concrete structure. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for analyzing the durability of ship lock concrete structures under the coupled effects of multiple erosion factors, so as to solve the above-mentioned problems existing in the prior art.

[0005] The application is as follows: A method for durability analysis of concrete structures in ship locks under the coupled effects of multiple erosion factors, the method comprising the following steps: The internal structure images of different erosion zones of the concrete specimen from the lock were obtained; the internal structure images of different erosion zones were identified and quantitatively analyzed by a deep learning image segmentation algorithm to generate an erosion structure factor set, which includes aggregate distribution, pore network and interface transition zone thickness. Based on a distributed optical fiber sensor network and an electrochemical sensor array, multi-factor environmental parameters of the lock concrete structure are monitored in real time. A multi-physics coupling inversion algorithm is used to calculate the environmental characteristics corresponding to the multi-factor environmental parameters. The multi-factor environmental parameters include at least dynamic water pressure, flow velocity, chloride / sulfate ion concentration, pH value and temperature gradient. The Sobol analysis method was used to perform sensitivity preprocessing analysis on the erosion structure factor set. Three-dimensional microstructure models of different erosion areas were constructed from the data after sensitivity preprocessing. An improved random aggregate delivery algorithm was used to optimize and remodel the constructed three-dimensional microstructure models. Dynamic simulations are performed on a three-dimensional microstructure model based on the environmental characteristics of different erosion zones. The three-dimensional microstructure model embeds a coupled phase-field damage equation and a spatiotemporally variable fractional diffusion equation. The coupled phase-field damage equation is used to identify crack propagation paths and generate structural defect evolution data corresponding to these paths. The spatiotemporally variable fractional diffusion equation is used to identify ion erosion fronts and generate structural defect evolution data corresponding to these fronts. A hypergraph neural network is used to perform spatiotemporal correlation analysis on the structural defect evolution data to predict the trends in crack propagation rate and ion penetration depth, generating prediction results. Based on the text information of the prediction results, keyword features are extracted from the text information using natural language processing (NLP) technology. The features of historical lock concrete structure erosion projects in the expert knowledge base are compared using the Euclidean distance similarity algorithm. Based on the comparison values, an early warning mechanism is established for construction and maintenance.

[0006] Furthermore, the acquisition of internal structural images of different erosion zones of the lock concrete specimen, and the identification and quantitative analysis of the internal structural images of different erosion zones using a deep learning image segmentation algorithm, includes: A full-range laser confocal 3D scanning imaging was performed on the concrete structure of the ship lock to be monitored, and a 3D image sequence of different erosion areas was obtained. Multi-directional projection 3D reconstruction is performed on 3D image sequences of different erosion areas to construct 3D reconstructed images; U-Net++ deep learning network is used to perform material visual semantic recognition on 3D reconstructed images to identify aggregate, cement paste, pores and interface transition zone components. The spatial distribution characteristics of each component are quantified by morphological algorithms to generate the spatial distribution characteristics of materials in different erosion zones. The spatial distribution characteristics include the volume fraction, gradation curve and orientation distribution of aggregates; the volume fraction, gradation curve and orientation distribution of cement paste; the connectivity, tortuosity and pore size distribution of pores; and the average thickness and microcrack density of the interface transition zone. The spatial distribution characteristics of materials in different erosion zones are analyzed by structural quantification to generate erosion structure factors for different erosion zones.

[0007] Furthermore, the multi-factor environmental parameters of the lock's concrete structure are monitored in real time based on a distributed optical fiber sensor network and an electrochemical sensor array; the environmental characteristics corresponding to the multi-factor environmental parameters are calculated using a multi-physics coupling inversion algorithm, including: A Kalman filter algorithm is used to fuse multi-sensor data to eliminate environmental noise and drift error. The multi-sensor includes the total number of sensor types in a distributed fiber optic sensor network and an electrochemical sensor array. The PTP protocol, a timestamp synchronization technology, ensures the time consistency of data acquired by the fiber optic sensor network and the electrochemical sensor array; the Kriging interpolation method is used to preprocess the discrete point data in the data acquired by the fiber optic sensor network and the electrochemical sensor array. The temperature field inversion algorithm of the multiphysics field coupling inversion algorithm is used to fit the temperature trend inversion algorithm of each region of the lock concrete temperature fluctuation data to obtain the temperature change curves of different erosion areas. The hydraulic scouring intensity data of the lock concrete was fitted and calculated using the fluid field inversion algorithm of the multiphysics field coupling inversion algorithm to obtain the hydraulic scouring intensity variation curves of different erosion areas. The fluid field inversion algorithm adopts the Navier-Stokes equation. The ion diffusion field inversion algorithm of the multiphysics field coupling inversion algorithm is used to fit the ion diffusion situation inversion algorithm of each region of the lock concrete ion diffusion data to obtain the ion change curves of different erosion areas. The ion change curves include chloride ion change curves and sulfate ion change curves. The ion diffusion field inversion algorithm adopts the Nernst-Planck equation. Environmental stress correlation mining was performed on the temperature change curves, hydraulic scour intensity change curves, and ion change curves of different erosion areas to generate the corresponding environmental characteristics of different erosion areas.

