Real-time early warning system of bridge structure health big data cloud platform

Abstract

The application discloses a bridge structure health big data cloud platform real-time early warning system and relates to the technical field of civil engineering structure health monitoring. The system collects relevant data, uniformly samples, filters and denoises, fills in missing values and aligns time sequences through an industrial internet of things gateway, and generates a structured data set. The cloud platform constructs an 8-dimensional feature matrix based on environmental excitation modal analysis, stress influence line variation coefficient calculation and crack propagation model, assesses the structure technical condition grade in combination with Bayesian dynamic correction and industry specifications. The health index is corrected by age, load and environmental reduction coefficient, and multi-level early warning threshold is set to trigger graded early warning. The system uses a non-inferior solution sorting genetic algorithm, takes the lowest safety risk and the optimal life cycle cost as double objectives, generates an optimal maintenance plan in combination with time, resource and safety constraints, realizes the intelligentization and closed loop of bridge health management, effectively improves the bridge safety life and reduces the operation and maintenance cost.

Description

Technical Field

[0001] This invention belongs to the field of civil engineering structural health monitoring technology, specifically relating to a real-time early warning system for a bridge structural health big data cloud platform. Background Technology

[0002] With the acceleration of urbanization and the expansion of infrastructure, the safety and durability of critical structures such as bridges are becoming increasingly prominent issues. Traditional bridge health monitoring methods mainly rely on manual inspections and local sensor data, which suffer from problems such as low data acquisition efficiency, poor real-time performance, and difficulty in fusing multi-source heterogeneous data, making it difficult to meet the needs of modern transportation infrastructure for high-precision and intelligent operation and maintenance.

[0003] In recent years, the rapid development of technologies such as the Internet of Things, big data analytics, and artificial intelligence has provided new solutions for bridge health monitoring. For example, real-time data acquisition technologies based on multi-type sensor networks (such as fiber optic grating sensors and GNSS positioning systems) can achieve high-precision dynamic monitoring of bridge parameters such as stress, displacement, and vibration.

[0004] However, existing systems mostly focus on isolated analysis of single parameters, lack the ability to comprehensively evaluate the structural status, and are difficult to achieve closed-loop management from monitoring data to maintenance decisions.

[0005] In terms of data processing, finite element modeling and model correction techniques have been widely used in bridge structural performance analysis. However, traditional methods often rely on static models and empirical formulas, making it difficult to dynamically adapt to changes in structural state.

[0006] Meanwhile, although real-time big data analytics can support the efficient processing of massive amounts of monitoring data, it still lacks targeted algorithms for bridge health status identification and early warning threshold setting, resulting in insufficient early warning accuracy.

[0007] In addition, existing maintenance strategies are mostly experience-driven and lack scientific decision support based on multi-objective optimization, making it difficult to balance safety risks, economic costs and resource constraints.

[0008] To address the aforementioned problems, this invention proposes a real-time early warning system for a bridge structure health big data cloud platform. Summary of the Invention

[0009] To overcome the shortcomings and deficiencies of the existing technology, the present invention adopts the following technical solution: The real-time early warning system of the bridge structural health big data cloud platform includes the following steps in its workflow: Step 1: Multi-source monitoring parameter acquisition and structured dataset construction: Monitoring parameters are acquired through a multi-type sensor network deployed on the bridge structure. Data is collected at a preset sampling frequency and then preprocessed to generate a structured bridge structure status dataset. Step 2: Extraction of core feature parameters and construction of feature matrix: Based on the structured bridge structural state dataset, extract bridge modal parameters, stiffness and mechanical behavior features and damage features, perform multi-dimensional feature correlation analysis, and form a bridge structural operation feature matrix; Step 3: Structural condition assessment based on standards and Bayesian inference: Based on the preset industry bridge technical condition assessment standards, a weighted scoring method is used to obtain a preliminary technical condition score, which is then corrected using a dynamic update model based on Bayesian inference to determine the structural technical condition level (BCI). Step 4: Health Index Calculation and Early Warning Threshold Setting: Calculate the health index HI based on the structural technical condition level, correct it with preset correction factors, set multi-level early warning thresholds, and generate health assessment results. Step 5: Life prediction and graded early warning triggering based on damage theory: Predict the remaining life of key components based on fatigue damage accumulation theory and crack propagation model, and trigger a graded early warning mechanism based on health assessment results and remaining life. Step Six: Generation and Output of Multi-Objective Maintenance Plan: Integrate maintenance decision data, construct a multi-objective maintenance optimization model, and generate a bridge structure maintenance plan after solving and selecting the optimal solution.

