Bridge structure anti-fatigue performance predictive maintenance method based on digital twinning

Abstract

The invention relates to the technical field of bridge structure health monitoring and maintenance, and discloses a bridge structure anti-fatigue performance predictive maintenance method based on digital twinning, and the method comprises the steps: constructing a bridge structure digital twinning model, and deploying a multi-mode sensing network; establishing a bridge fatigue mechanism knowledge graph, and structuring expert experience and historical cases; multi-source heterogeneous evidence fusion is realized based on a DS evidence theory, and evidence conflicts are eliminated; constructing a cause and effect inference engine based on a Bayesian network, and identifying a fatigue mechanism path; and a self-adaptive predictive maintenance decision-making system is realized, and an optimal maintenance strategy is generated. The problems that a traditional bridge fatigue analysis method is difficult to process multi-source heterogeneous data, the fatigue prediction precision is insufficient and the like are solved, and high-precision prediction and precise maintenance of the anti-fatigue performance of the bridge structure are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge structure health monitoring and maintenance. More specifically, it relates to a predictive maintenance method for the anti-fatigue performance of bridge structures based on digital twins. Background Art

[0002] As an important part of transportation infrastructure, bridge structures are constantly subjected to periodic actions such as vehicle loads, wind loads, and temperature changes during long-term service, resulting in the accumulation of fatigue damage inside the structure, which may ultimately lead to the degradation or even failure of the structural function. According to statistical data, fatigue damage has become one of the main reasons for the failure of large bridge structures, seriously threatening the safety and durability of bridge use.

[0003] The existing bridge fatigue analysis and maintenance mainly adopt the modes of regular inspection, empirical judgment, and passive maintenance. These methods usually rely on single types of evidence or simple correlation analysis. With the development of sensor technology and data analysis technology, some studies have begun to apply methods such as finite element analysis and machine learning for fatigue performance evaluation. However, these methods still have the following problems: multi-source heterogeneous data is difficult to effectively fuse, and multi-dimensional evidence such as sensing data, structural responses, environmental conditions, expert experience, and historical cases cannot be comprehensively considered, resulting in inaccurate identification of fatigue mechanisms; expert knowledge is fragmented, lacking a systematic knowledge organization and utilization mechanism, and it is difficult to support decision-making in complex scenarios; there are often conflicts between different evidence sources, lacking effective conflict resolution methods, reducing the reliability of judgment; fatigue prediction is mostly based on statistical models, making it difficult to reveal the underlying causal mechanisms and resulting in insufficient prediction accuracy; maintenance decision-making lacks a systematic optimization mechanism and mostly relies on artificial experience, leading to high maintenance costs and poor effects.

[0004] Therefore, there is an urgent need to develop a new method that can comprehensively utilize multi-source heterogeneous data, integrate expert experience, accurately predict fatigue performance, and optimize maintenance decision-making, providing reliable guarantee for the safe operation of bridge structures while reducing maintenance costs. Summary of the Invention

[0005] The present invention provides a predictive maintenance method for the anti-fatigue performance of bridge structures based on digital twins, which solves the technical problems in the prior art that bridge fatigue analysis is difficult to effectively process multi-source heterogeneous data, has insufficient fatigue prediction accuracy, and lacks an optimization mechanism for maintenance decision-making.

[0006] The present invention provides a predictive maintenance method for the anti-fatigue performance of bridge structures based on digital twins, including the following steps:

[0007] Construct a digital twin model of the bridge structure and deploy a multi-modal sensing network to achieve real-time data interaction between the physical structure and the virtual model;

[0008] Build a knowledge graph of bridge fatigue mechanism, transform expert experience and historical cases into structured knowledge, and form a fatigue mechanism knowledge base;

[0009] Based on the DS evidence theory, achieve the fusion of multi-source heterogeneous evidence, resolve the conflicts between evidence from different sources, and improve the accuracy of fatigue mechanism identification;

[0010] Construct a causal reasoning engine based on Bayesian network to identify the causal relationship path of fatigue damage development;

[0011] Implement an adaptive predictive maintenance decision-making system, and generate an optimal maintenance strategy based on the fatigue state prediction results.

[0012] Furthermore, the construction of the digital twin model of the bridge structure includes:

[0013] Adopt parametric modeling technology to construct a three-dimensional finite element model of the bridge;

[0014] Deploy a multi-modal sensing network at key parts of the bridge, including strain sensors, accelerometers, and displacement sensors;

[0015] Establish a mapping relationship between sensing data and digital models, and use the Bayesian calibration method to optimize the parameters of the finite element model;

[0016] Construct a bidirectional data channel based on edge computing technology to achieve real-time interaction between the physical space and the virtual space.

[0017] Furthermore, the establishment of the knowledge graph of bridge fatigue mechanism includes:

[0018] Define the core concepts, attributes, and relationships in the field of bridge fatigue based on the ontology method;

[0019] Extract structured knowledge from fatigue research literature, maintenance reports, and expert knowledge through natural language processing technology;

[0020] Construct knowledge reasoning rules to achieve automatic reasoning and expansion of knowledge;

[0021] Establish a case library, and classify and store historical fatigue cases according to failure modes, causes, and performance characteristics.

