Bridge distributed wireless intelligent networking and sensing system

Abstract

This invention relates to the field of bridge monitoring and discloses a distributed wireless intelligent network sensing system for bridges, comprising: a sensing node module for distributed acquisition of multimodal raw signals at predetermined locations on the bridge; an edge aggregation module for receiving the damage feature data, constructing evidence sources by combining multi-source correlation data, and outputting hierarchical decision results; a cloud decision module for classifying the hierarchical decision results according to working conditions, generating adaptive early warning thresholds based on probability correction and dynamic threshold optimization, and executing hierarchical early warnings; and a wireless network transmission module for data interaction through a hybrid wireless network. By refining damage feature data through signal decomposition and interference removal, effectively filtering interference data through multi-source evidence fusion and hierarchical decision-making, and generating adaptive early warning thresholds adapted to different working conditions through working condition classification, probability correction, and dynamic threshold optimization, accurate identification and reliable early warning of bridge structural damage are achieved.

Description

Technical Field

[0001] This invention relates to the field of bridge monitoring technology, specifically a bridge distributed wireless intelligent networking sensing system. Background Technology

[0002] As a core component of transportation infrastructure, bridges play a crucial role in connecting regional transportation and ensuring smooth traffic flow. Their structural health directly affects the safety of people's lives and property and the stable development of the social economy. With the increase in service life, the increasingly heavy traffic load, and the long-term erosion of the complex natural environment, bridge structures are prone to damage such as cracks, corrosion, and loosening. If these damages are not identified in a timely and accurate manner, they may lead to structural failure or even collapse.

[0003] In recent years, distributed wireless intelligent network sensing technology has gradually replaced traditional wired monitoring systems and become the mainstream technical solution for bridge health monitoring due to its advantages such as flexible deployment, low wiring cost, and wide coverage. However, existing distributed wireless intelligent network sensing systems for bridges often use simple filtering or fixed frequency band division to process raw signals in the signal processing stage. This makes it difficult to effectively distinguish the overlapping time-domain and frequency-domain characteristics of environmental interference such as temperature gradients, wind vibration, and traffic noise with early damage signals, frequently resulting in false damage misjudgments, leading to a high false alarm rate and insufficient reliability of early warning. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a bridge distributed wireless intelligent network sensing system, which solves the problem of frequent false damage misjudgments, resulting in a high false alarm rate for damage identification and insufficient reliability of early warning in existing bridge distributed wireless intelligent network sensing systems.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a bridge distributed wireless intelligent networking sensing system, comprising:

[0006] The sensing node module is used to collect multimodal raw signals at predetermined locations on the bridge in a distributed manner, and obtain damage feature data through signal decomposition and interference removal.

[0007] The edge convergence module is used to receive the damage feature data, construct evidence sources by combining multi-source correlation data, filter interfering data through evidence fusion and hierarchical decision-making, and output hierarchical decision results.

[0008] The cloud-based decision-making module is used to classify the working conditions of the hierarchical decision-making results, generate adaptive early warning thresholds based on probability correction and dynamic threshold optimization, and execute hierarchical early warnings.

[0009] The wireless networking transmission module is used for various data exchanges through hybrid wireless networking and employs a time synchronization protocol to ensure that the timing of various data is consistent.

[0010] By adopting the above technical solutions, damage feature data is purified through signal decomposition and interference removal, interference data is effectively filtered through multi-source evidence fusion and hierarchical decision-making, and adaptive warning thresholds are generated to suit different working conditions through working condition classification, probability correction, and dynamic threshold optimization. Combined with hybrid wireless networking and time synchronization, reliable data interaction and time sequence consistency are ensured. Thus, a false alarm suppression system of signal purification, interference filtering, dynamic warning, and transmission is constructed, realizing accurate identification and reliable early warning of bridge structural damage. This solves the problem of frequent false damage misjudgments in existing bridge distributed wireless intelligent network sensing, resulting in a high false alarm rate and insufficient reliability of early warning.

[0011] Preferably, the distributed acquisition of multimodal raw signals at predetermined locations on the bridge includes the following steps:

[0012] Sensing nodes are deployed at predetermined locations on the bridge, and the initial signals of the bridge are collected synchronously using the sensing nodes. The initial signals include strain signals, acoustic emission signals, and environmental parameters.

[0013] The initial signal is subjected to hardware low-pass filtering to remove high-frequency electromagnetic interference, thus obtaining the multimodal original signal.

[0014] Preferably, the step of obtaining damage feature data through signal decomposition and interference removal includes the following steps:

[0015] The original multimodal signal is subjected to empirical mode signal decomposition to obtain several intrinsic mode function components;

[0016] Time-frequency analysis is performed on each intrinsic mode function component to calculate the instantaneous amplitude and instantaneous frequency, and to divide the environmental interference frequency band and the damage-sensitive frequency band.

[0017] Interference removal is performed on components in the environmental interference frequency band, while components in the damage-sensitive frequency band are retained to form effective components;

[0018] Damage features are extracted from the effective components to form damage feature data.