[0010] Preferably, in step one, the multi-type sensor network includes fiber optic strain sensors, tilt sensors, distributed hydrostatic levels, triaxial displacement gauges, GPS / BeiDou dual-mode receivers, dynamic weighing sensors, and environmental sensors, which are deployed at key sections of the bridge main beam, piers, supports, bridge deck, and bridge site area, respectively. The preprocessing includes filtering and denoising, missing value completion, outlier screening, and time-series alignment. The filtering and denoising uses moving average filtering, wavelet denoising, or Kalman filtering algorithms. The missing value completion uses linear interpolation or polynomial interpolation. Outliers are screened based on statistical principles.

[0011] Preferably, in step two, the extraction of core feature parameters includes: identifying modal parameters of vibration data using environmental excitation modal analysis; calling a pre-stored finite element model and correcting the model based on parameter sensitivity ranking until the relative error between the calculated and measured values ​​meets the preset accuracy requirements; calculating stress influence lines and coefficients of variation based on moving load and strain data to assess structural stiffness changes; calculating the goodness of fit through load-strain linear regression analysis to identify nonlinear behavior of the structure; extracting crack propagation characteristics; calculating the correlation between modal parameters and stress distribution and the time-delay correlation between load and displacement response; and integrating these to form a bridge structure operation feature matrix.

[0012] Preferably, in step three, the weighted scoring method uses core feature parameters as a basis to assign scores to core components such as main beams, piers, supports, and connection nodes according to preset weights, and calculates deduction items in combination with indicators such as stress exceeding limits, modal frequency variation, crack propagation, and support slippage; the Bayesian inference dynamic update model uses the score distribution of technical conditions of similar bridges as the prior distribution and the current monitoring data as the likelihood function to iteratively calculate the posterior score until the preset convergence condition is met.

[0013] Preferably, in step four, the health index correction formula is: ,in This is the reduction factor for the number of years of service. This is the load reduction factor. The environmental reduction factor is used; the multi-level early warning thresholds include the lower limit of the early warning threshold corresponding to the third category of structural and technical conditions, the upper limit of the early warning threshold corresponding to the fourth category, and the emergency threshold corresponding to the fifth category.

[0014] Preferably, in step five, the fatigue damage accumulation theory adopts the linear cumulative damage criterion, and calculates the damage amount and remaining fatigue life by combining the stress cycle counting results and the preset SN curve; the crack propagation model adopts the crack propagation rate formula, calls the preset material constant according to the component material properties, and calculates the crack propagation trend and remaining life by combining the stress intensity factor amplitude.

[0015] Preferably, in step five, the graded early warning mechanism includes: when the structural technical condition level is Class III, triggering daily enhanced monitoring and early warning, increasing the sensor sampling frequency and adding data collection tasks for specific time periods; when the level is Class IV or the remaining lifespan is lower than the preset lifespan threshold, triggering planned maintenance early warning, and pushing damage information and evolution trends; when abnormal situations such as a sudden drop in modal parameters, a sudden change in displacement, or a sudden exceedance of strain occur, triggering a real-time safety alarm and simultaneously pushing emergency response instructions.

[0016] Preferably, in step six, the multi-objective maintenance optimization model has two objectives: minimizing safety risk and optimizing life cycle cost. Minimizing safety risk is quantified by maximizing the improvement in health index after maintenance, while optimizing life cycle cost covers initial maintenance cost, traffic interruption loss, and subsequent monitoring cost. The model solution adopts a genetic algorithm for sorting non-dominated solutions, and the solution process combines time window constraints, resource constraints, and safety constraints to select the optimal solution.

[0017] Preferably, the system includes: Multiple types of sensor networks are used to collect monitoring parameters such as stress, strain, displacement, vibration, load, and environment of bridge structures; Industrial IoT gateways are used to uniformly collect, preprocess, and align monitoring parameters to generate structured datasets. The cloud platform data processing module is used to extract core feature parameters, construct feature matrices, assess structural status based on norms and Bayesian inference, and calculate the corrected health index. The early warning decision module is used to predict remaining lifespan based on damage theory, trigger a graded early warning mechanism, and push early warning information. The maintenance plan generation module is used to build a multi-objective maintenance optimization model, solve for and output the bridge structure maintenance plan.

[0018] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: 1. This invention identifies 1-5 order vertical bending modal parameters based on environmental excitation modal analysis, and corrects the finite element model through Latin hypercube sampling, reducing the error between measured and calculated modal frequencies. An 8-dimensional feature matrix is ​​constructed by combining the stress influence line variation coefficient, goodness of fit, and crack propagation rate, achieving a full-chain diagnosis from "parameter monitoring - feature extraction - condition assessment," which improves efficiency and reduces the false negative rate compared to traditional manual inspection.

