Bridge crack repair effect prediction method and system based on performance simulation

Abstract

The invention relates to the technical field of bridge engineering detection and repair, in particular to a bridge crack repair effect prediction method and system based on performance simulation, and the method comprises the following steps: obtaining basic data and environmental load data of a target bridge, and matching at least two candidate repair schemes from a preset repair scheme library; constructing an initial performance model before repair, and calculating an initial bearing capacity loss rate and a crack propagation rate; quantifying the interaction influence of the multi-modal data through a coupling factor matrix, and iteratively calculating the repair condition of a set time period after repair to obtain an initial simulation result; and collecting real-time monitoring data in a set time period in the repairing process and after repairing, and inputting the real-time monitoring data into a preset LSTM adaptive correction model to obtain a corrected simulation result. According to the method, the long-term repair condition after repair can be accurately calculated, the accuracy of a prediction result and the environmental adaptability are remarkably improved, and a scientific basis is provided for long-term service performance pre-judgment.

Description

Technical Field

[0001] This invention relates to the field of bridge engineering inspection and repair technology, and in particular to a method and system for predicting the repair effect of bridge cracks based on performance simulation. Background Technology

[0002] As a core component of transportation infrastructure, the structural safety of bridges directly affects traffic safety and road network operational efficiency. During long-term service, reinforced concrete bridges are prone to cracking due to repeated traffic loads and environmental erosion (such as temperature changes, humidity fluctuations, and freeze-thaw cycles). If repairs are not timely or the repair methods are inappropriate, the cracks will further expand, reducing the bridge's load-bearing capacity and durability, and shortening its service life.

[0003] Existing technologies for predicting the effectiveness of bridge crack repair have significant shortcomings: First, most technologies focus only on the short-term performance of a single repair scheme (such as crack closure within a few days after repair), failing to fully consider the dynamic interaction between environmental loads and the properties of repair materials. For example, the bonding strength of epoxy resin grouting materials decays significantly faster in high humidity environments, and the risk of carbon fiber cloth delaminating from the concrete substrate increases under freeze-thaw cycles. This prediction method, which ignores the interaction, results in actual performance deviations of the same repair scheme exceeding 30% between rainy and dry areas, and between freeze-thaw and non-freeze-thaw areas, failing to meet the engineering needs of different regions. Second, existing prediction models mostly use fixed thresholds for parameter correction, i.e., a pre-set uniform error range triggering the correction mechanism. This makes it difficult to adapt to the error fluctuation patterns under different crack types (such as transverse independent cracks, longitudinal independent cracks, and intersecting cracks), different severity levels (mild, moderate, and severe), and different operating scenarios (high-frequency heavy loads and low-frequency light loads). The correction accuracy is limited, often resulting in "deviations between simulated and measured values ​​exceeding 5%", affecting the reliability of predictions. Third, the recommended solutions lack specificity. Most technologies rely on a single performance indicator (such as crack closure rate) to select solutions without considering the actual operational needs of the bridge. For example, high-frequency heavy-load bridges need to prioritize ensuring the load-bearing capacity retention rate, while low-frequency light-load bridges focus more on repair cost control. This single-indicator orientation can easily lead to "over-repair" or "under-repair," which wastes resources and may cause the bridge to fail in the short term after repair, failing to provide comprehensive support for engineering decisions.

[0004] Therefore, it is necessary to design a method and system for predicting the repair effect of bridge cracks based on performance simulation. Summary of the Invention

[0005] To address the technical deficiencies in the background technology, this invention proposes a method and system for predicting the repair effect of bridge cracks based on performance simulation, which solves the aforementioned technical problems and meets practical needs. The specific technical solution is as follows: A method for predicting the repair effect of bridge cracks based on performance simulation includes the following steps: Obtain basic data and environmental load data of the target bridge, and match at least two candidate repair schemes from a pre-set repair scheme library; Based on the basic data, an initial performance model before repair was constructed, and the initial bearing capacity loss rate and crack propagation rate were calculated through finite element analysis as benchmark data for predicting the repair effect. Based on candidate repair schemes, environmental load data, and baseline data, the interaction effects of multimodal data are quantified by the coupling factor matrix in the preset multi-field coupling simulation model. The repair status is iteratively calculated over a set time period after repair to obtain initial simulation results. Real-time monitoring data is collected during the repair process and within a set time period after the repair. The real-time monitoring data and the initial simulation results are input into the preset LSTM adaptive correction model to obtain the corrected simulation results.

[0006] Furthermore, after obtaining the corrected simulation results, the short-term and long-term performance indicators of each candidate repair scheme are output. Combined with the weight allocation rules of the bridge operation scenario, the comprehensive score of each candidate repair scheme is calculated, and the candidate repair scheme with the highest comprehensive score is selected as the optimal repair scheme.

[0007] Furthermore, the basic data includes bridge structural parameters and crack characteristic data. Specifically, the bridge structural parameters are the bridge span, the compressive strength of the concrete cube, and the cross-sectional reinforcement ratio. Specifically, the crack characteristic data are the initial crack width, crack length, and crack direction. The environmental load data includes long-term daily average maximum temperature, daily average minimum temperature, annual average relative humidity, annual freeze-thaw cycle count, daily average number of large vehicles, daily average number of medium-sized vehicles, and daily average number of small vehicles. The daily average maximum temperature and daily average minimum temperature need to be normalized using the following formula: ; in, This represents the standardized temperature parameter, where T is the actual monitored temperature at a certain moment. This is the highest average daily temperature. This is the lowest daily average temperature.

[0008] Furthermore, the preset repair scheme library is constructed in the following way: For the three types of cracks—horizontal independent cracks, longitudinal independent cracks, and intersecting cracks—repair schemes corresponding to the severity of cracks—mild, moderate, and severe—are matched respectively, forming nine basic repair scheme combinations. The crack type is determined by the crack direction. When the crack direction is ∈ [0°, 15°] or the crack direction is ∈ [165°, 180°], it is determined as a longitudinal independent crack. When the crack direction is ∈ [75°, 105°], it is determined as a transverse independent crack. When there are two or more cracks with an angle between their directions ∈ [30°, 150°] in the same monitoring area, they are determined as intersecting cracks. The severity of the crack is determined by the initial crack width: when the initial crack width is < 0.15 mm, it is determined to be a minor crack; when 0.15 mm ≤ initial crack width ≤ 0.3 mm, it is determined to be a moderate crack; when the initial crack width is > 0.3 mm, it is determined to be a severe crack.

[0009] Furthermore, the specific process of constructing the initial performance model before repair is as follows: A solid model of the target bridge was established using finite element analysis software. The geometric dimensions and material properties of the model were defined according to the bridge's structural parameters. Virtual cracks were set at corresponding locations on the model based on crack characteristic data. The width and length of the virtual cracks were consistent with the initial width and length of the cracks, respectively. The crack depth was set to 1.2 times the thickness of the concrete protective layer. The calculation process for the initial bearing capacity loss rate is as follows: Vehicle load standard values ​​are applied to both the intact bridge model without cracks and the model with virtual cracks. The ultimate bearing capacity of the two models is calculated by software. The difference between the two is calculated, and then the difference is divided by the ultimate bearing capacity of the intact bridge model. The result is converted into a percentage, which is the initial bearing capacity loss rate. The calculation process for the crack propagation rate is as follows: The stress intensity factor at the crack tip is extracted based on the finite element analysis results. Combined with the fracture toughness of concrete, the crack propagation rate is calculated using a simplified form of the Paris formula.

[0010] Furthermore, the expression for the coupling factor matrix in the multi-field coupling simulation model is as follows: ; in, The crack recurrence rate in year t after repair. Let be the crack propagation rate in year t after repair. The bond strength of the repair material in year t after repair. This represents the sensitivity coefficient of crack recurrence rate to standardized temperature. The sensitivity coefficient of crack recurrence rate to annual average relative humidity. The sensitivity coefficient of crack recurrence rate to the number of annual freeze-thaw cycles is given. These are the sensitivity coefficients of crack propagation rate to standardized temperature, annual average relative humidity, and annual number of freeze-thaw cycles, respectively. These are the sensitivity coefficients of bond strength to standardized temperature, annual average relative humidity, and annual number of freeze-thaw cycles, respectively. Let be the standardized temperature parameter for year t. Let be the annual average relative humidity in year t. Let t be the number of freeze-thaw cycles in year t. The iterative calculation uses a time step of 1 year. The formula for calculating the crack closure rate in year t after repair is as follows: ; in, Let be the crack closure rate in year t after repair. The initial closure rate 7 days after repair. The temperature decay coefficient of the material. Let be the standardized temperature difference between year t and year t-1. The humidity attenuation coefficient of the material. This represents the difference in annual average relative humidity between year t and year t-1.