[0019] Preferably, the interference removal operation on components within the low-frequency interference band includes the following steps:

[0020] Calculate the time average of each intrinsic mode function component within the low-frequency interference band;

[0021] Select the same frequency components collected by adjacent sensing nodes within the same monitoring area and calculate the ensemble average;

[0022] The difference between the time average and the set average is calculated, and components with a difference less than a preset threshold are identified as environmental interference and removed.

[0023] Preferably, the step of constructing evidence sources by combining multi-source correlation data and filtering out interfering data through evidence fusion and hierarchical decision-making includes the following steps:

[0024] Simultaneously collect UAV inspection image feature data and real-time environmental condition parameters, and construct a multi-source evidence source together with damage feature data;

[0025] For each source of evidence, the sensor reliability weight, feature matching degree weight, and environmental adaptability weight are calculated to obtain the credibility of each source of evidence.

[0026] Based on the credibility of each source of evidence, the degree of conflict between the sources of evidence is quantified, and the conflicting evidence is weighted and corrected to obtain the corrected sources of evidence.

[0027] Evidence sources are fused using evidence synthesis rules to obtain probability information, which includes damage probability, interference probability, and uncertainty probability.

[0028] Based on probabilistic information, hierarchical decision-making is performed, interfering data is filtered out, and predicted damage data is output to form hierarchical decision results.

[0029] Preferably, quantifying the degree of conflict among the various sources of evidence includes the following steps:

[0030] A framework for identifying structures with and without damage is established, and a basic trust assignment function for each source of evidence is constructed.

[0031] Based on the basic trust allocation function, the conflict coefficient between any two sets of evidence sources is calculated using the distance calculation method.

[0032] The conflict coefficient is compared with a preset conflict threshold to determine the conflict level between evidence sources.

[0033] Preferably, the step of generating an adaptive early warning threshold based on probability correction and dynamic threshold optimization includes the following steps:

[0034] Collect relevant parameters and use a classification algorithm to divide the bridge's operating status into several typical working conditions. The relevant parameters include vehicle speed, axle load, temperature and humidity, and wind speed.

[0035] A three-layer inference model based on working condition, feature, and damage is constructed based on Bayesian network. The hierarchical decision results and typical working condition parameters are input into the model to calculate the posterior damage probability.

[0036] A dynamic threshold optimization model is constructed, with the false alarm rate as the objective, to solve the warning threshold under various typical operating conditions and form an adaptive warning threshold.

[0037] Preferably, the method for determining the warning thresholds under various typical operating conditions includes the following steps:

[0038] Define typical working conditions as the state space and threshold adjustment amount as the action space, and construct a quintuple for the dynamic decision-making process;

[0039] Based on the five-tuple, a reward function including false positive rate and false negative rate is set to determine the discount factor;

[0040] An iterative algorithm is used to calculate the value function of various typical working conditions until the convergence condition is met, and the warning threshold corresponding to each typical working condition is output.

[0041] Preferably, the execution of tiered early warning includes the following steps:

[0042] The posterior damage probability is compared with the adaptive warning threshold to determine the warning level;

[0043] According to the warning level, corresponding warning response operations are performed, including marking abnormal locations, pushing warning information to designated personnel, and simulating the development trend of damage.

[0044] Preferably, the wireless networking transmission module includes the following units:

[0045] The short-range communication unit is used to construct a mesh topology between sensing nodes via a short-range wireless communication protocol and to perform multi-hop routing data transmission.

[0046] The long-range communication unit is used to construct a star topology between the sensing node module and the edge aggregation module, and between the edge aggregation module and the cloud decision module, through a long-range wireless communication protocol to carry out long-distance data transmission.

[0047] The backup communication unit is used to configure a backup communication link, which automatically switches when the communication link signal is interrupted.

[0048] This invention provides a bridge-distributed wireless intelligent networking sensing system. It has the following beneficial effects:

[0049] 1. This invention purifies damage feature data through signal decomposition and interference removal, effectively filters interference data through multi-source evidence fusion and hierarchical decision-making, generates adaptive early warning thresholds for different working conditions through working condition classification, probability correction, and dynamic threshold optimization, and ensures reliable data interaction and time sequence consistency by combining hybrid wireless networking and time synchronization. Thus, it constructs a false alarm suppression system of signal purification, interference filtering, dynamic early warning, and transmission, realizing accurate identification and reliable early warning of bridge structural damage. It solves the problem of frequent false damage misjudgments in existing bridge distributed wireless intelligent network sensing, resulting in a high false alarm rate and insufficient early warning reliability.

[0050] 2. This invention constructs a multi-source evidence system, introduces a triple weighting of sensor reliability, feature matching degree, and environmental adaptability to quantify the credibility of evidence, uses distance calculation methods to quantify the degree of evidence conflict and make targeted corrections, and then achieves scientific fusion of multi-source data through evidence synthesis rules. This effectively eliminates abnormal data interference caused by sensor drift, signal attenuation, etc., so that the hierarchical decision-making results can truly reflect the actual state of the bridge structure, avoid systemic false alarms caused by single node anomalies, make the judgment of suspected damage and highly suspicious damage more convincing, and improve the reliability of hierarchical decision-making.