[0019] 2. This invention constructs an adaptive risk prevention and control system through Bayesian dynamic correction and a multi-level early warning mechanism. The system adopts a two-level assessment model of "standardized initial assessment + Bayesian dynamic correction": setting deduction rules, combining normal prior distribution and the likelihood function of monitoring data, and reducing the posterior scoring error through three iterations to dynamically generate the BCI health index. Based on the corrected HI value, three-level thresholds are set and tiered responses are triggered. This mechanism realizes a closed loop of "monitoring-assessment-early warning-response," detecting risks earlier than fixed threshold early warnings and shortening emergency response time.

[0020] 3. This invention integrates the Miner fatigue damage accumulation model and the Paris crack propagation formula, calculates the remaining lifespan based on the annual average stress cycle count, and employs the NSGA-II genetic algorithm to optimize the maintenance scheme. The dual objective function encompasses both "lowest safety risk" and "optimal life-cycle cost," with constraints including project time window, resource limitations, and safety factors. An initial scheme is generated through real-number encoding, and the Pareto optimal solution is iteratively selected to reduce maintenance costs while extending the bridge's service life. This forms a scientific maintenance strategy of "preventive maintenance as the primary approach, supplemented by emergency response," improving the efficiency of asset life-cycle management. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A block diagram of the real-time early warning system of the bridge structure health big data cloud platform of the present invention is shown; Figure 2 A flowchart of the system of the present invention is shown; Figure 3 A flowchart of step six of the present invention is shown. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more exemplary embodiments. Numerous specific details are provided in the following description to give a full understanding of exemplary embodiments of this disclosure. However, those skilled in the art will recognize that the technical solutions of this disclosure can be practiced with one or more of the specific details omitted, or other methods, components, steps, etc., can be employed. In other instances, well-known structures, methods, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of this disclosure.

[0025] Example 1: See Figure 1 As shown in the figure, this embodiment discloses a real-time early warning system for a bridge structural health big data cloud platform. The system includes: The system includes multiple types of sensor networks, industrial IoT gateways, cloud platform data processing modules, early warning decision-making modules, and maintenance plan generation modules. Each module interacts with the other via Ethernet or 5G communication protocols.

[0026] Multiple types of sensor networks are used to collect monitoring parameters such as stress, strain, displacement, vibration, load, and environment of bridge structures; Industrial IoT gateways are used to uniformly collect, preprocess, and align monitoring parameters to generate structured datasets. The cloud platform data processing module is used to extract core feature parameters, construct feature matrices, assess structural status based on norms and Bayesian inference, and calculate the corrected health index. The early warning decision module is used to predict remaining lifespan based on damage theory, trigger a graded early warning mechanism, and push early warning information. The maintenance plan generation module is used to build a multi-objective maintenance optimization model, solve for and output the bridge structure maintenance plan.

[0027] The beneficial effects of this embodiment are as follows: the system achieves real-time monitoring of bridge status in all dimensions through multi-source sensor fusion and dynamic data preprocessing; it accurately assesses structural health and provides graded early warning by combining Bayesian inference and damage theory; and it reduces operation and maintenance costs and improves the safe life of bridges by generating maintenance plans through multi-objective optimization, thus forming an intelligent, closed-loop health management solution.

[0028] Example 2: See Figure 2 As shown in the figure, this embodiment discloses a real-time early warning system for a bridge structural health big data cloud platform. The specific workflow of the system is as follows: Step 1: Multi-source monitoring parameter acquisition and structured dataset construction.

[0029] S11, Sensor network deployment.

[0030] At key sections such as the mid-span, quarter-span, and supports of the main girder, a set of fiber optic strain sensors is arranged every 20cm along the section height direction. Each set contains two orthogonally arranged sensors to monitor the stress and strain distribution of the section. A dual-axis tilt sensor is installed 1.5m above the top surface of the foundation of the pier. A distributed static level is arranged every 3m along the pier axis to monitor the pier tilt and uneven settlement. A triaxial displacement gauge is installed at the upper and lower nodes of each support to monitor the X / Y slippage and Z-axis voidage of the support. A GPS / BeiDou dual-mode receiver is arranged every 50m on both sides of the bridge deck guardrail to synchronously collect the overall alignment and displacement data of the bridge deck. Dynamic weighing sensors are installed at the bridge deck lanes, and temperature and humidity sensors and wind speed sensors are deployed in the bridge site area.