[0022] Furthermore, in the implementation of multi-source heterogeneous evidence fusion based on the DS evidence theory, the evidence conflict coefficient is calculated by summing the products of the basic probability assignments between all mutually conflicting evidence sources, where mutual conflict means that the intersection of the propositions supported by different evidence sources is an empty set; the larger the value of this coefficient, the more serious the conflict between different evidence sources, and stronger evidence adjustment is required.

[0023] Further, the adjustment of evidence weights in the multi-source heterogeneous evidence fusion based on the DS evidence theory is achieved by calculating the average distance between each evidence source and other evidence sources, and converting this distance value into a weight value, such that the evidence with less difference from other evidence sources obtains a higher weight.

[0024] Further, the evidence fusion in the multi-source heterogeneous evidence fusion based on the DS evidence theory is achieved by calculating the sum of the product of probabilities of different evidence combinations supporting the same proposition and performing normalization processing to ensure that the final probability distribution satisfies the constraint condition that the sum of probabilities is 1.

[0025] Further, the construction of the causal inference engine based on the Bayesian network includes:

[0026] Constructing an initial Bayesian network structure based on the causal relationships in the knowledge graph;

[0027] Combining monitoring data and expert knowledge to learn the conditional probability table of the Bayesian network;

[0028] Constructing a posterior probability inference algorithm based on the MCMC method;

[0029] Creating a counterfactual intervention analysis module to calculate the intervention effects of different maintenance measures;

[0030] Creating a fatigue mechanism path identification algorithm to find the most likely fatigue development path.

[0031] Further, the calculation of the intervention effects of the counterfactual intervention analysis module is performed by weighted averaging the conditional probabilities under all possible covariate value conditions, where the weights are the marginal distribution probabilities of the covariates, so as to evaluate the causal impact of specific intervention measures on the outcome variable.

[0032] Further, the implementation of the adaptive predictive maintenance decision-making system includes:

[0033] Establishing a fatigue damage prediction model based on the LSTM neural network;

[0034] Defining a maintenance decision objective function that comprehensively considers safety reliability, maintenance cost, and structural life;

[0035] Constructing a maintenance strategy optimization algorithm based on reinforcement learning;

[0036] Creating a maintenance effect evaluation model;

[0037] Establishing a maintenance decision visualization interface.

[0038] Further, the predictive maintenance system for the anti-fatigue performance of bridge structures based on digital twin executes the above-mentioned predictive maintenance method for the anti-fatigue performance of bridge structures based on digital twin, including:

[0039] A digital twin model construction module for digital modeling of bridge structures and deployment of multi-modal sensing networks;

[0040] A fatigue knowledge graph construction module for integrating expert experience and historical cases to form a structured fatigue mechanism knowledge base;

[0041] A multi-source evidence fusion module for conflict resolution and fusion of fatigue evidence from different sources based on the DS evidence theory;

[0042] A causal reasoning engine module for identifying potential fatigue mechanism paths by applying Bayesian networks and causal analysis techniques;

[0043] An adaptive predictive maintenance decision-making module for generating the best maintenance strategy through multi-objective optimization techniques.

[0044] The beneficial effects of the present invention are as follows:

[0045] This solution realizes high-precision monitoring and simulation of the real-time state of the bridge by constructing a digital twin model of the bridge structure, providing a data basis for fatigue performance evaluation; establishing a knowledge graph of bridge fatigue mechanisms to realize the reuse of expert knowledge and automated reasoning, providing knowledge support for fatigue mechanism analysis; applying a multi-source heterogeneous evidence fusion mechanism based on the DS evidence theory to solve the evidence conflict problem and collaboratively analyze multi-dimensional evidence; revealing the causal relationship of fatigue damage with the help of a causal reasoning engine based on Bayesian networks, providing a mechanism support for precise maintenance; creating an adaptive predictive maintenance decision-making system to generate the optimal maintenance strategy based on fatigue state prediction, realizing the transformation from passive maintenance to active prevention. In addition, the improved DS evidence theory can dynamically adjust the evidence weight to improve the reliability of evidence fusion; through Bayesian networks and LSTM neural networks, the prediction accuracy of the fatigue state of the bridge structure is improved; the optimization of the maintenance strategy based on reinforcement learning can adaptively adjust the maintenance decision, reducing costs compared with traditional regular maintenance and improving the structural reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a flowchart of the method for predictive maintenance of the anti-fatigue performance of a bridge structure based on digital twin of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] Now, the subject matter described herein will be discussed with reference to exemplary embodiments. It should be understood that discussing these embodiments is only to enable those skilled in the art to better understand and thus implement the subject matter described herein, and the functions and arrangements of the elements discussed can be changed without departing from the protection scope of the content of this specification. Each example can omit, substitute, or add various processes or components as needed. Additionally, the features described in some examples can also be combined in other examples.