[0051] 3. This invention divides the bridge's operating state into various typical working conditions, constructs a working condition-feature-damage inference model based on a Bayesian network to correct the damage probability, and then constructs a dynamic threshold optimization model through a Markov decision process to solve for the optimal warning threshold under each working condition. This allows the warning threshold to be dynamically adjusted according to working condition parameters such as vehicle speed, axle load, temperature, humidity, and wind speed, which avoids false alarms caused by fluctuations in normal working conditions and prevents missed alarms for minor damage under complex working conditions. This improves the adaptability and reliability of the warning decision and balances the risks of false alarms and missed alarms. Attached Figure Description

[0052] Figure 1 This is an architecture diagram of a bridge distributed wireless intelligent networking sensing system proposed in this invention.

[0053] Figure 2 This is a flowchart of a bridge distributed wireless intelligent networking sensing method proposed in an embodiment of the present invention;

[0054] Figure 3 This is a data processing flowchart of a bridge distributed wireless intelligent networking sensing system proposed in an embodiment of the present invention. Detailed Implementation

[0055] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] Example 1:

[0057] In a first embodiment of the present invention, the present invention provides a bridge distributed wireless intelligent networking sensing system, such as... Figure 1 and Figure 3 As shown, it includes:

[0058] The sensing node module is used to collect multimodal raw signals at predetermined locations on the bridge in a distributed manner, and obtain damage feature data through signal decomposition and interference removal.

[0059] Furthermore, multimodal raw signals are collected in a distributed manner at predetermined locations on the bridge, including the following steps:

[0060] Sensing nodes are deployed at predetermined locations on the bridge, and the initial signals of the bridge are collected synchronously using the sensing nodes. The initial signals include strain signals, acoustic emission signals, and environmental parameters.

[0061] The initial signal is subjected to hardware low-pass filtering to remove high-frequency electromagnetic interference, thus obtaining the multimodal original signal.

[0062] Furthermore, damage characteristic data is obtained through signal decomposition and interference removal purification, including the following steps:

[0063] Empirical mode signal decomposition (EMS) is performed on the original multimodal signal to obtain several intrinsic mode function components.

[0064] Time-frequency analysis is performed on each intrinsic mode function component to calculate the instantaneous amplitude and instantaneous frequency, and to divide the environmental interference frequency band and the damage-sensitive frequency band.

[0065] Interference removal is performed on components in the environmental interference frequency band, while components in the damage-sensitive frequency band are retained to form effective components;

[0066] Damage features are extracted from the effective components to form damage feature data.

[0067] Furthermore, interference cancellation is performed on components within the low-frequency interference band, including the following steps:

[0068] Calculate the time average of each intrinsic mode function component within the low-frequency interference band;

[0069] Select the same frequency components collected by adjacent sensing nodes within the same monitoring area and calculate the ensemble average;

[0070] The difference between the time average and the set average is calculated, and components with a difference less than a preset threshold are identified as environmental interference and removed.

[0071] Specifically, the interference removal operation on the intrinsic mode function components within the environmental interference frequency band is based on the ergodicity theorem. It distinguishes between the stationary characteristics of the environmental interference signal and the non-stationary characteristics of the damage-related signal, thereby achieving precise removal of the interference signal.

[0072] Generally, the intrinsic mode function components within the environmental interference frequency band are mostly stationary random processes that satisfy the application conditions of the ergodicity theorem. Their time average and ensemble average are consistent. In contrast, damage-related signals are non-stationary processes, and there are significant differences between the two. This provides a theoretical basis for interference removal.

[0073] First, calculate the time average of each intrinsic mode function component within the environmental interference frequency band. Let a certain intrinsic mode function component be... The corresponding sampling duration is The formula for calculating the time average is: ,in, This represents the time average of the intrinsic mode function component. The first in the environmental interference frequency band The intrinsic mode function components at time 1 amplitude, The sampling duration of the signal is denoted by , which averages the overall stability of the component within the sampling period.

[0074] The sampling duration can be set according to the actual needs of bridge monitoring to ensure that the signal stability characteristics can be fully reflected. Generally, a duration that matches the natural vibration period of the bridge structure is selected to avoid time averaging deviation caused by too short a sampling duration.

[0075] After calculating the time average, the same-frequency components of adjacent sensing nodes within the same monitoring area are selected to calculate the ensemble average. The selection of adjacent sensing nodes is based on the deployment density of the monitoring area, and usually 3 to 5 adjacent nodes with the same-frequency components are selected to participate in the calculation of the ensemble average, ensuring that the ensemble average can effectively reflect the common characteristics of the signal within the area.

[0076] The formula for calculating the set average is: ,in, Indicates the first The set average of the components of the intrinsic mode function. This represents the number of neighboring sensing nodes involved in the calculation. For the first The data collected by the neighboring sensing nodes is the first The components of the same frequency intrinsic mode function at time 1 The amplitude of the set is averaged through collaborative verification of multi-node data, eliminating the random errors of a single node.