[0011] Furthermore, the network structure of the LSTM adaptive correction model includes an input layer, a hidden layer, and an output layer; The input layer has 4 neurons, which correspond to the crack width monitoring value, the structural strain monitoring value, the real-time standardized temperature, and the real-time relative humidity, respectively. The hidden layer contains two LSTM units, with 64 LSTM units in each layer. The forget gate, input gate, and output gate of the LSTM unit all use the tanh activation function, and the cell state update uses the sigmoid activation function. The number of neurons in the output layer is 2, corresponding to the correction amount of the material aging coefficient and the correction amount of the environmental sensitivity coefficient, respectively; The training loss function of the LSTM adaptive correction model uses the mean squared error, and the calculation formula is as follows: ; in, Mean squared error is the final calculated deviation index. When the value exceeds the preset value, parameter correction is triggered, where N is the total number of monitored samples. The simulated crack width value calculated by the performance simulation model during the i-th monitoring session. This refers to the actual crack width value collected by sensors such as crack gauges during the i-th monitoring.

[0012] Furthermore, the bridge operation scenarios are divided into three categories: high-frequency heavy-load scenarios, low-frequency light-load scenarios, and harsh environment scenarios. The criteria for determining the high-frequency heavy-load scenario are an average daily number of large vehicles passing through ≥50 vehicles, the criteria for determining the low-frequency light-load scenario are an average daily number of large vehicles passing through <10 vehicles, and the criteria for determining the harsh environment scenario are an annual freeze-thaw cycle count ≥50 times or an annual average relative humidity ≥85%. The weight allocation rule is as follows: When the scenario is determined to be a high-frequency heavy-load scenario, the weights for load-bearing capacity retention rate are ω1=0.5, crack recurrence rate is ω2=0.3, and repair cost is ω3=0.2. When the scenario is determined to be low-frequency and light-load, the weights for repair cost (ω3) and load-bearing capacity retention rate (ω1) are 0.3 and 0.2, respectively. When the scenario is determined to be a severe environmental condition, the weights for crack recurrence rate (ω2) and bearing capacity retention rate (ω1) are 0.4 and 0.1, respectively. The calculation logic for the overall score is as follows: Overall score = (bearing capacity retention rate × ω1) + ((1 - crack recurrence rate) × ω2) + ((1 - relative repair cost) × ω3).

[0013] Furthermore, the short-term performance indicators include the short-term crack closure rate and the short-term bearing capacity recovery rate after repair. The crack closure rate is the difference between the actual crack width and the initial width after repair, divided by the percentage of the initial width. The bearing capacity recovery rate is calculated by the difference between the ultimate bearing capacity of the repaired model and the ultimate bearing capacity of the intact model through finite element analysis, and expressed as a percentage. The long-term performance indicators include the crack recurrence rate in the mid-term after repair, the crack recurrence rate in the mid-to-long-term after repair, the crack recurrence rate in the long-term after repair, the bearing capacity retention rate in the mid-term after repair, the bearing capacity retention rate in the mid-to-long-term after repair, and the bearing capacity retention rate in the long-term after repair.

[0014] A bridge crack repair effect prediction system based on performance simulation includes a memory, a host computer, and a computer program stored in the memory and executable on the host computer. The computer program is configured to implement the steps of the bridge crack repair effect prediction method based on performance simulation as described above.

[0015] Compared with existing technologies, the bridge crack repair effect prediction method and system based on performance simulation provided by this invention has the following advantages: This invention integrates bridge foundation data and environmental load data, standardizes the environmental data, and combines it with the coupling factor matrix in a multi-field coupled simulation model to achieve a quantitative analysis of the interaction between crack state, repair material properties, and environmental load. This model can accurately calculate the crack closure rate, recurrence rate, and bearing capacity retention rate 1-5 years after repair, effectively solving the problem of large regional prediction deviations caused by neglecting the interaction of multiple factors in existing technologies. It controls the prediction error of repair effects in different environmental regions to within 3%, significantly improving the accuracy and environmental adaptability of the prediction results, and providing a scientific basis for long-term service performance prediction.

[0016] This invention introduces an LSTM adaptive correction model to replace the traditional fixed threshold correction mechanism. By collecting crack width, structural strain, and environmental parameters in real time, it dynamically optimizes the material aging coefficient and environmental sensitivity coefficient in the model with the goal of minimizing the mean square error. This correction method can autonomously learn the error fluctuation patterns under different crack types and severity levels without the need for manual preset correction thresholds. The model can adaptively adjust parameters based on actual monitoring data, further reducing the deviation between the simulation results and measured values ​​after correction. This solves the shortcomings of existing technologies, such as "limited correction accuracy and poor scenario adaptability," and significantly improves the dynamic adaptability and practical application value of the prediction model.

[0017] This invention establishes differentiated weight allocation rules based on bridge operation scenarios, calculates a comprehensive score by combining short-term and long-term performance indicators, and selects the optimal repair solution. This logic avoids both "over-repair" and "under-repair" driven by a single indicator, and balances repair effectiveness and cost according to actual operational needs—for example, prioritizing load-bearing capacity in high-frequency heavy-load scenarios, prioritizing cost control in low-frequency light-load scenarios, and prioritizing reducing recurrence risk in harsh environment scenarios. This effectively reduces the consumption of ineffective repair resources while ensuring the long-term service performance of the bridge after repair, providing comprehensive and targeted support for engineering decisions. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a method for predicting the repair effect of bridge cracks based on performance simulation, as described in this invention. Detailed Implementation

[0019] In the description of this invention, it should be understood that the terms "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "middle," and "inner," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, it should be noted that unless otherwise explicitly specified and limited, the terms "installed," "connected," and "joined" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention through specific circumstances.

[0020] The embodiments of the present invention will be described below with reference to the accompanying drawings and related examples. The embodiments of the present invention are not limited to the following examples, and the present invention relates to the relevant necessary components in this technical field, which should be regarded as well-known technology in this technical field and can be known and mastered by those skilled in this technical field.

[0021] See Figure 1 This invention provides a method for predicting the repair effect of bridge cracks based on performance simulation, comprising the following steps: Step S100: Obtain the basic data and environmental load data of the target bridge, and match at least two candidate repair schemes from the preset repair scheme library; The basic data of the target bridge is a set of static information reflecting the physical state and geometric characteristics of the bridge structure. It can be used to provide input parameters for constructing an initial performance model before repair, supporting subsequent analysis of load-bearing capacity and crack propagation. In this embodiment, the basic data of the target bridge can be obtained through methods such as reviewing design drawings, on-site surveys, and data collection using non-destructive testing equipment. Environmental load data is dynamic data of external natural conditions acting on the bridge structure and affecting its service performance. It can be used to quantify the impact of environmental factors such as temperature changes, humidity fluctuations, and freeze-thaw cycles on material properties and crack evolution. Furthermore, environmental load data can be collected and recorded by meteorological stations, environmental sensor networks, or historical climate databases. The repair scheme library is a collection of information storing various standardized bridge crack repair technologies and their parameter configurations. It can be used as a source of candidate solutions, supporting the matching of applicable repair methods according to bridge conditions. For example, the repair scheme library can be organized into a structured database based on engineering specifications, experimental data, and actual cases.

[0022] Step S200: Based on the basic data, construct the initial performance model before repair, and calculate the initial bearing capacity loss rate and crack propagation rate through finite element analysis as the benchmark data for predicting the repair effect. The initial performance model before repair is a numerical expression describing the structural mechanical behavior and damage state of a bridge before repair. It can be used to calculate the initial bearing capacity loss rate and crack propagation rate, serving as a reference benchmark for simulating subsequent repair effects. In this embodiment, the initial performance model before repair can use the basic data of the target bridge as input and generate a spatially discretized model through finite element modeling software. The initial bearing capacity loss rate is the proportion by which the ultimate bearing capacity of the bridge decreases relative to the design value due to the presence of cracks. It can be used to quantify the degree of structural damage and serves as one of the important indicators for evaluating the necessity and effectiveness of repair. In a specific embodiment, the initial bearing capacity loss rate can be obtained by simulating the structural response under standard load conditions through finite element analysis and comparing the maximum bearing capacity in the intact state with that in the damaged state. The crack propagation rate is the increase in crack length or width per unit time, reflecting the dynamic trend of damage development. It can be used to determine whether the crack is in a stable state and to help predict the deterioration process in the unrepaired state. Furthermore, the crack propagation rate can be iteratively solved by introducing the crack tip stress intensity factor into the finite element model based on the linear elastic fracture mechanics theory.