[0048] In at least one embodiment of the present invention, a predictive maintenance method for the anti-fatigue performance of bridge structures based on digital twins is disclosed. As Figure 1 shown, it includes the following steps:

[0049] Step 1: Construct a digital twin model of the bridge structure and deploy a multi-modal sensing network to achieve real-time data interaction between the physical structure and the virtual model;

[0050] Step 1.1: Adopt parametric modeling technology to construct a three-dimensional finite element model of the bridge according to the geometric parameters, material properties, and boundary conditions of the bridge structure. This model can accurately reflect the mechanical properties of the bridge structure;

[0051] Step 1.2: Deploy a multi-modal sensing network at key parts of the bridge, including strain sensors, accelerometers, displacement sensors, etc., to achieve real-time monitoring of the dynamic response of the bridge structure;

[0052] Step 1.3: Establish a mapping relationship between the sensing data and the digital model, and use the Bayesian calibration method to optimize the parameters of the finite element model so that the digital model can accurately reflect the dynamic characteristics of the actual structure;

[0053] Step 1.4: Configure data preprocessing algorithms, including outlier detection, missing value imputation, data denoising, etc., to ensure the quality of the acquired monitoring data;

[0054] Step 1.5: Construct a two-way data channel between the digital twin model and the physical bridge to achieve real-time interaction between the physical space and the virtual space. This channel is based on edge computing technology to ensure the real-time and reliability of data transmission.

[0055] Step 2: Establish a knowledge graph of bridge fatigue mechanism, transform expert experience and historical cases into structured knowledge, and form a fatigue mechanism knowledge base;

[0056] Step 2.1: Define the core concepts, attributes, and relationships in the field of bridge fatigue based on the ontology method, and establish a conceptual model of fatigue knowledge, including core elements such as material properties, environmental factors, load types, and fatigue damage modes;

[0057] Step 2.2: Extract structured knowledge from fatigue research literature, maintenance reports, and expert knowledge through natural language processing technology, and transform it into a triple form (subject - relationship - object) for storage;

[0058] Step 2.3: Construct knowledge reasoning rules, based on description logic and first-order predicate logic, to achieve automated reasoning and extension of knowledge, and enhance the expression ability of the knowledge graph;

[0059] Step 2.4, create a knowledge fusion algorithm to resolve the conflicts and overlaps among knowledge from different sources and ensure the consistency and integrity of the knowledge base;

[0060] Step 2.5, establish a case base, classify and store historical fatigue cases according to failure modes, causes, performance characteristics, etc., to form a queryable case knowledge base and support case-based reasoning.

[0061] Step 3, implement multi-source heterogeneous evidence fusion based on the DS evidence theory to resolve the conflicts among evidence from different sources and improve the accuracy of fatigue mechanism identification;

[0062] Step 3.1, establish an identification framework for bridge fatigue identification , where , , respectively represent the , , th possible fatigue mechanisms, represents the th possible fatigue mechanism, is the total number of fatigue mechanisms;

[0063] Step 3.2, assign a basic probability assignment function (BPA) to evidence from different sources (such as sensor data, expert knowledge, historical cases, etc.), denoted as , where represents the th evidence source;

[0064] Step 3.3, construct an evidence conflict detection algorithm to calculate the evidence conflict coefficient :

[0065] ;

[0066] where represents the evidence conflict coefficient, represents the th proposition supported by the first evidence source, represents the th proposition supported by the second evidence source, represents the basic probability assignment of the first evidence source to the proposition , represents that the intersection of the two propositions is an empty set, that is, the two propositions conflict with each other;

[0067] Step 3.4, design an adaptive weight assignment model based on the evidence distance to adjust the weights of conflicting evidence:

[0068] ;

[0069] Among them, represents the weight of the th evidence source, represents the average distance between the th evidence source and other evidence sources, represents the average distance between the th evidence source and other evidence sources, is the total number of evidence sources;

[0070] Step 3.5, apply the improved DS combination rule for evidence fusion to obtain the comprehensive BPA function :

[0071] ;

[0072] Among them, is the comprehensive BPA function. For any non-empty subset represents any non-empty subset in the frame of discernment, is the evidence conflict coefficient, represents that the intersection of two propositions is equal to the proposition , represents the th proposition supported by the first evidence source, represents the th proposition supported by the second evidence source, represents the basic probability assignment of the first evidence source to the proposition , represents the normalization factor to ensure that the final probability assignment satisfies the constraint that the sum of probabilities is 1;

[0073] Taking the fatigue damage identification of the bearing area of a bridge structure as an example, the specific application scenario is as follows:

[0074] Frame of discernment construction: First, according to the typical mechanism of bridge fatigue damage, establish the frame of discernment , among which, represents fatigue cracks, represents bearing corrosion, represents material aging, represents welding defects, represents overload damage;

[0075] Evidence source assignment of BPA: The system obtains data from three different evidence sources, namely: sensor monitoring data (denoted as ), expert evaluation (denoted as ), and historical case matching (denoted as ). Assume that the sensor data shows high-frequency micro-vibration characteristics in the support area. The expert assessment believes that there may be fatigue cracks or welding defects in this area, and historical case matching shows that most similar situations are caused by welding defects. Based on this information, the BPA distribution of each evidence source is as follows:

[0076] ;

[0077] ;

[0078] ;

[0079] Evidence conflict detection: Calculate the evidence conflict coefficient. Taking and as an example, the conflict coefficient between them is:

[0080] ;

[0081] Among them, represents and the conflict coefficient between, represents the empty set;

[0082] In this example, since there are no directly conflicting focal elements between the evidence sources (that is, and the intersection of the focal elements with assigned probabilities is not empty), so . However, in practical applications, if one evidence source indicates that the fault type is , while another indicates that it cannot be , a conflict will occur at this time.

