Urban bridge damage maintenance monitoring and analysis system

Abstract

The invention discloses an urban bridge damage maintenance monitoring analysis system, which relates to the technical field of intelligent monitoring crossing, and comprises an identification module for inputting a standardized data set into a convolutional neural network model, identifying a crack type, calculating a damage level, mapping a crack stress influence region based on a finite element grid discretization geometric structure, and calculating a damage degree; generating a damage feature vector; the reinforcement decision module is used for calculating a toughness evaluation value based on the damage state tensor, generating a reinforcement scheme according to the quantum annealing processor and outputting a dynamic maintenance instruction and a toughness evaluation report; and the report module is used for performing reinforcement operation according to the dynamic maintenance instruction, collecting a crack healing verification image through unmanned aerial vehicle inspection, performing maintenance effect evaluation in combination with a toughness evaluation report, and generating a three-dimensional visual evaluation report. And the toughness evaluation matrix is dynamically updated by using the conductivity change, so that the engineering management efficiency and transparency are improved. And the safety, the service life and the operation and maintenance efficiency of the bridge structure are obviously improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring technology, and in particular to a monitoring and analysis system for urban bridge damage maintenance. Background Technology

[0002] With the continuous expansion of urban infrastructure and the sustained growth of traffic loads, the safety and durability of bridge structures have received increasing attention. In the field of bridge engineering, traditional Structural Health Monitoring (SHM) technology mainly relies on static sensor networks, periodic manual inspections, and experience-based judgment, making it difficult to achieve real-time perception and dynamic assessment of damage status. In recent years, with the development of sensing technology, big data processing, and artificial intelligence algorithms, data-driven bridge damage identification and maintenance strategies have gradually become a research hotspot. Especially with the support of deep learning technology, models such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) have been widely applied to tasks such as crack identification, stress analysis, and damage classification. Furthermore, multi-source data fusion methods combined with finite element analysis (FEA) have also been used to improve the accuracy of damage location and severity assessment.

[0003] Existing bridge damage monitoring systems still have certain limitations when facing complex environmental interference and long-term service conditions. Most systems fail to adequately consider noise interference and the standardization of multi-source heterogeneous data during the data preprocessing stage, leading to error accumulation during feature extraction and affecting the accuracy of subsequent damage identification. Secondly, there are issues with the damage response mechanism process. More critically, existing technologies generally lack effective mechanisms. Traditional systems typically rely on manual intervention to formulate repair plans, lacking adaptive intelligent decision-making capabilities. This makes it difficult to achieve a closed-loop, automated maintenance resilience assessment system, and to quantify the structure's recovery capacity and disaster resistance performance at the system level, thus limiting their rapid response capabilities under sudden disaster events. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an urban bridge damage maintenance monitoring and analysis system to solve the problems of low bridge damage identification accuracy and delayed repair response.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides a monitoring and analysis system for urban bridge damage maintenance, comprising: a data acquisition module, which acquires the load magnitude, temperature changes, and crack location coordinates of the bridge, and performs noise filtering and normalization to obtain a standardized dataset; an identification module, which inputs the standardized dataset into a convolutional neural network model, identifies crack types and calculates damage levels, maps the stress-affected region of the crack based on a finite element mesh discretized geometric structure, and generates a damage feature vector; a repair module, which activates a healing agent in a microcapsule network within the concrete to fill the cracks when the damage level code in the damage feature vector reaches a damage threshold, records a repair log, and measures the change in conductivity before and after repair; an update module, which updates the three-dimensional toughness assessment matrix based on the change in conductivity before and after repair to obtain a damage state tensor; a reinforcement decision module, which calculates the toughness evaluation value based on the damage state tensor, generates a reinforcement scheme based on a quantum annealing processor, and outputs dynamic maintenance instructions and a toughness assessment report; and a reporting module, which performs reinforcement operations according to the dynamic maintenance instructions, collects crack healing verification images through UAV inspections, evaluates the maintenance effect in conjunction with the toughness assessment report, and generates a three-dimensional visualization assessment report.

[0008] As a preferred embodiment of the urban bridge damage maintenance monitoring and analysis system described in this invention, the following steps are taken: The system collects data on bridge load magnitude, temperature changes, and crack location coordinates, and performs noise filtering and normalization to obtain a standardized dataset.

[0009] High-frequency noise and low-frequency drift in the load magnitude are removed by using a bandpass filter, and short-term fluctuations in temperature change are eliminated by using a moving average filter.

[0010] The load magnitude and temperature variation are normalized, the bridge structure is divided into different three-dimensional mesh units according to the three-dimensional mesh, each three-dimensional mesh unit is assigned a unique mesh index, and the ratio of the mesh index to the total number of meshes in each direction of the three-dimensional mesh unit is calculated to obtain the proportional coordinates of the crack location.

[0011] By integrating the crack location scale coordinates, crack location coordinates, normalized load magnitude, and temperature change, a standardized dataset is obtained.

[0012] As a preferred embodiment of the urban bridge damage maintenance monitoring and analysis system described in this invention, the specific steps for inputting a standardized dataset into a convolutional neural network model to identify crack types and calculate damage levels are as follows:

[0013] The standardized dataset is converted into a three-dimensional feature tensor and input into a convolutional neural network. The feature map is abstracted through convolutional layers and spatial downsampling is performed using pooling layers to obtain a multi-scale feature tensor.