[0008] Furthermore, the sensitivity preprocessing analysis of the erosion structure factor set using the Sobol analysis method includes: Set the parameter ranges for aggregate distribution, pore network, and interface transition zone thickness, and sample each parameter according to the first variation sampling principle of the Sobol analysis method. The finite element method (FEM) is used to solve the problem based on the sampled parameters. The maximum stress of the internal structure in different erosion zones is calculated, and the sensitivity of each parameter, the strength of the nonlinear correlation effect between parameters, and the variable sensitivity discrimination weight coefficient are calculated. The mathematical model of Sobol analysis is expressed as follows: , , , in, Indicates sensitivity, This indicates the strength of the nonlinear correlation effect between parameters; Indicates the sensitivity discrimination weight coefficient of the variable; This represents the i-th output result of the Sobol model; This represents the (i+1)th output result of the Sobol model; This indicates the model's running result after the variables are modified according to a preset percentage; This represents the rate of change of the variable with respect to its initial value after the i-th run of the model; This represents the rate of change of the variable after the (i+1)th calculation of the model compared to the initial variable value; n is the number of calculations in the Sobol model. The influence of each parameter on the maximum stress of the internal structure and the influence between each parameter were analyzed. Based on the Sobol analysis results, weakly correlated factors were eliminated.

[0009] Furthermore, the construction of three-dimensional mesoscopic structural models of different erosion zones using the improved random aggregate delivery algorithm on the pre-processed data includes: Multi-frequency Gaussian filtering was applied to the internal structure image of the eroded region to obtain a Gaussian-filtered three-dimensional microstructure image. Based on the internal three-dimensional brightness and darkness differentiation distribution information, we perform fine-grained texture feature analysis on the Gaussian filtered three-dimensional microstructure image and extract the fine-grained texture features of the image. K-Means clustering analysis was performed on the fine-grained features of image texture to generate fine-grained clustering data of eroded area texture. The fine-grained clustering data of eroded area texture includes concrete porosity, crack fractal dimension and interface transition zone thickness. Based on the erosion structure factor set of different erosion regions, the erosion structure model of concrete pore tortuosity, crack fractal dimension and interface transition zone thickness is constructed to build a three-dimensional microstructure model. An improved random aggregate distribution algorithm was used to optimize the three-dimensional microstructure model.

[0010] Furthermore, the optimization of the three-dimensional mesostructure model using the improved random aggregate distribution algorithm includes: Reduce aggregate volume fraction in highly eroded areas; generate aggregate shape using hyperquadratic surface equation; use hierarchical bounding box (BVH) for aggregate location detection; construct AABB bounding box for each aggregate location; quickly retrieve neighboring aggregate locations using spatial hash table, and simultaneously construct an aggregate location topology map based on spatial hash table. Multi-scale porosity is generated based on porosity tortuosity; the crack fractal dimension path is constructed using the random midpoint displacement method; and the crack is optimized based on the crack fractal dimension path and aggregate position topology. A non-uniform interface transition zone thickness layer is generated on the aggregate surface based on fine-grained clustering data of erosion area texture and aggregate location topology map; the elastic modulus of the interface transition zone thickness is optimized by exponential decay from aggregate to slurry.

[0011] Furthermore, a spacetime-variable order fractional diffusion equation is used to analyze the non-Fickian diffusion behavior of chloride and sulfate ions: , exp(- ). , in, Indicates ion concentration. denoted by the ion diffusion coefficient, where j represents chloride or sulfate ions; (t,x) represents the time fraction order, 0 < t,x. (t,x)≤1 is used to describe the diffusion memory effect, which varies with time and location; β(t,x) represents the spatial fractional order, 1<β(t,x)≤2, which characterizes nonlocal diffusion behavior; t represents the monitoring time step, and x represents the coordinates of the crack propagation path; Indicates the diffusion activation energy. Indicates porosity. Indicates the initial porosity. The factor representing the pre-diffusion coefficient of ions is denoted by m, which represents the empirical coefficient. represents the vector differential operator used for partial derivative operations with respect to spatial variables; R represents the universal gas constant; T represents the absolute temperature; v represents the water flow velocity, which is solved by the Navier-Stokes equations. The coupled phase-field damage equation, which introduces a continuous phase-field variable s(x,t)∈[0,1] to describe crack propagation, is expressed as follows: , Where s represents the phase field variable, s=0 indicates no damage, and s=1 indicates complete fracture; g(s) =1- , g(s) represents the stiffness degradation function, which describes the reduction of the stiffness of concrete material by damage; Represents tensile elastic strain energy; This represents the compressive elastic strain energy; The critical fracture energy is used to describe the crack resistance of concrete materials; L represents the regularization length, which describes the width controlling crack propagation. This represents the phase field gradient.

[0012] Furthermore, the method of using a hypergraph neural network to perform spatiotemporal correlation analysis on structural defect evolution data to predict the trends in crack propagation rate and ion penetration depth includes: The structural defect evolution data corresponding to the crack propagation path is input into the first layer of the hypergraph neural network of the preset prediction model to obtain the first hypergraph vector; wherein, the first hypergraph vector represents the correlation between the structural defect evolution data features corresponding to the crack propagation path and the crack propagation rate; the preset prediction model is a model obtained based on the hypergraph neural network. The structural defect evolution data corresponding to the ion erosion front is input into the second layer of the preset prediction model's hypergraph neural network to obtain the second hypergraph vector; wherein, the second hypergraph vector represents the correlation between the structural defect evolution data corresponding to the ion erosion front to be processed and the trend of ion penetration depth change; The first hypergraph vector and the second hypergraph vector are concatenated and weighted to obtain the concatenated weighted feature vector to be processed. Based on the preset prediction model, the weighted feature vector to be processed is predicted to obtain the prediction result.

[0013] Furthermore, the text information based on the prediction results is used to extract keyword features through Natural Language Processing (NLP) technology. The features of historical lock concrete structure erosion projects in the expert knowledge base are compared using the Euclidean distance similarity algorithm. Based on the comparison values, an early warning mechanism is established for construction and maintenance, including: The text information of the prediction results is used to extract keyword features through natural language NLP technology to generate keyword feature vectors for the prediction results; the historical lock concrete structure erosion project in the expert knowledge base is also used to extract keyword features through natural language NLP technology to generate project keyword feature vectors. The similarity between the predicted keyword feature vector and the project keyword feature vector is calculated using the Euclidean distance similarity algorithm. A preset similarity threshold is used. If the similarity is greater than the preset threshold, the project that has already experienced erosion of the lock concrete structure is identified, and the predicted crack propagation rate, ion penetration depth, and corresponding solution for the project are output. If the similarity is less than or equal to the preset threshold, the project that has experienced erosion of the lock concrete structure with the closest similarity is identified, and the predicted crack propagation rate, ion penetration depth, and corresponding solution for the project are output. Construction and maintenance are then carried out, including at least preventive coating maintenance, local repair maintenance, and structural replacement maintenance.