[0031] S12. Data Acquisition and Preprocessing.

[0032] Sensor data is uniformly sampled through an industrial IoT gateway, with sampling frequencies set according to parameter type: 10Hz for fiber optic strain data, 20Hz for vibration acceleration data, 5Hz for GPS displacement data, 10Hz for load data, and 1Hz for environmental parameters. The collected strain data is denoised using a moving average filtering algorithm with a window size of 50 sampling points, the vibration data is processed using a 3-layer wavelet denoising algorithm based on db4 wavelet basis, and the GPS displacement data is processed using Kalman filtering. Equations of state: ; Observation equation: ; Where X(k) is the system state vector at time k (e.g., displacement, velocity); A is the state transition matrix (describing the state evolution over time); B is the control input matrix (connecting external inputs and state changes); U(k) is the control input vector at time k (e.g., external excitation); W(k) is the process noise (Gaussian white noise with a mean of 0); Z(k) is the observation vector at time k (sensor measurements); H is the observation matrix (establishing a linear relationship between the state and the observed quantities); and V(k) is the observation noise (Gaussian white noise with a mean of 0).

[0033] In bridge displacement monitoring applications, the state vector is taken as... The state transition matrix A is taken as follows: Based on a uniform motion model; since there is no external control input, the control input matrix B is set to 0; the observation matrix H is taken as... This indicates that only the position (displacement) quantity can be observed.

[0034] For data with ≤5 consecutive missing sampling points, linear interpolation is used for completion; for data with >5 consecutive missing sampling points, cubic polynomial interpolation is used for completion, based on the 3σ principle. Outliers are filtered out and marked as "to be verified" using the Network Time Protocol (NTP) to align various data to millisecond-level timestamps. A structured bridge structural status dataset is generated in the format of "sensor ID-acquisition time-parameter type-value-precision" and uploaded to the cloud platform via a 5G module.

[0035] Step 2: Extraction of core feature parameters and construction of feature matrix.

[0036] S21. Modal parameter identification and model correction: The random subspace method (SSI) in environmental excitation modal analysis (EMA) is used to identify modal parameters of vibration acceleration time history data, and obtain the frequency, mode shape (normalized) and damping ratio of the first to fifth vertical bending modes of the bridge. A pre-stored bridge finite element model (built using MidasCivil, element type: BEAM188 beam element) was called from the cloud platform. The identified modal frequencies were compared with the calculated values ​​from the model. The Latin hypercube sampling method was used to rank the sensitivity of parameters such as the elastic modulus of concrete and the moment of inertia of the cross section in the model. Parameters with a sensitivity > 0.7 were selected as correction variables. The least squares method was used for iterative optimization (objective function: minimizing the sum of squared errors of modal frequencies) until the relative error between the calculated frequency and the measured value was ≤ 5%. The sensitivity threshold > 0.7 was determined based on the cumulative contribution rate analysis of the Latin hypercube sampling results, which can screen out key parameters that have a total impact of more than 90% on the modal frequencies. The relative error ≤ 5% was set according to the commonly used accuracy requirements for model verification in the "Specifications for Load Testing of Highway Bridges" (JTG / TJ21-01-2015).

[0037] S22. Extraction of stiffness and mechanical behavior characteristics.

[0038] Based on the moving load time history collected by dynamic weighing sensors, and combined with strain sensor data at corresponding locations, the stress influence line at the mid-span section of the main beam is calculated using the convolution integral method, and the coefficient of variation is calculated using the following formula: Where σ is the standard deviation of the peak value of the stress influence line under the same load level, and μ is the mean value of the peak value; when CV > 15%, it is determined that there is a significant change in the structural stiffness.

[0039] Linear regression analysis was performed on the load-strain data (fitting equation: Where ε is the strain value, and the linear regression fitting equation is... In the linear regression, k is the dependent variable, b is the slope, and p is the load (the independent variable in the linear regression). The goodness of fit is calculated. (determination coefficient, (The larger the value, the better the fit). When the structure is found to exhibit nonlinear behaviors such as support detachment and component plastic deformation, the nonlinear initial load value is recorded simultaneously.

[0040] S23. Damage characteristics and correlation analysis.