[0077] The criteria for determining components with the same frequency are the consistency of instantaneous frequencies. That is, other node components whose instantaneous frequency difference with the current inherent mode function component is within a preset range are selected to ensure that the components participating in the ensemble averaging calculation have the same physical meaning.

[0078] Next, the difference between the time average and the set average is calculated. The formula for calculating the difference is: ,in, For the first The difference between the time average and the ensemble average of each intrinsic mode function component is the value of the difference. The magnitude of this difference directly reflects the stationary characteristics of the component.

[0079] The preset threshold is set based on the accuracy requirements of bridge monitoring, taking into account the measurement error of the sensor and the fluctuation range of environmental interference, to ensure that environmental interference and damage-related signals can be accurately distinguished.

[0080] The calculated difference Compare with a preset threshold; if If the value is less than a preset threshold, the intrinsic mode function component is determined to be an environmental interference signal and is removed; if... If the value is greater than or equal to a preset threshold, the component is determined to be a non-stationary damage-related signal and is retained.

[0081] The preset threshold can be determined through statistical analysis of historical monitoring data, combined with signal characteristics under a large number of environmental interference scenarios, to ensure the rationality and applicability of the threshold.

[0082] Through the above process, the collaborative advantages of distributed sensing nodes can be fully utilized to accurately eliminate environmental interference signals based on the ergodicity theorem, effectively eliminating the influence of irrelevant interferences such as temperature gradients and wind vibrations, providing high-quality data support for subsequent damage feature extraction, and ensuring the purity and reliability of damage features.

[0083] The edge convergence module is used to receive damage feature data, combine it with multi-source correlation data to construct evidence sources, filter out interfering data through evidence fusion and hierarchical decision-making, and output hierarchical decision results.

[0084] Furthermore, evidence sources are constructed by combining multi-source correlated data, and interfering data is filtered out through evidence fusion and hierarchical decision-making, including the following steps:

[0085] Simultaneously collect UAV inspection image feature data and real-time environmental condition parameters, and construct a multi-source evidence source together with damage feature data;

[0086] For each source of evidence, the sensor reliability weight, feature matching degree weight, and environmental adaptability weight are calculated to obtain the credibility of each source of evidence.

[0087] Based on the credibility of each source of evidence, the degree of conflict between the sources of evidence is quantified, and the conflicting evidence is weighted and corrected to obtain the corrected sources of evidence.

[0088] The modified evidence sources are fused using evidence synthesis rules to obtain probabilistic information, which includes the probability of damage, the probability of interference, and the probability of uncertainty.

[0089] Based on probabilistic information, hierarchical decision-making is performed, interfering data is filtered out, and predicted damage data is output to form hierarchical decision results.

[0090] Furthermore, quantifying the degree of conflict between different sources of evidence includes the following steps:

[0091] A framework for identifying structures with and without damage is established, and a basic trust assignment function for each source of evidence is constructed.

[0092] Based on the basic trust allocation function, the conflict coefficient between any two sets of evidence sources is calculated using the distance calculation method.

[0093] The conflict coefficient is compared with a preset conflict threshold to determine the conflict level between evidence sources.

[0094] Specifically, quantifying the degree of conflict between various sources of evidence is based on the DS evidence theory. By constructing an identification framework and a basic trust allocation function, a standardized distance calculation method is used to quantify the differences in support for the same proposition from different sources of evidence. This provides a basis for the weighted correction of conflicting evidence and ensures the accuracy of evidence fusion.

[0095] Generally, the identification framework should align with the requirements of bridge structural health monitoring, focusing on the question of whether structural damage exists, and avoiding redundant propositions that increase computational complexity. An identification framework should be established that includes both damaged and undamaged structures. ,in , The proposition indicates that the structure is damaged. Representing the proposition of structural integrity, this identification framework can accurately cover the core decision-making objectives of this system and provide a unified benchmark for the trust allocation of various evidence sources.

[0096] After constructing the identification framework, a basic trust assignment function needs to be built for each source of evidence. The basic trust assignment function describes the degree of support a single source of evidence provides for each proposition in the identification framework. It satisfies a normalization condition, meaning the sum of the trust assignment values ​​for all propositions is 1. The expression for the basic trust assignment function is: ,in, This indicates that a certain source of evidence is relevant to the identification framework. any subset Basic trust assignment value, Can be taken separately , and itself, Corresponding trust allocation value Represents the uncertainty of the source of evidence, reflecting the degree to which the source of evidence supports each proposition.

[0097] The basic trust assignment function can be constructed based on the feature matching results and confidence weights of the evidence source. For example, for damage feature data as an evidence source, if the extracted damage features have a high degree of matching with the historical damage feature database and the sensor reliability weight is large, then a higher trust assignment value can be assigned. Proposition: If the matching degree is low or environmental interference has a significant impact, then a higher trust assignment value should be given. Proposition or The proposition is to ensure that the allocation of trust is consistent with the actual reliability of the source of evidence.