[0014] The durability analysis of a ship lock concrete structure under multi-factor erosion coupling includes at least one processor and a memory; the memory stores computer execution instructions; the at least one processor executes the computer execution instructions stored in the memory, causing the at least one processor to perform any one of the methods described in the method for durability analysis of a ship lock concrete structure under multi-factor erosion coupling.

[0015] Compared with the prior art, the embodiments of the present invention achieve the following beneficial effects: This invention provides a method for acquiring internal structural images of different erosion zones of a ship lock concrete specimen; using a deep learning image segmentation algorithm to identify and quantify the internal structural images of different erosion zones to generate an erosion structure factor set; employing a multi-physics coupling inversion algorithm to calculate environmental characteristics corresponding to multi-factor environmental parameters; constructing a three-dimensional microstructural model of different erosion zones; performing dynamic simulation of the three-dimensional microstructural model based on environmental characteristics; using a hypergraph neural network to perform spatiotemporal correlation analysis on the simulation result data to generate prediction results; and comparing the textual information of the prediction results with an expert knowledge base using an Euclidean distance similarity algorithm and performing construction maintenance. This invention introduces high-precision three-dimensional multi-factor coupling simulation and improves the accuracy of coupling analysis of ship lock concrete structures through high-precision coupling influence prediction using a hypergraph neural network. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the durability analysis method for ship lock concrete structures under multi-factor erosion coupling provided in this embodiment of the invention. Detailed Implementation

[0017] The present invention will now be described in detail with reference to the accompanying drawings.

[0018] Example 1

[0019] This invention provides a method for durability analysis of ship lock concrete structures under multi-factor erosion coupling effects, such as... Figure 1 The method includes: S1. Obtain internal structure images of different erosion zones of the lock concrete specimen; use deep learning image segmentation algorithm to identify and quantify the internal structure images of different erosion zones to generate an erosion structure factor set, which includes aggregate distribution, pore network and interface transition zone thickness. S2. Based on a distributed optical fiber sensor network and an electrochemical sensor array, real-time monitoring of multi-factor environmental parameters of the lock concrete structure is conducted; a multi-physics field coupled inversion algorithm is used to calculate the environmental characteristics corresponding to the multi-factor environmental parameters; the multi-factor environmental parameters include at least dynamic water pressure, flow velocity, chloride / sulfate ion concentration, pH value and temperature gradient. S3. The Sobol analysis method is used to perform sensitivity preprocessing analysis on the erosion structure factor set. Three-dimensional microstructure models of different erosion areas are constructed based on the data after sensitivity preprocessing. An improved random aggregate delivery algorithm is used to optimize and remodel the constructed three-dimensional microstructure models. S4. Dynamic simulation is performed on the three-dimensional microstructure model based on the environmental characteristics of different erosion zones. The three-dimensional microstructure model embeds a coupled phase-field damage equation and a spatiotemporally variable fractional diffusion equation. The coupled phase-field damage equation is used to identify crack propagation paths and generate structural defect evolution data corresponding to the crack propagation paths. The spatiotemporally variable fractional diffusion equation is used to identify ion erosion fronts and generate structural defect evolution data corresponding to the ion erosion fronts. A hypergraph neural network is used to perform spatiotemporal correlation analysis on the structural defect evolution data to predict the crack propagation rate and ion penetration depth variation trends, and generate prediction results. It should be noted that the structural defect evolution data corresponding to the crack propagation path includes: crack area, maximum penetration depth of the crack along the thickness direction, fractal dimension, and vector of crack propagation direction; the structural defect evolution data corresponding to the ion erosion front includes: erosion penetration depth, erosion concentration gradient, porosity, and ion penetration enhancement coefficient of the interface transition zone. S5. Based on the text information of the prediction results, keyword features are extracted from the text information using natural language processing (NLP) technology. The features of historical lock concrete structure erosion projects in the expert knowledge base are compared using the Euclidean distance similarity algorithm. An early warning mechanism is established based on the comparison value for construction and maintenance.

[0020] Specifically, this invention involves acquiring internal structural images of different erosion zones of a ship lock concrete specimen; using a deep learning image segmentation algorithm to identify and quantify these images, generating an erosion structure factor set; employing a multi-physics coupling inversion algorithm to calculate environmental characteristics corresponding to multiple environmental parameters; constructing three-dimensional microstructural models of different erosion zones; dynamically simulating these models based on environmental characteristics; using a hypergraph neural network to perform spatiotemporal correlation analysis on the simulation results data, generating prediction results; and comparing the predicted results with an expert knowledge base using an Euclidean distance similarity algorithm and performing construction maintenance. This invention achieves high-precision three-dimensional multi-factor coupling simulation and improves the accuracy of coupling analysis of ship lock concrete structures through high-precision coupling influence prediction using a hypergraph neural network.