[0014] Based on multi-scale feature tensors, a fully connected weight calculation is performed in the crack type classification layer through a shared feature layer branch structure, and the crack type is obtained through a probability distribution activation function. In the damage level calculation layer, regression classification mapping is performed, and the damage level is obtained through discretization.

[0015] As a preferred embodiment of the urban bridge damage maintenance monitoring and analysis system of the present invention, the step of mapping the crack stress influence region based on the finite element mesh discretized geometric structure to generate a damage feature vector is as follows:

[0016] Based on crack type and damage level, a finite element mesh discretized geometry with crack topology is constructed, and the finite element mesh discretized geometry is configured through load condition boundary conditions. The stress and strain field of the structure is numerically solved to obtain the stress field of the structure.

[0017] Extract the principal stress tensors of nodal stresses within the crack location radius in the structural stress field, calculate the equivalent stress influence factor, and select the basic discrete elements of the geometric structure discretized by the finite element mesh to form the stress influence region and generate the region topology connectivity graph.

[0018] A convolutional neural network embedding process is applied to the topologically connected graph of the region to obtain node-level damage feature vectors. All node-level damage feature vectors are then integrated to obtain the damage feature vector.

[0019] As a preferred embodiment of the urban bridge damage maintenance monitoring and analysis system described in this invention, when the damage level code in the damage feature vector reaches the damage threshold, the healing agent in the microcapsule network within the concrete is activated to fill the cracks, and a repair log is recorded. The specific steps are as follows.

[0020] Collect the crack width-bearing capacity attenuation rate relationship calibrated by the three-point bending test of concrete beam to obtain the damage threshold. When the damage level code in the damage feature vector is greater than the damage threshold, generate an activation pulse according to the damage level code.

[0021] An activation pulse is applied to the target microcapsule network to break through the capsule wall and release the healing agent. The location coordinates of the crack, the amount of healing agent used, and the start time during the crack filling process are recorded to obtain a repair log.

[0022] As a preferred embodiment of the urban bridge damage maintenance monitoring and analysis system of the present invention, wherein: an AC excitation signal is applied to the copper electrode array pre-embedded on both sides of the crack, and the reference conductivity value before repair is measured by a high-precision bridge; after the crack is filled, the same AC excitation signal is applied at the same position, and the conductivity value after repair is collected.

[0023] As a preferred embodiment of the urban bridge damage maintenance monitoring and analysis system of the present invention, the specific steps for updating the three-dimensional toughness assessment matrix based on repair verification parameters to obtain the damage state tensor are as follows.

[0024] The damage level in the three-dimensional mesh cell corresponding to the damage feature vector is converted into a health index, and a three-dimensional toughness assessment matrix is ​​constructed based on the converted health index.

[0025] The conductivity recovery rate is calculated using the conductivity value after repair and the baseline conductivity value before repair, and the conductivity recovery rate is nonlinearly mapped to the repair efficiency factor using the Sigmoid function.

[0026] Based on the crack location coordinates in the repair log, locate the three-dimensional mesh element corresponding to the three-dimensional toughness assessment matrix, obtain the regional importance coefficient of the corresponding three-dimensional mesh element, update the three-dimensional toughness assessment matrix according to the repair effectiveness factor and the regional importance coefficient, and use it as the damage state tensor.

[0027] As a preferred embodiment of the urban bridge damage maintenance monitoring and analysis system of the present invention, the specific steps of calculating the toughness evaluation value based on the damage state tensor, generating a reinforcement scheme according to the quantum annealing processor, and outputting dynamic maintenance instructions and a toughness assessment report are as follows.

[0028] The damage state tensor and the weights of the three-dimensional mesh elements are used to perform spatial correlation index calculation to obtain the local toughness contribution value of each three-dimensional mesh element. The values ​​are then superimposed to generate a toughness evaluation value, and a toughness assessment report is generated by mapping through hierarchical rules.

[0029] With the optimization goal of maximizing safety benefits and minimizing material costs, a quantum annealing processor is used to transform safety benefits and material costs into competing terms of the energy function. The optimal solution is explored through the quantum tunneling effect of qubits, generating a hardening scheme and encoding it into dynamic maintenance instructions.

[0030] As a preferred embodiment of the urban bridge damage maintenance monitoring and analysis system of the present invention, the steps of performing reinforcement operations according to dynamic maintenance instructions and collecting crack healing verification images through drone inspections are as follows.

[0031] The reinforcement scheme parameters in the dynamic maintenance command are analyzed, the crack location coordinates are mapped to the bridge structure through GNSS positioning, the maintenance equipment is dispatched to the target location, and the bridge is repaired and reinforced according to the reinforcement scheme parameters.

[0032] A three-dimensional waypoint sequence is generated based on the crack location coordinates in the repair log, and a spatial navigation path is planned for the UAV path to obtain the flight route. Finally, the UAV is used to collect images to verify crack healing.

[0033] As a preferred embodiment of the urban bridge damage maintenance monitoring and analysis system of the present invention, the following steps are taken: the crack healing verification image is aligned with the three-dimensional grid cell corresponding to the three-dimensional toughness assessment matrix through a spatial registration algorithm, the apparent healing rate and internal healing quantification index of the crack are extracted, and the health change rate and stress distribution entropy parameter in the toughness assessment report are associated to calculate the three-dimensional toughness recovery rate, and the three-dimensional grid cell index is mapped to RGB color temperature value to generate a three-dimensional visualization assessment report.

