Intelligent maintenance decision-making method based on bridge big data

Abstract

The invention relates to the technical field of bridge maintenance, in particular to an intelligent maintenance decision-making method based on bridge big data, and the method comprises the following steps: collecting structure response data, environment data and traffic load data of a bridge through a sensor array, an inspection robot and a manual input system which are arranged at each key part of the bridge; the structure response data, the environment data and the traffic load data are subjected to cleaning, alignment and space-time fusion, a standardized bridge state big data set is formed, the bridge state big data set comprises historical and real-time data of the bridge, comprehensive and real-time state monitoring of the bridge is achieved, potential problems can be found in time, and the safety of the bridge is improved. The method is advantaged in that bridge performance degradation in the future can be predicted, preventive maintenance is realized, multiple optimized maintenance schemes can be generated, scientific decision is supported, bridge maintenance precision and efficiency are improved, bridge service life prolonging is facilitated, and full life cycle cost reduction is facilitated.

Description

Technical Field

[0001] This invention relates to the field of bridge maintenance technology, specifically to an intelligent decision-making method for bridge maintenance based on big data. Background Technology

[0002] As a core component of transportation infrastructure, the structural health and safe operation of bridges are of paramount importance. Traditional bridge maintenance and management rely heavily on regular manual inspections and limited static data, which presents several problems. First, manual inspections are highly subjective and inefficient, making it difficult to detect hidden damage within the structure. Second, existing decision-making methods are mostly based on experience thresholds, failing to deeply integrate and effectively mine massive, multi-source, and heterogeneous bridge monitoring big data, such as strain, deflection, vibration, crack images, environmental temperature and humidity, and traffic loads, thus hindering the transformation from "post-event maintenance" to "pre-event early warning" and "precise intervention."

[0003] However, with the development of IoT, big data and AI technologies, bridge health monitoring systems can now collect TB or even PB level data. However, in the field of bridge maintenance, there is still a lack of a holistic method that can systematically integrate this data and generate intelligent diagnosis, prediction and decision-making based on the integration results. Most existing decision-making technologies only stay at the level of data visualization and simple over-limit alarms, and fail to transform data into deep knowledge that can directly support maintenance actions and is actionable. Summary of the Invention

[0004] To overcome the shortcomings of existing technologies, this invention proposes an intelligent decision-making method for bridge maintenance based on big data. This method can promptly identify potential problems in bridges, predict future performance degradation, enable preventative maintenance, and generate multiple optimized maintenance plans to support scientific decision-making. This improves the accuracy and efficiency of bridge maintenance, helps extend the service life of bridges, and reduces the total life cycle cost.

[0005] The technical solution adopted by this invention to solve its technical problem is: a bridge maintenance intelligent decision-making method based on big data, comprising the following steps:

[0006] S1: By deploying sensor arrays, inspection robots, and manual data entry systems at key parts of the bridge, structural response data, environmental data, and traffic load data of the bridge are collected. The structural response data, environmental data, and traffic load data are then cleaned, aligned, and spatiotemporally fused to form a standardized bridge status big data set, which includes historical and real-time data of the bridge.

[0007] S2: Based on the standardized large dataset obtained in step S1, the current comprehensive health index of the bridge is calculated using a pre-trained deep learning evaluation model.

[0008] S3: Based on the historical and real-time data obtained in step S1, use the time series prediction algorithm to predict the degradation trajectory of key performance indicators of the bridge within a specific time window in the future, and obtain the performance degradation prediction results.

[0009] S4: Based on the comprehensive health index obtained in step S2 and the performance degradation prediction results obtained in step S3, combined with the pre-set maintenance knowledge base and decision rule base, the optimal maintenance strategy suggestion set is generated through a multi-objective optimization algorithm;

[0010] S5: Feed the implemented maintenance strategy and subsequent bridge condition change data back to the deep learning evaluation model in step S2 and the time series prediction algorithm in step S3 to achieve continuous optimization and iterative updates of the deep learning evaluation model.

[0011] Preferably, the formula for calculating the comprehensive health index in step S2 is as follows:

[0012] HHI=∑(W i ×NN(FFT(X i )))+b

[0013] Among them, X i Let be the standardized monitoring data vector of the i-th class, such as the strain, frequency, and deflection of a bridge; FFT is the Fast Fourier Transform, used to convert the time-domain signal to the frequency domain to extract features; NN() is a trained neural network model used to map frequency-domain features to local health scores; W i is the weight coefficient corresponding to the i-th type of data, and b is the model bias term.