[0083] Weight assignment: Calculate the weights of each evidence source based on the evidence distance. The evidence distance can be calculated by the Jousselme distance formula:

[0084] ;

[0085] Among them, is a similarity matrix related to the frame of discernment, represents the transpose operation.

[0086] Assume that the average distance between each evidence source calculated is: , , , then the weights of each evidence source are:

[0087] ;

[0088] ;

[0089] ;

[0090] Evidence fusion: Finally, fuse each evidence source using the improved DS combination rule:

[0091] ;

[0092] where, represents the BPA parameter after fusion, represents a subset in the frame of discernment;

[0093] For this example, the fused result may be:

[0094] ;

[0095] ;

[0096] ;

[0097] ;

[0098] This indicates that the welding defect ( ) is the most likely cause of fatigue damage. Compared with the judgments of each single evidence source, the fused result is more comprehensive and reliable.

[0099] Step 4, construct a causal inference engine based on the Bayesian network to identify the causal relationship path of fatigue damage development;

[0100] Step 4.1, construct an initial Bayesian network structure based on the causal relationships in the knowledge graph, and define nodes (variables) and edges (dependency relationships);

[0101] Step 4.2, combine the monitoring data and expert knowledge to learn the conditional probability table (CPT) of the Bayesian network, denoted as where, is the node variable, is the set of parent nodes of

[0102] Step 4.3, construct a posterior probability inference algorithm based on the MCMC (Markov Chain Monte Carlo) method to calculate the probability of fatigue mechanism under given observed evidence;

[0103] Step 4.4, create a counterfactual intervention analysis module to calculate the intervention effect through do-calculus:

[0104] ;

[0105] where, represents that under the intervention variable takes the value of When the conditional probability of the result variable is the value of the intervention variable, is the covariate, is the intervention variable, is the result variable,

[0106] Step 4.5: Create a fatigue mechanism path recognition algorithm to find the most likely fatigue development path based on the maximum a posteriori probability (MAP) criterion, providing a mechanistic explanation for predictive maintenance.

[0107] Taking the causal analysis of fatigue damage of bridge girders as an example, the specific application process is as follows:

[0108] Bayesian network structure construction: In the scenario of main girder fatigue analysis, first extract relevant nodes and relationships from the knowledge graph to construct the following Bayesian network structure:

[0109] Root nodes (without parent nodes): Load type (LT), Environmental condition (EC), Material quality (MQ);

[0110] Intermediate nodes: Stress concentration (SC), Corrosion degree (CD), Structural response (SR);

[0111] Leaf node (result node): Fatigue crack (FC);

[0112] The dependencies between these nodes are represented as directed edges. For example: Load type → Stress concentration, Environmental condition → Corrosion degree, Material quality → Stress concentration, Stress concentration → Structural response, Corrosion degree → Structural response, Structural response → Fatigue crack, etc.

[0113] Conditional probability table learning: Use historical data and expert knowledge collected from the digital twin model to calculate the conditional probability table for each node. For example, for the "Structural response" node, its conditional probability table can be expressed as shown in Table 1:

[0114] Table 1: Conditional probability table :

[0115]

[0116] Posterior probability inference: When the digital twin model monitors the observed values of certain variables, the system uses the MCMC method for posterior probability inference. For example, assume that the sensor monitors: Load type = Heavy vehicles passing repeatedly (LT = H), Environmental condition = Humid (EC = W), Structural response = High-frequency vibration (SR = H), and it is necessary to infer the probability of fatigue crack occurrence .

[0117] The MCMC sampling process is as follows:

[0118] Initialize the Markov chain: randomly assign initial values to the unobserved variables;

[0119] Perform Gibbs sampling for each unobserved variable and update its state according to its conditional probability distribution;

[0120] Repeat the sampling step multiple times (e.g., 10,000 times) and collect samples;

[0121] Statistically analyze the state distribution of the fatigue crack variable in the samples to obtain the posterior probability;

[0122] After calculation, the obtained posterior probability may be:

[0123] ,

[0124] indicating that under the given observation conditions, the probability of fatigue crack occurrence is 78%.

[0125] Counterfactual intervention analysis: To evaluate the effectiveness of different maintenance measures, the system conducts counterfactual intervention analysis. For example, assume that in the current state, the system estimates that the probability of fatigue crack occurrence is 78%. Now, we want to evaluate the effect of increasing structural reinforcement (reducing stress concentration). Use do-calculus to calculate:

[0126] ;

[0127] where, is the conditional probability after intervention, and the distribution before intervention remains unchanged. After calculation, it may be obtained that , indicating that by reducing the stress concentration to a low level through reinforcement measures, the probability of fatigue crack occurrence can be reduced from 78% to 35%.

[0128] Fatigue mechanism path identification: Based on the constructed Bayesian network and the calculated conditional probabilities of each node, the system uses the maximum posterior probability criterion to identify the most likely fatigue development path. In the above example, the possible maximum posterior probability path is:

[0129] Heavy vehicles passing repeatedly (LT = H) → High stress concentration (SC = H) → High structural response (SR = H) → Fatigue crack occurrence (FC = Y);

[0130] The joint probability of this path is the highest among all possible paths, providing the most likely mechanistic explanation for the occurrence of fatigue damage and serving as a basis for targeted maintenance decisions.