[0023] Step S300: Based on candidate repair schemes, environmental load data and benchmark data, the interaction effects of multimodal data are quantified by the coupling factor matrix in the preset multi-field coupling simulation model, and the repair situation after a set time period is iteratively calculated to obtain the initial simulation results. The baseline data is a set of key performance indicators output from the initial performance model before repair through finite element analysis. It can be used as the initial conditions for multi-field coupled simulations, ensuring the comparability of performance changes after repair. The multi-field coupled simulation model is a numerical simulation system integrating the interaction mechanisms of multiple physical fields such as mechanics, chemistry, and thermodynamics. It can be used to simulate the long-term interaction behavior of repair materials and concrete matrix under complex environments, predicting the performance evolution after repair. In this embodiment, the multi-field coupled simulation model can use a system of partial differential equations to describe the evolution of each physical field, and achieve inter-field coupling solutions through numerical algorithms. The coupling factor matrix is ​​a numerical matrix representing the intensity of mutual influence between different physical field variables. It can be used to quantify the nonlinear interaction between environmental loads, material properties, and crack states, improving the simulation realism. In a specific embodiment, the coupling factor matrix can be determined based on experimental calibration and theoretical derivation to form a multi-dimensional tensor structure. The material parameters, process parameters, standardized environmental parameters, and baseline data are input into a multi-field coupled simulation model. The interaction between crack state, repair material properties, and environmental load is quantified using a coupling factor matrix. The crack closure rate, crack recurrence rate, and bearing capacity retention rate are iteratively calculated for 1-5 years post-repair to obtain initial simulation results. These initial simulation results are preliminary predictions output by the multi-field coupled simulation model without real-time feedback. They can be used as one of the inputs to the LSTM adaptive correction model for comparison with measured values ​​and to drive parameter optimization.

[0024] Step S400: Collect real-time monitoring data during the repair process and within a set time period after repair, input the real-time monitoring data and initial simulation results into the preset LSTM adaptive correction model, and obtain the corrected simulation results.

[0025] Real-time monitoring data is collected during the repair process and within 30 days after repair. This real-time monitoring data, along with the initial simulation results, is input into the LSTM adaptive correction model. By minimizing the error between the simulated and measured values, the parameter correction amount is calculated, and the material aging coefficient and environmental sensitivity coefficient in the multi-field coupled simulation model are dynamically updated to obtain the corrected simulation results. The LSTM adaptive correction model is a dynamic parameter adjustment system based on a long short-term memory neural network, which can be used to optimize the material aging coefficient and environmental sensitivity coefficient online, reducing the deviation between simulation and measurement. In one specific embodiment, the LSTM adaptive correction model can utilize time series learning capabilities to correlate historical simulation errors with current monitoring data and update the model parameters in reverse. The corrected simulation results are updated prediction data obtained by rerunning the simulation after parameter optimization by the LSTM adaptive correction model, which can be used to more closely reflect actual service performance and improve the reliability of long-term performance predictions.

[0026] Taking a highway bridge in a high-humidity area as an example, a T-beam bridge on a highway in a rainy region of southern China developed transverse cracks. The system first acquired its structural dimensions, reinforcement information, and temperature and humidity records for the past five years to construct an initial performance model. Finite element analysis confirmed that the load-bearing capacity loss rate reached 12%. Two candidate solutions, epoxy resin grouting and carbon fiber reinforcement, were matched from the repair solution library. Using a multi-field coupled simulation model combined with a humidity-bond strength attenuation coefficient matrix, it was predicted that the epoxy solution would experience rapid degradation of bond performance and a high risk of recurrence in a high-humidity environment. After deploying an LSTM correction model and integrating on-site crack width and strain monitoring data, the material aging coefficient was dynamically adjusted, revealing that the measured closure rate was lower than the initial prediction. Considering that this bridge operates under a high-frequency, heavy-load scenario, a differentiated weighting rule was implemented, assigning a higher weight to the load-bearing capacity retention rate. The final comprehensive score showed that the carbon fiber reinforcement solution was superior and was recommended.

[0027] Real-time monitoring data and initial simulation results are input into the LSTM adaptive correction model to obtain corrected simulation results. The material aging coefficient and environmental sensitivity coefficient are dynamically optimized with the goal of minimizing the mean square error, and the parameters are adaptively adjusted by autonomously learning the error fluctuation law. Differential weight allocation rules are formulated based on the bridge operation scenario, and a comprehensive score is calculated by combining short-term and long-term performance indicators to select the optimal repair scheme. The system first obtains the basic data and environmental load data of the target bridge to construct the initial performance model before repair, and quantifies the initial bearing capacity loss rate and crack propagation rate through finite element analysis to form benchmark data for comparison. Based on this, for multiple candidate schemes matched from the repair scheme library, combined with environmental load and benchmark data, a multi-field coupled simulation model and its internal coupling factor matrix are used to analyze the crack... The interaction between crack condition, repair material properties, and environmental factors is modeled, and the performance evolution over a set time period after repair is iteratively calculated to output initial simulation results. Subsequently, real-time monitoring data such as crack width, structural strain, and environmental parameters during the repair process are collected and input along with the initial simulation results into an LSTM adaptive correction model. This model aims to minimize the mean square error and dynamically optimizes the material aging coefficient and environmental sensitivity coefficient to adaptively adjust the simulation parameters, outputting corrected simulation results that are closer to reality. Finally, based on the bridge's operational scenario, corresponding differentiated weight allocation rules are applied to weight and merge short-term and long-term performance indicators into a comprehensive score, which is used to select the optimal repair scheme. The entire process realizes a closed loop from static modeling to dynamic correction and scenario-based decision-making, improving the accuracy of prediction and engineering adaptability.

[0028] In one embodiment of the present invention, after obtaining the corrected simulation results, the short-term performance index and long-term performance index of each candidate repair scheme are output. The comprehensive score of each candidate repair scheme is calculated in combination with the weight allocation rules of the bridge operation scenario, and the candidate repair scheme with the highest comprehensive score is taken as the optimal repair scheme.

[0029] The criteria for determining the optimal repair scheme are as follows: First, candidate schemes with a bearing capacity retention rate of <80% after 5 years of repair are excluded; for the remaining candidate schemes, they are ranked from highest to lowest according to their comprehensive scores, and the candidate scheme ranked first is recommended; if there are candidate schemes with the same comprehensive score, the scheme with the lower repair cost per unit length is given priority.

[0030] The short-term performance indicators (STIs) of each candidate repair scheme are a set of quantitative parameters extracted from the corrected simulation results, reflecting the initial effects of the repair (typically several days to several months). These STIs can be used to evaluate the immediate response capability and initial stability of different candidate schemes after implementation. In this embodiment, the STIs of each candidate repair scheme can be obtained by sampling and statistically analyzing the performance curves of the corrected simulation results in the early time period. The long-term performance indicators (LTIs) of each candidate repair scheme are a set of quantitative parameters extracted from the corrected simulation results, reflecting the long-term (typically more than one year) service performance of the repair. These LTIs can be used to evaluate the durability and reliability of different candidate schemes under complex environments and continuous loads. In a specific embodiment, the LTIs of each candidate repair scheme can be obtained by performing integral, extreme value, or slope analysis based on the evolution trend of the corrected simulation results over a multi-year time span. The weighting allocation rule is a weighting mechanism that sets the importance of performance indicators according to different operational scenarios. This can be used to make the comprehensive score more closely match actual needs and avoid a single indicator dominating the decision. In an exemplary embodiment, the weighting allocation rule can be implemented by establishing a scenario-weight mapping table based on expert experience and historical data analysis. Furthermore, the weight allocation rules are coordinated with the bridge operation scenario: the rules activate corresponding weight configurations based on the specific scenario; simultaneously, the weight allocation rules are coordinated with the comprehensive score of each candidate repair scheme: the weights directly affect the contribution ratio of each indicator to the total score. The comprehensive score is an overall evaluation value calculated by integrating the short-term and long-term performance indicators of each candidate repair scheme and weighting them according to their respective weights. It can be used to compare the overall merits of multiple candidate schemes horizontally, supporting scientific decision-making. In this embodiment, the comprehensive score of each candidate repair scheme can be obtained by multiplying the normalized performance indicators by their corresponding weights and then summing them to generate a scalar score.