[0021] In the above embodiments, specifically, acquiring internal structural images of different erosion zones of the lock concrete specimen; and identifying and quantifying the internal structural images of different erosion zones using a deep learning image segmentation algorithm includes: A full-range laser confocal 3D scanning imaging was performed on the concrete structure of the ship lock to be monitored, and a 3D image sequence of different erosion areas was obtained. Multi-directional projection 3D reconstruction is performed on 3D image sequences of different erosion areas to construct 3D reconstructed images; U-Net++ deep learning network is used to perform material visual semantic recognition on 3D reconstructed images to identify aggregate, cement paste, pores and interface transition zone components. The spatial distribution characteristics of each component are quantified by morphological algorithms to generate the spatial distribution characteristics of materials in different erosion zones. The spatial distribution characteristics include the volume fraction, gradation curve and orientation distribution of aggregates; the volume fraction, gradation curve and orientation distribution of cement paste; the connectivity, tortuosity and pore size distribution of pores; and the average thickness and microcrack density of the interface transition zone. The spatial distribution characteristics of materials in different erosion zones are analyzed by structural quantification to generate erosion structure factors for different erosion zones. It should be noted that the specific process for performing structural quantitative analysis on the spatial distribution characteristics of materials in different erosion regions to generate erosion structure factors for different erosion regions is as follows: Based on the spatial distribution characteristics of materials in the eroded area, a quantitative structural analysis is performed to generate the proportion characteristics of concrete aggregate and cement paste. Material morphology characteristics are analyzed in the eroded area to generate material morphology features; Quantitative analysis of concrete erosion structure was conducted on the proportion characteristics of concrete aggregates and cement paste, as well as the material morphology characteristics, to generate a set of erosion structure factors for different erosion zones.

[0022] In the above embodiments, specifically, the real-time monitoring of multi-factor environmental parameters of the lock concrete structure based on a distributed optical fiber sensor network and an electrochemical sensor array; and the calculation of the environmental characteristics corresponding to the multi-factor environmental parameters using a multi-physics coupling inversion algorithm, including: A Kalman filter algorithm is used to fuse multi-sensor data to eliminate environmental noise and drift error. The multi-sensor includes the total number of sensor types in a distributed fiber optic sensor network and an electrochemical sensor array. The PTP protocol, a timestamp synchronization technology, ensures the time consistency of data acquired by the fiber optic sensor network and the electrochemical sensor array; the Kriging interpolation method is used to preprocess the discrete point data in the data acquired by the fiber optic sensor network and the electrochemical sensor array. The temperature field inversion algorithm of the multiphysics field coupling inversion algorithm is used to fit the temperature trend inversion algorithm of each region of the lock concrete temperature fluctuation data to obtain the temperature change curves of different erosion areas. The hydraulic scouring intensity data of the lock concrete was fitted and calculated using the fluid field inversion algorithm of the multiphysics field coupling inversion algorithm to obtain the hydraulic scouring intensity variation curves of different erosion areas. The fluid field inversion algorithm adopts the Navier-Stokes equation. The ion diffusion field inversion algorithm of the multiphysics field coupling inversion algorithm is used to fit the ion diffusion situation inversion algorithm of each region of the lock concrete ion diffusion data to obtain the ion change curves of different erosion areas. The ion change curves include chloride ion change curves and sulfate ion change curves. The ion diffusion field inversion algorithm adopts the Nernst-Planck equation. Environmental stress correlation mining was performed on the temperature change curves, hydraulic scour intensity change curves, and ion change curves of different erosion areas to generate the corresponding environmental characteristics of different erosion areas.

[0023] It should be noted that the environmental characteristics specifically include: 1. Characteristics of thermo-mechanical-chemical coupling strength: Temperature-scour synergy coefficient, which characterizes the accelerating effect of temperature gradient on hydraulic scour; Ion transport enhancement factor, which characterizes the enhancement of ion diffusion coefficient by hydraulic shear force; Freeze-thaw-erosion coupling index, which characterizes the synergistic damage risk of low temperature and chloride ion concentration. 2. Spatiotemporal distribution characteristics: hotspot area coordinates, representing the coordinate location of the erosion concentration area; erosion front advance rate, representing the predicted remaining service life; periodic fluctuation spectrum, representing the cyclical characteristics of environmental loads; 3. Material response characteristics: microcrack density, characterizing the crack length per unit volume; effective porosity, characterizing the proportion of connected pore volume; The environmental stress correlation mining specifically includes: using the Granger causal analysis algorithm to test whether temperature / hydraulic scouring is a causal factor for ion diffusion; ranking the importance of generated environmental features using random forest feature importance ranking; and updating the joint probability distribution of each influencing factor corresponding to the environmental features in real time using a dynamic Bayesian network. The control method for the temperature field inversion algorithm includes: , in, Indicates the density of concrete. Let represent the specific heat capacity of concrete, k represent the thermal conductivity of concrete, and Q represent the heat released during hydration of concrete. This represents the rate of change of temperature over time. This represents the energy transport caused by heat conduction, calculated using the finite element / finite volume method.

[0024] In the above embodiments, specifically, the sensitivity preprocessing analysis of the erosion structure factor set using the Sobol analysis method includes: Set the parameter ranges for aggregate distribution, pore network, and interface transition zone thickness, and sample each parameter according to the first variation sampling principle of the Sobol analysis method. The finite element method (FEM) is used to solve the problem based on the sampled parameters. The maximum stress of the internal structure in different erosion zones is calculated, and the sensitivity of each parameter, the strength of the nonlinear correlation effect between parameters, and the variable sensitivity discrimination weight coefficient are calculated. The mathematical model of Sobol analysis is expressed as follows: , , , in, Indicates sensitivity, This indicates the strength of the nonlinear correlation effect between parameters; Indicates the sensitivity discrimination weight coefficient of the variable; This represents the i-th output result of the Sobol model; This represents the (i+1)th output result of the Sobol model; This indicates the model's running result after the variables are modified according to a preset percentage; This represents the rate of change of the variable with respect to its initial value after the i-th run of the model; This represents the rate of change of the variable after the (i+1)th calculation of the model compared to the initial variable value; n is the number of calculations in the Sobol model. The influence of each parameter on the maximum stress of the internal structure and the influence between each parameter were analyzed. Based on the Sobol analysis results, weakly correlated factors were eliminated.

