Intelligent Maintenance Decision-making System and Decision-making Method for Highway Bridges

Abstract

The present invention provides an intelligent maintenance decision-making system and decision-making method for highway bridges. The system includes data acquisition, processing and storage, bridge condition assessment, maintenance decision-making, implementation and management, monitoring and feedback modules. The data acquisition module collects various types of data such as bridges, environment, and traffic through multiple means such as fiber Bragg gratings and vibration sensors. The processing and storage module uses the DBSCAN algorithm to clean the data, one-hot encoding to process the categorical data, and constructs a distributed database based on the blockchain for storage. The assessment module uses the Isomap algorithm to evaluate the technical condition and predict diseases. The decision-making module uses the particle swarm optimization-simulated annealing algorithm to generate solutions and optimizes them using the analytic hierarchy process-grey clustering analysis. The maintenance implementation is managed with the help of a BIM model. The monitoring module collects data in real time to feedback and adjust strategies. This system can collect data in real time with multiple sensors, process the data accurately, evaluate diseases accurately, optimize maintenance decisions, dynamically adjust strategies, and ensure the safety of bridges and extend their service life.

Description

Technical Field

[0001] The present invention relates to the field of intelligent maintenance of bridges, and particularly to an intelligent maintenance decision-making system and method for highway bridges. Background Art

[0002] As a key component of transportation infrastructure, the structural safety and service performance of highway bridges directly affect the smoothness and safety of transportation. However, over time and with the continuous increase in traffic flow, highway bridges will inevitably develop various diseases, such as cracks, structural aging, and degradation of material properties. These diseases not only reduce the bearing capacity and service life of the bridges but may also trigger serious safety accidents. Traditional bridge maintenance methods mainly rely on manual regular inspections and experience-based judgments, suffering from problems such as long inspection cycles, strong subjectivity, and difficulty in real-time monitoring of bridge conditions, making it impossible to detect and handle bridge diseases in a timely manner and difficult to meet the requirements of modern highway bridge maintenance.

[0003] Although there are already some related methods applied in the field of bridge maintenance, there are still some problems. For example, the fusion and processing of multi-sensor data are not efficient enough, and the accuracy and reliability of the data need to be improved; the adaptability and generalization ability of the maintenance decision-making model are insufficient, making it difficult to handle bridges of different types and working conditions; the adjustment of maintenance strategies lacks real-time and pertinence, and cannot be optimized in a timely manner according to the dynamic changes of bridge conditions. Therefore, it is of great practical significance to develop an intelligent maintenance decision-making system and method that can achieve real-time monitoring, accurate assessment, decision-making, and dynamic adjustment of highway bridges. Summary of the Invention

[0004] The main object of the present invention is to provide an intelligent maintenance decision-making system and method for highway bridges, so as to solve the problems that traditional bridge maintenance methods mainly rely on manual regular inspections and experience-based judgments, with long inspection cycles, strong subjectivity, and difficulty in real-time monitoring of bridge conditions, making it impossible to detect and handle bridge diseases in a timely manner and difficult to meet the requirements of modern highway bridge maintenance.

[0005] To solve the above technical problems, the technical solution adopted by the present invention is: an intelligent maintenance decision-making system for highway bridges, which includes a data acquisition module, a data processing and storage module, a bridge condition assessment module, a maintenance decision-making module, a maintenance implementation and management module, and a maintenance monitoring and feedback module;

[0006] The bridge condition assessment module includes a technical condition module of the bridge and a module for predicting the development trend of diseases;

[0007] The maintenance implementation and management module includes a BIM-based three-dimensional model module of the bridge;

[0008] The monitoring and feedback module is set on the bridge. The monitoring and feedback module includes fiber Bragg grating sensors, vibration sensors, and high-definition cameras. The data monitored by the fiber Bragg grating sensors, vibration sensors, and high-definition cameras is transmitted to the data processing and storage module. The data in the data processing and storage module is input into the bridge condition assessment module to conduct a technical condition assessment of the bridge and predict the development trend of diseases.

[0009] In the preferred solution, the acquisition module includes fiber Bragg grating sensors, vibration sensors, high-definition cameras, drone detection equipment, and manual inspection methods for acquisition;

[0010] The acquisition module acquires condition data, environmental data, traffic data, and historical maintenance record data of the highway intranet and the Internet.

[0011] In the preferred solution, the data processing and storage module uses the DBSCAN algorithm to clean the acquired data, encodes the categorical data using the one-hot encoding method, and constructs a distributed database based on blockchain technology to store the processed data.

[0012] In the preferred solution, both the technical condition module and the module for predicting the development trend of diseases in the bridge condition assessment module use the Isomap algorithm to evaluate the technical condition of the bridge.

[0013] In the preferred solution, the method includes:

[0014] S1. Use fiber Bragg grating sensors, vibration sensors, high-definition cameras, drone detection equipment, and manual inspection methods to collect various data of the bridge, and at the same time collect bridge condition data, environmental data, traffic data, and historical maintenance record data through the highway intranet and the Internet;

[0015] S2. The data processing and storage module uses the DBSCAN algorithm to clean the data collected in step S1, encodes the categorical data using the one-hot encoding method, and stores the processed data in a distributed database based on blockchain technology;

[0016] S3. The bridge condition assessment module extracts the data stored in the database, uses the Isomap algorithm to evaluate the technical condition of the bridge for the extracted data, predicts the development of diseases, and determines the health level of the bridge and the priority of maintenance requirements;

[0017] S4. Evaluate according to the technical condition obtained in step S3, predict the disease development data, generate multiple maintenance plans using the particle swarm optimization simulated annealing hybrid algorithm, evaluate and optimize the plans using the method combining the analytic hierarchy process and grey clustering analysis, and select the optimal maintenance plan;

[0018] S5. According to the selected maintenance plan, organize construction in accordance with construction specifications and safety standards with the help of BIM technology, and manage the construction progress and quality;

[0019] S6. The data collected by the fiber Bragg grating sensors, vibration sensors and high-definition cameras in the monitoring and feedback module are used to monitor the bridge status in real time;

[0020] The monitoring data is then transmitted to the bridge condition assessment module in S3, and the maintenance effectiveness is re-evaluated regularly and the results are fed back to adjust the maintenance strategy.

[0021] In the preferred solution, in step S1, the fiber Bragg grating sensors are installed at the mid-span of the bridge, around the bearings, and at the connection between the pier and the beam;

[0022] Special glue is applied to the installation positions at the mid-span of the bridge and around the bearings, and the fiber Bragg grating sensors are pasted on the bridge surface;

[0023] Embedded installation is carried out at the connection between the pier and the beam. A groove is chiseled at the connection between the pier and the beam, the fiber Bragg grating sensor is placed in the groove, and the groove is sealed with concrete slurry;

[0024] The fiber Bragg grating sensors monitor the strain and temperature of the bridge structure at all times;

[0025] High-definition cameras and UAV detection equipment obtain the external images of the bridge;

[0026] Manual inspections record the detailed locations and forms of the diseases.

[0027] In the preferred solution, the vibration sensors analyze the vibration information of the bridge when vehicles pass by. The specific steps include:

[0028] A1. First, use the time-frequency analysis algorithm to reveal the frequency characteristics of the signal at different times, including using the short-time Fourier transform or wavelet transform to expand the signal in the time and frequency dimensions to reveal the frequency characteristics of the signal at different times. Utilize the multi-resolution analysis characteristics of the wavelet transform to automatically adjust the window size according to the signal frequency and capture the transient changes in the bridge vibration signal;

[0029] A2. Then use the wavelet packet decomposition algorithm to achieve fine decomposition of the signal, obtain sub-signals in different frequency bands, analyze the energy distribution and amplitude change characteristics of the sub-signals, and compare the characteristic differences of the sub-signals in the normal and abnormal states to judge whether there are damages or abnormalities in the bridge structure;

[0030] A3. Then, adaptively process non-linear and non-stationary signals using the empirical mode decomposition algorithm. Decompose the bridge vibration signal into multiple intrinsic mode functions using the empirical mode decomposition algorithm, extract the main characteristic components related to the bridge structure vibration, remove noise and interference components, calculate the energy and frequency parameters of the intrinsic mode function components, and observe their changing trends over time, so as to evaluate the stability and health status of the bridge structure;

[0031] A4. Finally, identify abnormal vibration patterns using the support vector machine algorithm. Conduct abnormal vibration pattern recognition with the support vector machine algorithm, extract amplitude, frequency, and energy characteristic parameters from the vibration signal as the input of the support vector machine. Use a large amount of bridge vibration data in normal and abnormal states to train the support vector machine, enabling it to learn the characteristic differences between normal and abnormal vibration patterns. When actually monitoring, input the characteristics of the real-time collected vibration signal into the trained support vector machine model to determine whether the current vibration state is normal and the type and severity of the abnormality.