[0014] Preferably, the weighting coefficient W i Determined through the following optimization formula:

[0015] min∑(HHI e -(∑W i ×S i )+b) 2 +λ∑|W i |

[0016] Among them, HHI e S is the benchmark value for bridge health as determined manually. i This represents the data of class i processed by NN(FFT(X)). i The calculated local score, where λ is the regularization coefficient.

[0017] Preferably, the time series prediction algorithm in step S3 is a spatiotemporal graph convolutional network, whose formula includes:

[0018] Graph convolution operations:

[0019] Where H(I) is the input feature of the I-th layer, and H(I+1) is the output feature. W(I) is a normalized bridge structure graph adjacency matrix, representing the spatial topological relationship between different sensor measurement points. W(I) is a trainable parameter weight matrix, and σ is a nonlinear activation function.

[0020] Preferably, the objective function of the multi-objective optimization algorithm described in step S4 is:

[0021] minF = [C(M), -E(M), R(M)]

[0022] HHI p ≥HHI d T(M)≤T{max}

[0023] Where M represents a candidate maintenance strategy, C(M) represents the total cost target of the strategy, E(M) represents the expected improvement target of the strategy, R(M) represents the traffic disruption target during the implementation of the strategy, and HHI p The predicted health level after the strategy is implemented must be greater than the safety threshold HHI. d T(M) represents the time required for the strategy, which must be less than the maximum allowed time T(max).

[0024] Preferably, a non-dominated sorting genetic algorithm is used to solve the objective function of the multi-objective optimization algorithm to obtain a set of Pareto optimal solutions, which serves as the optimal maintenance strategy suggestion set.

[0025] Preferably, the data cleaning in step S1 includes outlier detection and removal of the collected raw data based on the isolated forest algorithm.

[0026] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, performs the steps of the method described above.

[0027] A computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the method as described above.

[0028] The beneficial effects of this invention are as follows:

[0029] This invention enables comprehensive and real-time condition monitoring of bridges, timely detection of potential problems, prediction of future performance degradation, and preventative maintenance. It also generates multiple optimized maintenance plans to support scientific decision-making, improves the accuracy and efficiency of bridge maintenance, helps extend the service life of bridges, and reduces the total life cycle cost. Attached Figure Description

[0030] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0031] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0032] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0033] like Figure 1 As shown, in one embodiment of the present invention, a smart maintenance decision-making method based on bridge big data is provided, including the following steps:

[0034] S1: By deploying sensor arrays, inspection robots, and manual data entry systems at key locations on the bridge, structural response data, environmental data, and traffic load data of the bridge are collected. The structural response data, environmental data, and traffic load data are then cleaned, aligned, and spatiotemporally fused to form a standardized bridge status dataset. The bridge status dataset includes historical and real-time data of the bridge. The data cleaning in step S1 includes outlier detection and removal of the collected raw data based on the isolated forest algorithm.

[0035] Specifically, this invention deploys sensor arrays at key bridge components, including main beams, cables, and piers. These sensors, including strain sensors, acceleration sensors, displacement sensors, and temperature and humidity sensors, collect structural response data, environmental data, and traffic load data. Structural response data includes the bridge's strain, vibration frequency, deflection, and crack width. Environmental data includes the bridge's temperature, humidity, wind speed, and corrosive medium concentration. Traffic load data includes traffic flow, vehicle type, and axle load. Simultaneously, an inspection robot periodically collects high-definition images and crack data of the bridge's underside and piers. Furthermore, a manual data entry terminal is provided for inputting daily inspection records, historical maintenance records, and other manually collected data, facilitating real-time understanding of the bridge's health status. An isolated forest algorithm is used to detect and remove outliers from the collected raw data, resulting in a large-scale bridge status dataset. This dataset includes both historical and real-time bridge status data, forming the basis for intelligent bridge maintenance decisions.

[0036] S2: Based on the standardized large dataset obtained in step S1, the current comprehensive health index of the bridge is calculated using a pre-trained deep learning evaluation model.

[0037] S3: Based on the historical and real-time data obtained in step S1, use the time series prediction algorithm to predict the degradation trajectory of key performance indicators of the bridge within a specific time window in the future, and obtain the performance degradation prediction results.