[0025] In the above embodiments, specifically, the step of constructing three-dimensional mesoscopic structural models of different erosion regions using the improved random aggregate delivery algorithm on the data after sensitivity preprocessing includes: Multi-frequency Gaussian filtering was applied to the internal structure image of the eroded region to obtain a Gaussian-filtered three-dimensional microstructure image. Based on the internal three-dimensional brightness and darkness differentiation distribution information, we perform fine-grained texture feature analysis on the Gaussian filtered three-dimensional microstructure image and extract the fine-grained texture features of the image. K-Means clustering analysis was performed on the fine-grained features of image texture to generate fine-grained clustering data of eroded area texture. The fine-grained clustering data of eroded area texture includes concrete porosity, crack fractal dimension and interface transition zone thickness. Based on the erosion structure factor set of different erosion regions, the erosion structure model of concrete pore tortuosity, crack fractal dimension and interface transition zone thickness is constructed to build a three-dimensional microstructure model. An improved random aggregate distribution algorithm was used to optimize the three-dimensional microstructure model.

[0026] It should be noted that the three-dimensional microstructure model in this embodiment is designed using Neper software. Neper software generates the concrete microstructure based on Voronoi tessellation and supports the placement of non-spherical aggregates. The multiphysics phase field platform in this embodiment adopts the MOOSE platform. The geometric model is generated by Neper software and combined with the coupled phase field damage equation and the spatiotemporal variable order fractional diffusion equation solver of the MOOSE platform. The MOOSE platform has a built-in phase field module.

[0027] In the above embodiments, specifically, the optimization of the three-dimensional mesostructure model using the improved random aggregate distribution algorithm includes: Reduce aggregate volume fraction in highly eroded areas; generate aggregate shape using hyperquadratic surface equations; use hierarchical bounding box (BVH) for aggregate location detection; construct AABB bounding boxes for each aggregate location; quickly retrieve neighboring aggregate locations using a spatial hash table, and simultaneously construct an aggregate location topology map based on the spatial hash table; this step is used for adaptive aggregate placement in eroded areas. Multi-scale pores are generated based on pore tortuosity; crack fractal dimension paths are constructed using the random midpoint displacement method; cracks are optimized based on crack fractal dimension paths and aggregate position topology; this step is used for pore-crack network optimization. A non-uniform interface transition zone thickness layer is generated on the aggregate surface based on fine-grained clustering data of erosion zone texture and aggregate location topology map; the elastic modulus of the interface transition zone thickness is optimized by exponential decay from aggregate to slurry; this step is used for interface transition zone thickness gradient control.

[0028] In the above embodiments, specifically, a spatiotemporally variable fractional diffusion equation is used to analyze the non-Fickian diffusion behavior of chloride and sulfate ions: , exp(- ). , in, Indicates ion concentration. denoted by the ion diffusion coefficient, where j represents chloride or sulfate ions; (t,x) represents the time fraction order, 0 < t,x. (t,x)≤1 is used to describe the diffusion memory effect, which varies with time and location; β(t,x) represents the spatial fractional order, 1<β(t,x)≤2, which characterizes nonlocal diffusion behavior; t represents the monitoring time step, and x represents the coordinates of the crack propagation path; Indicates the diffusion activation energy. Indicates porosity. Indicates the initial porosity. The factor representing the pre-diffusion coefficient of ions is denoted by m, which represents the empirical coefficient. represents the vector differential operator used for partial derivative operations with respect to spatial variables; R represents the universal gas constant; T represents the absolute temperature; v represents the water flow velocity, which is solved by the Navier-Stokes equations. The hydraulic erosion wear equation, used to supplement the identification of crack propagation paths, is expressed as follows: , Where h represents the depth of wear on the concrete surface. This represents the wear coefficient of the concrete surface. Indicates water flow density. Represents shear stress. This represents the critical shear stress. Represents the Heaviside step function; The coupled phase-field damage equation, which introduces a continuous phase-field variable s(x,t)∈[0,1] to describe crack propagation, is expressed as follows: , Where s represents the phase field variable, s=0 indicates no damage, and s=1 indicates complete fracture; g(s) =1- , g(s) represents the stiffness degradation function, which describes the reduction of the stiffness of concrete material by damage; Represents tensile elastic strain energy; Represents compressive elastic strain energy; The critical fracture energy is used to describe the crack resistance of concrete materials; L represents the regularization length, which describes the width controlling crack propagation. This represents the phase field gradient.

[0029] In the above embodiments, specifically, the step of using a hypergraph neural network to perform spatiotemporal correlation analysis on structural defect evolution data and predict the trends of crack propagation rate and ion penetration depth includes: The structural defect evolution data corresponding to the crack propagation path is input into the first layer of the hypergraph neural network of the preset prediction model to obtain the first hypergraph vector; wherein, the first hypergraph vector represents the correlation between the structural defect evolution data features corresponding to the crack propagation path and the crack propagation rate; the preset prediction model is a model obtained based on the hypergraph neural network. The structural defect evolution data corresponding to the ion erosion front is input into the second layer of the preset prediction model's hypergraph neural network to obtain the second hypergraph vector; wherein, the second hypergraph vector represents the correlation between the structural defect evolution data corresponding to the ion erosion front to be processed and the trend of ion penetration depth change; The first hypergraph vector and the second hypergraph vector are concatenated and weighted to obtain the concatenated weighted feature vector to be processed. Based on the preset prediction model, the weighted feature vector to be processed is predicted to obtain the prediction result.