[0032] In the preferred solution, the steps of the bridge condition assessment method using the Isomap algorithm are as follows:

[0033] S31. Extract bridge data from the blockchain-based distributed database, including bridge structure strain, displacement, crack size, traffic data, environmental parameters, and past maintenance records. Process missing data using the multiple imputation method and detect outliers using the local outlier factor algorithm;

[0034] S32. Use the ReliefF algorithm for feature selection, and use the selected features as dimensions to construct a high-dimensional dataset, providing a comprehensive information basis for bridge technical condition assessment and disease development prediction and capturing complex relationships;

[0035] S33. Use the Mahalanobis distance to measure the distance between data points, use the kd-tree to determine the k nearest neighbor points of each data point, and construct a neighborhood graph to reflect the local similarity of data points and provide a structural basis for subsequent calculations;

[0036] S34. Use the BellmanFord algorithm to calculate the shortest path between any two points in the neighborhood graph as the geodesic distance, reflecting the true structural relationship of bridge samples, and use it for low-dimensional embedding and feature extraction to provide key information for assessment and prediction;

[0037] S35. Adopt the kernel Isomap method, introduce the Laplacian kernel function, perform eigenvalue decomposition after centering the kernel matrix, and project the data into a low-dimensional space; use the spectral clustering algorithm to cluster bridge samples in the low-dimensional space, and assign a technical condition score to bridge samples according to the clustering results and related factors;

[0038] S36. Adopt a long short-term memory network, use the low-dimensional space feature sequence as the input, and train the network to predict the disease development trend based on historical data;

[0039] S37. Use the fuzzy comprehensive evaluation method to divide the bridge health level according to the technical condition score and the disease development prediction result, and combine the bridge importance factor to determine the maintenance priority using the data envelopment analysis method.

[0040] In the preferred solution, the generation of multiple maintenance methods based on the particle swarm optimization simulated annealing hybrid algorithm in step S4 is as follows:

[0041] S41. Obtain the bridge technical condition assessment and disease development prediction data from step S3, clarify the maintenance goal, determine the decision variables, where the decision variables include the type of maintenance measures, the maintenance time node, and the amount of maintenance resource allocation, and set the upper limit of maintenance cost and the upper limit of maintenance time constraint conditions;

[0042] Set the particle swarm size , randomly initialize the position vector and velocity vector of each particle, where is the number of decision variables;

[0043] Define the fitness function as , where , , are the weight coefficients and , represents the maintenance effect, represents the maintenance cost, represents the maintenance time;

[0044] In each iteration, calculate the fitness value of each particle, update the historical optimal position of the particle itself and the global optimal position of the entire particle swarm, and update the particle velocity and position through the formulas and , where is the inertia weight, and are the learning factors, and are random numbers uniformly distributed in the interval [0, 1], represents the current iteration number;

[0045] If , accept the new position; otherwise accept the new position with probability , where is the current temperature, according to cool down, is the cooling coefficient;

[0046] After each update of the particle position, check whether the constraint conditions are satisfied. If not, correct the particle position to make it return to the feasible solution space;

[0047] When the preset maximum number of iterations or the fitness value changes less than the set threshold in several consecutive iterations stop the iteration, record the position of the better particle and generate multiple maintenance plans accordingly;

[0048] The steps of evaluating and optimizing by combining the analytic hierarchy process and grey clustering analysis include:

[0049] S42. Construct a hierarchical structure model, including selecting the optimal maintenance plan, maintenance effect, cost-benefit, implementation difficulty, environmental impact and the multiple maintenance plans generated above; construct a judgment matrix and assign values; calculate the maximum eigenvalue of the judgment matrix and its corresponding eigenvector , normalize to obtain the weight vector, and at the same time calculate the consistency index and the random consistency ratio , when the judgment matrix is considered to have satisfactory consistency, otherwise readjust;

[0050] Among them, represents the order of the judgment matrix, and RI represents the random consistency index;

[0051] Weight the weights of each criterion in the criterion layer and the weights of each plan in the plan layer under the corresponding criterion and sum them to obtain the comprehensive weight of each plan;

[0052] Grey clustering analysis is used to cluster and evaluate the maintenance plans: The clustering evaluation of the maintenance plans includes the improvement rate of the maintenance effect, the cost reduction rate, the construction period shortening rate and the whitenization weight function;

[0053] For each maintenance plan and each clustering index , according to the index value calculate its clustering coefficient belonging to each grey class , where is the number of clustering indexes, is the weight of the index ; divide each maintenance plan into the corresponding grey class according to the clustering coefficient, and sort the maintenance plans in combination with the comprehensive weight obtained by the analytic hierarchy process;

[0054] Determine the optimal solution by combining expert experience with actual situation: Further evaluate the top-ranked solutions in combination with expert experience and actual engineering situation, and finally select the optimal bridge maintenance solution.

[0055] In the preferred solution, the BIM technology manages the construction progress and quality by constructing a three-dimensional model of the bridge and associating the construction progress information.

[0056] In the preferred solution, the solutions in S6 include:

[0057] S61. The data collected by fiber Bragg grating sensors, vibration sensors, and high-definition cameras are transmitted to the data processing center in real time using wireless transmission technology and adopting a data encryption and retransmission mechanism, and stored in a distributed time series database; the sensor data is corrected for errors through the sensor error correction formula where is the original measured value, is the true value, and are the coefficients obtained by least squares fitting;

[0058] Linear interpolation formula is used for data preprocessing to fill in the missing data values, where is the time point where the missing value is located, and are the adjacent time points before and after it, and are the corresponding measured values;

[0059] The outliers are detected and removed using the local outlier factor method; the maximum-minimum normalization is used to normalize the data of different types of sensors;

[0060] S62. The preprocessed data is sent to the bridge condition assessment module in S3, and the current state of the bridge is evaluated based on the Isomap algorithm model and the relevant evaluation index system; calculate the change value of the bridge health condition score and the crack propagation speed change rate ;

[0061] where is the current score, is the score before maintenance, where is the current bridge damage propagation speed, is the bridge damage propagation speed before maintenance;

[0062] Construct a comprehensive evaluation index for the maintenance effect where , , are the corresponding weights and , is the actual maintenance cost, is the budgeted maintenance cost;

[0063] S63. According to the comprehensive evaluation index of maintenance effectiveness and the set high threshold and low threshold make a comparison and judgment;

[0064] If , maintain the current maintenance strategy;

[0065] If , re-analyze the causes and development trends of bridge diseases, use the particle swarm optimization simulated annealing hybrid algorithm to generate a new maintenance plan, and use the method combining analytic hierarchy process and grey clustering analysis to evaluate and optimize the new plan and select the optimal strategy;

[0066] If , make appropriate fine-tuning of the implementation frequency of maintenance measures and the allocation ratio of maintenance resources.

[0067] The present invention provides a highway bridge intelligent maintenance decision-making system and decision-making method. The highway bridge intelligent maintenance decision-making system and decision-making method can realize real-time collection of bridge state data by multiple sensors, efficiently fuse and process multi-source data, improve the accuracy and reliability of data, and provide a solid support for bridge condition assessment and disease prediction. The system can accurately capture bridge diseases, timely discover potential problems, effectively reduce the probability of safety accidents, and ensure the smoothness and safety of transportation. Based on advanced data analysis and evaluation models, combined with the construction of high-dimensional data space, the construction of neighborhood graphs, and the calculation of geodesic distances, the complex relationships of data can be analyzed more accurately, and the accurate assessment of bridge technical conditions and the reliable prediction of disease development can be realized. Using the method combining particle swarm optimization simulated annealing hybrid algorithm and analytic hierarchy process grey clustering analysis, the optimal maintenance decision-making combination can be searched under the consideration of constraints such as maintenance cost and time, the maintenance plan can be evaluated and optimized, the utilization efficiency of maintenance resources can be improved, and the maintenance cost can be reduced. The maintenance strategy can be adjusted in real time according to the dynamic changes of the bridge state, enhancing the adaptability and pertinence of decision-making, ensuring that the bridge is always in good use condition, and extending the service life of the bridge. Brief Description of the Drawings

[0068] The following further describes the present invention in conjunction with the drawings and embodiments:

[0069] Figure 1 is the flow chart of the maintenance decision-making system of the present invention;

[0070] Figure 2 is the block diagram of the data collection module of the present invention;

[0071] Figure 3 is the block diagram of the data processing and storage module of the present invention;

[0072] Figure 4 is the block diagram of the bridge condition assessment module of the present invention;

[0073] Figure 5 is the block diagram of the maintenance decision-making module of the present invention;

[0074] Figure 6 is the block diagram of the maintenance implementation and management module of the present invention;

[0075] Figure 7 is the block diagram of the maintenance monitoring and feedback module of the present invention. Detailed implementation manners

[0076] Embodiment 1

[0077] As Figures 1-7 shown, a smart maintenance decision-making system for highway bridges, the decision-making system includes a data acquisition module, a data processing and storage module, a bridge condition assessment module, a maintenance decision-making module, a maintenance implementation and management module, and a maintenance monitoring and feedback module;

[0078] The bridge condition assessment module includes a technical condition module of the bridge and a module for predicting the development trend of diseases;

[0079] The maintenance implementation and management module includes a BIM-based three-dimensional model module of the bridge;

[0080] The monitoring and feedback module is set on the bridge. The monitoring and feedback module includes fiber Bragg grating sensors, vibration sensors, and high-definition cameras. The data monitored by the fiber Bragg grating sensors, vibration sensors, and high-definition cameras is transmitted to the data processing and storage module. The data in the data processing and storage module is input into the bridge condition assessment module to evaluate the technical condition of the bridge and predict the development trend of diseases.