[0034] The beneficial effects of this invention are as follows: by collecting and standardizing load, temperature, and crack information, data quality and model generalization ability are improved; the introduction of a microcapsule self-healing mechanism automatically triggers repair when a damage threshold is reached, enhancing the durability and safety of the structure; and the dynamic updating of the toughness assessment matrix using conductivity changes makes structural condition assessment more scientific and accurate, improving engineering management efficiency and transparency. The overall system possesses high intelligence, automation, and scalability, significantly improving the safety, service life, and maintenance efficiency of bridge structures. Attached Figure Description

[0035] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 This is a schematic diagram of an urban bridge damage maintenance monitoring and analysis system.

[0037] Figure 2 The flowchart for generating damage feature vectors.

[0038] Figure 3 The flowchart for generating the damage state tensor.

[0039] Figure 4 A flowchart generated for a 3D visualization report. Detailed Implementation

[0040] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0041] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0042] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0043] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides an urban bridge damage maintenance monitoring and analysis system, comprising the following steps:

[0044] The data acquisition module collects data on the bridge's load magnitude, temperature changes, and crack location coordinates, and performs noise filtering and normalization to obtain a standardized dataset.

[0045] A bandpass filter is used to remove high-frequency noise and low-frequency drift in the load magnitude, and a moving average filter is used to eliminate short-term fluctuations in temperature.

[0046] Specifically, the acquired load magnitude signal is subjected to noise removal using a bandpass filter to remove high-frequency noise components (e.g., the frequency of the example value is higher than 5Hz) and low-frequency drift components (e.g., the frequency of the example value is lower than 0.01Hz). Then, median filtering is used to eliminate impulse noise with a window size (e.g., 5 seconds in the example value). Temperature change data is processed using a moving average filter, and short-term fluctuations are eliminated by setting a sliding window (e.g., 10 minutes in the example value) while retaining the slow trend of change. All data that have undergone noise filtering are used to obtain the load magnitude and temperature change.

[0047] The load magnitude and temperature variation are normalized, and the bridge structure is divided into different three-dimensional mesh units according to the three-dimensional mesh. Each three-dimensional mesh unit is assigned a unique mesh index, and the ratio of the mesh index to the total number of meshes in each direction of the three-dimensional mesh unit is calculated to obtain the proportional coordinates of the crack location.

[0048] Specifically, after the load magnitude noise filtering process is completed to obtain the load magnitude correction value, and the temperature change noise filtering process is completed to obtain the temperature change trend value, the load magnitude correction value is calculated using the historical minimum and maximum load magnitudes through the minimum-maximum normalization method to obtain the load magnitude normalization value. The temperature change trend value is calculated using the historical minimum and maximum temperatures through the minimum-maximum normalization method to obtain the temperature change normalization value. The bridge structure is divided into regular three-dimensional mesh units, and a unique mesh index is assigned to each three-dimensional mesh unit. Based on the crack location coordinates, the mesh index of the three-dimensional mesh unit to which it belongs is determined. The ratio of the mesh index to the total number of meshes in the x-direction of the three-dimensional mesh is calculated to obtain the x-axis proportional coordinate value of the crack location. The ratio of the mesh index to the total number of meshes in the y-direction is calculated to obtain the y-axis proportional coordinate value, and the ratio of the mesh index to the total number of meshes in the z-direction is calculated to obtain the z-axis proportional coordinate value, thus generating the crack location proportional coordinates.

[0049] By integrating the crack location scale coordinates, crack location coordinates, normalized load magnitude, and temperature change, a standardized dataset is obtained.

[0050] Specifically, the crack location coordinates, crack location scale coordinates, normalized load values, and normalized temperature change values ​​are combined in a fixed order: the normalized load values ​​are arranged as the first dimension of the data entries, the normalized temperature change values ​​are arranged as the second dimension, and the crack location scale coordinates are arranged as the third, fourth, and fifth dimensions, respectively. The values ​​of all dimensions form a five-dimensional feature vector, and the five-dimensional feature vectors at each time point are stacked in time series order to form a row and column structure data table, generating a standardized dataset.

[0051] The identification module inputs a standardized dataset into a convolutional neural network model to identify crack types and calculate damage levels. It then maps the stress-affected region of the crack based on the discretized geometric structure of the finite element mesh, generating a damage feature vector.

[0052] The standardized dataset is converted into a three-dimensional feature tensor and input into a convolutional neural network. The feature map is abstracted through convolutional layers and spatial downsampling is performed using pooling layers to obtain a multi-scale feature tensor.

[0053] Specifically, the standardized dataset is reconstructed into a three-dimensional feature tensor according to the spatiotemporal alignment principle. Based on the bridge's three-dimensional grid cell index, standardized data entries from three consecutive time points in the time series (each entry contains five-dimensional data including normalized load values, normalized temperature change values, and crack location ratio coordinates) are stacked along the time dimension to form a three-dimensional tensor. The three-dimensional tensor is then input into a convolutional neural network model. The convolutional layers of the convolutional neural network model use convolutional kernels to perform feature map extraction operations (e.g., the example has 64 kernels), and nonlinear mapping is performed using the ReLU activation function. Max pooling is performed on the feature map output by the convolutional layers to achieve spatial downsampling. After repeating the convolution-pooling operation three times, a four-level abstract multi-scale feature tensor is obtained.