[0030] It should be noted that, for each type of structural defect evolution data, its expansion characteristics under different conditions (such as crack propagation rate and ion penetration depth variation trend) are analyzed. Each type of structural defect evolution data is analyzed individually, its expansion process is simulated, and the expansion range and speed are recorded. Finite element analysis or other numerical simulation methods are used to evaluate the expansion impact of structural defect evolution, generating structural defect evolution data corresponding to the crack propagation path and structural defect evolution data corresponding to the ion erosion front. The structural defect evolution data corresponding to the crack propagation path and the structural defect evolution data corresponding to the ion erosion front reflect the expansion evolution characteristics of all types of structural defects. Based on the generated structural defect evolution, the defects are located, and the three-dimensional microstructural model is marked with defects, indicating their location and type to ensure clear visualization. Based on the obtained expansion evolution characteristic data, a hypergraph neural network is constructed to perform spatiotemporal correlation analysis on the mutual influence between the structural defect evolution data corresponding to the crack propagation path and the structural defect evolution data corresponding to the ion erosion front, as well as their comprehensive impact on the performance of the concrete erosion structure. The coupling effects between each defect are recorded, and the final prediction analysis results are obtained, generating a prediction result analysis report with text information. The prediction result analysis report with text information includes crack propagation prediction text, ion erosion prediction text, and multi-factor coupling analysis text; The crack propagation prediction text specifically includes: crack location coordinates, representing precise location of high-risk areas; current maximum width, representing whether emergency repair is triggered; future propagation rate, representing the predicted remaining lifespan; main propagation direction, representing guidance for reinforcement scheme design; and fractal dimension change, representing assessment of crack branching risk. The ion erosion prediction text includes: chloride ion penetration depth; erosion front gradient, which characterizes the location of weak areas in the seepage prevention system; pore filling state, which characterizes the assessment of the degree of chemical erosion damage; and time-varying curve of diffusion coefficient, which characterizes the quantification of damage-transport coupling effect. The multi-factor coupling analysis text includes the water pressure-fracture synergy coefficient, which characterizes the contribution of hydraulic scouring; the temperature-erosion coupling index, which characterizes the formulation of seasonal maintenance strategies; and the stress concentration-permeability correlation, which characterizes the identification of high-risk areas of mechanical-chemical interaction.

[0031] In the above embodiments, specifically, the text information based on the prediction results is used to extract keyword features through Natural Language Processing (NLP) technology, and the features of historical lock concrete structure erosion projects in the expert knowledge base are compared using the Euclidean distance similarity algorithm. Based on the comparison values, an early warning mechanism is established for construction and maintenance, including: The text information of the prediction results is used to extract keyword features through natural language NLP technology to generate keyword feature vectors for the prediction results; the historical lock concrete structure erosion project in the expert knowledge base is also used to extract keyword features through natural language NLP technology to generate project keyword feature vectors. The similarity between the predicted keyword feature vector and the project keyword feature vector is calculated using the Euclidean distance similarity algorithm. A preset similarity threshold is used. If the similarity is greater than the preset threshold, the project that has already experienced erosion of the lock concrete structure is identified, and the predicted crack propagation rate, ion penetration depth, and corresponding solution for the project are output. If the similarity is less than or equal to the preset threshold, the project that has experienced erosion of the lock concrete structure with the closest similarity is identified, and the predicted crack propagation rate, ion penetration depth, and corresponding solution for the project are output. Construction and maintenance are then carried out, including at least preventive coating maintenance, local repair maintenance, and structural replacement maintenance.

[0032] It should be noted that the Euclidean distance similarity algorithm is used. = , in , These are the k-th vectors corresponding to the keyword feature vectors of the prediction results and the keyword feature vectors of the project, respectively.

[0033] The technical highlight of this invention is that it combines Natural Language Processing (NLP) technology with Euclidean distance similarity algorithm to quickly find matching solutions by searching historical lock concrete structure erosion projects in the expert knowledge base. This technology, which is different from existing technologies, is used to solve the problem of not being able to take accurate and rapid maintenance measures for lock concrete structures when the risk of predicted concrete structure erosion is high.

[0034] Example 2 A durability analysis system for ship lock concrete structures under multi-factor erosion coupling includes at least one processor and a memory; the memory stores computer execution instructions; the at least one processor executes the computer execution instructions stored in the memory, causing the at least one processor to perform any one of the methods for durability analysis of ship lock concrete structures under multi-factor erosion coupling.

[0035] It should be understood that the above embodiments are one or more embodiments of the present invention, and there are many other embodiments and variations based on the present invention; any variations and modifications made by those skilled in the art through the present invention without making pioneering innovations are all within the protection scope of the present invention.

Claims

1. A method for durability analysis of concrete structures in ship locks under multi-factor erosion coupling, characterized in that, The method includes the following steps: The internal structure images of different erosion zones of the concrete specimen from the lock were obtained; the internal structure images of different erosion zones were identified and quantitatively analyzed by a deep learning image segmentation algorithm to generate an erosion structure factor set, which includes aggregate distribution, pore network and interface transition zone thickness. Based on a distributed optical fiber sensor network and an electrochemical sensor array, multi-factor environmental parameters of the lock concrete structure are monitored in real time. A multi-physics coupling inversion algorithm is used to calculate the environmental characteristics corresponding to the multi-factor environmental parameters. The multi-factor environmental parameters include at least dynamic water pressure, flow velocity, chloride / sulfate ion concentration, pH value and temperature gradient. The Sobol analysis method was used to perform sensitivity preprocessing analysis on the erosion structure factor set. Three-dimensional microstructure models of different erosion areas were constructed from the data after sensitivity preprocessing. An improved random aggregate delivery algorithm was used to optimize and remodel the constructed three-dimensional microstructure models. Dynamic simulations are performed on a three-dimensional microstructure model based on the environmental characteristics of different erosion zones. The three-dimensional microstructure model embeds a coupled phase-field damage equation and a spatiotemporally variable fractional diffusion equation. The coupled phase-field damage equation is used to identify crack propagation paths and generate structural defect evolution data corresponding to these paths. The spatiotemporally variable fractional diffusion equation is used to identify ion erosion fronts and generate structural defect evolution data corresponding to these fronts. A hypergraph neural network is used to perform spatiotemporal correlation analysis on the structural defect evolution data to predict the trends in crack propagation rate and ion penetration depth, generating prediction results. Based on the text information of the prediction results, keyword features are extracted from the text information using natural language processing (NLP) technology. The features of historical lock concrete structure erosion projects in the expert knowledge base are compared using the Euclidean distance similarity algorithm. Based on the comparison values, an early warning mechanism is established for construction and maintenance.