[0081] The data acquisition module collects bridge-related information through various means, including using fiber Bragg grating sensors, vibration sensors, and high-definition cameras to monitor the bridge status in real time, and using unmanned aerial vehicle detection equipment and manual inspections to obtain more comprehensive data. At the same time, data such as bridge conditions, environment, traffic, and historical maintenance records are collected from the highway intranet and the Internet.

[0082] The data processing and storage module processes the collected data. The DBSCAN algorithm is used to clean the data to remove noise and outliers. The one-hot encoding method is used to convert categorical data into numerical form for subsequent processing. Then, a distributed database is constructed based on blockchain technology to store the data, ensuring the security, immutability, and data backup and synchronization between different nodes.

[0083] The bridge condition assessment module includes a technical condition module and a disease development trend prediction module. By receiving data from the data processing and storage module, the Isomap algorithm is used to assess the technical condition of the bridge, predict the development of diseases, and then determine the health level of the bridge and the priority of maintenance needs.

[0084] The maintenance decision-making module generates a variety of maintenance plans based on the bridge condition assessment results using a particle swarm optimization simulated annealing hybrid algorithm, and evaluates and optimizes the plans through a combination of analytic hierarchy process and grey clustering analysis, ultimately selecting the optimal maintenance plan.

[0085] The maintenance implementation and management module uses BIM to build a three-dimensional bridge model module, organizes construction according to construction specifications and safety standards, and effectively manages construction progress and quality.

[0086] The maintenance monitoring and feedback module is installed on the bridge. Its fiber grating sensors, vibration sensors and high-definition cameras collect data in real time and transmit it to the data processing and storage module. The bridge condition assessment module then regularly evaluates the maintenance results and adjusts the maintenance strategy based on the feedback results, forming a complete intelligent maintenance closed-loop management system.

[0087] In the preferred solution, the acquisition module includes fiber grating sensors, vibration sensors, high-definition cameras, drone detection equipment, and manual inspection methods;

[0088] The collection module collects status data, environmental data, traffic data and historical maintenance record data from the highway intranet and the Internet.

[0089] The data collection module of the highway bridge intelligent maintenance decision-making system uses a variety of methods and equipment to collect multiple types of data. The objects collected include bridge condition data, environmental data, traffic data and historical maintenance record data on the highway intranet and the Internet.

[0090] In terms of specific data collection methods, fiber grating sensors are installed at key stress-bearing parts of the bridge, such as the mid-span, around the support, and at the connection between the pier and the beam. In the mid-span and around the support, they are installed by applying special glue to ensure that the sensor fits tightly to the bridge surface and can effectively monitor the strain and temperature of the structure; in the connection between the pier and the beam, embedded installation is adopted, first chiseling out a groove to place the sensor, and then sealing it with concrete slurry. This method can ensure that the sensor works stably in complex stress areas and obtains accurate data.

[0091] High-definition cameras and drone detection devices give full play to their respective advantages to obtain bridge appearance images, providing an intuitive basis for evaluating the bridge appearance condition and helping to detect surface diseases such as cracks and spalling. As an important supplementary means, manual inspection can record the detailed location and form of diseases with the experience and careful observation of professional personnel, and capture some subtle diseases that are difficult to directly detect by instruments.

[0092] Vibration sensors are specifically used to analyze the vibration information when vehicles pass over the bridge. This information is crucial for evaluating the dynamic response and stability of the bridge structure. By deeply analyzing the vibration information, the working state of the bridge under different traffic loads can be understood, and potential structural problems can be detected in a timely manner. The cooperation of multiple acquisition methods provides comprehensive and accurate data support for subsequent bridge condition assessment, maintenance decision-making, etc.

[0093] In the preferred solution, the data processing and storage module uses the DBSCAN algorithm to clean the collected data, encodes the categorical data using the one-hot encoding method, and constructs a distributed database based on blockchain technology to store the processed data.

[0094] The data processing and storage module undertakes the key tasks of data processing and preservation in the intelligent maintenance decision-making system for highway bridges. It first uses the DBSCAN algorithm to clean the collected data.

[0095] The DBSCAN algorithm is a density-based spatial clustering algorithm that can automatically identify and remove noise points and outliers in the data, effectively improving the quality and reliability of the data, and ensuring that subsequent analysis is based on a more accurate data basis. Then, the one-hot encoding method is used to process the categorical data.

[0096] For categorical data such as bridge types and disease types, one-hot encoding maps each category to a unique binary vector, converting non-numerical categorical data into a numerical form suitable for computer processing, which is convenient for subsequent data analysis and model training. Finally, a distributed database is constructed based on blockchain technology to store the processed data. Blockchain technology has characteristics such as decentralization, immutability, and traceability, making the data more secure and reliable during storage and transmission. The distributed storage method ensures that the data is backed up on multiple nodes to prevent data loss, and the data synchronization between different nodes can also ensure data consistency, providing stable data support for the entire intelligent maintenance decision-making system.

[0097] In the preferred solution, both the technical condition module and the module for predicting the development trend of diseases in the bridge condition assessment module use the Isomap algorithm to evaluate the technical condition of the bridge.

[0098] In the bridge condition assessment module, both the technical condition module and the predicted disease development trend module use the Isomap algorithm to evaluate the technical condition of the bridge. The Isomap algorithm, namely the isometric mapping algorithm, can construct a high-dimensional data space based on the bridge structure characteristics and multi-source monitoring data. By calculating the geodesic distance between data points, it maps the high-dimensional data to a low-dimensional space, thereby effectively extracting the inherent change characteristics of the bridge structure.

[0099] Using this algorithm, the current technical condition of the bridge can be accurately evaluated, and the performance of the bridge structure can be observed from multi-dimensional data such as strain, displacement, and cracks. It can also predict the development trend of diseases, analyze the future development direction of diseases, and thus provide key basis for determining the health level of the bridge and the priority of maintenance needs, helping to formulate scientific and reasonable maintenance decisions in the follow-up.

[0100] Example 2

[0101] Further described in combination with Example 1, as Figures 1-7 shown, the method includes: S1. Using fiber Bragg grating sensors, vibration sensors, high-definition cameras, unmanned aerial vehicle detection equipment, and manual inspection methods to collect various data of the bridge, and at the same time collecting bridge condition data, environmental data, traffic data, and historical maintenance record data through the highway intranet and the Internet;

[0102] S2. The data processing and storage module uses the DBSCAN algorithm to clean the data collected in step S1, encodes the categorical data using the one-hot encoding method, and stores the processed data in a distributed database based on blockchain technology;

[0103] S3. The bridge condition assessment module extracts the data stored in the database, uses the Isomap algorithm to evaluate the technical condition of the bridge, predicts the development of diseases, and determines the health level of the bridge and the priority of maintenance needs;

[0104] S4. According to the technical condition obtained in step S3, evaluate and predict the disease development data, use the particle swarm optimization simulated annealing hybrid algorithm to generate multiple maintenance plans, and use the method combining analytic hierarchy process and grey clustering analysis to evaluate and optimize the plans, and select the optimal maintenance plan;

[0105] S5. According to the selected maintenance plan, use BIM technology to organize construction in accordance with construction specifications and safety standards, and manage the construction progress and quality;

[0106] S6. The data collected by the fiber Bragg grating sensors, vibration sensors, and high-definition cameras in the monitoring and feedback module are used to monitor the bridge state in real time;

[0107] The monitored data is then transmitted to the bridge condition assessment module in S3, and the maintenance effectiveness is re-evaluated regularly and the results are fed back to adjust the maintenance strategy.

[0108] In the preferred solution, in step S1, the fiber Bragg grating sensors are installed at the mid-span of the bridge, around the bearings, and at the connection between the pier and the beam.

[0109] Special glue is applied to the installation positions at the mid-span of the bridge and around the bearings, and the fiber Bragg grating sensors are pasted on the bridge surface.

[0110] For the embedded installation at the connection between the pier and the beam, a groove is chiseled at the connection between the pier and the beam, the fiber Bragg grating sensor is placed into the groove, and the groove is sealed with concrete slurry.

[0111] The fiber Bragg grating sensors monitor the strain and temperature of the bridge structure.

[0112] The high-definition cameras and UAV detection equipment acquire the appearance images of the bridge.

[0113] The manual inspection records the detailed positions and forms of the diseases.

[0114] In the preferred solution, the BIM technology manages the construction progress and quality by constructing a three-dimensional model of the bridge and associating the construction progress information.

[0115] Embodiment 3

[0116] Further described in combination with Embodiment 2, the vibration sensors analyze the vehicle-passing vibration information of the bridge. The specific steps include:

[0117] A1. First, the time-frequency analysis algorithm is used to reveal the frequency characteristics of the signal at different times, including using the short-time Fourier transform or wavelet transform to expand the signal in the time and frequency dimensions to reveal the frequency characteristics of the signal at different times. Utilizing the multi-resolution analysis characteristic of the wavelet transform, the window size is automatically adjusted according to the signal frequency to capture the transient changes in the bridge vibration signal.