[0098] Based on the fundamental trust allocation function of each evidence source, the Jousselme distance method is used to quantify the degree of conflict between any two sets of evidence sources. The Jousselme distance effectively reflects the difference in trust distribution between two evidence sources within the identification framework; the greater the difference, the higher the degree of conflict. The calculation formula is as follows: ,in, Indicates the first Basic trust assignment function for each source of evidence With the Basic trust assignment function for each source of evidence The Jousselme distance between the two sources of evidence is the conflict coefficient between them. For identification framework Similarity matrix, matrix elements , and All are recognition frameworks A subset of these is used to characterize the similarity between different propositions; and All are based on recognition frameworks The basic trust assignment function is constructed in column vector form, with each element corresponding to the trust assignment value of each proposition.

[0099] Similarity matrix The construction of [the framework] needs to be combined with the propositional relationships of the identification framework, for example, when [the framework is incomplete]. = and = hour, ;when = and = hour, ;when = and = hour, ( , )= This process continues to ensure that the similarity matrix accurately reflects the logical connections between propositions.

[0100] When calculating the conflict coefficient, the basic trust assignment function for each source of evidence is first converted into a column vector form. For example, for each source of evidence... Its basic trust assignment vector is Then, substitute the values ​​into the Jousselme distance formula and perform matrix operations to obtain the conflict coefficients of the two sets of evidence sources. The range of the conflict coefficient is [0,1]. The closer the value is to 1, the higher the degree of conflict between the two sets of evidence sources. The closer the value is to 0, the stronger the consistency between the two sets of evidence sources.

[0101] The setting of the preset conflict threshold needs to be based on a large amount of experimental data and engineering experience, taking into account factors such as the type of evidence source, sensor accuracy, and environmental interference intensity, to ensure accurate differentiation between minor, moderate, and severe conflicts. In some embodiments, the preset conflict threshold can be set to 0.6. When the conflict coefficient is less than 0.6, it is judged as a minor conflict, which does not require significant correction; when the conflict coefficient is between 0.6 and 0.8, it is judged as a moderate conflict, which requires appropriate weighted correction; when the conflict coefficient is greater than 0.8, it is judged as a severe conflict, which requires significant weighted correction to avoid fusion distortion.

[0102] The calculated conflict coefficients are compared one by one with preset conflict thresholds. Based on the comparison results, the conflict level between each pair of evidence sources is determined. This conflict level directly guides the weighted correction strategy for subsequent conflicting evidence, ensuring that the corrected evidence sources can more accurately reflect the actual health status of the bridge structure. Through the above process, the degree of conflict among evidence sources is quantified and graded, providing support for the rationality and reliability of subsequent evidence fusion.

[0103] The cloud-based decision-making module is used to classify the working conditions of the hierarchical decision-making results, generate adaptive early warning thresholds based on probability correction and dynamic threshold optimization, and execute hierarchical early warnings.

[0104] Furthermore, an adaptive early warning threshold is generated based on probability correction and dynamic threshold optimization, including the following steps:

[0105] Collect relevant parameters and use a classification algorithm to divide the bridge's operating status into several typical working conditions. The relevant parameters include vehicle speed, axle load, temperature and humidity, and wind speed.

[0106] A three-layer inference model based on working condition, feature, and damage is constructed based on Bayesian network. The hierarchical decision results and typical working condition parameters are input into the model to calculate the posterior damage probability.

[0107] A dynamic threshold optimization model is constructed, with the false alarm rate as the objective, to solve the warning threshold under various typical operating conditions and form an adaptive warning threshold.

[0108] Furthermore, the early warning thresholds for various typical operating conditions are determined, including the following steps:

[0109] Define typical working conditions as the state space and threshold adjustment amount as the action space, and construct a quintuple for the dynamic decision-making process;

[0110] Based on the five-tuple, a reward function including false positive rate and false negative rate is set to determine the discount factor;

[0111] An iterative algorithm is used to calculate the value function of various typical working conditions until the convergence condition is met, and the warning threshold corresponding to each typical working condition is output.

[0112] Furthermore, implementing tiered early warning systems includes the following steps:

[0113] The posterior damage probability is compared with the adaptive warning threshold to determine the warning level;

[0114] According to the warning level, corresponding warning response operations are performed, including marking abnormal locations, pushing warning information to designated personnel, and simulating the development trend of damage.

[0115] Specifically, solving the early warning thresholds under various typical working conditions and implementing graded early warnings involves constructing a dynamic threshold optimization model based on Markov decision processes, obtaining the optimal early warning thresholds adapted to different working conditions through iterative algorithms, and then combining the posterior damage probability to achieve accurate graded early warnings.

[0116] Generally, the construction of a dynamic decision-making process needs to revolve around the actual operating scenario of the bridge, linking elements such as working conditions, threshold adjustments, and reward feedback to form a closed-loop optimization logic. First, a typical working condition is defined as the state space. state space This includes various typical scenarios for bridge operation. These scenarios are classified using a classification algorithm based on parameters such as vehicle speed, axle load, temperature and humidity, and wind speed. In some embodiments, typical operating conditions can be divided into six categories: unloaded sunny day, heavily loaded sunny day, unloaded rainy day, heavily loaded rainy day, strong wind, and low temperature. Each state ∈ It corresponds to a specific operating environment and load combination, ensuring that the state space can fully cover the actual operating conditions of the bridge.