2. The method for durability analysis of ship lock concrete structures under multi-factor erosion coupling as described in claim 1, characterized in that, The internal structure images of different erosion zones of the ship lock concrete specimen were obtained; The internal structure of different erosion regions was identified and quantitatively analyzed using deep learning image segmentation algorithms, including: A full-range laser confocal 3D scanning imaging was performed on the concrete structure of the ship lock to be monitored, and a 3D image sequence of different erosion areas was obtained. Multi-directional projection 3D reconstruction is performed on 3D image sequences of different erosion areas to construct 3D reconstructed images; U-Net++ deep learning network is used to perform material visual semantic recognition on 3D reconstructed images to identify aggregate, cement paste, pores and interface transition zone components. The spatial distribution characteristics of each component are quantified by morphological algorithms to generate the spatial distribution characteristics of materials in different erosion zones. The spatial distribution characteristics include the volume fraction, gradation curve and orientation distribution of aggregates; the volume fraction, gradation curve and orientation distribution of cement paste; the connectivity, tortuosity and pore size distribution of pores; and the average thickness and microcrack density of the interface transition zone. The spatial distribution characteristics of materials in different erosion zones are analyzed by structural quantification to generate erosion structure factors for different erosion zones.

3. The method for durability analysis of ship lock concrete structures under multi-factor erosion coupling as described in claim 1, characterized in that, The system, based on a distributed optical fiber sensor network and an electrochemical sensor array, monitors multi-factor environmental parameters of the lock's concrete structure in real time. The environmental characteristics corresponding to multi-factor environmental parameters are calculated using a multi-physics coupled inversion algorithm, including: A Kalman filter algorithm is used to fuse multi-sensor data to eliminate environmental noise and drift error. The multi-sensor includes the total number of sensor types in a distributed fiber optic sensor network and an electrochemical sensor array. The PTP protocol, a timestamp synchronization technology, ensures the time consistency of data acquired by the fiber optic sensor network and the electrochemical sensor array; the Kriging interpolation method is used to preprocess the discrete point data in the data acquired by the fiber optic sensor network and the electrochemical sensor array. The temperature field inversion algorithm of the multiphysics field coupling inversion algorithm is used to fit the temperature trend inversion algorithm of each region of the lock concrete temperature fluctuation data to obtain the temperature change curves of different erosion areas. The hydraulic scouring intensity data of the lock concrete was fitted and calculated using the fluid field inversion algorithm of the multiphysics field coupling inversion algorithm to obtain the hydraulic scouring intensity variation curves of different erosion areas. The fluid field inversion algorithm adopts the Navier-Stokes equation. The ion diffusion field inversion algorithm of the multiphysics field coupling inversion algorithm is used to fit the ion diffusion situation inversion algorithm of each region of the lock concrete ion diffusion data to obtain the ion change curves of different erosion areas. The ion change curves include chloride ion change curves and sulfate ion change curves. The ion diffusion field inversion algorithm adopts the Nernst-Planck equation. Environmental stress correlation mining was performed on the temperature change curves, hydraulic scour intensity change curves, and ion change curves of different erosion areas to generate the corresponding environmental characteristics of different erosion areas.

4. The method for durability analysis of ship lock concrete structures under multi-factor erosion coupling as described in claim 1, characterized in that, The sensitivity preprocessing analysis of the erosion structure factor set using the Sobol analysis method includes: Set the parameter ranges for aggregate distribution, pore network, and interface transition zone thickness, and sample each parameter according to the first variation sampling principle of the Sobol analysis method. The finite element method (FEM) is used to solve the problem based on the sampled parameters. The maximum stress of the internal structure in different erosion zones is calculated, and the sensitivity of each parameter, the strength of the nonlinear correlation effect between parameters, and the variable sensitivity discrimination weight coefficient are calculated. The mathematical model of Sobol analysis is expressed as follows: ， ， ， in, Indicates sensitivity, This indicates the strength of the nonlinear correlation effect between parameters; Indicates the sensitivity discrimination weight coefficient of the variable; This represents the i-th output result of the Sobol model; This represents the (i+1)th output result of the Sobol model; This indicates the model's running result after the variables are modified according to a preset percentage; This represents the rate of change of the variable with respect to its initial value after the i-th run of the model; This represents the rate of change of the variable after the (i+1)th calculation of the model compared to the initial variable value; n is the number of calculations in the Sobol model. The influence of each parameter on the maximum stress of the internal structure and the influence between each parameter were analyzed. Based on the Sobol analysis results, weakly correlated factors were eliminated.

5. The method for durability analysis of ship lock concrete structures under multi-factor erosion coupling as described in claim 1, characterized in that, The method of constructing three-dimensional mesoscopic structural models of different erosion zones from the pre-processed data using an improved random aggregate delivery algorithm includes: Multi-frequency Gaussian filtering was applied to the internal structure image of the eroded region to obtain a Gaussian-filtered three-dimensional microstructure image. Based on the internal three-dimensional brightness and darkness differentiation distribution information, we perform fine-grained texture feature analysis on the Gaussian filtered three-dimensional microstructure image and extract the fine-grained texture features of the image. K-Means clustering analysis was performed on the fine-grained features of image texture to generate fine-grained clustering data of eroded area texture. The fine-grained clustering data of eroded area texture includes concrete porosity, crack fractal dimension and interface transition zone thickness. Based on the erosion structure factor set of different erosion regions, the erosion structure model of concrete pore tortuosity, crack fractal dimension and interface transition zone thickness is constructed to build a three-dimensional microstructure model. An improved random aggregate distribution algorithm was used to optimize the three-dimensional microstructure model.