[0031] After obtaining the corrected simulation results optimized by the LSTM adaptive correction model, this invention extracts the short-term response performance and long-term durability evolution data of each candidate repair scheme, forming short-term and long-term performance indicators. Combining the bridge's operational scenario with appropriate weighting rules, different indicators are assigned differentiated levels of importance, thereby calculating the comprehensive score for each candidate scheme. This process realizes a shift from single-indicator judgment to multi-dimensional, contextualized evaluation. Finally, the scheme with the highest comprehensive score is determined as the optimal repair scheme, ensuring that the recommended results meet both structural safety requirements and actual operational constraints, thus improving the rationality of repair decisions and the technical effectiveness of resource utilization efficiency.

[0032] In one embodiment of the present invention, the basic data includes bridge structural parameters and crack characteristic data. Specifically, the bridge structural parameters are bridge span, concrete cube compressive strength and cross-sectional reinforcement ratio, and the crack characteristic data are crack initial width, crack length and crack direction. Bridge structural parameters are key design indicators describing the geometry and material properties of the main bridge structure. They can be used to characterize the overall structural capacity of the bridge and support the construction of initial performance models. Furthermore, bridge structural parameters can be obtained by consulting construction drawings, inspection reports, or on-site measurements. Crack characteristic data is a set of damage information reflecting the physical morphology and spatial distribution characteristics of cracks. It can be used to quantify the initial state of cracks, serving as a basis for assessing propagation trends and repair needs. Furthermore, crack characteristic data can be obtained through visual inspection, laser scanning, or digital image correlation techniques. Bridge span refers to the horizontal distance between the centers of two supports (for simply supported beam bridges) or the horizontal distance between the centers of two adjacent piers (for continuous beam bridges). Concrete cube compressive strength refers to the compressive strength measured after curing a standard concrete specimen with a side length of 150mm for 28 days under standard curing conditions (temperature 20±2℃, relative humidity ≥95%). Section reinforcement ratio refers to the ratio of the total area of ​​longitudinal reinforcement in the tension zone of a bridge to the effective cross-sectional area of ​​the member. The initial width of a crack refers to the horizontal distance at the widest point of the crack (the cross-sectional dimension perpendicular to the crack direction); the crack length refers to the straight-line distance between the two ends of the crack (the dimension along the crack direction); the crack direction refers to the angle between the centerline of the crack and the longitudinal axis of the bridge (along the bridge direction).

[0033] The environmental load data includes long-term daily average maximum temperature, daily average minimum temperature, annual average relative humidity, annual freeze-thaw cycle count, daily average number of large vehicles, daily average number of medium-sized vehicles, and daily average number of small vehicles. The daily average maximum temperature and daily average minimum temperature need to be normalized using the following formula: ; in, This represents the standardized temperature parameter, where T is the actual monitored temperature at a certain moment. This is the highest average daily temperature. This is the lowest daily average temperature.

[0034] The average daily maximum temperature refers to the arithmetic mean of the daily maximum temperatures over the past 5 years; the average daily minimum temperature refers to the arithmetic mean of the daily minimum temperatures over the past 5 years; the average annual relative humidity refers to the arithmetic mean of the relative humidity over the past 5 years; the annual number of freeze-thaw cycles refers to the number of times the "freeze-thaw cycle standard" is met each year over the past 5 years (freeze-thaw cycle: the temperature rises from ≤-5℃ to ≥5℃ within 1 day, and the duration is ≥4 hours); the average daily number of large vehicles passing through refers to the total number of large vehicles (load ≥10 tons) passing through each day over the past 5 years (the average value over 5 years is calculated after annual statistics); the average daily number of medium-sized vehicles passing through refers to the total number of medium-sized vehicles (load 3-10 tons) passing through each day over the past 5 years (the average value over 5 years is calculated after annual statistics); the average daily number of small vehicles passing through refers to the total number of small vehicles (load <3 tons) passing through each day over the past 5 years (the average value over 5 years is calculated after annual statistics).

[0035] For example, in the scenario of an urban overpass in a region with significant seasonal temperature differences, a city overpass in North China experiences an average daily maximum temperature of 35°C in summer and an average daily minimum temperature as low as -10°C in winter, undergoing approximately 60 freeze-thaw cycles annually. The system collected data on the bridge's cracks, with an initial width of 0.25 mm, a length of 3 m, and a transverse orientation. During the simulation, a normalization formula was used to convert the measured temperature of 32°C on a summer day into (32 + 10) / (35 + 10) = 0.93, and the winter temperature of -8°C into (-8 + 10) / (35 + 10) = 0.04, achieving a balanced expression of temperature variables under large temperature differences. After these standardized parameters were input into a multi-field coupled model, they, along with factors such as humidity and freeze-thaw cycles, act on the aging process of the carbon fiber fabric bonding layer, accurately predicting that the risk of interfacial delamination is concentrated during the period of drastic temperature changes in spring and autumn, providing a scientific basis for selecting the maintenance window.

[0036] This invention normalizes the daily average maximum and minimum temperatures and calculates standardized temperature parameters using a normalization formula. By substituting actual monitored temperatures, daily average maximum, and daily average minimum temperatures into the temperature normalization formula, it achieves precise characterization of structural load-bearing capacity and spatial quantification of damage state by clarifying the basic data composition. It comprehensively reflects the dual effects of climate and transportation by refining environmental load inputs, eliminates modeling biases caused by absolute temperature differences between geographical regions by normalizing temperature parameters, and improves the modeling consistency of environmental-material interaction effects in multi-field coupled simulations through the generation and integration of standardized temperature parameters. This technology can support accurate prediction of crack closure rate, recurrence rate, and load-bearing capacity retention rate 1-5 years after repair, enhancing the model's adaptability and predictive stability under multi-regional conditions, and laying a reliable data foundation for subsequent adaptive correction and scheme optimization.

[0037] It should be noted that the preset repair scheme library is constructed in the following way: For the three types of cracks—horizontal independent cracks, longitudinal independent cracks, and intersecting cracks—repair schemes corresponding to the severity of cracks—mild, moderate, and severe—are matched respectively, forming nine basic repair scheme combinations. Transverse independent cracks are crack morphologies that develop independently along the transverse principal stress direction of the bridge and can be used as a typical crack type, corresponding to specific repair techniques. In one specific embodiment, the direction angle of transverse independent cracks can be determined using image recognition technology to determine whether they fall within a preset range. Longitudinal independent cracks are crack morphologies that extend along the longitudinal axis of the bridge and do not intersect. They can be used to reflect longitudinal stress or construction joint problems in the structure, guiding longitudinal sealing or grouting treatment. Intersecting cracks are a network-like damage pattern formed by multiple cracks intersecting at a certain angle within the same monitoring area. They can be used to characterize structural deterioration under complex stress states and require a composite repair strategy. Nine basic repair scheme combinations are formed by matching three types of cracks with three crack severity levels. This involves pairwise combining transverse, longitudinal, and intersecting cracks with three levels of severity (mild, moderate, and severe), generating nine unique combinations. Furthermore, this operation can be achieved by manually entering scheme configurations based on expert experience or by automatically generating the optimal parameter set by combining experimental data and simulation results, thereby establishing a structured scheme classification system and improving the systematicness and completeness of scheme organization.

[0038] The crack type is determined by the crack direction. When the crack direction is ∈ [0°, 15°] or the crack direction is ∈ [165°, 180°], it is determined as a longitudinal independent crack. When the crack direction is ∈ [75°, 105°], it is determined as a transverse independent crack. When there are two or more cracks with an angle between their directions ∈ [30°, 150°] in the same monitoring area, they are determined as intersecting cracks. The severity of the crack is determined by the initial crack width: when the initial crack width is < 0.15 mm, it is determined to be a minor crack; when 0.15 mm ≤ initial crack width ≤ 0.3 mm, it is determined to be a moderate crack; when the initial crack width is > 0.3 mm, it is determined to be a severe crack.

[0039] A crack observation instrument with angle measurement function is used to scan the cracks of the target bridge in the whole area, focusing on recording the direction of each crack and marking the monitoring area where the crack is located. The measured crack direction is then compared with the preset angle range.

[0040] Taking the post-inspection repair planning of a highway bridge as an example, a highway bridge was found to have cracks in multiple areas during routine inspections. The system divided the beam into several monitoring areas. A crack with a direction of 88° and a width of 0.22mm was found on the bottom surface of a mid-span section. According to the rules, it was determined to be a transverse independent crack with moderate damage. Two cracks with directions of 40° and 110° were found near the supports, with an included angle of 70°, falling within the range of [30°, 150°], and were therefore identified as intersecting cracks. One of them had a width of 0.35mm, and was determined to be severe. Based on this, the system matched a low-pressure slow-speed epoxy grouting scheme corresponding to "transverse + moderate" and a carbon fiber mesh embedding + interface reinforcement composite repair scheme corresponding to "intersecting + severe" from the repair scheme library, forming two candidate paths for subsequent performance simulation. This classification mechanism ensures that different types and levels of cracks receive differentiated treatment suggestions.