[0117] Define the threshold adjustment amount as the action space. Action space Each action in The adjustment range representing the warning threshold is generally set in conjunction with the accuracy requirements of bridge monitoring, such as the action. The range of values ​​is That is, through actions This allows the current warning threshold to be adjusted ± from the baseline value. The adjustment ensures that the threshold adjustment can adapt to parameter fluctuations under different operating conditions.

[0118] Constructing state transition probabilities State transition probability Indicates the state Next action Then, the bridge's operational status will be transferred to... The probability. This probability is obtained based on statistical analysis of historical monitoring data, such as in the state. Perform actions under no-load, clear weather conditions Threshold adjustment Then, transition to the state. Probability of heavy load in sunny weather conditions This reflects the switching patterns between different operating conditions, providing a realistic basis for dynamic optimization.

[0119] A reward function based on the quintuple is used to include both false positive and false negative rates. The core of the reward function is to balance false positives and false negatives, prioritizing reducing the false positive rate to meet the core requirements of the system. The formula for calculating the reward function is: ,in, Indicates the state Next action The immediate reward value afterward, This represents the false alarm rate under this combination of operating conditions and actions, i.e., the probability that a non-damaging signal is judged as a damage. This represents the false negative rate for this combination, i.e., the probability that the damage signal is not detected. This is the weighting coefficient, which ranges from 0 to 1. Generally, We set the value to 0.7 to prioritize the optimization objective with the lowest false positive rate, ensuring that the reward value accurately reflects the rationality of the action.

[0120] Determine the discount factor Discount factor The weight used to balance current rewards and future cumulative rewards is generally set to 0.9. This value considers both the immediate effect of the current action and the cumulative benefits of long-term optimization, avoiding the threshold optimization from getting stuck in a local optimum due to excessive focus on short-term rewards.

[0121] Thus, the five-tuple for the dynamic decision-making process is now complete. This provides a complete mathematical model to support the solution of the early warning threshold.

[0122] An iterative algorithm is used to calculate the value function for various typical working conditions until the convergence condition is met. In some embodiments, a value iteration algorithm is selected as the iterative algorithm, and the value function... Indicates the state The iterative formula for the long-term cumulative reward that can be obtained is as follows: ,in, Indicates the first The value function after the nth iteration Indicates the first State after the next iteration The value function, This indicates selecting the action that maximizes the long-term cumulative reward. .

[0123] The iterative process is initialized with the value function of all states. Then, the value function of each state is updated iteratively according to the above formula. Each iteration traverses all state and action combinations and calculates the corresponding long-term cumulative reward. The convergence condition is set as follows: the maximum difference between the value functions of two adjacent iterations is less than a preset convergence threshold. ,Right now , Generally, 10 is taken. -3 This ensures that the value function converges to a stable value.

[0124] Once the iteration meets the convergence condition, the optimal action for each typical working condition is determined based on the optimal value function. The optimal action The corresponding threshold is the warning threshold for that operating condition. The optimal warning thresholds for all operating conditions are aggregated to form an adaptive warning threshold set.

[0125] After obtaining the adaptive warning threshold, a tiered warning operation is executed. The tiered warning is based on the comparison between the posterior damage probability and the adaptive warning threshold. The posterior damage probability is calculated by a three-layer Bayesian network inference model of working condition-feature-damage, reflecting the possibility of damage to the bridge structure.

[0126] The posterior damage probability is compared one by one with the adaptive early warning threshold for the corresponding typical working conditions to determine the early warning level. As an option, the early warning level is divided into three levels. When the posterior damage probability is in the first interval, it is determined to be a suspected early warning level. At this time, only the abnormal location is marked in the digital twin model, and no early warning information is pushed to relevant personnel to avoid unnecessary maintenance responses. When the posterior damage probability is in the second interval, it is determined to be a confirmed early warning level. Early warning information is pushed to maintenance personnel, and a drone inspection task is associated, requiring on-site verification to be completed within a specified time. When the posterior damage probability is in the third interval, it is determined to be an emergency early warning level. Emergency early warning information is immediately pushed to the emergency response team, and the damage development trend is simulated based on the bridge finite element mechanical model to provide technical support for emergency response.

[0127] The first interval is defined as a posterior probability of damage greater than or equal to 0.3 and less than 0.5; the second interval is greater than or equal to 0.5 and less than 0.8; and the third interval is greater than or equal to 0.8. The interval division can be adjusted according to the importance of the bridge and the required monitoring accuracy. Through this tiered early warning mechanism, differentiated measures can be taken based on different damage risks, ensuring timely response to major risks while avoiding resource waste caused by minor anomalies, thus improving the practicality and accuracy of the early warning system.

[0128] The wireless networking transmission module is used for various data exchanges through hybrid wireless networking and employs a time synchronization protocol to ensure that the timing of various data is consistent.

[0129] Furthermore, the wireless networking transmission module includes the following units:

[0130] The short-range communication unit is used to construct a mesh topology between sensing nodes via a short-range wireless communication protocol and to perform multi-hop routing data transmission.