[0118] For the vehicle-passing vibration signal of the bridge collected by the vibration sensors denoising processing is performed to remove the possible DC component and high-frequency noise. The moving average filtering method can be adopted. Let the filtering window size be , then the filtered signal is:

[0119] ;

[0120] Here, the wavelet transform is selected for time-frequency analysis. The wavelet transform analyzes the signal through the dilation and translation of a basic wavelet function For the signal , its continuous wavelet transform Defined as:

[0121] ;

[0122] Where is the scale parameter, controlling the dilation of the wavelet function; is the translation parameter, controlling the translation of the wavelet function; is the complex conjugate of

[0123] Utilize the multi - resolution analysis characteristic of wavelet transform to automatically adjust the window size according to the signal frequency. As the scale increases, the wavelet function becomes wider, and the resolution for low - frequency signals improves; as the scale decreases, the wavelet function becomes narrower, and the resolution for high - frequency signals improves. By analyzing the wavelet coefficients at different scales and positions, reveal the frequency characteristics of the signal at different moments, and capture the transient changes in the bridge vibration signal.

[0124] A2. Then, use the wavelet packet decomposition algorithm to achieve fine decomposition of the signal, obtain sub - signals in different frequency bands, and by analyzing the energy distribution and amplitude change characteristics of the sub - signals, compare the characteristic differences of the sub - signals in normal and abnormal states, thereby judging whether there are damages or abnormalities in the bridge structure;

[0125] Perform wavelet packet decomposition on the signal after time - frequency analysis processing. Wavelet packet decomposition is based on wavelet decomposition and further decomposes the high - frequency sub - bands, decomposing the signal into finer frequency bands. Let the wavelet packet coefficient of the signal at the th layer and the th node be , then the recurrence formula for wavelet packet decomposition is:

[0126] ;

[0127] ;

[0128] Where and are the low - pass and high - pass filter coefficients respectively.

[0129] Calculate the energy and amplitude of each sub - signal. The energy of the sub - signal is defined as:

[0130] ;

[0131] The amplitude of the sub - signal can take the maximum value of its absolute value, that is .

[0132] Establish a database of the energy and amplitude characteristics of each sub-signal under the normal state of the bridge. During actual monitoring, compare the energy and amplitude of the current sub-signals with the normal characteristics in the database, and calculate the characteristic difference index and :

[0133] ;

[0134] ;

[0135] where and are respectively the energy and amplitude of the sub-signal of the th layer and the rd node under the normal state. When or exceeds the set threshold, it is judged that there may be damage or abnormality in the bridge structure.

[0136] A3. Then, use the empirical mode decomposition algorithm to adaptively process non-linear and non-stationary signals. Decompose the bridge vibration signal into multiple intrinsic mode functions by using the empirical mode decomposition algorithm, extract the main characteristic components related to the bridge structure vibration, remove the noise and interference components, calculate the energy and frequency parameters of the intrinsic mode function components, and observe their changing trends over time, so as to evaluate the stability and health status of the bridge structure;

[0137] Perform empirical mode decomposition on the signal after wavelet packet decomposition, and decompose the signal into multiple intrinsic mode functions (IMFs) and a residual component , that is, . The specific steps of EMD are as follows:

[0138] Determine all the local maximum and minimum points of the signal , and use cubic spline curves to fit the maximum and minimum points respectively to obtain the upper envelope and the lower envelope .

[0139] Calculate the average value of the upper and lower envelopes .

[0140] Subtract the average value of the envelope from the signal to obtain a new signal .

[0141] Judge whether meets the conditions of the IMF (that is, the number of extreme points and zero-crossing points of the signal is equal or at most differs by one, and the upper and lower envelopes of the signal are locally symmetric about the time axis). If it does not meet the conditions, then Repeat the above steps for the new signal until the condition is met to obtain the first IMF .

[0142] Subtract from the original signal to obtain the residual signal . Take as the new signal and repeat the above steps to successively obtain until the residual component becomes a monotonic function or a constant.

[0143] Analyze the correlation between each IMF component and the vibration of the bridge structure, and select the main IMF components related to the vibration of the bridge structure . The correlation can be judged by calculating the correlation coefficient between the IMF component and the original signal :

[0144] ;

[0145] where and are the means of and respectively. Select the IMF component with a larger correlation coefficient as the main feature component.

[0146] Calculate the energy and frequency of the main IMF component. The energy calculation formula is , and the frequency can be obtained through Fourier transform. Observe the changing trends of the energy and frequency parameters over time. When abnormal changes occur in the energy or frequency, it indicates that the stability and health condition of the bridge structure may have changed.

[0147] A4. Finally, use the support vector machine algorithm to identify abnormal vibration patterns. With the support vector machine algorithm for abnormal vibration pattern recognition, extract the amplitude, frequency, and energy characteristic parameters from the vibration signal as the input of the support vector machine. Use a large amount of bridge vibration data under normal and abnormal states to train the support vector machine so that it learns the characteristic differences between normal and abnormal vibration patterns. During actual monitoring, input the characteristics of the real-time collected vibration signal into the trained support vector machine model to judge whether the current vibration state is normal and the type and severity of the abnormality.

[0148] Extract characteristic parameters such as amplitude, frequency, and energy from the main IMF components after empirical mode decomposition processing to form a feature vector , where is the amplitude, is the frequency, is the energy.

[0149] Collect a large number of bridge vibration signals in normal and abnormal states, extract characteristic parameters according to the above method, and construct a training data set , where is the feature vector of the -th sample, is the label of the sample, indicating an abnormal state, indicating a normal state.

[0150] Select a suitable kernel function (such as the radial basis function ), and use the training data set to train the support vector machine. The goal of the support vector machine is to find an optimal hyperplane such that samples of different classes can be separated to the greatest extent. The training process can be achieved by solving the following optimization problem:

[0151] ;

[0152] ;

[0153] where is the penalty factor, which controls the balance between classification error and margin size; is the slack variable, which allows some samples to violate the classification boundary.

[0154] During actual monitoring, extract the characteristic parameters of the real-time collected vibration signal according to the above steps to obtain the feature vector , input it into the trained support vector machine model, and judge whether the current vibration state is normal according to the output of the model. If , it is judged as an abnormal state, and further judge the type and severity of the abnormality according to the decision boundary of the model and the position of the feature vector.

[0155] Example 4

[0156] Combined with Example 2 for further illustration, the steps of the bridge condition assessment method using the Isomap algorithm are as follows:

[0157] S31. Extract bridge data from the blockchain-based distributed database, including bridge structure strain, displacement, crack size, traffic data, environmental parameters, and past maintenance records. Process the missing data using the multiple imputation method, and detect outliers using the local outlier factor algorithm;

[0158] Extract data such as bridge structure strain, displacement, crack size, traffic data, environmental parameters, and past maintenance records from a blockchain-based distributed database. These data are scattered and stored on different nodes, and the consensus mechanism of the blockchain ensures the consistency and integrity of the data. Use database query statements to filter out the required data according to conditions such as the time range of the data and the bridge number.

[0159] For data columns with missing values, the multiple imputation method is used for processing. First, use a regression model to predict the missing values. For each missing value, establish a regression equation based on other relevant variables. Suppose we want to fill in the missing value of variable and select variables with relatively high correlation with to establish a linear regression model , and estimate the regression coefficients by the least squares method. Then, repeat this process multiple times to generate multiple imputed datasets, and finally combine and process these datasets to reduce the imputation error.

[0160] Calculate the local outlier factor for each data point. For each point in the dataset, first determine its set of nearest neighbor points , calculate the average distance from point to its nearest neighbor points , define the local reachability density of point , where is the Euclidean distance from point to point . Finally, calculate the local outlier factor of point . When is greater than the set threshold, the point is considered an outlier and removed from the dataset.

[0161] S32. Apply the ReliefF algorithm for feature selection, and use the selected features as dimensions to construct a high-dimensional dataset, providing a comprehensive information basis for bridge technical condition assessment and disease development prediction and capturing complex relationships;

[0162] Initialize the weight of each feature , where represents the feature. For each sample in the dataset, find its nearest neighbor sample from the same class (the nearest hit sample), and find nearest neighbor samples from different classes (the nearest miss samples). For each feature , update its weight :

[0163] ;

[0164] where is the sample and on the feature , is the number of samples in the dataset. Repeat the above process multiple times, and finally sort according to the feature weights and select the features with higher weights.

[0165] Use the filtered features as dimensions to construct a high-dimensional dataset. Each sample corresponds to a point in the high-dimensional space, and its coordinates are determined by the values of the selected features.

[0166] S33. Use the Mahalanobis distance to measure the distance between data points, use the kd-tree to determine the k nearest neighbor points of each data point, construct a neighborhood graph, reflect the local similarity of data points and provide a structural basis for subsequent calculations;

[0167] For any two data points and in the high-dimensional dataset, calculate their Mahalanobis distance , where is the covariance matrix of the dataset. The Mahalanobis distance takes into account the correlation and scale of the data and can measure the distance between data points more accurately.

[0168] Use the kd-tree data structure to efficiently determine the nearest neighbor points of each data point. The kd-tree is a binary search tree that reduces the time complexity of the nearest neighbor search by recursively dividing the data space into two half-spaces.

[0169] According to the nearest neighbor points of each data point, construct a neighborhood graph. If the data point is one of the nearest neighbor points of the data point , then add an edge from to in the neighborhood graph, and the weight of the edge is their Mahalanobis distance.