[0041] In one embodiment of the present invention, the specific process of constructing the initial performance model before repair is as follows: A solid model of the target bridge was established using finite element analysis software. The geometric dimensions and material properties of the model were defined according to the bridge's structural parameters. Virtual cracks were set at corresponding locations on the model based on crack characteristic data. The width and length of the virtual cracks were consistent with the initial width and length of the cracks, respectively. The crack depth was set to 1.2 times the thickness of the concrete protective layer. A solid model of the target bridge is established using general-purpose finite element analysis software (such as Abaqus and ANSYS). First, based on the defined bridge structural parameters (bridge span, concrete cube compressive strength, and cross-sectional reinforcement ratio), the model's geometric dimensions (e.g., a simply supported beam with a span of 20m, beam height of 1.2m, and web thickness of 0.3m) and material properties (e.g., concrete elastic modulus of 34.5GPa, Poisson's ratio of 0.2, and steel yield strength of 400MPa) are defined in the software to ensure that the model's material properties are consistent with the actual bridge. According to the crack characteristic data (initial crack width, crack length, and crack direction) in claim 3, virtual cracks are set at corresponding locations on the model (e.g., at the bottom of the beam, mid-span, or near the supports). The virtual crack depth is set to "1.2 times the thickness of the concrete cover" (e.g., if the concrete cover thickness is 50mm, then the virtual crack depth = 50mm × 1.2 = 60mm). This setting is based on empirical values ​​verified by a large amount of engineering data, balancing "simulation accuracy" and "computational efficiency," and avoiding structural stress simulation deviations caused by excessively shallow or deep crack depths. The model is meshed, with a fine mesh used in the area near the crack to accurately capture the stress state at the crack tip, and a conventional mesh used in other areas to reduce the computational load. Boundary conditions are set according to the actual support form of the bridge (e.g., the two ends of a simply supported beam are set as hinged supports and rolling supports, respectively), to ensure that the stress on the model is consistent with the actual working conditions.

[0042] The calculation process for the initial bearing capacity loss rate is as follows: Vehicle load standard values ​​are applied to both the intact bridge model without cracks and the model with virtual cracks. The ultimate bearing capacity of the two models is calculated by software. The difference between the two is calculated, and then the difference is divided by the ultimate bearing capacity of the intact bridge model. The result is converted into a percentage, which is the initial bearing capacity loss rate. In the same finite element software, a bridge model with integrity and a model with virtual cracks were constructed respectively. The standard values ​​of vehicle loads in the "General Specifications for Design of Highway Bridges and Culverts" (such as highway-I lane load, uniformly distributed load of 10.5kN / m, concentrated load of 360kN) were applied to the two models respectively. The ultimate bearing capacity of the two models was calculated by the nonlinear analysis module of the software (such as the analysis module considering plastic damage of concrete). The criterion for determining the ultimate bearing capacity is: when the stress of a certain section of the model reaches the design value of the compressive strength of concrete, or when the structure shows obvious plastic deformation (such as the beam deflection reaching 1 / 500 of the span), the corresponding load value is the ultimate bearing capacity, which is recorded as the ultimate bearing capacity of the intact model and the ultimate bearing capacity of the cracked model respectively.

[0043] Calculate the difference between the ultimate bearing capacity of the intact model and the ultimate bearing capacity of the cracked model, divide it by the ultimate bearing capacity of the intact model, and then convert it into a percentage. This percentage is the initial bearing capacity loss rate.

[0044] The calculation process for the crack propagation rate is as follows: The stress intensity factor at the crack tip is extracted based on the finite element analysis results. Combined with the fracture toughness of concrete, the crack propagation rate is calculated using a simplified form of the Paris formula.

[0045] Crack propagation rate is a key indicator for predicting the "natural development trend of cracks." It is calculated using "finite element stress extraction + material mechanics formulas," specifically as follows: Based on the finite element analysis results with a virtual crack model, the stress post-processing function of the software is used to extract the Type I stress intensity factor at the crack tip (Type I is an open crack, which is the main type of bridge crack). The extraction location is the midpoint of the leading edge line of the crack tip to ensure that the data can reflect the stress concentration at the crack tip. The fracture toughness of the target bridge concrete is obtained through laboratory tests (such as the fracture toughness of C50 concrete, which is usually 1.5-1.8 MPa・m^(1 / 2)). If test data cannot be obtained, the empirical values ​​recommended in the "Standard for Durability Design of Concrete Structures" can be used to ensure the reliability of material parameters. The crack propagation rate is calculated using a simplified formula based on the Paris equation, as follows: ; in, This represents the crack propagation rate. Wield's material constant for concrete. Stress intensity factor denoted as fracture toughness, and m as crack propagation index.

[0046] This calculation process combines finite element analysis with material mechanics formulas, avoiding the calculation deviations caused by the traditional empirical formulas that "ignore the structural stress state," thus improving the calculation accuracy of crack propagation rate by more than 20%.

[0047] This invention establishes a physical model of the target bridge using finite element analysis software. The geometric dimensions and material properties of the model are defined based on the bridge's structural parameters. Virtual cracks are set at corresponding locations on the model based on crack characteristic data, with the width and length of the virtual cracks consistent with measured values, and the depth set to 1.2 times the thickness of the concrete protective layer. By applying standard vehicle load values ​​to both a crack-free, intact bridge model and a model with virtual cracks, the difference in their ultimate bearing capacities is calculated and normalized to a percentage. Furthermore, the stress intensity factor at the crack tip is extracted based on the finite element analysis results. Combined with the fracture toughness of concrete, a simplified form of the Paris equation is used to calculate the crack propagation rate. This gives the digital model physical realism, reasonably reflects the downward propagation tendency of cracks, and enables a quantitative assessment of structural performance degradation. Simultaneously, the fracture mechanics mechanism is integrated into the initial state modeling, accurately depicting the current damage level and predicting crack evolution trends. This provides reliable initial boundary conditions and performance benchmarks for subsequent multi-field coupled simulations, enhancing the mechanistic support for predicting repair effects.

[0048] It should be noted that the expression for the coupling factor matrix in the multi-field coupling simulation model is as follows: ; in, The crack recurrence rate in year t after repair is the proportion of cracks that reappear or expand to their initial width in year t after repair. This represents the crack propagation rate in year t after repair, i.e., the average annual increase in crack width in year t. The bond strength of the repair material in year t after repair is defined as the ability of the repair material to bond with the concrete substrate. The sensitivity coefficient of crack recurrence rate to standardized temperature represents the effect of a 1°C change in standardized temperature on the recurrence rate. The sensitivity coefficient of crack recurrence rate to annual average relative humidity. The sensitivity coefficient of crack recurrence rate to the number of annual freeze-thaw cycles is given. These are the sensitivity coefficients of crack propagation rate to standardized temperature, annual average relative humidity, and annual number of freeze-thaw cycles, respectively. These are the sensitivity coefficients of bond strength to standardized temperature, annual average relative humidity, and annual number of freeze-thaw cycles, respectively. The standardized temperature parameters for year t are obtained by normalizing the daily average maximum / minimum temperatures. Let be the annual average relative humidity in year t. Let t be the number of freeze-thaw cycles in year t.

[0049] Environmental parameters deviate during monitoring due to sensor errors and environmental interference. These deviations need to be corrected before being input into the multi-field coupling model. Temperature monitoring deviations mainly originate from sensor drift (systematic error) and environmental interference (random error). A combination of "calibration coefficient correction + moving average filtering" is used, with the correction formula as follows: ; in, This is the corrected actual temperature. For sensor calibration coefficients, The actual temperature measured by the sensor. This is the sensor zero-point deviation compensation term, where n is the moving average window size. The measured temperature within the i-th window. The average standard temperature within the i-th window.

[0050] Humidity monitoring deviations mainly stem from sensor hysteresis and temperature cross-interference. A solution of "temperature compensation + drift correction" is employed, as shown in the following formula: ; in, This is the corrected actual relative humidity. The relative humidity is measured by the sensor. This is the temperature cross-interference coefficient. This is the corrected actual temperature. For sensor drift rate, This refers to the continuous operating time of the sensor.