[0131] Step 5: Implement an adaptive predictive maintenance decision-making system to generate an optimal maintenance strategy based on the fatigue state prediction results;

[0132] Step 5.1: Based on the fused fatigue evidence and causal reasoning results, establish a fatigue damage prediction model, and use deep learning methods (such as LSTM networks) to predict the fatigue damage trend of key parts;

[0133] Step 5.2: Define the maintenance decision objective function, comprehensively considering safety reliability, maintenance cost, and structural life:

[0134] ;

[0135] where, represents the objective function, represents the reliability index, represents the cost index, represents the life index, , , are the corresponding weight coefficients respectively;

[0136] Step 5.3: Construct an optimization algorithm for maintenance strategies based on reinforcement learning, define the state space, action space, and reward function, and learn the optimal maintenance strategy by interacting with the environment;

[0137] Step 5.4: Create a maintenance effect evaluation model, simulate the effects of different maintenance plans based on the digital twin model, and update the maintenance strategy in real time;

[0138] Step 5.5: Establish a visualization interface for maintenance decisions, intuitively display the fatigue status, prediction trend, and recommended maintenance plans, and support engineers in making decisions.

[0139] Taking the fatigue maintenance of the main cable joint area of a large suspension bridge as an example, the specific application process is as follows:

[0140] Construction of the fatigue damage prediction model: Based on the real-time monitoring data and historical data collected by the digital twin model, establish an LSTM neural network model to predict the fatigue damage evolution of the main cable joint area. The LSTM network structure includes:

[0141] Input layer: Receive various sensor data (strain, displacement, temperature, humidity, etc.) and traffic load information;

[0142] LSTM layer 1: Contains 64 neurons, and the activation function is tanh;

[0143] Dropout layer: Randomly discard 20% of the neurons to prevent overfitting;

[0144] LSTM layer 2: Contains 32 neurons, and the activation function is tanh;

[0145] Fully connected layer: Map the output of the LSTM layer to the prediction target;

[0146] Output layer: Predict the fatigue damage indicators at different future time points (1 day, 7 days, 30 days, 90 days).

[0147] For model training, 90% of the historical data is used as the training set, 10% as the validation set. The Adam optimizer is used with a learning rate of 0.001, a batch size of 64, and 500 training rounds. The average absolute error (MAE) of the trained model on the validation set is less than 5%, and it can accurately predict the development trend of fatigue damage.

[0148] Definition of the maintenance decision objective function: According to the requirements of bridge operation and management, define the indicators of the objective function:

[0149] Reliability index : Based on the structural reliability index, in the range [0,1], the larger the value, the safer the structure;

[0150] Cost index : The normalized representation of the maintenance cost, in the range [0,1], the smaller the value, the lower the cost;

[0151] Life index : The normalized representation of the remaining service life, in the range [0,1], the larger the value, the longer the life;

[0152] The weight coefficients of each index are determined according to the importance of the bridge and the management strategy. For example: (Safety is the primary consideration), , .

[0153] Implementation of the maintenance strategy optimization algorithm: Implement the reinforcement learning algorithm based on the Deep Q-Network (DQN), where:

[0154] State space: Includes multi-dimensional state vectors such as fatigue damage degree, damage development rate, remaining structural life, and last maintenance time;

[0155] Action space: Includes different maintenance strategies, such as: no operation, inspection and monitoring, local strengthening, replacement of important components, comprehensive repair, etc.;

[0156] Reward function: Based on the objective function Calculate, that is, the value of the objective function corresponding to the state after performing a certain action.

[0157] The Q-network structure includes:

[0158] Input layer: Receive the state vector (the dimension is the size of the state space);

[0159] Hidden layer 1: 128 neurons, ReLU activation function;

[0160] Hidden layer 2: 64 neurons, ReLU activation function;

[0161] Output layer: The size of the action space, outputting the Q-value for each action;

[0162] The experience replay mechanism and the target network are used in the training process to improve the learning stability. After 10,000 rounds of training, the Q-network can recommend the optimal maintenance strategy according to the bridge state.

[0163] Specific application case: The decision-making process after detecting signs of fatigue damage in the main cable joint area:

[0164] The digital twin model collects strain and displacement data of the joint area in real time through sensors.

[0165] The DS evidence theory fuses the sensing data, inspection records, and historical cases to determine that the fatigue mechanism is "stress concentration caused by local corrosion of steel wires", with a confidence level of 83%.

[0166] The Bayesian causal network identifies the fatigue development path: Environmental corrosion (EC = H) → Material deterioration (MD = M) → Stress concentration (SC = H) → Fatigue crack development (FC = Y).

[0167] The LSTM prediction model predicts that if no measures are taken, the fatigue damage index will increase from the current 0.32 to 0.47 within 30 days and may reach 0.68 (exceeding the warning value of 0.65) within 90 days.