[0054] It should be noted that, for training the convolutional neural network model, the standardized dataset was divided into training, validation, and test sets (in an example ratio of 7:2:1). A stochastic gradient descent optimizer was used, with an initial learning rate of 0.001 and a momentum coefficient of 0.9. The error between the predicted output and the labeled data was calculated using the cross-entropy loss function. The convolutional kernel weight parameters were updated layer by layer using the backpropagation algorithm. After each training epoch, the classification accuracy and root mean square error were calculated on the validation set. An early stopping mechanism was triggered when the validation set loss function did not decrease for 10 consecutive epochs to save the optimal weights. Performance was evaluated on the test set, and a confusion matrix and receiver operating characteristic (ROC) curve were generated, resulting in the trained convolutional neural network model.

[0055] Based on multi-scale feature tensors, a fully connected weight calculation is performed in the crack type classification layer through a shared feature layer branch structure, and the crack type is obtained through a probability distribution activation function. In the damage level calculation layer, regression classification mapping is performed, and the damage level is obtained through discretization.

[0056] Specifically, the multi-scale feature tensor input to the shared feature layer branch structure performs feature dimensionality reduction processing to obtain a low-dimensional feature vector; the crack type classification layer branch performs fully connected weight calculation processing (e.g., the number of neurons in the example fully connected layer is 128), and inputs it into the Softmax function for probability distribution activation, taking the highest probability category index to match the crack type label table (e.g., index 0 corresponds to "transverse through crack", index 1 corresponds to "oblique shear crack") to output the crack type; simultaneously, the damage level calculation layer branch performs fully connected regression processing (16 neurons in the example), outputs continuous damage values, and then obtains the damage level through a piecewise discretization function (e.g., continuous values ​​1.0-3.0 are mapped to discrete damage level 1, and 3.1-5.0 is level 2).

[0057] Based on crack type and damage level, a finite element mesh discretized geometry with crack topology is constructed. The finite element mesh discretized geometry is configured by load condition boundary conditions, and the structural stress-strain field is numerically solved to obtain the structural stress field.

[0058] Specifically, based on specific parameters of crack type and damage level (e.g., crack type example is tensile crack or shear crack, damage level example is level 1 to level 5), a 3D solid geometry of the bridge is generated using CAD geometric modeling tools; crack topology features are created on the 3D solid geometry of the bridge (specifically, mesh node separation or contact element definition at the crack location); a finite element mesh discretization geometry containing crack topology is generated using finite element mesh generation software (e.g., ANSYS Meshing software) (e.g., element size example is 2mm to ensure crack edge resolution); according to the actual load condition boundary conditions, the load and constraint conditions are directly mapped to the corresponding nodes of the finite element mesh; the structural stress and strain field is numerically solved in a finite element solver (e.g., ABAQUS Standard solver engine software) (the analysis type is set to static linear solution, and the relative error of the iterative convergence criterion example is ≤1%); the structural stress field distribution results are exported, and the structural stress field is generated.

[0059] The principal stress tensors of the nodes within the radius of the crack location are extracted from the structural stress field, and the equivalent stress influence factor is calculated. The basic discrete elements of the geometric structure are selected by screening the finite element mesh to form the stress influence region and generate the topological connectivity graph of the region.

[0060] Specifically, based on the structural stress field, a spherical region (with an example radius of 0.5 mm) is determined at the crack tip of the finite element mesh discretized geometry containing crack topology. The principal stress tensors of all nodes within the spherical region are extracted, and the Frobenius norm is obtained based on this stress tensor. The product of the Frobenius norm and the logarithmic distance term is the equivalent stress influence factor. The equivalent stress influence factor of each node is obtained based on the maximum shear stress criterion. Nodes with an equivalent stress influence factor greater than the critical threshold (in the example, ≥ 85% of the standard value of concrete tensile strength) are screened and mapped back to the basic discrete elements in the finite element mesh discretized geometry containing crack topology. All elements containing the screened nodes are marked as constituting the stress influence region. Topological analysis is performed on the elements within the stress influence region to establish an element node adjacency matrix (with an example dimension of n×n, where n is the total number of elements). Connectivity is determined based on the element node sharing relationship (two elements are considered connected when they share ≥ 2 nodes). The set of connected elements is traversed using a depth-first search algorithm, and the topological connectivity graph of the region is output.

[0061] A convolutional neural network embedding process is applied to the topologically connected graph of the region to obtain node-level damage feature vectors. All node-level damage feature vectors are then integrated to obtain the damage feature vector.

[0062] Specifically, the topological connectivity graph of the region is input into the trained convolutional neural network model. The convolutional neural network model performs node feature aggregation operation, calculates the average principal stress tensor of all nodes in the first-order neighborhood of each node, and concatenates it with the node's own principal stress tensor to form a 128-dimensional feature vector. Through two layers of graph convolution operation (e.g., the example layer outputs a 64-dimensional dimension) and processing with the LeakyReLU activation function, node-level damage feature vectors are generated (each node outputs a 64-dimensional vector). Global max pooling is performed on all node-level damage feature vectors to obtain global feature descriptors, which are then fused through a 128-dimensional fully connected layer to generate damage feature vectors.