[0041] Crack data was collected using a crack gauge, and the crack width versus time curve was fitted using the least squares method to calculate the crack propagation rate v. The Pearson correlation coefficient r was used to calculate the correlation between modal frequencies and stress distribution (threshold: The modal frequencies are determined to be strongly correlated with stress. The cross-correlation function is used to analyze the time-delay correlation between load and displacement response (time delay ≤ 0.5s is synchronous response). The modal parameters (3 dimensions), stress influence line variation coefficient (1 dimension), goodness of fit R² (1 dimension), crack propagation rate (1 dimension), and correlation coefficient (2 dimensions) are integrated to form an 8-dimensional bridge structure operation characteristic matrix.

[0042] Step 3: Structural state assessment based on norms and Bayesian inference.

[0043] S31. Preliminary weighted score calculation.

[0044] Based on the "Standard for Technical Condition Assessment of Highway Bridges" (JTG / TH21-2011), the weights of each core component are determined as follows: main beam 0.4, pier 0.3, bearing 0.2, and connection node 0.1. According to Table 5.1.2 of the specification, the basic score is set at 100 points, and deductions are calculated according to the following standards: Stress exceeding limits: 5 points will be deducted when the measured stress is 10% to 20% of the design stress, and 10 points will be deducted when the measured stress is more than 20% of the design stress. Modal frequency variation: 3 points will be deducted if the measured frequency decreases by 3% to 5% compared to the calculated value of the corrected model, and 8 points will be deducted if the decrease is greater than 5%. Crack propagation: Deduct 4 points if the propagation rate is 0.1~0.3 mm / year, and deduct 9 points if the propagation rate is >0.3 mm / year; Support slippage: Deduct 2 points when the slippage is 5~10mm, and deduct 6 points when the slippage is >10mm.

[0045] According to the formula Calculations were performed to obtain a preliminary technical condition score.

[0046] S32, Bayesian dynamic correction.

[0047] The technical condition score distribution of similar bridges recommended by the standards is used as the prior distribution (set as a normal distribution). ,in , , , for The mean and standard deviation are used to set the likelihood function of the current monitoring data (preliminary score) as a normal distribution. The mean BCI is the true score to be estimated, and the standard deviation σ is set to 5 based on the noise level of the monitoring data. The posterior score is calculated using the Bayesian formula: posterior distribution ∝ prior distribution × likelihood function. This process is iterated three times until the difference between two consecutive scores is ≤0.5, yielding the final Structural Technical Condition Level (BCI). Scores of 90 and above are classified into five categories: 80-89, 70-79, 60-69, and below 60.

[0048] Step 4: Health Index Calculation and Early Warning Threshold Setting.

[0049] S41, Health Index (HI) Calculation.

[0050] Define HI = 100 - BCI deduction value, and make corrections based on the bridge's service life t, design load level L, and environmental corrosion degree C (C = 1 when the chloride ion content in the concrete, as quantitatively analyzed by core drilling, exceeds 0.10% of the cementitious material mass, otherwise C = 0), and design load level L.

[0051] Correction formula: ; This is the reduction factor for the number of years of service. The design service life is calculated as 100 years; t represents the service life.

[0052] is the load reduction factor, and L is the load level factor. The value of L is 1.0 for Highway Class I and 0.9 for Highway Class II.

[0053] C is the environmental reduction factor, and C is the corrosion degree identifier.

[0054] S42. Multi-level early warning threshold setting: Set the threshold according to the corrected HI value. Lower limit of warning threshold: =70 (corresponding to three types of states); Warning threshold upper limit: =60 (corresponding to four types of states); Emergency threshold: <50 (corresponding to five types of states).

[0055] Will Compare with the threshold to generate a list containing "technical condition level", The health assessment report, which includes "value, abnormal parameter name, and deviation amount", is stored in the cloud platform database.

[0056] Step 5: Lifetime prediction and graded early warning triggering based on damage theory.

[0057] S51. Prediction of remaining life of key components.

[0058] (1) Calculation of cumulative fatigue damage.

[0059] The rainflow counting method was used to perform stress cycle counting on the strain time history data of the main beam to obtain the stress amplitude Δσ at each level and the number of cycles N. The damage was then calculated according to the Miner linear cumulative damage criterion. ;in, Actual stress amplitude The number of stress cycles under the given conditions; For the corresponding stress amplitude The fatigue life is calculated according to the SN curve in the "Design Specification for Highway Steel Structure Bridges" (JTGD64-2015). ); Stress amplitude (the difference between the maximum stress and the minimum stress); This is the recommended value in the "Design Code for Highway Steel Structure Bridges".

[0060] (2) Crack propagation prediction.