[0168] The reinforcement learning algorithm evaluates the Q-values of various maintenance plans based on the current state:

[0169] No operation: Q-value = -25.6;

[0170] Continue monitoring: Q-value = -12.3;

[0171] Local anti-corrosion treatment: Q-value = 18.5;

[0172] Sling clip replacement: Q-value = 15.2;

[0173] Main cable section replacement: Q-value = -8.7;

[0174] The system recommends the "local anti-corrosion treatment" plan (with the highest Q-value) and predicts that after implementation:

[0175] The reliability index will increase from 0.78 to 0.92;

[0176] The maintenance cost is 350,000 yuan (normalized value is 0.25);

[0177] The expected life of this area will be extended by 7.5 years;

[0178] The comprehensive objective function value will be increased from 0.52 to 0.83;

[0179] Maintain the decision visualization interface to intuitively display the above analysis results, including the fatigue status heat map, trend prediction curve, and comparison charts of each maintenance plan, to support engineers in making final decisions.

[0180] In practical applications, the predictive maintenance system updates data after each maintenance and continuously adjusts and optimizes its decision-making strategy through reinforcement learning, forming a closed-loop adaptive maintenance management mechanism, effectively improving the anti-fatigue performance and safety level of the bridge structure, while reducing the maintenance cost throughout the life cycle.

[0181] This implementation method integrates digital twin technology, knowledge graph, and causal reasoning to provide a systematic solution for the predictive maintenance of the anti-fatigue performance of bridge structures, presenting many technical advantages: through multi-source evidence fusion and causal reasoning, the accuracy of fatigue mechanism identification is improved, and the misjudgment rate is reduced, laying a solid foundation for maintenance decisions; with the help of digital twin and adaptive maintenance decision-making system, accurate prediction of fatigue damage trends is realized, and the optimal maintenance plan is generated, promoting the transformation of maintenance decisions from experience-driven to data-driven, and reducing human judgment bias; using the improved DS evidence theory, effectively resolving the conflict of multi-source heterogeneous evidence, ensuring the reliability of the fusion results, and making the fatigue analysis more comprehensive and objective; establishing a fatigue knowledge graph, structurally storing expert experience and historical cases, realizing knowledge reuse and accumulation, and continuously empowering fatigue analysis; adopting a predictive maintenance mode, compared with traditional periodic maintenance and breakdown maintenance, reducing the maintenance cost, extending the service life of the bridge, and improving social and economic benefits; using a causal reasoning engine to reveal the causal relationship and development path of fatigue damage, providing interpretable analysis results, and enhancing the credibility and transparency of maintenance decisions.

[0182] In an embodiment of the present invention, an example of the aforementioned predictive maintenance method for the anti-fatigue performance of bridge structures based on digital twin is provided;

[0183] Application scenario description: This implementation method is applied to the predictive maintenance of a large cross-sea cable-stayed bridge. The bridge is 1680 meters long, with a main span of 580 meters. It has been in use for 12 years, with an average daily traffic volume of about 45,000 vehicle trips, and the proportion of heavy vehicles is about 18%. The bridge is located in a coastal environment and faces severe challenges of marine corrosion and typhoon loads. In particular, there are potential fatigue damage hazards in the bearing area and the cable-stayed cable anchorage area.

[0184] The bridge management unit faces the following challenges:

[0185] Traditional regular inspections are difficult to detect early signs of fatigue damage in a timely manner;

[0186] Maintenance decisions mainly rely on the experience judgment of engineers and lack reliable quantitative basis;

[0187] There are conflicts among multi-source information (such as detection data, inspection records, historical cases, etc.), making it difficult to integrate;

[0188] It is impossible to accurately predict the development trend of fatigue damage, resulting in blindness in the selection of maintenance timing and plans;

[0189] Based on the above challenges, the management unit decides to apply the digital twin-based predictive maintenance method for bridge structure anti-fatigue performance proposed in this implementation method to conduct intelligent maintenance management of the bridge.

[0190] The specific steps of the implementation process are as follows:

[0191] First, according to the design drawings and measured geometric data of the bridge, a refined three-dimensional model of the bridge was constructed. Through finite element analysis software, the model was divided into approximately 87,500 elements and 98,600 nodes, and a mechanical model of the bridge structure was established. Subsequently, a multi-modal sensing network was deployed at key parts of the bridge, as shown in Table 2:

[0192] Table 2: Layout of the sensing network for the cable-stayed bridge:

[0193]

[0194] Through edge computing devices and 5G networks, real-time collection of sensing data and dynamic update of the digital twin model were achieved. Using the Bayesian parameter calibration method, the finite element model was optimized with 6 months of monitoring data, making the error between the dynamic characteristics of the digital model and the actual structural response less than 5%.

[0195] To establish a bridge fatigue knowledge graph, the team defined 328 core concepts and 563 relationships through literature analysis and expert interviews, covering the key knowledge areas of bridge fatigue damage. Some core entities and relationships are shown in Table 3:

[0196] Table 3: Some core entities and relationships of the fatigue knowledge graph:

[0197]

[0198] Structured knowledge was extracted from 287 research literatures and 152 maintenance reports through natural language processing technology, forming approximately 25,600 knowledge triples. In addition, through the construction of knowledge inference rules, automatic derivation of implicit knowledge was achieved, enriching the content of the knowledge graph.