[0131] The long-range communication unit is used to construct a star topology between the sensing node module and the edge aggregation module, and between the edge aggregation module and the cloud decision module, through a long-range wireless communication protocol to carry out long-distance data transmission.

[0132] The backup communication unit is used to configure a backup communication link, which automatically switches when the communication link signal is interrupted.

[0133] Specifically, the wireless networking transmission module achieves reliable data interaction and time sequence consistency at all levels through a hybrid topology and time synchronization protocol, adapting to the signal transmission requirements of distributed bridge monitoring.

[0134] Generally, short-range communication units employ short-range wireless communication protocols to construct a mesh topology among sensing nodes, relying on multi-hop routing to overcome obstruction limitations and ensure signal interconnection among densely deployed nodes. Alternatively, long-range communication units employ long-range wireless communication protocols to construct a star topology between sensing node modules and edge aggregation modules, and between edge aggregation modules and cloud decision-making modules, meeting the requirements for long-distance data transmission.

[0135] The backup communication unit is configured with a backup communication link, which automatically switches when the primary link signal is interrupted to avoid data transmission interruption. Specifically, a precise time synchronization protocol is used to achieve timing alignment, and the clock offset correction formula is: ,in This is the clock offset. The time when the synchronization message is received from the node. The time when the master node sends the synchronization message. The physical distance between nodes. The local clock is adjusted using this formula to determine the signal propagation speed, ensuring consistent timing of multi-source data.

[0136] Example 2:

[0137] In a second embodiment of the present invention, the present invention provides a bridge distributed wireless intelligent networking sensing method, such as... Figure 2 As shown, it includes the following steps:

[0138] Multimodal raw signals were collected in a distributed manner at predetermined locations on the bridge, and damage characteristic data were obtained through signal decomposition and interference removal.

[0139] The system receives damage characteristic data, constructs evidence sources by combining multi-source correlation data, filters out interfering data through evidence fusion and hierarchical decision-making, and outputs hierarchical decision-making results.

[0140] The hierarchical decision-making results are classified according to working conditions, and adaptive early warning thresholds are generated based on probability correction and dynamic threshold optimization, and hierarchical early warnings are executed.

[0141] Various data exchanges are conducted through a hybrid wireless network, and a time synchronization protocol is used to ensure that the timing of various data is consistent.

[0142] In the long-term health monitoring scenario of long-span bridges on mountainous highways, the bridges are located in complex terrain and are constantly affected by factors such as strong winds, temperature variations, and frequent passage of heavy-duty trucks, resulting in significant environmental interference. Existing distributed sensing systems struggle to distinguish between environmental interference and damage signals, and their fixed thresholds cannot adapt to dynamic operating conditions, leading to a high false alarm rate in damage identification. This forces maintenance teams to conduct frequent but ineffective inspections, increasing manpower costs and potentially overlooking real damage hazards, threatening bridge traffic safety and structural durability. To address these issues, this invention employs a bridge distributed wireless intelligent networking sensing method, the process of which is as follows: Figure 2 As shown. The specific implementation process of this method is as follows:

[0143] First, strain signals, acoustic emission signals, and temperature and humidity parameters are collected simultaneously at predetermined locations such as the mid-span, quarter-span, top of piers, and supports of the main beam of the bridge. Then, signal decomposition and interference removal are achieved through empirical mode decomposition and ergodicity theorem to purify and obtain damage characteristic data.

[0144] Subsequently, damage feature data is received, and multi-source evidence sources are constructed by combining UAV inspection image features and real-time environmental condition parameters. After credibility quantification, conflict degree analysis and evidence synthesis rules are fused, interference data is filtered and graded decision results of suspected damage or highly suspected damage are output.

[0145] Next, the working conditions are classified according to the hierarchical decision results. The damage probability is corrected based on the three-layer inference model of working condition-feature-damage constructed based on Bayesian network. A dynamic threshold optimization model is constructed through Markov decision process to generate adaptive warning thresholds that are suitable for different working conditions and to execute hierarchical warnings.

[0146] Finally, the wireless network transmission module achieves data interaction at all levels through a hybrid network of short-range mesh topology and long-range star topology. It adopts a precise time synchronization protocol to ensure data timing consistency, configures backup communication links to avoid transmission interruption, and ensures reliable transmission of monitoring data and accurate push of early warnings.