[0061] For cracks wider than 0.1 mm, use the Paris formula. Calculate the expansion trend, where, The value represents the crack propagation rate. The material constants C and m in the Paris formula need to be determined based on the actual materials of the bridge components. For Q345 steel, C = 3.2 × 10⁻¹¹, m = 3.0; for C50 concrete, C = 1.65 × 10⁻¹¹ can be used. 9 m=1.75; the constant is stored in the cloud platform database and is automatically retrieved according to the material properties of the damaged component; stress intensity factor amplitude α is the shape factor; a is the current crack length. We take 1.12; a is the current crack length. According to the "Specification for Testing and Evaluation of Bearing Capacity of Highway Bridges" (JTG / TJ21-2011), determine the critical crack propagation size. The Paris formula is solved by numerical integration to calculate the crack size from the current dimension. Expand to critical size Required number of stress cycles The remaining life is then calculated based on the average number of stress cycles per year. .

[0062] S52, Implementation of the tiered early warning mechanism.

[0063] Strengthen daily monitoring and early warning: When BCI is category three ( When the cloud platform automatically sends instructions to the strain and vibration sensors, it increases the sampling frequency to 20Hz, adds a daily no-load data collection task from 2:00 to 4:00 AM, and generates a daily monitoring report. Planned maintenance early warning: When BCI is Category IV ( When the remaining life is less than 20% of the design life, the early warning module pushes early warning information containing "damaged component ID, remaining life estimate, risk level (medium risk)" to the maintenance unit terminal, including a damage evolution trend chart. Real-time security alert: An emergency alert will be triggered immediately when any of the following occurs: ① The modal frequency drops by more than 5% within 1 hour; ②The displacement of the pier top changes abruptly by more than 20 mm within 10 minutes; ③ The instantaneous exceedance of the strain limit is greater than 30% of the design value and lasts for more than 30 seconds. The alarm will be triggered simultaneously by the audible and visual warning device (installed in the duty room at the bridgehead) and the emergency response instruction of "temporary load limit of 15t, prohibition of large vehicles" will be pushed to the traffic management system.

[0064] Step 6: Generation and output of multi-objective maintenance plan.

[0065] See Figure 3 As shown, the specific process for step six is ​​as follows: S61. The maintenance decision data integration links early warning information with the "bridge structural drawings, historical maintenance records, and material inventory data" pre-stored on the cloud platform to generate decision data: The location of the damaged component corresponds to the drawing number, the average repair time in history (±1 day), and the required material type.

[0066] S62. Cost and performance impact analysis calculates the cost of different maintenance schemes: Traffic interruption losses are calculated as "average daily traffic volume × hourly delay cost × maintenance period"; in this embodiment, maintenance costs are calculated as "labor + materials + equipment rental", and a cost estimation table is generated; Construct a performance impact factor matrix, with factors including "HI improvement after maintenance" and "life extension years", each with a weight of 0.5.

[0067] S63. Maintenance plan optimization and output.

[0068] A maintenance optimization model with the dual objectives of "lowest safety risk and optimal life cycle cost" is constructed. Based on the constraints, a non-dominated genetic algorithm (NSGA-II) is used to solve for the optimal maintenance scheme, as detailed below: S631, Construction of dual objective function.

[0069] (1) Minimize safety risk objective function to maintain post-structural health index Maximizing the value increase is the core quantification method for reducing security risks; the function expression is: ;in, This is the current health index correction value. The expected health index correction value after maintenance (based on historical similar maintenance cases: the average HI score increases by 15-20 points after crack repair and by 20-25 points after support replacement).

[0070] (2) Optimal objective function for life cycle cost.

[0071] The comprehensive calculation of the entire maintenance lifecycle cost, encompassing initial maintenance costs, traffic interruption losses, and subsequent monitoring costs, is expressed by the following function: ;in, The initial maintenance cost (labor + materials + equipment rental) can be calculated as follows: "3 people / team of construction team × 200 yuan / person / day × construction period + material unit price × usage + number of equipment units × 1000 yuan / unit / day × construction period". Losses due to traffic disruptions can be calculated, for example, as "average daily traffic volume × hourly delay cost × construction period"; The cost of subsequent monitoring is calculated based on the sensor operation and data processing costs for new monitoring tasks added within 6 months after maintenance.

[0072] S632, Setting Constraints.

[0073] Time window constraint: Maintenance duration T satisfies ( For the shortest construction period of similar maintenance, such as crack repair Heaven, support replacement sky; The maintenance period is limited to the material's shelf life or the critical period for risk diffusion, with a maximum of 15 days. Furthermore, the maintenance start date must avoid dates with an average daily traffic volume of ≥5000 vehicles within the past 3 months and statutory holidays.