[0038] S4: Based on the comprehensive health index obtained in step S2 and the performance degradation prediction results obtained in step S3, combined with the pre-set maintenance knowledge base and decision rule base, the optimal maintenance strategy suggestion set is generated through a multi-objective optimization algorithm;

[0039] S5: Feed the implemented maintenance strategy and subsequent bridge condition change data back to the deep learning evaluation model in step S2 and the time series prediction algorithm in step S3 to achieve continuous optimization and iterative updates of the deep learning evaluation model.

[0040] Specifically, the formula for calculating the comprehensive health index mentioned in step S2 is as follows:

[0041] HHI=∑(W i ×NN(FFT(X i )))+b

[0042] Among them, X i Let be the standardized monitoring data vector of the i-th class, such as the strain, frequency, and deflection of a bridge; FFT is the Fast Fourier Transform, used to convert the time-domain signal to the frequency domain to extract features; NN() is a trained neural network model used to map frequency-domain features to local health scores; W i is the weight coefficient corresponding to the i-th type of data, and b is the model bias term.

[0043] Specifically, the weighting coefficient W i Determined through the following optimization formula:

[0044] min∑(HHI e -(∑W i ×S i )+b) 2 +λ∑|W i |

[0045] Among them, HHI e S is the benchmark value for bridge health as determined manually. i This represents the data of class i processed by NN(FFT(X)). i The calculated local score, where λ is the regularization coefficient.

[0046] Specifically, the time series prediction algorithm mentioned in step S3 is a spatiotemporal graph convolutional network, and its formula includes:

[0047] Graph convolution operations:

[0048] Where H(I) is the input feature of the I-th layer, and H(I+1) is the output feature. W(I) is a normalized bridge structure graph adjacency matrix, representing the spatial topological relationship between different sensor measurement points. W(I) is a trainable parameter weight matrix, and σ is a nonlinear activation function.

[0049] Specifically, the objective function of the multi-objective optimization algorithm described in step S4 is:

[0050] minF = [C(M), -E(M), R(M)]

[0051] HHI p ≥HHI d T(M)≤T{max}

[0052] Where M represents a candidate maintenance strategy, C(M) represents the total cost objective of the strategy, E(M) represents the expected improvement objective of the strategy (therefore, a negative sign is used to minimize it), and R(M) represents the traffic disruption objective (such as road closure time) during the implementation of the strategy. HHI p The predicted health level after the strategy is implemented must be greater than the safety threshold HHI. d T(M) represents the time required for the strategy, which must be less than the maximum allowed time T(max).

[0053] Specifically, a non-dominated sorting genetic algorithm is used to solve the objective function of the multi-objective optimization algorithm to obtain a set of Pareto optimal solutions, which serves as the optimal maintenance strategy suggestion set.

[0054] More specifically, this invention uses Fast Fourier Transform to convert time-domain signals to frequency domain to extract features. Then, a trained neural network model maps the frequency domain features to local health scores, and finally, a weighted sum is obtained to obtain the HHI. This effectively captures the dynamic characteristics of bridge structures and improves the accuracy of health assessment. By minimizing the error between the HHI and the human assessment benchmark, the weight coefficients of various data types are determined, which can balance the importance of different types of data and improve the reliability of comprehensive assessment.

[0055] Furthermore, a spatiotemporal graph convolutional network is introduced for time series prediction. This network can simultaneously consider the spatial topological relationships and time series characteristics of the bridge structure, better capturing the mutual influence between different parts of the bridge and improving prediction accuracy. Finally, through multi-objective optimization, considering multiple objectives such as cost, effectiveness improvement, and traffic interference of maintenance strategies, and setting health and time constraints, a set of Pareto optimal solutions is obtained using a non-dominated sorting genetic algorithm as the optimal maintenance strategy suggestion. This achieves a balance among multiple objectives and provides decision-makers with multiple bridge maintenance options. Therefore, this invention realizes the entire process from data collection, analysis, prediction to decision generation, improving the efficiency and accuracy of bridge maintenance management.

[0056] For example, suppose a large bridge that has been in use for 10 years needs maintenance decisions. In the current technology, the approach is for engineers to conduct manual inspections every quarter, mainly relying on visual inspections and simple instrument measurements. After discovering problems, maintenance work is arranged based on experience and fixed maintenance cycles. For example, if cracks are found in the bridge deck, repairs are scheduled for the next quarter. If the strain of the main beam exceeds a preset threshold, reinforcement is planned during the annual overhaul. This method is lagging, makes it difficult to discover potential problems in a timely manner, and cannot formulate the optimal maintenance plan based on the actual condition of the bridge.