[0199] For the fatigue state assessment of the cable-stayed cable anchorage area, the system collected information from different evidence sources. The multi-source evidence and its basic probability assignment in a certain fatigue analysis are shown in Table 4:

[0200] Table 4: Example data of multi-source evidence fusion:

[0201]

[0202] By calculating the evidence distance matrix, the system identifies an obvious conflict between evidence sources. In particular, the conflict degree between the sensing data and the matching of historical cases is as high as 0.43. The evidence weights are calculated by applying the evidence distance, and the weights of each evidence source are obtained as follows: , , , .

[0203] The result after fusion shows that the credibility of "joint stress concentration" is 0.42, the credibility of "anchoring bolt loosening" is 0.28, the credibility of "anchor plate weld fatigue" is 0.21, and the uncertainty is reduced to 0.09.

[0204] Based on the knowledge graph, a Bayesian causal network of the fatigue mechanism in the stay cable anchorage zone is constructed, including key variables such as environmental factors, load characteristics, manufacturing quality, material properties, stress state, and fatigue performance.

[0205] Using the monitoring data and inspection results, the system calculates the conditional probability table between variables. When high-frequency vibration and small displacement changes in the anchorage zone are observed, the system conducts posterior probability inference, and the inference results are shown in Table 5:

[0206] Table 5: Calculation results of posterior probability of the causal inference engine:

[0207]

[0208] Through counterfactual intervention analysis, the system evaluates the effects of different maintenance measures. For example, it is calculated that the intervention of adding dampers can reduce the probability of the main fatigue path from 0.62 to 0.28, optimizing the connection structure can reduce the probability to 0.35, and replacing the anchoring device can reduce the probability to 0.08.

[0209] Based on the fused evidence and causal inference results, the system constructs an LSTM fatigue damage prediction model to predict the fatigue damage in the stay cable anchorage zone.

[0210] Meanwhile, based on the reinforcement learning algorithm, the system compares and evaluates various maintenance strategies, and the evaluation results are shown in Table 6:

[0211] Table 6: Evaluation results of maintenance strategies:

[0212]

[0213] Based on the objective function evaluation, the system recommends the "installation of dampers" maintenance strategy, which achieves the best balance between cost and benefit. The system displays the recommended solution and its expected effects through a visual interface to support managers in making the final decision.

[0214] Verification of technical effects: After the predictive maintenance system has been operating on this cross-sea cable-stayed bridge for 18 months, the maintenance effect of the anti-fatigue performance of the bridge structure has been significantly improved, which is mainly reflected in the following two core technical effects:

[0215] Significant improvement in the accuracy of fatigue mechanism identification: Through multi-source evidence fusion and causal reasoning technology, this implementation method has greatly improved the accuracy of fatigue mechanism identification. Before and after the system was put into use, the identification results of 28 known fatigue damage cases were compared, as shown in Table 7:

[0216] Table 7: Comparison of the accuracy of fatigue mechanism identification:

[0217]

[0218] The data shows that compared with the traditional method based on single evidence analysis, the accuracy of fatigue mechanism identification of this method has increased from 57.1% to 85.7%, with an average increase of 28.6%. Especially for the fatigue of the support area and the anchorage area with higher complexity, the accuracy improvement rate exceeds 28%.

[0219] Significant reduction in maintenance costs: The predictive maintenance mode adopted in this implementation method has significantly reduced the maintenance costs of the bridge compared with the traditional planned maintenance and passive maintenance modes. The comparison of the maintenance costs for 18 months before and after the system implementation is shown in Table 8:

[0220] Table 8: Comparative analysis of maintenance costs (unit: 10,000 yuan):

[0221]

[0222] As can be seen from Table 8, although the operation and maintenance cost of the monitoring system has increased by 82.2%, the costs of other maintenance types have been significantly reduced, especially the emergency repair cost has been reduced by 63.9%. Generally speaking, within 18 months, the total bridge maintenance cost has been reduced from 9.16 million yuan to 6.25 million yuan, a reduction of 2.91 million yuan, with a decrease rate of 31.8%.

[0223] In addition, the system has also brought other quantitative benefits, including:

[0224] The average remaining life of the key components of the bridge is expected to be extended by 25%;

[0225] The maintenance decision response time has been shortened from an average of 3.5 days to 0.8 days, with a 76.7% improvement;

[0226] The risk of bridge safety accidents is reduced by approximately 45%;

[0227] The life-cycle cost of the bridge is expected to be reduced by about 28%.

[0228] In summary, the present embodiment has achieved remarkable technical effects in practical applications. It not only improves the accuracy of fatigue mechanism identification, but also significantly reduces the maintenance cost, extends the service life of the bridge, and provides a strong guarantee for the safe operation of the bridge structure.

[0229] The embodiments of the present invention have been described above, but the present embodiment is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present embodiment, those of ordinary skill in the art can also make more equivalent embodiments in various forms, all of which fall within the protection scope of the present embodiment.