[0074] Resource constraints: The number of construction teams ≤ 5 (the maximum number of teams available to the maintenance unit); the inventory of key materials must meet the demand (e.g., the inventory of bearings ≥ the required replacement quantity, and the inventory of epoxy resin grouting materials ≥ the crack volume × 1.2 times the loss coefficient).

[0075] Safety constraints: The load-bearing capacity of the temporary support system during maintenance must be greater than or equal to the structure's self-weight plus the construction load.

[0076] S633, non-dominated solution sorting genetic algorithm solution.

[0077] The NSGA-II algorithm was used for iterative solution. The algorithm parameters were set as follows: population size 50, number of iterations 100, crossover probability 0.8, mutation probability 0.05, and crowding distance threshold 0.1.

[0078] The solution process is as follows: Population initialization: 50 maintenance plans are randomly generated, each containing decision variables; real numbers are used for encoding. Decision variables include: maintenance start date (represented as the number of days from the current date, ranging from...). Construction period (number of days, scope) ), Number of work teams (integer, range) Material usage (continuous value, range) ) Fitness evaluation: Calculate the fitness value of each group of solutions according to the biobjective function, and eliminate infeasible solutions (such as solutions with insufficient resources or time conflicts) in combination with constraints. Non-dominated sorting: The population is divided into different levels by fast non-dominated sorting, and the crowding distance is calculated to maintain population diversity; Selection, crossover, and mutation: The tournament selection method is used to select the parent generation, and binary crossover and polynomial mutation are simulated to generate the offspring generation. After merging the parent and offspring generations, the sorting and selection are repeated until the iteration ends.

[0079] S634, Optimal Solution Selection and Output.

[0080] From the final Pareto optimal solution set, according to " Partial and The optimal solution is selected based on the criteria of “…”, and the maintenance time window is determined. Operation guidelines are generated according to the damage type: crack repair adopts the process of “epoxy resin grouting + carbon fiber cloth bonding”, and bearing replacement adopts the process of “synchronous jacking - old bearing removal - new bearing installation”. The integrated maintenance plan includes “maintenance time, construction team, equipment list, and traffic diversion route” and is pushed to the maintenance unit terminal in PDF format.

[0081] The beneficial effects of this embodiment are as follows: the system achieves real-time monitoring of the bridge in all dimensions through multi-source sensor fusion, accurately assesses the structural state by combining Bayesian dynamic correction and damage theory, responds to risks in a timely manner through a graded early warning mechanism, and generates maintenance plans through multi-objective optimization, effectively improving the safe life of the bridge, reducing operation and maintenance costs, and forming an intelligent health management closed loop.

[0082] All formulas in this invention are dimensionless and calculated numerically. The preset parameters in the formulas can be set by those skilled in the art according to the actual situation.

[0083] The weighting coefficients of this invention are used to measure the degree of influence of different factors or variables on a certain outcome or decision. The weighting coefficient is defined as the numerical value assigned to each factor when comparing and evaluating multiple factors, reflecting their importance or priority. These weighting coefficients can be determined according to specific circumstances and needs, and are usually jointly formulated and confirmed by professionals or relevant stakeholders. By reasonably setting the weighting coefficients, programs or systems can be helped to make decisions or predictions more accurately.

[0084] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

[0085] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to specific implementations. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A real-time early warning system for a bridge structural health big data cloud platform, characterized in that, The system workflow includes the following steps: Step 1: Multi-source monitoring parameter acquisition and structured dataset construction: Monitoring parameters are acquired through a multi-type sensor network deployed on the bridge structure. Data is collected at a preset sampling frequency and then preprocessed to generate a structured bridge structure status dataset. Step 2: Core Feature Parameter Extraction and Feature Matrix Construction: Based on the structured bridge structural state dataset, extract bridge modal parameters, stiffness and mechanical behavior characteristics, and damage characteristics, perform multi-dimensional feature correlation analysis, and form a bridge structural operation feature matrix. The core feature parameter extraction includes: using environmental excitation modal analysis to identify modal parameters of vibration data, calling a pre-stored finite element model and correcting the model based on parameter sensitivity ranking until the relative error between the calculated and measured values ​​meets the preset accuracy requirements; calculating stress influence lines and coefficients of variation based on moving load and strain data to assess structural stiffness changes; calculating the goodness of fit through load-strain linear regression analysis to identify nonlinear behavior of the structure; extracting crack propagation characteristics, calculating the correlation between modal parameters and stress distribution, and the time-delay correlation between load and displacement response, and integrating them to form the bridge structural operation feature matrix. Step 3: Structural condition assessment based on standards and Bayesian inference: Based on the preset industry bridge technical condition assessment standards, a weighted scoring method is used to obtain a preliminary technical condition score, which is then corrected using a dynamic update model based on Bayesian inference to determine the structural technical condition level (BCI). Step 4: Health Index Calculation and Early Warning Threshold Setting: Calculate the health index HI based on the structural technical condition level, and then adjust it using a preset correction factor. The health index correction formula is as follows: ,in This is the reduction factor for the number of years of service. This is the load reduction factor. Assign environmental reduction factors, set multi-level early warning thresholds, and generate health assessment results; Step 5: Life Prediction and Graded Early Warning Triggering Based on Damage Theory: Based on fatigue damage accumulation theory and crack propagation model, the remaining life of key components is predicted, and a graded early warning mechanism is triggered by combining the health assessment results and remaining life. The fatigue damage accumulation theory adopts the linear cumulative damage criterion, and calculates the damage amount and remaining fatigue life by combining the stress cycle counting results and the preset SN curve. The crack propagation model adopts the crack propagation rate formula, calls the preset material constant according to the component material properties, and calculates the crack propagation trend and remaining life by combining the stress intensity factor amplitude. Step Six: Generation and Output of Multi-Objective Maintenance Plan: Integrate maintenance decision data, construct a multi-objective maintenance optimization model, and generate a bridge structure maintenance plan after solving and selecting the optimal solution.