[0057] However, by employing this invention, sensor arrays can be deployed at key parts of the bridge, including the main beam, cables, and piers. These sensors include strain sensors, acceleration sensors, displacement sensors, and temperature and humidity sensors, used to collect structural response data, environmental data, and traffic load data. The structural response data includes the bridge's strain, vibration frequency, deflection, and crack width; the environmental data includes the bridge's temperature, humidity, wind speed, and corrosive medium concentration; and the traffic load data includes the bridge's traffic flow, vehicle type, and axle load. Simultaneously, an inspection robot is deployed to periodically collect high-definition images and crack data of the bridge's underside and piers. Furthermore, a manual data entry terminal is provided for inputting daily inspection records, historical maintenance records, and other manually collected data. An outlier detection and removal are performed on the collected raw data using an isolated forest algorithm, thereby obtaining a large dataset of bridge status data, which includes both historical and real-time bridge status data.

[0058] Then, using a pre-trained deep learning model, the fused data was analyzed, and the current comprehensive health index of the bridge was calculated to be 0.82, with a maximum score of 1. Based on historical data and the current state, a spatiotemporal graph convolutional network was used to predict the degradation trend of key performance indicators of the bridge over the next 5 years. The prediction results showed that if no measures were taken, the deflection of the main beam might exceed the safety limit in 3 years. Therefore, the system combined the prediction results, maintenance knowledge base, and decision rules to generate multiple maintenance plans through a multi-objective optimization algorithm: Plan 1: Immediately reinforce the main beam, with an estimated cost of 5 million yuan and a construction period of 2 months, increasing the comprehensive health index to 0.90; Plan 2: Implement in two phases, first repairing the bridge deck, and then reinforcing the main beam one year later, with a total cost of 4.5 million yuan and a total construction period of 3 months, ultimately achieving a comprehensive health index of 0.88; Plan 3: Only perform preventative maintenance, with a cost of 2 million yuan and a construction period of 1 month, slightly increasing the comprehensive health index to 0.85.

[0059] Next, these solutions are presented to decision-makers, along with a multi-dimensional comparative analysis of each solution's cost, effectiveness, traffic impact, and other factors. Decision-makers can choose the most suitable solution based on the actual situation and budget. Meanwhile, after the selected solution is implemented, this invention continues to monitor changes in the bridge's condition and feeds the actual effects back to the evaluation model and prediction algorithm for continuous optimization and updates.

[0060] Therefore, this invention enables comprehensive and real-time monitoring of bridge conditions, timely detection of potential problems, prediction of future bridge performance degradation, and preventative maintenance. Furthermore, it can generate multiple optimized maintenance plans, support scientific decision-making, improve the accuracy and efficiency of bridge maintenance, help extend the service life of bridges, and reduce the total life cycle cost.

[0061] In another embodiment of the present invention, an electronic device is also provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it performs the following steps: collecting structural response data, environmental data, and traffic load data of the bridge through sensor arrays deployed at key parts of the bridge, inspection robots, and a manual data entry system; cleaning, aligning, and spatiotemporally fusing the structural response data, environmental data, and traffic load data to form a standardized bridge state dataset, which includes historical and real-time bridge data; and based on the obtained standardized dataset... Based on the data set, a pre-trained deep learning evaluation model is used to calculate the current comprehensive health index of the bridge. Based on the obtained historical and real-time data, a time series prediction algorithm is used to predict the degradation trajectory of key performance indicators of the bridge within a specific future time window, obtaining performance degradation prediction results. Based on the obtained comprehensive health index and performance degradation prediction results, combined with a pre-set maintenance knowledge base and decision rule base, a set of optimal maintenance strategy suggestions is generated through a multi-objective optimization algorithm. The implemented maintenance strategies and subsequent bridge status change data are fed back to the deep learning evaluation model and the time series prediction algorithm to achieve continuous optimization and iterative updates of the deep learning evaluation model.

[0062] In another embodiment of the present invention, a computer-readable storage medium is also provided, on which a computer program is stored. When executed by a processor, the program performs the following steps: collecting structural response data, environmental data, and traffic load data of the bridge through sensor arrays, inspection robots, and manual data entry systems deployed at key parts of the bridge; cleaning, aligning, and spatiotemporally fusing the structural response data, environmental data, and traffic load data to form a standardized bridge state dataset, which includes historical and real-time data of the bridge; calculating the current comprehensive health index of the bridge based on the obtained standardized dataset using a pre-trained deep learning evaluation model; predicting the degradation trajectory of key performance indicators of the bridge within a specific future time window using a time series prediction algorithm based on the obtained historical and real-time data, obtaining performance degradation prediction results; generating an optimal maintenance strategy suggestion set based on the obtained comprehensive health index and performance degradation prediction results, combined with a pre-set maintenance knowledge base and decision rule base, using a multi-objective optimization algorithm; and feeding back the executed maintenance strategy and subsequent bridge state change data to the deep learning evaluation model and the time series prediction algorithm to achieve continuous optimization and iterative updates of the deep learning evaluation model.