[0170] S34. Use the BellmanFord algorithm to calculate the shortest path between any two points in the neighborhood graph as the geodesic distance, reflect the true structural relationship of the bridge samples, be used for low-dimensional embedding and feature extraction, and provide key information for evaluation and prediction;

[0171] Application of the BellmanFord algorithm: For any two nodes and , use the BellmanFord algorithm to calculate the shortest path between them. Initialize the distance estimates of all nodes to infinity, where is the starting node, . Then perform iterations. For each edge in the neighborhood graph, if , then update , where is the weight of the edge . Finally, check if there is a negative-weight cycle. If there is, the algorithm fails; otherwise is the shortest path length from to .

[0172] Repeat the above process to calculate the shortest path between any two points in the neighborhood graph and construct the geodesic distance matrix , where represents the geodesic distance between data points and .

[0173] S35. Adopt the kernel Isomap method, introduce the Laplacian kernel function, perform eigenvalue decomposition after centering the kernel matrix, and project the data into a low-dimensional space; use the spectral clustering algorithm to cluster the bridge samples in the low-dimensional space, and assign a technical condition score to the bridge samples according to the clustering results and relevant factors;

[0174] Adopt the kernel Isomap method: Introduce the Laplacian kernel function to calculate the kernel matrix , where is the bandwidth parameter of the kernel function. Center the kernel matrix to obtain . Perform eigenvalue decomposition on the centered kernel matrix to obtain the eigenvalues and eigenvectors . Select the eigenvectors corresponding to the first largest eigenvalues, project the data into a low-dimensional space, and the projected coordinates are .

[0175] In the low-dimensional space, construct the similarity matrix , where . Calculate the degree matrix , where . Construct the Laplacian matrix . Perform eigenvalue decomposition on the Laplacian matrix and select the first The eigenvectors corresponding to the smallest eigenvalues form a matrix . For each row of the matrix , perform normalization to obtain the matrix . Use the mean algorithm to cluster the rows of the matrix , and divide the bridge samples into classes.

[0176] According to factors such as the clustering results, bridge structure characteristics, and disease history, assign a technical condition score to each bridge sample. For example, for samples clustered into the category with better health conditions, assign a higher score; for samples clustered into the category with more serious diseases, assign a lower score.

[0177] S36. Adopt a long short-term memory network, use the low-dimensional space feature sequence as the input, and train the network to predict the disease development trend based on historical data;

[0178] Construct an LSTM network, including an input layer, an LSTM layer, and an output layer. The input layer receives the low-dimensional space feature sequence. The LSTM layer contains multiple LSTM units, which are used to process sequence data and capture long-term dependencies. The output layer outputs the predicted value of the disease development.

[0179] Use the low-dimensional space feature sequence as the input and the corresponding disease development data as the output to train the LSTM network. Use the backpropagation algorithm to update the weight parameters of the network and minimize the loss function between the predicted value and the true value, such as the mean square error loss function , where is the true value, is the predicted value.

[0180] Use the trained LSTM network, input the current low-dimensional space feature sequence, and obtain the predicted result of the disease development.

[0181] S37. Use the fuzzy comprehensive evaluation method to divide the bridge health level according to the technical condition score and the predicted result of the disease development. Combine the bridge importance factor and use the data envelopment analysis method to determine the maintenance priority.

[0182] Determine the evaluation factor set , such as the technical condition score, disease development trend, etc.; determine the comment set , such as healthy, sub-healthy, diseased, severely diseased, etc. Establish a fuzzy relation matrix , where represents the membership degree of the th evaluation factor to the th comment. Determine the weight vector of each evaluation factor, and through the fuzzy transformation Obtain the comprehensive evaluation result vector , and determine the health level of the bridge according to the principle of maximum membership degree.

[0183] Considering the importance factors of the bridge, such as traffic flow, economic impact, etc., construct decision-making units (DMUs). Each DMU includes input indicators (such as maintenance cost, maintenance time, etc.) and output indicators (such as improvement of bridge health level, slowdown of disease development, etc.). Use the DEA model to calculate the efficiency value of each DMU. The higher the efficiency value, the higher the output of the DMU under a certain input. Sort the bridges according to the efficiency value to determine the maintenance priority.

[0184] Capture complex relationships: The technical condition and disease development of the bridge are comprehensively affected by various factors, and there are complex non-linear relationships among these factors. The high-dimensional data space can integrate these factors together, so that these complex relationships can be captured in this space. When conducting evaluation and prediction, the model constructed based on such a high-dimensional space can better simulate the actual situation and improve the accuracy of evaluation and prediction.

[0185] Relationship between neighborhood graph construction and bridge technical condition evaluation and disease development prediction: Reflect local similarity: Neighborhood graph construction is based on the distance relationship between data points, and constructs a graph structure by determining the k nearest neighbor points of each data point. In bridge condition evaluation, the neighborhood graph can reflect the local similarity between data points. If two bridge samples are in a near-neighbor relationship in the neighborhood graph, it means that they are similar in multiple features, and then their technical conditions and disease development trends may also be similar. This helps to refer to the situation of similar samples during evaluation and prediction, and improve the reliability of evaluation and prediction.

[0186] Provide a structural basis for subsequent calculations: The structure of the neighborhood graph provides a basis for subsequent geodesic distance calculations, and the geodesic distance plays a key role in evaluation and prediction. At the same time, the topological structure of the neighborhood graph can also be used as a kind of feature information and applied to the evaluation and prediction model to further improve the performance of the model.

[0187] Relationship between geodesic distance calculation and bridge technical condition assessment and disease development prediction: Reflecting the true structural relationship: Geodesic distance takes into account the global structure of data points in high-dimensional space and is obtained by calculating the shortest path length between any two points in the neighborhood graph. In bridge condition assessment, geodesic distance can more accurately reflect the true structural relationship between bridge samples. Compared with simple Euclidean distance, it can better reflect the similarities and differences of bridges under complex environments and multiple factors. When evaluating the technical condition of bridges, bridge samples can be classified according to geodesic distance, and samples with similar distances may be in similar health states. In disease development prediction, based on geodesic distance, the association of disease development between different bridge samples can be better analyzed, so as to more accurately predict the spread and development trend of diseases.

[0188] For low-dimensional embedding and feature extraction: Geodesic distance is the key input for low-dimensional embedding. By preserving the geodesic distance between high-dimensional data points in the low-dimensional space, the intrinsic features of the data can be extracted. These low-dimensional features can more effectively represent the technical condition and disease development information of bridges, providing more representative and discriminative inputs for subsequent evaluation and prediction models, thereby improving the accuracy of evaluation and prediction.

[0189] Example 5

[0190] Further illustrated in conjunction with Example 2, the various maintenance methods generated based on the particle swarm optimization simulated annealing hybrid algorithm in step S4 are as follows:

[0191] S41. Obtain bridge technical condition assessment and disease development prediction data from step S3, clarify the maintenance objectives, determine the decision variables, where the decision variables include the type of maintenance measures, the time nodes of maintenance, the allocation amount of maintenance resources, and set the upper limit of maintenance cost and the upper limit of maintenance time as constraint conditions;

[0192] Set the particle swarm size , randomly initialize the position vector and velocity vector of each particle, where is the number of decision variables;

[0193] Define the fitness function as , where , , are weight coefficients and , represents the maintenance effect, represents the maintenance cost, represents the maintenance time;

[0194] In each iteration, calculate the fitness value of each particle , update the historical optimal position of the particle itself and the global optimal position of the entire particle swarm , through the formula and update the particle velocity and position, where is the inertia weight, and are learning factors, and are random numbers uniformly distributed in the interval [0, 1], represents the current iteration number;

[0195] If , accept the new position; otherwise accept the new position with probability , where is the current temperature, and cool down according to , is the cooling coefficient;

[0196] After each update of the particle position, check whether the constraint conditions are met. If not, correct the particle position to make it return to the feasible solution space;

[0197] When the preset maximum number of iterations or the fitness value changes less than the set threshold in several consecutive iterations , stop the iteration and record the relatively optimal particle position to generate multiple maintenance plans;

[0198] The steps of evaluating and optimizing using the method combining the analytic hierarchy process and grey clustering analysis include:

[0199] S42. Construct a hierarchical structure model, including selecting the optimal maintenance plan, maintenance effect, cost-benefit, implementation difficulty, environmental impact, and the multiple maintenance plans generated above; construct a judgment matrix and assign values; calculate the maximum eigenvalue of the judgment matrix and its corresponding eigenvector , normalize to obtain the weight vector, and at the same time calculate the consistency index and the random consistency ratio . When , it is considered that the judgment matrix has satisfactory consistency, otherwise readjust;

[0200] Among them, represents the order of the judgment matrix, and RI represents the random consistency index;

[0201] Parameter : It represents the order of the judgment matrix, that is, the number of rows or columns of the elements in the judgment matrix. When constructing the judgment matrix of the analytic hierarchy process, its size depends on the number of elements in the criterion layer or the alternative layer. For example, when evaluating bridge maintenance plans from four criteria: maintenance effect, cost - benefit, implementation difficulty, and environmental impact, the constructed judgment matrix is of this order, and at this time . It plays a key role in the calculation of the consistency index CI. It affects the value of CI, and further affects the judgment of the consistency of the judgment matrix. CI is used to measure the degree of deviation of the judgment matrix from consistency. The smaller the CI value, the better the consistency of the judgment matrix.