Claims

1. A method for predictive maintenance of the anti-fatigue performance of a bridge structure based on digital twins, characterized in that It includes the following steps: Construct a digital twin model of the bridge structure and deploy a multi-modal sensing network to achieve real-time data interaction between the physical structure and the virtual model; Establish a knowledge graph of bridge fatigue mechanism, transform expert experience and historical cases into structured knowledge, and form a fatigue mechanism knowledge base; Based on the DS evidence theory, achieve multi-source heterogeneous evidence fusion, resolve conflicts between evidence from different sources, and improve the accuracy of fatigue mechanism identification; Construct a causal reasoning engine based on the Bayesian network to identify the causal relationship path of fatigue damage development; Implement an adaptive predictive maintenance decision-making system to generate an optimal maintenance strategy based on the fatigue state prediction results.

2. The method for predictive maintenance of the anti-fatigue performance of a bridge structure based on digital twin according to claim 1, characterized in that The construction of the digital twin model of the bridge structure includes: Construct a three-dimensional finite element model of the bridge using parametric modeling technology; Deploy a multi-modal sensing network at key parts of the bridge, including strain sensors, accelerometers, and displacement sensors; Establish a mapping relationship between the sensing data and the digital model, and use the Bayesian calibration method to optimize the parameters of the finite element model; Construct a bidirectional data channel based on edge computing technology to achieve real-time interaction between the physical space and the virtual space.

3. The method for predictive maintenance of the anti-fatigue performance of a bridge structure based on digital twin according to claim 1, characterized in that The establishment of the knowledge graph of bridge fatigue mechanism includes: Define the core concepts, attributes, and relationships in the field of bridge fatigue based on the ontology method; Extract structured knowledge from fatigue research literature, maintenance reports, and expert knowledge through natural language processing technology; Construct knowledge inference rules to achieve automated reasoning and extension of knowledge; Establish a case library, and classify and store historical fatigue cases according to failure modes, causes, and performance characteristics.

4. The method for predictive maintenance of the anti-fatigue performance of a bridge structure based on digital twins according to claim 1, wherein In the implementation of multi-source heterogeneous evidence fusion based on the DS evidence theory, the evidence conflict coefficient is calculated by summing the products of the basic probability assignments between all mutually conflicting evidence sources, where mutual conflict means that the intersection of the propositions supported by different evidence sources is an empty set; the larger the value of this coefficient, the more serious the conflict between different evidence sources, and stronger evidence adjustment is required.

5. The predictive maintenance method for the anti-fatigue performance of a bridge structure based on digital twin according to claim 1, characterized in that In the implementation of multi-source heterogeneous evidence fusion based on the DS evidence theory, the evidence weight adjustment is achieved by calculating the average distance between each evidence source and other evidence sources, and converting this distance value into a weight value, so that the evidence with less difference from other evidence sources obtains a higher weight.

6. The predictive maintenance method for the anti-fatigue performance of a bridge structure based on digital twins according to claim 1, characterized in that, In the implementation of multi-source heterogeneous evidence fusion based on the DS evidence theory, the evidence fusion is achieved by calculating the sum of the products of the probabilities of different evidence combinations supporting the same proposition and performing normalization processing to ensure that the final probability distribution satisfies the constraint condition that the sum of probabilities is 1.

7. The method for predictive maintenance of the anti-fatigue performance of a bridge structure based on digital twins according to claim 1, wherein The construction of the causal reasoning engine based on the Bayesian network includes: Construct an initial Bayesian network structure based on the causal relationships in the knowledge graph; Combine the monitoring data and expert knowledge to learn the conditional probability table of the Bayesian network; Construct a posterior probability inference algorithm based on the MCMC method; Create a counterfactual intervention analysis module to calculate the intervention effects of different maintenance measures; Create a fatigue mechanism path identification algorithm to find the most likely fatigue development path.

8. The predictive maintenance method for anti-fatigue performance of bridge structures based on digital twins according to claim 7, wherein The intervention effect calculation of the counterfactual intervention analysis module is to perform a weighted average of the conditional probabilities under all possible covariate value conditions, where the weights are the marginal distribution probabilities of the covariates, so as to evaluate the causal impact of a specific intervention measure on the outcome variable.

9. The method for predictive maintenance of the anti-fatigue performance of a bridge structure based on digital twins according to claim 1, characterized in that The implementation of the adaptive predictive maintenance decision system includes: Establish a fatigue damage prediction model based on the LSTM neural network; Define a maintenance decision objective function that comprehensively considers safety reliability, maintenance cost, and structural life; Construct a maintenance strategy optimization algorithm based on reinforcement learning; Create a maintenance effect evaluation model; Establish a maintenance decision visualization interface.

10. A predictive maintenance system for the anti-fatigue performance of a bridge structure based on digital twins, which executes the predictive maintenance method for the anti-fatigue performance of a bridge structure based on digital twins according to any one of claims 1-9, characterized in that, Including: A digital twin model construction module for digital modeling of bridge structures and deployment of multimodal sensing networks; A fatigue knowledge graph construction module for integrating expert experience and historical cases to form a structured fatigue mechanism knowledge base; A multi-source evidence fusion module for conflict resolution and fusion of fatigue evidence from different sources based on the DS evidence theory; A causal reasoning engine module for identifying potential fatigue mechanism paths by applying Bayesian networks and causal analysis techniques; An adaptive predictive maintenance decision module for generating the best maintenance strategy through multi-objective optimization techniques.

Images