[0063] The repair module activates the healing agent in the microcapsule network within the concrete to fill cracks when the damage level code in the damage feature vector reaches the damage threshold, records the repair log, and measures the change in conductivity before and after repair.

[0064] The relationship between crack width and bearing capacity attenuation rate calibrated by three-point bending test of concrete beam is collected to obtain the damage threshold. When the damage level code in the damage feature vector is greater than the damage threshold, an activation pulse is generated according to the damage level code.

[0065] Specifically, damage level codes generated by a convolutional neural network model based on crack width grading standards are extracted from the damage feature vector. Based on the crack width-bearing capacity attenuation rate relationship calibrated by the three-point bending test of concrete beam, the damage threshold is determined to be level 3 (example damage level code value). When the damage level code in the damage feature vector is greater than the damage threshold level 3 (example damage level code level 4 or 5), the corresponding voltage amplitude activation pulse is generated according to the damage level code mapping to the voltage amplitude control parameter table (example mapping relationship: level 4 → pulse voltage 18V, level 5 → 24V).

[0066] An activation pulse is applied to the target microcapsule network to break through the capsule wall and release the healing agent. The location coordinates of the crack, the amount of healing agent used, and the start time during the crack filling process are recorded to obtain a repair log.

[0067] Specifically, based on the voltage amplitude control parameter table mapped by the damage level code (e.g., a voltage of 18V corresponding to damage level 4), an activation pulse (example pulse width 10 seconds) is applied to the target microcapsule network through a piezoelectric exciter. The pulsed electric field breaks down the capsule wall to release the healing agent. At the same time, during the filling of the crack, the current crack position coordinates are collected, the actual amount of healing agent used is measured (example value of 0.5mL measured by flow meter) and the start timestamp (UTC millisecond level) is recorded, and the data is encapsulated into a repair log in the order of the fields.

[0068] An AC excitation signal was applied to the copper electrode arrays pre-embedded on both sides of the crack, and the baseline conductivity value before repair was measured by a high-precision bridge. After the crack was filled, the same AC excitation signal was applied at the same location, and the conductivity value after repair was collected.

[0069] Specifically, before filling the crack, a fixed frequency AC excitation signal (e.g., 1 kHz frequency and 5 V amplitude in the example) is applied to the copper electrode array pre-embedded on both sides of the crack using a constant current source. The baseline conductivity value between the electrodes before repair is measured and stored using a high-precision bridge (accuracy example 0.1 μS). After the healing agent fills the crack, the exact same AC excitation signal (1 kHz frequency and 5 V amplitude) is applied to the same physical location of the same set of copper electrode arrays immediately. The conductivity value after repair is acquired and recorded using the same high-precision bridge.

[0070] The update module updates the three-dimensional toughness assessment matrix based on the repair verification parameters to obtain the damage state tensor.

[0071] The damage level in the three-dimensional mesh cell corresponding to the damage feature vector is converted into a health index, and a three-dimensional toughness assessment matrix is ​​constructed based on the converted health index.

[0072] Specifically, the damage level code of the target 3D mesh element, which maps the crack location proportional coordinates, is converted into a health index through a preset discrete-continuous mapping relationship (example: damage level code 1 → health 0.8, level 2 → 0.6, level 3 → 0.4); a 3D array completely consistent with the initial 3D mesh element index system is constructed, and the initial value of the global element health is set to the example value of 1.0. Based on the 3D mesh element index coordinates, the converted health index value is overlaid on the corresponding position to generate a 3D toughness assessment matrix.

[0073] The conductivity recovery rate is calculated using the conductivity value after repair and the baseline conductivity value before repair, and the conductivity recovery rate is nonlinearly mapped to the repair efficiency factor using the Sigmoid function.

[0074] Specifically, the conductivity recovery rate is calculated by comparing the repaired conductivity value with the baseline conductivity value before repair, resulting in a numerical representation of the conductivity recovery rate. The Sigmoid function is used to process the conductivity recovery rate value. The repair efficiency factor represents the degree of conductivity recovery and ranges between 0 and 1, achieving a non-linear mapping. For example, in practical implementation, when the conductivity recovery rate is 0.95, the Sigmoid function outputs a repair efficiency factor of approximately 0.73, or when the conductivity recovery rate is 1.0, the output is approximately 0.88, thus obtaining the repair efficiency factor.

[0075] It should be noted that the formula for calculating the conductivity recovery rate is:

[0076]

[0077] Where A represents the conductivity recovery rate, Q represents the conductivity value after repair, and T represents the baseline conductivity value before repair.

[0078] The formula for calculating the repair efficacy factor is:

[0079]

[0080] Where E represents the repair efficacy factor, h represents a coefficient that controls the steepness of the curve, and c is an offset used to adjust the center position of the Sigmoid function.

[0081] Based on the crack location coordinates in the repair log, locate the three-dimensional mesh element corresponding to the three-dimensional toughness assessment matrix, obtain the regional importance coefficient of the corresponding three-dimensional mesh element, update the three-dimensional toughness assessment matrix according to the repair effectiveness factor and the regional importance coefficient, and use it as the damage state tensor.