[0063] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A bridge maintenance intelligent decision-making method based on big data, characterized in that: Includes the following steps: S1: By deploying sensor arrays, inspection robots, and manual data entry systems at key parts of the bridge, structural response data, environmental data, and traffic load data of the bridge are collected. The structural response data, environmental data, and traffic load data are then cleaned, aligned, and spatiotemporally fused to form a standardized bridge status big data set, which includes historical and real-time data of the bridge. S2: Based on the standardized large dataset obtained in step S1, the current comprehensive health index of the bridge is calculated using a pre-trained deep learning evaluation model. S3: Based on the historical and real-time data obtained in step S1, use the time series prediction algorithm to predict the degradation trajectory of key performance indicators of the bridge within a specific time window in the future, and obtain the performance degradation prediction results. S4: Based on the comprehensive health index obtained in step S2 and the performance degradation prediction results obtained in step S3, combined with the pre-set maintenance knowledge base and decision rule base, the optimal maintenance strategy suggestion set is generated through a multi-objective optimization algorithm; S5: Feed the implemented maintenance strategy and subsequent bridge condition change data back to the deep learning evaluation model in step S2 and the time series prediction algorithm in step S3 to achieve continuous optimization and iterative updates of the deep learning evaluation model.

2. The intelligent maintenance decision-making method based on bridge big data according to claim 1, characterized in that: The formula for calculating the comprehensive health index mentioned in step S2 is as follows: HHI=∑(W i ×NN(FFT(X i )))+b Among them, X i Let be the standardized monitoring data vector of the i-th class, such as the strain, frequency, and deflection of a bridge; FFT is the Fast Fourier Transform, used to convert the time-domain signal to the frequency domain to extract features; NN() is a trained neural network model used to map frequency-domain features to local health scores; W i is the weight coefficient corresponding to the i-th type of data, and b is the model bias term.

3. The intelligent maintenance decision-making method based on bridge big data according to claim 2, characterized in that: The weighting coefficient W i Determined through the following optimization formula: min∑(HHI e -(∑W i ×S i )+b) 2 +λ∑|W i | Among them, HHI e S is the benchmark value for bridge health as determined manually. i This represents the data of class i processed by NN(FFT(X)). i The calculated local score, where λ is the regularization coefficient.

4. The intelligent maintenance decision-making method based on bridge big data according to claim 1, characterized in that: The time series prediction algorithm described in step S3 is a spatiotemporal graph convolutional network, and its formula includes: Graph convolution operations: Where H(I) is the input feature of the I-th layer, and H(I+1) is the output feature. W(I) is a normalized bridge structure graph adjacency matrix, representing the spatial topological relationship between different sensor measurement points. W(I) is a trainable parameter weight matrix, and σ is a nonlinear activation function.

5. The intelligent maintenance decision-making method based on bridge big data according to claim 1, characterized in that: The objective function of the multi-objective optimization algorithm described in step S4 is: minF = [C(M), -E(M), R(M)] HHI p ≥HHI d ,T(M)≤T{max} Where M represents a candidate maintenance strategy, C(M) represents the total cost target of the strategy, E(M) represents the expected improvement target of the strategy, R(M) represents the traffic disruption target during the implementation of the strategy, and HHI p The predicted health level after the strategy is implemented must be greater than the safety threshold HHI. d T(M) represents the time required for the strategy, which must be less than the maximum allowed time T(max).

6. The intelligent maintenance decision-making method based on bridge big data according to claim 5, characterized in that: A non-dominated sorting genetic algorithm is used to solve the objective function of the multi-objective optimization algorithm to obtain a set of Pareto optimal solutions, which serves as the optimal maintenance strategy suggestion set.

7. The intelligent maintenance decision-making method based on bridge big data according to claim 1, characterized in that: The data cleaning described in step S1 includes outlier detection and removal from the collected raw data based on the isolated forest algorithm.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, it implements the steps of the method as described in any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by the processor, it implements the steps of the method as described in any one of claims 1 to 7.

Images