6. The durability analysis method for ship lock concrete structures under multi-factor erosion coupling as described in claim 5, characterized in that, The optimization of the three-dimensional mesostructure model using the improved random aggregate distribution algorithm includes: Reduce aggregate volume fraction in highly eroded areas; generate aggregate shape using hyperquadratic surface equation; use hierarchical bounding box (BVH) for aggregate location detection; construct AABB bounding box for each aggregate location; quickly retrieve neighboring aggregate locations using spatial hash table, and simultaneously construct an aggregate location topology map based on spatial hash table. Multi-scale porosity is generated based on porosity tortuosity; the crack fractal dimension path is constructed using the random midpoint displacement method; and the crack is optimized based on the crack fractal dimension path and aggregate position topology. A non-uniform interface transition zone thickness layer is generated on the aggregate surface based on fine-grained clustering data of erosion area texture and aggregate location topology map; the elastic modulus of the interface transition zone thickness is optimized by exponential decay from aggregate to slurry.

7. The method for durability analysis of ship lock concrete structures under multi-factor erosion coupling as described in claim 1, characterized in that, Spatiotemporally variable fractional diffusion equations are used to analyze the non-Fickian diffusion behavior of chloride and sulfate ions: ， exp(- ). , in, Indicates ion concentration. denoted by the ion diffusion coefficient, where j represents chloride or sulfate ions; (t,x) represents the time fraction order, 0 < t,x. (t,x)≤1 is used to describe the diffusion memory effect, which varies with time and location; β(t,x) represents the spatial fractional order, 1<β(t,x)≤2, which characterizes nonlocal diffusion behavior; t represents the monitoring time step, and x represents the coordinates of the crack propagation path; Indicates the diffusion activation energy. Indicates porosity. Indicates the initial porosity. The factor representing the pre-diffusion coefficient of ions is denoted by m, which represents the empirical coefficient. represents the vector differential operator used for partial derivative operations with respect to spatial variables; R represents the universal gas constant; T represents the absolute temperature; v represents the water flow velocity, which is solved by the Navier-Stokes equations. The coupled phase-field damage equation, which introduces a continuous phase-field variable s(x,t)∈[0,1] to describe crack propagation, is expressed as follows: ， Where s represents the phase field variable, s=0 indicates no damage, and s=1 indicates complete fracture; g(s) =1- , g(s) represents the stiffness degradation function, which describes the reduction of the stiffness of concrete material by damage; Represents tensile elastic strain energy; This represents the compressive elastic strain energy; The critical fracture energy is used to describe the crack resistance of concrete materials; L represents the regularization length, which describes the width controlling crack propagation. This represents the phase field gradient.

8. The method for durability analysis of ship lock concrete structures under multi-factor erosion coupling as described in claim 1, characterized in that, The method of using a hypergraph neural network to perform spatiotemporal correlation analysis on structural defect evolution data to predict the trends of crack propagation rate and ion penetration depth includes: The structural defect evolution data corresponding to the crack propagation path is input into the first layer of the hypergraph neural network of the preset prediction model to obtain the first hypergraph vector; wherein, the first hypergraph vector represents the correlation between the structural defect evolution data features corresponding to the crack propagation path and the crack propagation rate; the preset prediction model is a model obtained based on the hypergraph neural network. The structural defect evolution data corresponding to the ion erosion front is input into the second layer of the preset prediction model's hypergraph neural network to obtain the second hypergraph vector; wherein, the second hypergraph vector represents the correlation between the structural defect evolution data corresponding to the ion erosion front to be processed and the trend of ion penetration depth change; The first hypergraph vector and the second hypergraph vector are concatenated and weighted to obtain the concatenated weighted feature vector to be processed. Based on the preset prediction model, the weighted feature vector to be processed is predicted to obtain the prediction result.

9. The method for durability analysis of ship lock concrete structures under multi-factor erosion coupling as described in claim 1, characterized in that, The text information based on the prediction results is processed using Natural Language Processing (NLP) technology to extract keyword features. The Euclidean distance similarity algorithm is used to compare the features of historical lock concrete structure erosion projects in the expert knowledge base. Based on the comparison values, an early warning mechanism is established for construction and maintenance, including: The text information of the prediction results is used to extract keyword features through natural language NLP technology to generate keyword feature vectors for the prediction results; the historical lock concrete structure erosion project in the expert knowledge base is also used to extract keyword features through natural language NLP technology to generate project keyword feature vectors. The similarity between the predicted keyword feature vector and the project keyword feature vector is calculated using the Euclidean distance similarity algorithm. A preset similarity threshold is used. If the similarity is greater than the preset threshold, the project that has already experienced erosion of the lock concrete structure is identified, and the predicted crack propagation rate, ion penetration depth, and corresponding solution for the project are output. If the similarity is less than or equal to the preset threshold, the project that has experienced erosion of the lock concrete structure with the closest similarity is identified, and the predicted crack propagation rate, ion penetration depth, and corresponding solution for the project are output. Construction and maintenance are then carried out, including at least preventive coating maintenance, local repair maintenance, and structural replacement maintenance.

10. A durability analysis system for ship lock concrete structures under multi-factor erosion coupling, characterized in that, The method includes at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, causing the at least one processor to perform the method according to any one of claims 1-9.

Images