[0082] Specifically, based on the crack location coordinates (including longitude, latitude, and elevation values) in the repair log, the coordinate system of the 3D toughness assessment matrix is ​​matched to determine the 3D mesh cell number where the crack location coordinates are located. The regional importance coefficient of the 3D mesh cell is read, and a weighted recovery value is generated based on the repair effectiveness factor and the regional importance coefficient. The weighted recovery value is used to update the original value of the corresponding 3D mesh cell in the 3D toughness assessment matrix. For example, when the repair effectiveness factor is 0.83 and the regional importance coefficient is 1.2, the weighted recovery value is calculated to be 0.996, and this value overwrites the original value of the target cell in the 3D toughness assessment matrix. The updated 3D toughness assessment matrix is ​​then used as the damage state tensor.

[0083] The hardening decision module calculates the toughness evaluation value based on the damage state tensor, generates a hardening scheme according to the quantum annealing processor, and outputs dynamic maintenance instructions and toughness assessment reports.

[0084] The damage state tensor and the weights of the three-dimensional mesh elements are used to perform spatial correlation index calculation to obtain the local toughness contribution value of each three-dimensional mesh element. These values ​​are then superimposed to generate a toughness evaluation value, and a toughness assessment report is generated through hierarchical rule mapping.

[0085] Specifically, the system automatically identifies and classifies structural regions based on BIM component labels, assigning zoning weight coefficients (1.5 for critical load-bearing zones, 1.0 for secondary load-transmitting zones, and 0.7 for non-load-bearing zones). Load correction coefficients are then superimposed based on load categories. The distance from the center point of each 3D mesh element to the structural boundary is obtained, and a boundary distance factor formula is applied for hierarchical weighting. The final weight of each 3D mesh element is the product of the zoning weight coefficient, the load correction coefficient, and the boundary distance factor, resulting in the configured 3D mesh element weight. Based on the value of each 3D mesh element in the damage state tensor and the configured 3D mesh element weight, the local toughness contribution value of each 3D mesh element is obtained. The local toughness contribution values ​​of all 3D mesh elements are accumulated, and the superimposed result serves as the toughness evaluation value. The toughness evaluation value is divided into four levels (e.g., a toughness evaluation value of 82.6 is mapped to level B). A structural toughness assessment report containing the toughness level and a heatmap of the local toughness contribution value distribution is output.

[0086] It should be noted that the formula for calculating the local toughness contribution value is:

[0087] C(i,j,k)=D(i,j,k)·W(i,j,k)·S(i,j,k);

[0088] Where C(i,j,k) represents the local toughness contribution value at the three-dimensional mesh coordinate (i,j,k), D(i,j,k) represents the damage state at the three-dimensional mesh coordinate (i,j,k), W(i,j,k) represents the weight at the three-dimensional mesh coordinate (i,j,k), and S(i,j,k) represents the spatial correlation index factor at the three-dimensional mesh coordinate (i,j,k).

[0089] With the optimization goal of maximizing safety benefits and minimizing material costs, a quantum annealing processor is used to transform safety benefits and material costs into competing terms of the energy function. The optimal solution is explored through the quantum tunneling effect of qubits, generating a hardening scheme and encoding it into dynamic maintenance instructions.

[0090] Specifically, the reinforcement scheme is optimized through a quantum annealing processor. The safety benefit objective is transformed into a negative term of the energy function, and the material cost objective is transformed into a positive term of the energy function. The total energy function is constructed, and the transverse and longitudinal magnetic field strengths are adjusted by the quantum annealing processor to make the Hamiltonian evolve and converge to the ground state, outputting a binary optimal solution. According to the coordinate transformation rules, the binary optimal solution is mapped to dynamic maintenance instructions (for example, the binary sequence "101" is converted into an instruction set: the coordinates of the bottom of the carbon fiber cloth reinforced beam and the coordinates of the epoxy resin injected into the floor slab, and the output is a dynamic maintenance instruction containing material type and three-dimensional coordinates).

[0091] The reporting module performs reinforcement operations based on dynamic maintenance instructions, collects crack healing verification images through drone inspections, evaluates the maintenance effect by combining toughness assessment reports, and generates a 3D visualization assessment report.

[0092] The reinforcement scheme parameters in the dynamic maintenance command are analyzed, the crack location coordinates are mapped to the bridge structure through GNSS positioning, the maintenance equipment is dispatched to the target location, and the bridge is repaired and reinforced according to the reinforcement scheme parameters.

[0093] Specifically, the material type identifier, three-dimensional coordinate parameters, and construction sequence number in the dynamic maintenance instructions are analyzed; Beidou BDS-3 and GPS signals are received through a dual-frequency GNSS receiver, and the three-dimensional coordinates are converted into WGS-84 geodetic coordinates and matched using RTK positioning technology to obtain the specific structural component number (e.g., pier number PD-02); the maintenance equipment resource library is called according to the construction sequence number, the carbon fiber cloth laying robot is dispatched to the bottom coordinate of the beam, and the epoxy resin injection truck is simultaneously dispatched to the floor slab coordinate; 24 hours after construction, the bond strength is tested using a pull-out tester and the crack filling rate is verified using a crack microscope.