2. The real-time early warning system of the bridge structural health big data cloud platform according to claim 1, characterized in that, In step one, the multi-type sensor network includes fiber optic strain sensors, tilt sensors, distributed hydrostatic levels, triaxial displacement gauges, GPS / BeiDou dual-mode receivers, dynamic weighing sensors, and environmental sensors, which are deployed at key sections of the bridge main beam, piers, supports, bridge deck, and bridge site area, respectively. The preprocessing includes filtering and denoising, missing value completion, outlier screening, and time-series alignment. The filtering and denoising uses moving average filtering, wavelet denoising, or Kalman filtering algorithms. The missing value completion uses linear interpolation or polynomial interpolation. Outliers are screened based on statistical principles.

3. The real-time early warning system of the bridge structural health big data cloud platform according to claim 1, characterized in that, In step three, the weighted scoring method uses core feature parameters as a basis to assign scores to the main beam, piers, supports, and connection nodes according to preset weights, and calculates deduction items by combining stress exceeding limits, modal frequency variation, crack propagation, and support slippage indices; the dynamic update model of the Bayesian inference uses the score distribution of technical conditions of similar bridges as the prior distribution and the current monitoring data as the likelihood function to iteratively calculate the posterior score until the preset convergence condition is met.

4. The real-time early warning system of the bridge structural health big data cloud platform according to claim 1, characterized in that, In step five, the graded early warning mechanism includes: when the structural technical condition level is Class III, triggering daily enhanced monitoring and early warning, increasing the sensor sampling frequency and adding data collection tasks for specific time periods; when the level is Class IV or the remaining lifespan is lower than the preset lifespan threshold, triggering planned maintenance early warning, and pushing damage information and evolution trends; when there is a sudden drop in modal parameters, a sudden change in displacement, or an abnormal situation of instantaneous strain exceeding the limit, triggering a real-time safety alarm and simultaneously pushing emergency response instructions.

5. The real-time early warning system of the bridge structural health big data cloud platform according to claim 1, characterized in that, In step six, the multi-objective maintenance optimization model takes the lowest safety risk and the optimal life cycle cost as its dual objectives. The lowest safety risk is quantified by maximizing the improvement of the health index after maintenance, while the optimal life cycle cost covers the initial maintenance cost, traffic interruption loss and subsequent monitoring cost. The model is solved using a genetic algorithm that sorts non-dominated solutions. The solution process combines time window constraints, resource constraints and safety constraints to select the optimal solution.

6. The real-time early warning system of the bridge structural health big data cloud platform according to claim 1, characterized in that, The system includes: A multi-type sensor network is used to collect monitoring parameters of the bridge structure, including stress-strain, displacement, vibration, load, and environmental parameters. Industrial IoT gateways are used to uniformly collect, preprocess, and align monitoring parameters to generate structured datasets. The cloud platform data processing module is used to extract core feature parameters, construct feature matrices, assess structural status based on norms and Bayesian inference, and calculate the corrected health index. The early warning decision module is used to predict remaining lifespan based on damage theory, trigger a graded early warning mechanism, and push early warning information. The maintenance plan generation module is used to build a multi-objective maintenance optimization model, solve for and output the bridge structure maintenance plan.

Images