[0051] Sensitivity coefficients in multi-field coupled simulations need to be calibrated using a two-dimensional approach: "indoor testing + fitting of engineering measured data." This ensures that the coefficients accurately reflect the quantitative correlation between environmental parameters and performance indicators, avoiding simulation biases caused by subjective settings. The specific calibration steps are as follows: The test factors were "standardized temperature, annual average relative humidity, and annual freeze-thaw cycles". Each factor was set with 3-5 levels to cover different environmental scenarios. Concrete specimens of the same strength grade as the target bridge were used to prefabricate cracks with the same characteristics as the actual cracks (width, length, and direction matching). The mainstream repair materials (epoxy resin, carbon fiber cloth, etc.) from the scheme library were used for repair. Three parallel specimens were set up for each group of tests. The specimens were placed in an artificial climate chamber with corresponding environmental parameters and continuously monitored for 1-3 years. Changes in crack recurrence rate, crack propagation rate, and bonding strength of repair materials were recorded periodically, while the measured values ​​of environmental parameters were collected during the same period. Outliers were removed, and the experimental data were standardized to ensure consistency in the dimensions of each environmental parameter. A multiple linear regression model was constructed with performance indicators as the dependent variable and environmental parameters as the independent variables. ; in, For performance indicators, , , For environmental parameters, , , This is the sensitivity coefficient to be calibrated. For constant terms, This is random error; The regression model is solved using the least squares method to minimize the sum of squared errors and obtain the initial value of the sensitivity coefficient.

[0052] Collect over 100 sets of measured data (including environmental parameters and performance index monitoring data) from the repair projects of similar bridges. Substitute the calibrated sensitivity coefficients into the multi-field coupled simulation model and calculate the mean square error (MES) between the simulated and measured values. If MES > 0.02 (when the performance index is a percentage), use the gradient descent method to adjust the sensitivity coefficients. Iterate and optimize until MES ≤ 0.02, and finally determine a stable sensitivity coefficient value.

[0053] This invention first constructs a coupling factor matrix that includes the correlation between crack recurrence rate, crack propagation rate, bonding strength of repair materials, and environmental parameters (standardized temperature, humidity, freeze-thaw cycles), transforming the interaction of multiple factors into quantifiable mathematical relationships. Then, with a time step of one year, the performance status of each year after repair is iteratively derived through the crack closure rate calculation formula, forming initial simulation results covering 1-5 years after repair. This solves the problem of large regional prediction deviations and inaccurate long-term performance predictions caused by the "ignoring of the dynamic interaction between environment and materials" in existing technologies, and provides accurate initial data support for subsequent LSTM adaptive correction.

[0054] In this matrix, each row corresponds to the relationship between a core performance indicator and environmental parameters: the first row quantifies the interaction between crack recurrence rate and environmental parameters, the second row quantifies the interaction between crack propagation rate and environmental parameters, and the third row quantifies the interaction between the bonding strength of repair materials and environmental parameters, thus realizing the mathematical and calculable analysis of the influence of multiple factors.

[0055] The iterative calculation uses a time step of 1 year. The formula for calculating the crack closure rate in year t after repair is as follows: ; in, Let be the crack closure rate in year t after repair. The initial closure rate 7 days after repair. The temperature decay coefficient of the material. Let be the standardized temperature difference between year t and year t-1. The humidity attenuation coefficient of the material. This represents the difference in annual average relative humidity between year t and year t-1.

[0056] This invention employs a coupling factor matrix with explicit mathematical expressions in a multi-field coupled simulation model to establish quantitative relationships between the post-repair crack recurrence rate, crack propagation rate, and repair material bond strength, and standardized temperature parameters, annual average relative humidity, and annual freeze-thaw cycle count, respectively. It also introduces various environmental sensitivity coefficients to accurately describe the independent and cumulative effects of different climatic factors on various performance indicators. Furthermore, it standardizes environmental data to eliminate modeling biases caused by differences in absolute climatic values ​​between regions. It iterative calculations with a one-year time step update state variables annually and track the performance evolution path 1–5 years after repair. Finally, using the initial closure rate 7 days after repair as a benchmark, combined with standardized temperature and annual average relative humidity differences between adjacent years, the crack closure rate is dynamically corrected using material temperature decay coefficients and material humidity decay coefficients to reflect the continuous impact of environmental fluctuations on the repair effect. This approach enhances the physical interpretability and cross-regional adaptability of the simulation, provides high-quality initial prediction results for subsequent LSTM adaptive correction, and supports the accuracy and stability of the entire prediction process.

[0057] In one embodiment of the present invention, the network structure of the LSTM adaptive correction model includes an input layer, a hidden layer, and an output layer; The input layer has 4 neurons, which correspond to the crack width monitoring value, the structural strain monitoring value, the real-time standardized temperature, and the real-time relative humidity, respectively. Crack width monitoring values ​​are collected through a crack gauge, directly reflecting the actual evolution of the crack; structural strain monitoring values ​​are collected through strain gauges, reflecting the influence of structural stress state on the crack; real-time standardized temperature is calculated using a temperature normalization formula to eliminate regional temperature differences and ensure the consistency of environmental parameters; real-time relative humidity is collected through a temperature and humidity sensor, reflecting the attenuation effect of humidity on the bonding performance of repair materials.

[0058] The hidden layer contains two LSTM units, with 64 LSTM units in each layer. The forget gate, input gate, and output gate of the LSTM unit all use the tanh activation function, and the cell state update uses the sigmoid activation function. Each LSTM unit contains three gating mechanisms: a "forget gate," an "input gate," and an "output gate," which control the "forgetting of historical information," the "updating of cell state," and the "output of current information," respectively, adapting to the "long-range dependency" characteristics of time-series data.

[0059] The number of neurons in the output layer is 2, corresponding to the correction amount of the material aging coefficient and the correction amount of the environmental sensitivity coefficient, respectively; The aging coefficient correction is used to correct the aging rate of the repair materials (such as epoxy resin and carbon fiber cloth) in the multi-field coupling model; the environmental sensitivity coefficient correction is used to correct the sensitivity of the material to environmental factors.

[0060] The training loss function of the LSTM adaptive correction model uses the mean squared error, and the calculation formula is as follows: ; in, Mean squared error is the final calculated deviation index. When the value exceeds the preset value, parameter correction is triggered, where N is the total number of monitored samples. The simulated crack width value calculated by the performance simulation model during the i-th monitoring session. This refers to the actual crack width value collected by sensors such as crack gauges during the i-th monitoring.

[0061] The LSTM model parameters were scientifically determined based on actual engineering needs and data characteristics: the input layer has 4 neurons, corresponding to crack width monitoring values, structural strain monitoring values, real-time standardized temperature, and real-time relative humidity, comprehensively covering the three core dimensions of "crack state - structural stress - environmental influence". Correlation analysis showed that the correlation coefficients between these four parameters and the simulation error are all >0.7, effectively covering the main influencing factors of error fluctuations. The hidden layer uses 2 layers of LSTM units with 64 neurons per layer. This 2-layer structure balances fitting and generalization capabilities, avoiding insufficient feature extraction from a single layer or overfitting from 3 or more layers. Meta-validation achieves rapid error convergence (convergence after 500 iterations) without overfitting, balancing accuracy and computational efficiency. It offers more comprehensive feature extraction compared to 32 units and is more cost-effective than 128 units. For activation functions, the forget gate, input gate, and output gate use the tanh activation function to adapt to the positive and negative fluctuations in time-series data and avoid gradient vanishing. Cell state updates use the sigmoid activation function to precisely control the proportion of historical information forgotten and adapt to the long-range dependence of crack evolution. The loss function chosen is mean squared error (MSE), which can amplify the impact of larger deviations to prioritize the correction of key deviations. The correction threshold is set to 0.02 mm. 2The threshold is set based on actual engineering conditions, corresponding to a sum of squared deviations of multiple samples ≤ 0.02 and a single sample average deviation ≤ 0.14 mm, which meets the accuracy requirements of the "Technical Specification for Crack Detection of Highway Bridges". It has been verified by 10 sets of engineering data to effectively trigger the necessary parameter correction and avoid over-correction.

[0062] In one embodiment of the present invention, the bridge operation scenarios are divided into three categories: high-frequency heavy-load scenarios, low-frequency light-load scenarios, and harsh environment scenarios. The criteria for determining the high-frequency heavy-load scenarios are an average daily number of large vehicles passing through ≥50 vehicles, the criteria for determining the low-frequency light-load scenarios are an average daily number of large vehicles passing through <10 vehicles, and the criteria for determining the harsh environment scenarios are an annual freeze-thaw cycle count ≥50 times or an annual average relative humidity ≥85%. The average daily number of large vehicles passing through the area comes from the "Highway Traffic Flow Monitoring Platform" of the traffic management department, taking the average daily data of the past 3 years (excluding abnormal periods, a total of 1095 data points were collected); the annual number of freeze-thaw cycles comes from the "Climate Observation Annual Report" of the local meteorological station, taking the average of the past 5 years; the annual average relative humidity comes from the "Annual Climate Statistics Report" of the meteorological station, taking the average of the past 5 years.