[0094] A three-dimensional waypoint sequence is generated based on the crack location coordinates in the repair log, and a spatial navigation path is planned for the UAV path to obtain the flight route. Finally, the UAV is used to collect images to verify crack healing.

[0095] Specifically, the coordinates of the crack location in the repair log are extracted and converted into a three-dimensional waypoint sequence in the WGS-84 geodetic coordinate system (each waypoint has positioning accuracy requirements, such as ±8mm for plane and ±15mm for elevation); based on the three-dimensional waypoint sequence, the A algorithm is used to plan the spatial navigation path and generate an obstacle avoidance flight route; a UAV equipped with a 40MP optical lens is controlled to fly along the flight route and acquire crack healing verification images at a height of 1m above the crack surface at a 45-degree downward angle.

[0096] The crack healing verification image is aligned with the three-dimensional mesh cell corresponding to the three-dimensional toughness assessment matrix using a spatial registration algorithm. The apparent healing rate and internal healing quantification index of the crack are extracted and correlated with the health change rate and stress distribution entropy parameter in the toughness assessment report. The three-dimensional toughness recovery rate is calculated and mapped to RGB color temperature values ​​according to the three-dimensional mesh cell index to generate a three-dimensional visualization assessment report.

[0097] Specifically, the iterative nearest-point algorithm is used to register the pixel coordinates of the crack healing verification image with the spatial coordinates of the three-dimensional grid cells in the three-dimensional toughness assessment matrix, forming a coordinate mapping relationship. Based on the registered area, the apparent healing rate (the percentage reduction in the pixel grayscale variance of the original crack area) of the crack healing verification image is calculated. At the same time, the structural stiffness change rate of the corresponding three-dimensional grid cell in the three-dimensional toughness assessment matrix is ​​extracted as the internal healing degree (determined according to the JTG / TJ23-2008 ultrasonic difference method). The apparent healing rate, internal healing degree, and the health change rate (health difference before and after repair / initial health) and stress distribution entropy (Shannon entropy value of the three-dimensional toughness assessment matrix) in the toughness assessment report are mapped to a color temperature value of 4500K-6500K (example: recovery rate 0.388 is mapped to 5500K). Using the three-dimensional grid cell index as the red channel reference, the green channel component of the apparent healing rate and the blue channel component of the internal healing degree are superimposed to generate RGB color temperature values, and a three-dimensional visualization assessment report is output.

[0098] In summary, this invention improves data quality and model generalization ability through: the collection and standardized processing of load, temperature, and crack information; the introduction of a microcapsule self-healing mechanism, which automatically triggers repair when a damage threshold is reached, enhancing the durability and safety of the structure; and the dynamic updating of the toughness assessment matrix using conductivity changes, making structural condition assessment more scientific and accurate, and improving engineering management efficiency and transparency. The overall system possesses high intelligence, automation, and scalability, significantly improving the safety, service life, and maintenance efficiency of bridge structures.

[0099] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A monitoring and analysis system for damage maintenance of urban bridges, characterized in that: include, The data acquisition module collects data on the bridge's load magnitude, temperature changes, and crack location coordinates, and performs noise filtering and normalization to obtain a standardized dataset. The identification module inputs a standardized dataset into a convolutional neural network model to identify crack types and calculate damage levels. It also maps the stress-affected region of cracks based on the discretized geometric structure of the finite element mesh and generates a damage feature vector. The repair module activates the healing agent in the microcapsule network within the concrete to fill the cracks when the damage level code in the damage feature vector reaches the damage threshold, records the repair log, and measures the change in conductivity before and after repair. The update module updates the three-dimensional toughness assessment matrix based on the change in conductivity before and after repair, and obtains the damage state tensor. The hardening decision module calculates the toughness evaluation value based on the damage state tensor, generates a hardening scheme according to the quantum annealing processor, and outputs dynamic maintenance instructions and toughness assessment reports. The reporting module performs reinforcement operations based on dynamic maintenance instructions, collects crack healing verification images through drone inspections, evaluates the maintenance effect by combining toughness assessment reports, and generates a 3D visualization assessment report.

2. The urban bridge damage maintenance monitoring and analysis system as described in claim 1, characterized in that: The data collection process involves gathering information on the bridge's load magnitude, temperature variations, and crack location coordinates, followed by noise filtering and normalization to obtain a standardized dataset. The specific steps are as follows: High-frequency noise and low-frequency drift in the load magnitude are removed by using a bandpass filter, and short-term fluctuations in temperature change are eliminated by using a moving average filter. The load magnitude and temperature variation are normalized, the bridge structure is divided into different three-dimensional mesh units according to the three-dimensional mesh, each three-dimensional mesh unit is assigned a unique mesh index, and the ratio of the mesh index to the total number of meshes in each direction of the three-dimensional mesh unit is calculated to obtain the proportional coordinates of the crack location. By integrating the crack location scale coordinates, crack location coordinates, normalized load magnitude, and temperature change, a standardized dataset is obtained.