[0147] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A bridge distributed wireless intelligent networking and sensing system, characterized in that, include: The sensing node module is used to collect multimodal raw signals at predetermined locations on the bridge in a distributed manner, and obtain damage feature data through signal decomposition and interference removal. The edge convergence module is used to receive the damage feature data, construct evidence sources by combining multi-source correlation data, filter interfering data through evidence fusion and hierarchical decision-making, and output hierarchical decision results. The cloud-based decision-making module is used to classify the working conditions of the hierarchical decision-making results, generate adaptive early warning thresholds based on probability correction and dynamic threshold optimization, and execute hierarchical early warnings. The wireless networking transmission module is used for various data exchanges through hybrid wireless networking and employs a time synchronization protocol to ensure that the timing of various data is consistent. 2.The bridge distributed wireless intelligent networking and sensing system of claim 1, wherein: The distributed acquisition of multimodal raw signals at predetermined locations on the bridge includes the following steps: Sensing nodes are deployed at predetermined locations on the bridge, and the initial signals of the bridge are collected synchronously using the sensing nodes. The initial signals include strain signals, acoustic emission signals, and environmental parameters. The initial signal is subjected to hardware low-pass filtering to remove high-frequency electromagnetic interference, thus obtaining the multimodal original signal. 3.The bridge distributed wireless intelligent networking and sensing system of claim 1, wherein: The process of obtaining damage feature data through signal decomposition and interference removal includes the following steps: The original multimodal signal is subjected to empirical mode signal decomposition to obtain several intrinsic mode function components; Time-frequency analysis is performed on each intrinsic mode function component to calculate the instantaneous amplitude and instantaneous frequency, and to divide the environmental interference frequency band and the damage-sensitive frequency band. Interference removal is performed on components in the environmental interference frequency band, while components in the damage-sensitive frequency band are retained to form effective components; Damage features are extracted from the effective components to form damage feature data.

4. The bridge distributed wireless intelligent networking and sensing system of claim 3, wherein: The interference removal operation for components in the low-frequency interference band includes the following steps: Calculate the time average of each intrinsic mode function component within the low-frequency interference band; Select the same frequency components collected by adjacent sensing nodes within the same monitoring area and calculate the ensemble average; The difference between the time average and the set average is calculated, and components with a difference less than a preset threshold are identified as environmental interference and removed.

5. The bridge distributed wireless intelligent networking and sensing system of claim 1, wherein: The process of constructing evidence sources by combining multi-source correlation data, and filtering out interfering data through evidence fusion and hierarchical decision-making, includes the following steps: Simultaneously collect UAV inspection image feature data and real-time environmental condition parameters, and construct a multi-source evidence source together with damage feature data; For each source of evidence, the sensor reliability weight, feature matching degree weight, and environmental adaptability weight are calculated to obtain the credibility of each source of evidence. Based on the credibility of each source of evidence, the degree of conflict between the sources of evidence is quantified, and the conflicting evidence is weighted and corrected to obtain the corrected sources of evidence. Evidence sources are fused using evidence synthesis rules to obtain probability information, which includes damage probability, interference probability, and uncertainty probability. Based on probabilistic information, hierarchical decision-making is performed, interfering data is filtered out, and predicted damage data is output to form hierarchical decision results.

6. The bridge distributed wireless intelligent networking and sensing system of claim 5, wherein: The quantification of the degree of conflict between various sources of evidence includes the following steps: A framework for identifying structures with and without damage is established, and a basic trust assignment function for each source of evidence is constructed. Based on the basic trust allocation function, the conflict coefficient between any two sets of evidence sources is calculated using the distance calculation method. The conflict coefficient is compared with a preset conflict threshold to determine the conflict level between evidence sources.

7. The bridge distributed wireless intelligent networking and sensing system of claim 1, wherein: The process of generating adaptive warning thresholds based on probability correction and dynamic threshold optimization includes the following steps: Collect relevant parameters and use a classification algorithm to divide the bridge's operating status into several typical working conditions. The relevant parameters include vehicle speed, axle load, temperature and humidity, and wind speed. A three-layer inference model based on working condition, feature, and damage is constructed based on Bayesian network. The hierarchical decision results and typical working condition parameters are input into the model to calculate the posterior damage probability. A dynamic threshold optimization model is constructed, with the false alarm rate as the objective, to solve the warning threshold under various typical operating conditions and form an adaptive warning threshold.

8. The bridge distributed wireless intelligent networking and sensing system of claim 7, wherein: The method for determining the warning thresholds under various typical operating conditions includes the following steps: Define typical working conditions as the state space and threshold adjustment amount as the action space, and construct a quintuple for the dynamic decision-making process; Based on the five-tuple, a reward function including false positive rate and false negative rate is set to determine the discount factor; An iterative algorithm is used to calculate the value function of various typical working conditions until the convergence condition is met, and the warning threshold corresponding to each typical working condition is output.

9. The bridge distributed wireless intelligent networking and sensing system of claim 1, wherein: The implementation of tiered early warning includes the following steps: The posterior damage probability is compared with the adaptive warning threshold to determine the warning level; According to the warning level, corresponding warning response operations are performed, including marking abnormal locations, pushing warning information to designated personnel, and simulating the development trend of damage.

10. The bridge distributed wireless intelligent networking and sensing system of claim 1, wherein: The wireless networking transmission module includes the following units: The short-range communication unit is used to construct a mesh topology between sensing nodes via a short-range wireless communication protocol and to perform multi-hop routing data transmission. The long-range communication unit is used to construct a star topology between the sensing node module and the edge aggregation module, and between the edge aggregation module and the cloud decision module, through a long-range wireless communication protocol to carry out long-distance data transmission. The backup communication unit is used to configure a backup communication link, which automatically switches when the communication link signal is interrupted.

Images