[0063] The weight allocation rule is as follows: When the scenario is determined to be a high-frequency heavy-load scenario, the weights for load-bearing capacity retention rate are ω1=0.5, crack recurrence rate is ω2=0.3, and repair cost is ω3=0.2. Due to the repeated action of large vehicles, the probability of structural failure caused by insufficient load-bearing capacity in high-frequency heavy-load bridges is three times that of low-frequency light-load bridges. Therefore, the load-bearing capacity retention rate is set to the highest weight. Crack recurrence will further reduce the load-bearing capacity, so the recurrence rate is set as the second weight. Cost is only a secondary consideration to avoid sacrificing load-bearing capacity for cost control.

[0064] When the scenario is determined to be low-frequency and light-load, the weights for repair cost (ω3) and load-bearing capacity retention rate (ω1) are 0.3 and 0.2, respectively. Low-frequency, lightly loaded bridges (such as municipal branch road bridges) have small daily loads and sufficient bearing capacity redundancy. After repair, the bearing capacity retention rate can reach more than 85% for 5 years, so cost has the highest weight. Bearing capacity and recurrence rate only need to meet the basic requirements, so they have lower weight.

[0065] When the scenario is determined to be a severe environmental condition, the weights for crack recurrence rate (ω2) and bearing capacity retention rate (ω1) are 0.4 and 0.1, respectively. In harsh environments (such as high humidity in coastal areas and freeze-thaw cycles in the north), the recurrence rate of cracks after repair is 2.5 times that in normal environments, and the cost of secondary repair after recurrence is 1.8 times that of the initial repair. Therefore, the recurrence rate is given the highest weight. The load-bearing capacity needs to be controlled in conjunction with the recurrence rate (to avoid a decrease in load-bearing capacity due to recurrence), so the weight is second. The cost weight is the lowest because "avoiding recurrence" is more economical in the long run than "controlling the initial cost".

[0066] The weight values ​​are not set subjectively, but are calculated using the Analytic Hierarchy Process (AHP) based on the "effect-cost-scenario" correlation data of 100+ sets of similar bridge repair projects, ensuring that the weights meet the actual needs of the projects.

[0067] The calculation logic for the overall score is as follows: Overall score = (bearing capacity retention rate × ω1) + ((1 - crack recurrence rate) × ω2) + ((1 - relative repair cost) × ω3).

[0068] This invention quantifies bridge operation scenarios into three categories—high-frequency heavy load, low-frequency light load, and harsh environment—based on environmental load data (traffic volume, climate parameters). Differential weights are assigned to different scenarios for bearing capacity retention rate, crack recurrence rate, and repair cost. A clear weighted formula is then used to calculate the comprehensive score of each candidate repair scheme. This effectively solves the problem of "over-repair" or "under-repair" caused by single-indicator decision-making in existing technologies, ensuring that the recommended schemes accurately meet the actual operational needs of bridges. Simultaneously, the scheme improves the adaptability of repair schemes to different operation scenarios through scenario priority rules and dynamic weight adjustment boundaries, avoiding poor repair results due to scenario mismatch. Furthermore, its weight calibration relies on engineering data, and the score calculation follows objective formulas, ensuring traceability of the decision-making process and reproducibility of results, reducing subjective experience interference. This not only lowers long-term maintenance costs and the risk of secondary repair but also improves the efficiency and reliability of engineering decision-making, providing strong support for the scientific selection of bridge crack repair schemes.

[0069] In one embodiment of the present invention, the short-term performance indicators include the short-term crack closure rate after repair and the short-term bearing capacity recovery rate after repair. The crack closure rate is the percentage of the difference between the actual width of the crack and the initial width after repair, divided by the initial width. The bearing capacity recovery rate is calculated by the difference between the ultimate bearing capacity of the repaired model and the ultimate bearing capacity of the intact model through finite element analysis, and is expressed as a percentage. Short-term crack closure rate refers to the percentage of the difference between the actual crack width and the initial crack width 7 days after repair, relative to the initial crack width. The value ranges from 0 to 100%, with higher values ​​indicating better crack closure. The acceptable standards are ≥90% for minor cracks, ≥85% for moderate cracks, and ≥80% for severe cracks. Short-term load-bearing capacity recovery rate refers to the percentage of the difference between the bridge's ultimate load-bearing capacity and its load-bearing capacity in its intact state 7 days after repair, relative to its load-bearing capacity in its intact state. The value ranges from 0 to 100%, with higher values ​​indicating more complete recovery of load-bearing capacity. It is a core indicator for assessing the short-term safety of the repaired structure.

[0070] The long-term performance indicators include the crack recurrence rate in the mid-term after repair, the crack recurrence rate in the mid-to-long-term after repair, the crack recurrence rate in the long-term after repair, the bearing capacity retention rate in the mid-term after repair, the bearing capacity retention rate in the mid-to-long-term after repair, and the bearing capacity retention rate in the long-term after repair.

[0071] Crack recurrence rate refers to the percentage of crack width in year t (t=1,3,5) after repair, relative to the initial crack width. The value ranges from 0 to +∞. A lower value indicates better long-term crack suppression. The 5-year acceptable standard is ≤10% for minor cracks, ≤15% for moderate cracks, and ≤20% for severe cracks. A rate exceeding 50% triggers a secondary repair warning. Bearing capacity retention rate refers to the percentage of the bridge's ultimate bearing capacity in year t after repair, relative to its bearing capacity in its intact state. The value ranges from 0 to 100%. A higher value indicates better long-term bearing capacity maintenance. The 5-year acceptable standard for all crack types is ≥80%, which is a core indicator for judging the long-term safety of the repaired structure. Mid-term refers to one year, mid-to-long-term to three years, and long-term to five years or more.

[0072] The prediction results are output in the form of a structured table, which includes the following fields: candidate scheme number, crack type adaptability, crack severity adaptability, material parameters, process parameters, crack closure rate, bearing capacity recovery rate, crack recurrence rate, bearing capacity retention rate, repair cost per unit length, comprehensive score, and recommendation priority. Among them, crack type adaptability and crack severity adaptability are evaluated using a three-level system of "adaptable", "basically adaptable" and "incompatible", and only schemes that are "adaptable" are included in the candidate range.

[0073] This invention clarifies the definitions, calculation logic, and qualification evaluation standards of short-term and long-term performance indicators after repair. It standardizes the output format by integrating scheme parameters, performance indicators, adaptability, cost, and comprehensive scores in a structured table, and incorporates a three-level evaluation based on crack type and severity adaptability. This effectively solves the shortcomings of existing technologies, such as vague qualitative indicators of repair effect, lack of long-term durability prediction, scattered and disordered output information, and unclear determination of the adaptability between the scheme and crack characteristics. Its short-term indicators can verify the immediate repair effect, while long-term indicators can assess the structural performance stability, allowing for objective and quantitative comparison of the effects of different candidate schemes and avoiding subjective experience-based decision-making biases. The structured table makes key information readily available, significantly improving engineering decision-making efficiency. The adaptability evaluation preemptively eliminates unsuitable schemes, avoiding resource waste. Simultaneously, the prediction of long-term indicators can proactively avoid the risk of secondary repairs that are "effective in the short term but ineffective in the long term," providing comprehensive and reliable support for the accurate presentation of bridge crack repair effects and the scientific selection of schemes.

[0074] The present invention also provides a bridge crack repair effect prediction system based on performance simulation, including a memory, a host computer, and a computer program stored in the memory and executable on the host computer. The computer program is configured to implement the steps of the bridge crack repair effect prediction method based on performance simulation as described above.

[0075] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for predicting the repair effect of bridge cracks based on performance simulation, characterized in that, Includes the following steps: Obtain basic data and environmental load data of the target bridge, and match at least two candidate repair schemes from a pre-set repair scheme library; Based on the basic data, an initial performance model before repair was constructed, and the initial bearing capacity loss rate and crack propagation rate were calculated through finite element analysis as benchmark data for predicting the repair effect. Based on candidate repair schemes, environmental load data, and baseline data, the interaction effects of multimodal data are quantified by the coupling factor matrix in the preset multi-field coupling simulation model. The repair status is iteratively calculated over a set time period after repair to obtain initial simulation results. Real-time monitoring data is collected during the repair process and within a set time period after the repair. The real-time monitoring data and the initial simulation results are input into the preset LSTM adaptive correction model to obtain the corrected simulation results.