3. The urban bridge damage maintenance monitoring and analysis system as described in claim 2, characterized in that: The process of inputting a standardized dataset into a convolutional neural network model to identify crack types and calculate damage levels involves the following steps: The standardized dataset is converted into a three-dimensional feature tensor and input into a convolutional neural network. The feature map is abstracted through convolutional layers and spatial downsampling is performed using pooling layers to obtain a multi-scale feature tensor. Based on multi-scale feature tensors, a fully connected weight calculation is performed in the crack type classification layer through a shared feature layer branch structure, and the crack type is obtained through a probability distribution activation function. In the damage level calculation layer, regression classification mapping is performed, and the damage level is obtained through discretization.

4. The urban bridge damage maintenance monitoring and analysis system as described in claim 3, characterized in that: The process of mapping the crack stress influence region based on the finite element mesh discretization geometry and generating a damage feature vector involves the following steps: Based on crack type and damage level, a finite element mesh discretized geometry with crack topology is constructed, and the finite element mesh discretized geometry is configured by load condition boundary conditions. The stress-strain field of the structure is numerically solved to obtain the stress field of the structure. Extract the principal stress tensors of nodal points within the crack radius in the structural stress field, calculate the equivalent stress influence factor, and select the basic discrete elements of the geometric structure discretized by the finite element mesh to form the stress influence region and generate the region topology connectivity graph. A convolutional neural network embedding process is applied to the topologically connected graph of the region to obtain node-level damage feature vectors. All node-level damage feature vectors are then integrated to obtain the damage feature vector.

5. The urban bridge damage maintenance monitoring and analysis system as described in claim 4, characterized in that: When the damage level code in the damage feature vector reaches the damage threshold, the healing agent in the microcapsule network within the concrete is activated to fill the cracks, and a repair log is recorded. The specific steps are as follows. Collect the crack width-bearing capacity attenuation rate relationship calibrated by the three-point bending test of concrete beam to obtain the damage threshold. When the damage level code in the damage feature vector is greater than the damage threshold, generate an activation pulse according to the damage level code. An activation pulse is applied to the target microcapsule network to break through the capsule wall and release the healing agent. The location coordinates of the crack, the amount of healing agent used, and the start time during the crack filling process are recorded to obtain a repair log.

6. The urban bridge damage maintenance monitoring and analysis system as described in claim 1, characterized in that: An AC excitation signal is applied to the copper electrode array pre-embedded on both sides of the crack, and the reference conductivity value before repair is measured by a high-precision bridge. After the crack is filled, the same AC excitation signal is applied at the same position, and the conductivity value after repair is collected.

7. The urban bridge damage maintenance monitoring and analysis system as described in claim 6, characterized in that: The process of updating the three-dimensional toughness assessment matrix based on repair verification parameters to obtain the damage state tensor is as follows: The damage level in the three-dimensional mesh cell corresponding to the damage feature vector is converted into a health index, and a three-dimensional toughness assessment matrix is ​​constructed based on the converted health index. The conductivity recovery rate is calculated using the conductivity value after repair and the baseline conductivity value before repair, and the conductivity recovery rate is nonlinearly mapped to the repair efficiency factor using the Sigmoid function. Based on the crack location coordinates in the repair log, locate the three-dimensional mesh element corresponding to the three-dimensional toughness assessment matrix, obtain the regional importance coefficient of the corresponding three-dimensional mesh element, update the three-dimensional toughness assessment matrix according to the repair effectiveness factor and the regional importance coefficient, and use it as the damage state tensor.

8. The urban bridge damage maintenance monitoring and analysis system as described in claim 7, characterized in that: The specific steps for calculating the toughness evaluation value based on the damage state tensor, generating a hardening scheme according to the quantum annealing processor, and outputting dynamic maintenance instructions and a toughness assessment report are as follows. The damage state tensor and the weights of the three-dimensional mesh elements are used to perform spatial correlation index calculation to obtain the local toughness contribution value of each three-dimensional mesh element. The values ​​are then superimposed to generate a toughness evaluation value, and a toughness assessment report is generated by mapping through hierarchical rules. With the optimization goal of maximizing safety benefits and minimizing material costs, a quantum annealing processor is used to transform safety benefits and material costs into competing terms of the energy function. The optimal solution is explored through the quantum tunneling effect of qubits, generating a hardening scheme and encoding it into dynamic maintenance instructions.

9. The urban bridge damage maintenance monitoring and analysis system as described in claim 8, characterized in that: The reinforcement operation is carried out according to dynamic maintenance instructions, and the crack healing verification images are collected by drone inspection. The specific steps are as follows. The reinforcement scheme parameters in the dynamic maintenance command are analyzed, the crack location coordinates are mapped to the bridge structure through GNSS positioning, the maintenance equipment is dispatched to the target location, and the bridge is repaired and reinforced according to the reinforcement scheme parameters. A three-dimensional waypoint sequence is generated based on the crack location coordinates in the repair log, and a spatial navigation path is planned for the UAV path to obtain the flight route. Finally, the UAV is used to collect images to verify crack healing.

10. The urban bridge damage maintenance monitoring and analysis system as described in claim 1, characterized in that: The process involves aligning the crack healing verification image with the corresponding 3D mesh cells of the 3D toughness assessment matrix using a spatial registration algorithm, extracting the apparent crack healing rate and internal healing quantification indicators, as well as associating them with the health change rate and stress distribution entropy parameters in the toughness assessment report, calculating the 3D toughness recovery rate, and mapping it to RGB color temperature values ​​according to the 3D mesh cell index to generate a 3D visualization assessment report.

Images