[0202] Parameter RI: It represents the random - consistency index, which is a reference index introduced to test the consistency of the judgment matrix. It is the average consistency index calculated from a large number of randomly generated judgment matrices. Different - order judgment matrices correspond to different $RI$ values, and these values are obtained through a large number of previous experiments and statistical analyses. For example, when , $RI$ is approximately 0.58; , RI is approximately 0.90. When calculating the random - consistency ratio CR, $RI$ is used as the denominator, and the calculated consistency index CI is divided by it to obtain the CR value. When , it is considered that the judgment matrix has satisfactory consistency; if , it indicates that the consistency of the judgment matrix is poor, and the element values in the judgment matrix need to be readjusted to improve its consistency to ensure that the weight results obtained by the analytic hierarchy process are reliable and reasonable.

[0203] The comprehensive weight of each alternative is obtained by weighted - summing the weights of each criterion in the criterion layer and the weights of each alternative in the corresponding criterion in the alternative layer;

[0204] Grey - clustering analysis is used to cluster - evaluate the maintenance plans: The clustering evaluation of the maintenance plans includes the improvement rate of maintenance effect, the cost - reduction rate, the construction - period shortening rate, and the whitenization weight function;

[0205] For each maintenance plan and each clustering index , according to the index value , calculate its clustering coefficient belonging to each grey class , where is the number of clustering indexes, is the weight of index ; According to the clustering coefficients, each maintenance plan is classified into the corresponding grey class, and combined with the comprehensive weight obtained by the analytic hierarchy process, the maintenance plans are sorted;

[0206] Determine the optimal solution by combining expert experience with actual situations: Further evaluate the top-ranked solutions by combining expert experience and actual engineering situations, and finally select the optimal bridge maintenance plan.

[0207] The calculation steps of the particle swarm optimization simulated annealing hybrid algorithm are as follows:

[0208] 1. Initialize the particle swarm and set the particle swarm size , and randomly initialize the position vector of each particle and the velocity vector .

[0209] 2. Define the fitness function , and determine the weight coefficients , , according to the actual requirements of bridge maintenance.

[0210] 3. In each iteration, calculate the fitness value of each particle, and update the historical optimal position of the particle itself and the global optimal position of the entire particle swarm .

[0211] 4. Update the particle velocity and position according to the formula, and use the simulated annealing algorithm to accept worse solutions with a certain probability to avoid falling into local optima.

[0212] 5. After each update of the particle position, check whether the constraint conditions are satisfied (such as the upper limit of maintenance cost , the upper limit of maintenance time ), and if not, correct the particle position to make it return to the feasible solution space.

[0213] 6. When the preset maximum number of iterations is reached or the fitness value changes less than the set threshold in several consecutive iterations, stop the iteration and record the positions of the better particles to generate multiple maintenance plans.

[0214] Construct a hierarchical structure model for scheme evaluation and decision-making

[0215] 1. Construct a hierarchical structure model, construct a judgment matrix and assign values.

[0216] 2. Calculate the maximum eigenvalue of the judgment matrix and its corresponding eigenvector , normalize to obtain the weight vector, and at the same time calculate the consistency index CI and the random consistency ratio CR to ensure that the judgment matrix has satisfactory consistency.

[0217] 3. The weight of each criterion at the criterion level and the weight of each solution at the solution level under the corresponding criterion are weighted and summed to obtain the comprehensive weight of each solution.

[0218] 4. Use grey cluster analysis to cluster and evaluate the maintenance schemes, calculate the clustering coefficient of each maintenance scheme belonging to each grey class, and classify each maintenance scheme into the corresponding grey class according to the clustering coefficient.

[0219] 5. The maintenance plans are ranked according to the comprehensive weights obtained by the hierarchical analysis method. The top-ranked plans are further evaluated based on expert experience and actual engineering conditions, and the optimal bridge maintenance plan is finally selected.

[0220] The implementation cases are as follows:

[0221] After the evaluation in the previous step S3, it was found that the bridge had a certain degree of damage, including structural cracks, concrete spalling, etc., and it was predicted that the damage would develop further. In order to ensure the safe use of the bridge, a scientific and reasonable maintenance plan needs to be formulated.

[0222] Generate maintenance plan based on particle swarm optimization and simulated annealing hybrid algorithm:

[0223] 1. Data acquisition and target determination

[0224] From step S3, the technical condition assessment data of the bridge (such as crack width, depth, structural strain, etc.) and disease development prediction data (such as crack expansion speed in the next year, etc.) are obtained. The maintenance goal is to maximize the maintenance effect of the bridge and extend the service life of the bridge under the premise of meeting cost and time constraints. The decision variables are determined as follows:

[0225] Types of maintenance measures, including crack repair, structural reinforcement, etc.;

[0226] Maintenance time nodes, including the month in which maintenance should be started;

[0227] The amount of maintenance resources allocated, including material usage, man-hours, etc.;

[0228] Also set a maintenance cost cap Ten thousand yuan, upper limit of maintenance time Months.

[0229] 2. Particle swarm initialization

[0230] Setting the particle swarm size , the number of decision variables (Corresponding to the three decision variables mentioned above). Randomly initialize the position vector of each particle and the velocity vector For example, particles The position vector of may be expressed as The velocity vector represents the rate of change of these decision variables.

[0231] 3. Define the fitness function

[0232] Define the fitness function and determine the weight coefficients according to the actual situation , , . Among them, is calculated through a professional bridge evaluation model, which comprehensively considers factors such as the improvement degree of bridge diseases and the enhancement of structural performance; is the maintenance cost, which is calculated according to different types of maintenance measures and the amount of resource allocation; is the maintenance time, which is determined by the complexity of the maintenance measures and the resource input situation.

[0233] 4. Iterative calculation

[0234] In each iteration: Calculate the fitness value of each particle .

[0235] Update the historical optimal position of the particle itself and the global optimal position of the entire particle swarm .

[0236] Update the particle velocity and position through the formulas and . Set the inertia weight , the learning factors , and are random numbers uniformly distributed in the interval $[0, 1]$, represents the current iteration number.

[0237] If , accept the new position; otherwise accept the new position with probability . The initial temperature , the temperature reduction coefficient , and cool down according to .

[0238] After each update of the particle position, check whether the constraint conditions are satisfied (the maintenance cost does not exceed , the maintenance time does not exceed ). If not, correct the particle position to make it return to the feasible solution space. For example, if the maintenance cost corresponding to a certain particle exceeds , then reduce the amount of resource allocation.

[0239] 5. Stop the iteration

[0240] When the preset maximum number of iterations $T_{maxIter}=200$ is reached or the fitness value changes less than the set threshold $\epsilon = 0.01$ in 10 consecutive iterations, stop the iteration. Record the positions of the better particles and generate multiple maintenance plans accordingly, such as:

[0241] Plan 1: Adopt crack repair and local reinforcement measures, start maintenance in the 3rd month, invest 60 tons of materials, with an estimated cost of 1.8 million yuan, and the maintenance time is 5 months.

[0242] Plan 2: Conduct comprehensive structural reinforcement, start maintenance in the 1st month, invest 80 tons of materials, with an estimated cost of 2 million yuan, and the maintenance time is 6 months.

[0243] Use a method combining the analytic hierarchy process and grey clustering analysis to evaluate and optimize the plans

[0244] 1. Construct a hierarchical structure model

[0245] Construct a hierarchical structure model. The target layer is to select the optimal maintenance plan, the criterion layer includes maintenance effect, cost - benefit, implementation difficulty, and environmental impact, and the plan layer is the multiple maintenance plans (Plan 1, Plan 2) generated above.

[0246] 2. Construct judgment matrices and calculate weights

[0247] Construct judgment matrices And assign values. For example, for the judgment matrix of the criterion layer:

[0248] ;

[0249] Calculate the maximum eigenvalue of the judgment matrix And its corresponding eigenvector , and normalize it to obtain the weight vector. Calculate the consistency index , the random consistency ratio , for , , then , and it is considered that the judgment matrix has satisfactory consistency.

[0250] For the plan layer, judgment matrices are also constructed and weights are calculated respectively under each criterion. For example, under the maintenance effect criterion:

[0251] ;

[0252] The calculated weights of Plan 1 and Plan 2 under the maintenance effect criterion are respectively .

[0253] The comprehensive weight of each solution is obtained by weighted summation of the weights of each criterion in the criterion layer and the weights of each solution in the solution layer under the corresponding criterion.

[0254] 3. Grey clustering analysis

[0255] Select the improvement rate of maintenance effect, cost reduction rate, and construction period shortening rate as clustering indicators, and determine the weights of each indicator , , . Determine the whitenization weight function according to the actual data.

[0256] For each maintenance solution and each clustering indicator , calculate its clustering coefficient belonging to each grey class . . Assume that it is divided into three grey classes (excellent, good, medium), calculate the clustering coefficients of Solution 1 and Solution 2, and classify them into the corresponding grey classes.

[0257] The clustering coefficient of Solution 1 is , , , then Solution 1 belongs to the excellent class; the clustering coefficient of Solution 2 is , , , then Solution 2 belongs to the good class.

[0258] Rank the maintenance solutions by combining the comprehensive weights obtained by the analytic hierarchy process. Solution 1 has a higher comprehensive score and ranks ahead.

[0259] 4. Determine the optimal solution

[0260] Combined with expert experience and the actual engineering situation, the experts believe that Solution 1 can ensure the maintenance effect while having reasonable cost and time control, relatively low implementation difficulty, and less environmental impact. Finally, Solution 1 is selected as the optimal bridge maintenance solution.