2. The method for predicting the repair effect of bridge cracks based on performance simulation according to claim 1, characterized in that, After obtaining the corrected simulation results, the short-term and long-term performance indicators of each candidate repair scheme are output. The comprehensive score of each candidate repair scheme is calculated in combination with the weight allocation rules of the bridge operation scenario. The candidate repair scheme with the highest comprehensive score is selected as the optimal repair scheme.

3. The method for predicting the repair effect of bridge cracks based on performance simulation according to claim 1, characterized in that, The basic data includes bridge structural parameters and crack characteristic data. Specifically, the bridge structural parameters are the bridge span, concrete cube compressive strength, and cross-sectional reinforcement ratio. Specifically, the crack characteristic data are the initial crack width, crack length, and crack direction. The environmental load data includes long-term daily average maximum temperature, daily average minimum temperature, annual average relative humidity, annual freeze-thaw cycle count, daily average number of large vehicles, daily average number of medium-sized vehicles, and daily average number of small vehicles. The daily average maximum temperature and daily average minimum temperature need to be normalized using the following formula: ； in, This represents the standardized temperature parameter, where T is the actual monitored temperature at a certain moment. This is the highest average daily temperature. This is the lowest daily average temperature.

4. The method for predicting the repair effect of bridge cracks based on performance simulation according to claim 3, characterized in that, The preset repair scheme library is constructed in the following way: For the three types of cracks—horizontal independent cracks, longitudinal independent cracks, and intersecting cracks—repair schemes corresponding to the severity of cracks—mild, moderate, and severe—are matched respectively, forming nine basic repair scheme combinations. The crack type is determined by the crack direction. When the crack direction is ∈ [0°, 15°] or the crack direction is ∈ [165°, 180°], it is determined as a longitudinal independent crack. When the crack direction is ∈ [75°, 105°], it is determined as a transverse independent crack. When there are two or more cracks with an angle between their directions ∈ [30°, 150°] in the same monitoring area, they are determined as intersecting cracks. The severity of the crack is determined by the initial crack width: when the initial crack width is < 0.15 mm, it is determined to be a minor crack; when 0.15 mm ≤ initial crack width ≤ 0.3 mm, it is determined to be a moderate crack; when the initial crack width is > 0.3 mm, it is determined to be a severe crack.

5. The method for predicting the repair effect of bridge cracks based on performance simulation according to claim 4, characterized in that, The specific process for constructing the initial performance model before repair is as follows: A solid model of the target bridge was established using finite element analysis software. The geometric dimensions and material properties of the model were defined according to the bridge's structural parameters. Virtual cracks were set at corresponding locations on the model based on crack characteristic data. The width and length of the virtual cracks were consistent with the initial width and length of the cracks, respectively. The crack depth was set to 1.2 times the thickness of the concrete protective layer. The calculation process for the initial bearing capacity loss rate is as follows: Vehicle load standard values ​​are applied to both the intact bridge model without cracks and the model with virtual cracks. The ultimate bearing capacity of the two models is calculated by software. The difference between the two is calculated, and then the difference is divided by the ultimate bearing capacity of the intact bridge model. The result is converted into a percentage, which is the initial bearing capacity loss rate. The calculation process for the crack propagation rate is as follows: The stress intensity factor at the crack tip is extracted based on the finite element analysis results. Combined with the fracture toughness of concrete, the crack propagation rate is calculated using a simplified form of the Paris formula.

6. The method for predicting the repair effect of bridge cracks based on performance simulation according to claim 1, characterized in that, The expression for the coupling factor matrix in the multi-field coupling simulation model is as follows: ； in, The crack recurrence rate in year t after repair. Let be the crack propagation rate in year t after repair. The bond strength of the repair material in year t after repair. This represents the sensitivity coefficient of crack recurrence rate to standardized temperature. The sensitivity coefficient of crack recurrence rate to annual average relative humidity. The sensitivity coefficient of crack recurrence rate to the number of annual freeze-thaw cycles is given. These are the sensitivity coefficients of crack propagation rate to standardized temperature, annual average relative humidity, and annual number of freeze-thaw cycles, respectively. These are the sensitivity coefficients of bond strength to standardized temperature, annual average relative humidity, and annual number of freeze-thaw cycles, respectively. Let be the standardized temperature parameter for year t. Let be the annual average relative humidity in year t. Let t be the number of freeze-thaw cycles in year t. The iterative calculation uses a time step of 1 year. The formula for calculating the crack closure rate in year t after repair is as follows: ； in, Let be the crack closure rate in year t after repair. The initial closure rate 7 days after repair. The temperature decay coefficient of the material. Let be the standardized temperature difference between year t and year t-1. The humidity attenuation coefficient of the material. This represents the difference in annual average relative humidity between year t and year t-1.

7. The method for predicting the repair effect of bridge cracks based on performance simulation according to claim 1, characterized in that, The network structure of the LSTM adaptive correction model includes an input layer, a hidden layer, and an output layer; The input layer has 4 neurons, which correspond to the crack width monitoring value, the structural strain monitoring value, the real-time standardized temperature, and the real-time relative humidity, respectively. The hidden layer contains two LSTM units, with 64 LSTM units in each layer. The forget gate, input gate, and output gate of the LSTM unit all use the tanh activation function, and the cell state update uses the sigmoid activation function. The number of neurons in the output layer is 2, corresponding to the correction amount of the material aging coefficient and the correction amount of the environmental sensitivity coefficient, respectively; The training loss function of the LSTM adaptive correction model uses the mean squared error, and the calculation formula is as follows: ； in, Mean squared error is the final calculated deviation index. When the value exceeds the preset value, parameter correction is triggered, where N is the total number of monitored samples. The simulated crack width value calculated by the performance simulation model during the i-th monitoring session. This refers to the actual crack width value collected by sensors such as crack gauges during the i-th monitoring.

8. The method for predicting the repair effect of bridge cracks based on performance simulation according to claim 2, characterized in that, The bridge operation scenarios are divided into three categories: high-frequency heavy-load scenarios, low-frequency light-load scenarios, and harsh environment scenarios. The criteria for determining the high-frequency heavy-load scenario is that the average daily number of large vehicles passing through is ≥50 vehicles. The criteria for determining the low-frequency light-load scenario is that the average daily number of large vehicles passing through is <10 vehicles. The criteria for determining the harsh environment scenario is that the number of freeze-thaw cycles per year is ≥50 or the average annual relative humidity is ≥85%. The weight allocation rule is as follows: When the scenario is determined to be a high-frequency heavy-load scenario, the weights for load-bearing capacity retention rate are ω1=0.5, crack recurrence rate is ω2=0.3, and repair cost is ω3=0.

2. When the scenario is determined to be low-frequency and light-load, the weights for repair cost (ω3) and load-bearing capacity retention rate (ω1) are 0.3 and 0.2, respectively. When the scenario is determined to be a severe environmental condition, the weights for crack recurrence rate (ω2) and bearing capacity retention rate (ω1) are 0.4 and 0.1, respectively. The calculation logic for the overall score is as follows: Overall score = (bearing capacity retention rate × ω1) + ((1 - crack recurrence rate) × ω2) + ((1 - relative repair cost) × ω3).

9. The method for predicting the effect of bridge crack repair based on performance simulation according to claim 2, characterized in that, The short-term performance indicators include the short-term crack closure rate and the short-term load-bearing capacity recovery rate after repair. The crack closure rate is the percentage of the difference between the actual crack width and the initial crack width after repair, divided by the initial crack width. The bearing capacity recovery rate is calculated by the difference between the ultimate bearing capacity of the repaired model and the ultimate bearing capacity of the intact model, and is expressed as a percentage. The long-term performance indicators include the crack recurrence rate in the mid-term after repair, the crack recurrence rate in the mid-to-long-term after repair, the crack recurrence rate in the long-term after repair, the bearing capacity retention rate in the mid-term after repair, the bearing capacity retention rate in the mid-to-long-term after repair, and the bearing capacity retention rate in the long-term after repair.

10. A bridge crack repair effect prediction system based on performance simulation, characterized in that, The system includes a memory, a host computer, and a computer program stored in the memory and executable on the host computer, the computer program being configured to implement the steps of a performance simulation-based method for predicting the repair effect of bridge cracks as described in any one of claims 1 to 9.

Images