[0261] Through the above embodiments, it shows how to use the particle swarm optimization simulated annealing hybrid algorithm to generate multiple bridge maintenance solutions, and use the method combining the analytic hierarchy process and grey clustering analysis to evaluate and optimize the solutions, so as to determine the optimal maintenance solution.

[0262] Embodiment 6

[0263] Combined with Embodiment 2 for further illustration, the solutions in S6 include: S61. The data collected by fiber Bragg grating sensors, vibration sensors, and high-definition cameras are transmitted to the data processing center in real time using wireless transmission technology and adopting data encryption and retransmission mechanisms, and stored in a distributed time series database; through the sensor error correction formula Perform error correction on the sensor data, where is the original measurement value, is the true value, and are the coefficients obtained by least squares fitting;

[0264] The data preprocessing adopts the linear interpolation formula to fill in the missing data values, where is the time point where the missing value is located, and are the adjacent time points before and after it, and are the corresponding measurement values;

[0265] Detect and remove outliers using the method based on local outlier factor; perform normalization processing on different types of sensor data using min-max normalization;

[0266] S62. Deliver the preprocessed data to the bridge condition assessment module in S3, and evaluate the current state of the bridge based on the Isomap algorithm model and the relevant evaluation index system; calculate the change value of the bridge health condition score and the change rate of crack propagation speed ;

[0267] where is the current score, is the score before maintenance, where is the current bridge damage expansion speed, is the bridge damage expansion speed before maintenance;

[0268] Construct a comprehensive evaluation index for maintenance effectiveness where , , are the corresponding weights and , is the actual maintenance cost, is the budgeted maintenance cost;

[0269] S63. Compare and judge according to the comprehensive evaluation index for maintenance effectiveness with the set high threshold and the low threshold ;

[0270] If , maintain the current maintenance strategy;

[0271] If , re-analyze the causes and development trends of bridge diseases, use the particle swarm optimization simulated annealing hybrid algorithm to generate a new maintenance plan, and adopt the method combining analytic hierarchy process and grey clustering analysis to evaluate and optimize the new plan and select the optimal strategy;

[0272] If , make appropriate fine-tuning to the implementation frequency of maintenance measures and the allocation ratio of maintenance resources.

[0273] Specific embodiment: Evaluation of bridge maintenance effectiveness and strategy adjustment

[0274] A cross-river bridge has adopted an advanced monitoring and maintenance management system, and installed devices such as fiber Bragg grating sensors, vibration sensors, and high-definition cameras to monitor the bridge status in real time. After a period of maintenance, it is necessary to evaluate the maintenance effectiveness and adjust the maintenance strategy according to the evaluation results.

[0275] Step S61: Data collection, transmission, processing and preprocessing

[0276] 1. Data collection and transmission

[0277] The fiber Bragg grating sensors, vibration sensors, and high-definition cameras respectively collect data on the structural strain, vehicle-passing vibration information, and appearance condition of the bridge. Using wireless transmission technology, the collected data is transmitted to the data processing center in real time. To ensure the security and reliability of data transmission, data encryption and retransmission mechanisms are adopted. For example, when packet loss occurs during data transmission, the system will automatically retransmit the lost data. The collected data is finally stored in a distributed time series database for subsequent analysis and processing.

[0278] 2. Sensor error correction

[0279] For the original data collected by the sensors, use the sensor error correction formula to correct the errors. First, obtain the coefficients and by least squares fitting. Assume that the fiber Bragg grating sensor measures strain data, collect a certain number of original measurement values and the corresponding true value samples (obtained through high-precision calibration equipment), and calculate , using the least squares method. If the original measurement value of the fiber Bragg grating sensor at a certain moment, then the corrected true value .

[0280] 3. Data preprocessing

[0281] When data is missing, use the linear interpolation formula to fill it. The vibration sensor is at Data missing at a moment, and the adjacent time points before and after it The corresponding measured values , The corresponding measured values , then The interpolation result at the moment is .

[0282] Use the method based on local outlier factor to detect and remove outliers. Calculate the local outlier factor of each data point. When the local outlier factor exceeds the set threshold, determine that the data point is an outlier and remove it from the dataset.

[0283] Adopt the maximum - minimum normalization to normalize the data of different types of sensors, and map the data to the interval $[0, 1]$. For the strain data of fiber Bragg grating sensors, let its maximum value be , the minimum value be , and the value of a certain data point be , then the normalized value is .

[0284] Step S62: Calculation of comprehensive evaluation index for bridge condition assessment and maintenance effectiveness

[0285] 1. Bridge condition assessment

[0286] Send the pre - processed data to the bridge condition assessment module in S3, and evaluate the current state of the bridge based on the Isomap algorithm model and related evaluation index system. Assume that the current health condition score of the bridge obtained through this evaluation model is , and the score before maintenance is , then the change value of the bridge health condition score .

[0287] Meanwhile, measure the current crack propagation speed of the bridge mm / month, and the crack propagation speed of the bridge before maintenance mm / month, then the change rate of the crack propagation speed .

[0288] 2. Calculation of comprehensive evaluation index for maintenance effectiveness

[0289] Set the weights , , , the actual maintenance cost ten thousand yuan, and the budgeted maintenance cost ten thousand yuan. Then the comprehensive evaluation index for maintenance effectiveness .

[0290] Step S63: Adjustment of maintenance strategy

[0291] 1. Threshold Setting

[0292] Set the high threshold , and the low threshold .

[0293] 2. Strategy Judgment and Adjustment

[0294] Since , it meets , so the implementation frequency of maintenance measures and the allocation ratio of maintenance resources are appropriately fine-tuned. For example, the original monthly regular inspection is adjusted to once every half month, and at the same time, the allocation ratio of material resources for crack repair is increased from to , and the allocation ratio of resources for structural reinforcement is decreased from to .

[0295] The above embodiments are only the preferred technical solutions of the present invention and should not be regarded as limitations on the present invention. The protection scope of the present invention should be the technical solutions recorded in the claims, including equivalent replacement solutions of the method features in the technical solutions recorded in the claims. That is, equivalent replacement improvements within this scope are also within the protection scope of the present invention.

Claims

1. A highway bridge intelligent maintenance decision-making system, characterized by: The decision-making system includes a data acquisition module, a data processing and storage module, a bridge condition assessment module, a maintenance decision-making module, a maintenance implementation and management module, and a maintenance monitoring and feedback module; The bridge condition assessment module includes the bridge technical condition module and the disease development trend prediction module; The technical condition module of the bridge in the bridge condition assessment module uses the Isomap algorithm to assess the technical condition of the bridge; The module for predicting the development trend of diseases uses the Isomap algorithm to predict the development of bridge diseases; The maintenance decision-making module generates multiple maintenance plans based on the bridge condition assessment results using a particle swarm optimization simulated annealing hybrid algorithm, and evaluates and optimizes the plans using a combination of analytic hierarchy process and grey clustering analysis to ultimately select the optimal maintenance plan. The maintenance implementation and management module includes a BIM bridge 3D model module; The maintenance monitoring and feedback module is set on the bridge. The maintenance monitoring and feedback module includes a fiber grating sensor, a vibration sensor and a high-definition camera. The data monitored by the fiber grating sensor, the vibration sensor and the high-definition camera are transmitted to the data processing and storage module. The data in the data processing and storage module is input into the bridge condition assessment module to assess the technical condition of the bridge and predict the development trend of the disease. The steps of the bridge condition assessment method using the Isomap algorithm are as follows: S31. Extract bridge data from a distributed database based on blockchain, including bridge structure strain, displacement, crack size, traffic data, environmental parameters, and past maintenance records. Use multiple filling methods to process missing data and use the local outlier factor algorithm to detect outliers. S32. Use the ReliefF algorithm for feature selection and use the selected features as dimensions to construct a high-dimensional data set, providing a comprehensive information basis for bridge technical condition assessment and disease development prediction and capturing complex relationships; S33, using Mahalanobis distance to measure the distance between data points, using kd tree to determine the k nearest neighbor points of each data point, constructing a neighborhood graph, reflecting the local similarity of data points and providing a structural basis for subsequent calculations; S34. Use BellmanFord algorithm to calculate the shortest path between any two points in the neighborhood graph as the geodesic distance, which reflects the real structural relationship of the bridge samples and is used for low-dimensional embedding and feature extraction, providing key information for evaluation and prediction; S35. Using the kernel Isomap method, the Laplace kernel function is introduced to perform eigendecomposition after centering the kernel matrix, and the data is projected into a low-dimensional space. The spectral clustering algorithm is used to cluster the bridge samples in the low-dimensional space, and the technical condition scores are assigned to the bridge samples based on the clustering results and related factors. S36, using long short-term memory network, taking low-dimensional spatial feature sequence as input, and predicting the disease development trend based on historical data by training the network; S37. Use fuzzy comprehensive evaluation method to divide bridge health level according to technical condition score and disease development prediction results. Combined with bridge importance factors, use data envelopment analysis method to determine maintenance priority.

2. According to claim 1, a highway bridge intelligent maintenance decision-making system is characterized by: The acquisition modules include fiber grating sensors, vibration sensors, high-definition cameras, drone detection equipment, and manual inspection methods; The collection module collects status data, environmental data, traffic data and historical maintenance record data from the highway intranet and the Internet.

3. The intelligent highway bridge maintenance decision-making system according to claim 1 is characterized by: The data processing and storage module uses the DBSCAN algorithm to clean the collected data, uses the one-hot encoding method to encode the classified data, and builds a distributed database based on blockchain technology to store the processed data.

4. A decision-making method for a highway bridge intelligent maintenance decision-making system, applied to the system according to any one of claims 1 to 3, characterized in that: The method includes: S1. Collect various data of bridges using fiber grating sensors, vibration sensors, high-definition cameras, drone detection equipment and manual inspections. At the same time, collect bridge condition data, environmental data, traffic data and historical maintenance record data through the highway intranet and the Internet; S2, the data processing and storage module uses the DBSCAN algorithm to clean the data collected in step S1, encodes the classified data using the one-hot encoding method, and stores the processed data in a distributed database based on blockchain technology; S3, the bridge condition assessment module extracts the data stored in the database, uses the Isomap algorithm to assess the technical condition of the bridge, predicts the development of the disease, and determines the health level and maintenance priority of the bridge; S4. Evaluate the technical conditions obtained in step S3. The bridge condition assessment module predicts the disease development data. Then, the particle swarm optimization simulated annealing hybrid algorithm in the maintenance decision-making module is used to generate multiple maintenance plans. The hierarchical analysis method combined with the grey clustering analysis method is used to evaluate and optimize the plans and select the optimal maintenance plan. S5. Based on the selected maintenance plan, the maintenance implementation and management module uses BIM technology to organize construction in accordance with construction specifications and safety standards, and manage construction progress and quality; BIM technology manages construction progress and quality by building a three-dimensional model of the bridge and associating it with construction progress information; S6, the data collected by the fiber grating sensor, vibration sensor and high-definition camera in the maintenance monitoring and feedback module monitors the bridge status in real time; The monitoring data is then transmitted to the bridge condition assessment module in S3, which regularly re-evaluates the maintenance effectiveness and provides feedback to adjust the maintenance strategy.

5. According to claim 4, a decision-making method for a highway bridge intelligent maintenance decision-making system is characterized by: step In S1, FBG sensors are installed in the mid-span of the bridge, around the support, and at the connection between the pier and the beam; Apply special glue to the installation position in the middle of the bridge span and around the support, and stick the fiber grating sensor on the bridge surface; The connection between the pier and the beam is embedded, a groove is chiseled out at the connection between the pier and the beam, the fiber grating sensor is placed in the groove, and the groove is sealed with concrete slurry; Fiber Bragg grating sensors monitor the strain and temperature of bridge structures; High-definition cameras and drone inspection equipment obtain images of the bridge's exterior; Manual inspections were conducted to record the detailed location and morphology of the disease.

6. The decision-making method of the highway bridge intelligent maintenance decision-making system according to claim 4 is characterized by: The vibration sensor analyzes the vibration information of vehicles passing through the bridge. The specific steps include: A1. First, use the time-frequency analysis algorithm to reveal the frequency characteristics of the signal at different times, including using short-time Fourier transform or wavelet transform to expand the signal in two dimensions of time and frequency to reveal the frequency characteristics of the signal at different times. Use the multi-resolution analysis characteristics of wavelet transform to automatically adjust the window size according to the signal frequency to capture the transient changes in the bridge vibration signal; A2. Then use the wavelet packet decomposition algorithm to achieve fine signal decomposition to obtain sub-signals in different frequency bands. By analyzing the energy distribution and amplitude change characteristics of the sub-signals, and comparing the characteristic differences of the sub-signals under normal and abnormal conditions, it is possible to determine whether the bridge structure is damaged or abnormal. A3. Use the empirical mode decomposition algorithm to adaptively process nonlinear and non-stationary signals. Use the empirical mode decomposition algorithm to decompose the bridge vibration signal into multiple intrinsic mode functions, extract the main characteristic components related to the vibration of the bridge structure, remove noise and interference components, calculate the energy and frequency parameters of the intrinsic mode function components and observe their changing trends over time, and then evaluate the stability and health of the bridge structure; A4. Finally, the support vector machine algorithm is used to identify abnormal vibration patterns. The support vector machine algorithm is used to identify abnormal vibration patterns. The amplitude, frequency, and energy characteristic parameters are extracted from the vibration signal as the input of the support vector machine. A large amount of bridge vibration data under normal and abnormal conditions is used to train the support vector machine so that it can learn the characteristic differences between normal and abnormal vibration patterns. During actual monitoring, the vibration signal features collected in real time are input into the trained support vector machine model to determine whether the current vibration state is normal and the type and severity of the abnormality.

7. The decision-making method of the highway bridge intelligent maintenance decision-making system according to claim 4 is characterized by: In step S4, multiple maintenance methods are generated based on the particle swarm optimization simulated annealing hybrid algorithm: S41. Obtain the bridge technical condition assessment and disease development prediction data from step S3, clarify the maintenance objectives, determine the decision variables, including the type of maintenance measures, maintenance time nodes, maintenance resource allocation, and set the upper limit of the maintenance cost. , Maintenance time limit Constraints; Define the fitness function as ,in , , is the weight coefficient and , Indicates the maintenance effect. represents the maintenance cost, Indicates maintenance time; In each iteration, the fitness value of each particle is calculated , update the particle's own historical optimal position and the global optimal position of the entire particle swarm , initialize the position vector of each particle and the velocity vector , through the formula and Update particle velocity and position, where is the inertia weight, and is the learning factor, and is a random number uniformly distributed in the interval [0, 1]. Indicates the current iteration number; like , accept the new position; otherwise, with probability Accept new position where For the current temperature, press Cooling, is the temperature reduction coefficient; After each update of the particle position, check whether the constraint conditions are met. If not, correct the particle position to return to the feasible solution space; When the preset maximum number of iterations is reached Or the fitness value changes less than the set threshold in several consecutive iterations When , the iteration is stopped, and the optimal particle position is recorded to generate multiple maintenance plans; The evaluation and optimization steps using the combined method of analytic hierarchy process and grey clustering analysis include: S42. Construct a hierarchical model, including selecting the optimal maintenance plan, maintenance effect, cost-effectiveness, implementation difficulty, environmental impact and the multiple maintenance plans generated above; construct a judgment matrix And assign values; calculate the maximum eigenvalue of the judgment matrix and its corresponding eigenvector , normalize to get the weight vector, and calculate the consistency index and random consistency ratio ,when When the judgment matrix is ​​considered to have satisfactory consistency, otherwise it will be readjusted; in, represents the order of the judgment matrix, RI represents the random consistency index; The weight of each criterion at the criterion level and the weight of each solution at the solution level under the corresponding criterion are weighted and summed to obtain the comprehensive weight of each solution; Grey cluster analysis is used to cluster the maintenance schemes for evaluation: the cluster evaluation of the maintenance schemes includes the improvement rate of maintenance effect, the rate of cost reduction, the rate of shortening the construction period and the whitening weight function; For each maintenance plan And each clustering index , according to the indicator value Calculate its belonging to each gray class The clustering coefficient ,in is the number of clustering indicators, For indicators weight; divide each maintenance plan into corresponding gray categories according to the clustering coefficient, and rank the maintenance plans based on the comprehensive weight obtained by the hierarchical analysis method; Determine the best solution by combining expert experience with actual conditions: Further evaluate the top-ranked solutions based on expert experience and actual project conditions, and ultimately select the best bridge maintenance solution.

8. The decision-making method of the highway bridge intelligent maintenance decision-making system according to claim 4 is characterized by: The solutions in S6 include: The data collected by S61, fiber grating sensor, vibration sensor and high-definition camera are transmitted to the data processing center in real time by wireless transmission technology and data encryption and retransmission mechanism, and stored in the distributed time series database; the sensor error correction formula is used to correct the data. Perform error correction on sensor data, where is the original measurement value, is the true value, and are the coefficients obtained by least squares fitting; Data preprocessing uses linear interpolation formula Fill in missing data values, where is the time point where the missing value is located, and For the adjacent time points before and after it, and is the corresponding measured value; The method based on local outlier factor is used to detect and remove outliers; the maximum and minimum normalization is used to normalize the sensor data of different types; S62: The pre-processed data is transmitted to the bridge condition assessment module in S3, and the current state of the bridge is assessed based on the Isomap algorithm model and the relevant assessment index system; the change value of the bridge health status score is calculated. and crack growth rate ; in is the current rating, is the pre-maintenance score, where is the current bridge damage expansion rate, is the bridge damage expansion rate before maintenance; Constructing comprehensive evaluation indicators for maintenance effectiveness ,in , , is the corresponding weight and , is the actual maintenance cost, is the budgeted maintenance cost; S63. Comprehensive evaluation index based on maintenance effectiveness With the high threshold set and low threshold Make comparative judgments; like , maintain the current maintenance strategy; like , re-analyze the causes and development trends of bridge diseases, use the particle swarm optimization simulated annealing hybrid algorithm to generate new maintenance plans, and use the method of combining hierarchical analysis and grey clustering analysis to evaluate and optimize the new plans to select the optimal strategy; like , and make appropriate fine-tuning to the frequency of implementation of maintenance measures and the proportion of maintenance resource allocation.

Images