Bridge state monitoring method and system based on edge cloud collaboration

Abstract

The invention discloses a bridge state monitoring method and system based on edge cloud collaboration, and relates to the technical field of bridge safety, and the method comprises the steps: collecting real-time multi-source heterogeneous data; screening to obtain effective multi-source heterogeneous data; performing feature extraction and damage sensitivity index calculation on the effective multi-source heterogeneous data through a damage sensitivity index formula; transmitting the data to a cloud server through dynamic bandwidth allocation; constructing a bridge state evaluation model in the cloud server; obtaining a bridge overall state score according to the bridge state evaluation model; predicting the remaining service life and risk level of the bridge according to the overall state score of the bridge; and performing multi-level notification and alarm on a management platform, a supervision mechanism and a bridge site according to the risk level. According to the method, transmission delay and bandwidth occupation are effectively reduced through edge cloud collaboration and grading processing. And a multi-scale feature fusion model and a special calculation formula are adopted to realize accurate evaluation of the bridge state and reliable prediction of the residual life. And full-process linkage is realized through a closed-loop early warning mechanism.

Description

Technical Field

[0001] This invention relates to the field of bridge safety technology, and in particular to a bridge condition monitoring method and system based on edge-cloud collaboration. Background Technology

[0002] As a core component of transportation infrastructure, bridges play a crucial role in the cross-regional flow of people and goods, and their structural safety is directly related to public safety and socio-economic stability. With increasing service life, heavier traffic loads, and more frequent extreme weather events, bridge structures are prone to fatigue damage, crack propagation, and stiffness degradation. Failure to monitor and warn in a timely manner can lead to structural failure or even collapse, causing significant casualties and property damage. Therefore, achieving real-time, accurate, and efficient monitoring of bridge structural conditions and early identification of potential safety hazards is of paramount practical significance for ensuring bridge operational safety, extending service life, and optimizing maintenance decisions.

[0003] Current bridge monitoring technologies are mainly divided into two categories: traditional manual monitoring and automated monitoring. Traditional manual monitoring relies on professionals carrying testing equipment to conduct on-site surveys, load tests, or non-destructive testing. This method is not only labor-intensive, with long monitoring cycles and high costs, but also difficult to achieve continuous monitoring around the clock, and prone to missing instantaneous damage signals, failing to meet the real-time early warning needs of large bridges. Automated monitoring technology has developed rapidly in recent years, mainly including single-point monitoring systems based on single sensors, centralized monitoring systems based on wired networks, and preliminary edge-cloud collaborative monitoring frameworks. Among these issues, single-sensor monitoring systems can only acquire local physical quantity data, making it difficult to reflect the overall structural state of the bridge; centralized monitoring systems transmit all data to the cloud for processing, resulting in high bandwidth consumption, high response latency, and susceptibility to network outages; while the preliminary edge-cloud collaborative framework has achieved a division of labor between the edge and the cloud, it has significant shortcomings: data preprocessing uses a fixed window algorithm, which easily misclassifies dynamic load signals as noise; the edge side lacks targeted damage-sensitive feature indicators, resulting in low feature extraction efficiency; cloud-based evaluation models mostly use single convolutional neural networks or time-series models, making it difficult to simultaneously capture local microscopic damage and global long-term degradation features; data transmission does not consider priority differences, resulting in wasted bandwidth resources; the correlation between remaining life prediction and real-time status is weak, the early warning mechanism lacks a closed-loop design, and it cannot quickly respond to emergency scenarios.

[0004] Although existing automated monitoring technologies have made significant progress compared to traditional manual monitoring, there are still some shortcomings. Data preprocessing accuracy is insufficient; fixed-window algorithms struggle to adapt to dynamic load changes in bridges, resulting in a high false positive rate for abnormal data; lightweight and sensitive edge feature extraction is difficult to balance—high feature dimensionality leads to high computational costs, while low feature dimensionality results in low damage identification sensitivity; data transmission lacks a dynamic adaptation mechanism, preventing the priority transmission of high-risk data and impacting emergency response speed; cloud-based assessment models have limited feature capture capabilities, making it difficult for a single model to cover the multi-scale damage characteristics of bridges; and the remaining service life prediction is disconnected from real-time monitoring status, leading to insufficient reliability of prediction results. Summary of the Invention

[0005] This invention provides a bridge condition monitoring method and system based on edge-cloud collaboration to address the following issues in existing technologies: insufficient accuracy of data preprocessing during bridge monitoring; fixed window algorithms' inability to adapt to dynamic load changes in bridges, resulting in high misjudgment rates of abnormal data; difficulty in balancing lightweight and sensitivity in edge feature extraction (high feature dimensionality leads to high computational cost, while low feature dimensionality results in low damage identification sensitivity); lack of dynamic adaptation mechanisms in data transmission, preventing priority transmission of high-risk data and impacting emergency response speed; limited feature capture capabilities of cloud-based evaluation models, making it difficult for a single model to cover multi-scale damage characteristics of bridges; and disconnect between remaining service life prediction and real-time monitoring status, leading to insufficient reliability of prediction results.

[0006] On one hand, embodiments of the present invention provide a bridge status monitoring method based on edge-cloud collaboration, including: S1. Real-time multi-source heterogeneous data of the bridge site is collected at the bridge site using acceleration sensors, strain sensors, displacement sensors, temperature and humidity sensors and camera equipment; S2. On the edge computing device at the bridge site, the real-time multi-source heterogeneous data is filtered using an improved sliding window algorithm to obtain effective multi-source heterogeneous data; S3. On the edge computing device at the bridge site, feature extraction and damage sensitivity index calculation are performed on the effective multi-source heterogeneous data using the damage sensitivity index formula to obtain feature packets and damage sensitivity indices; the damage sensitivity index formula is as follows: ; in, It is a damage sensitivity indicator, with a value of 0-1; It is the real-time strain value. It is the maximum allowable strain designed according to the parameters of the bridge design specifications. This is the real-time displacement deviation value. It is the allowable displacement value specified in the standard. It is real-time vibration acceleration. It is the safe vibration acceleration threshold. These are weighting coefficients determined based on the bridge structure type. ; S4. Perform dynamic bandwidth allocation based on the damage sensitivity index; transmit the feature packet and the effective multi-source heterogeneous data to the cloud server through the dynamic bandwidth allocation; S5. On the cloud server, a bridge condition assessment model based on multi-scale feature fusion is constructed based on historical valid multi-source heterogeneous data, the feature package, and the damage sensitivity index; the overall bridge condition score is obtained based on the bridge condition assessment model based on multi-scale feature fusion and the valid multi-source heterogeneous data; the input of the bridge condition assessment model based on multi-scale feature fusion is the valid multi-source heterogeneous data, the feature package, and the damage sensitivity index, and the output of the bridge condition assessment model based on multi-scale feature fusion is the overall bridge condition score from 0 to 100; S6. The remaining service life and risk level of the bridge are predicted by the cloud server based on historical degradation data, design service life and the overall condition score of the bridge. S7. The cloud server sends multi-level notifications and alarms to the management platform, regulatory agencies, and edge computing devices at the bridge site based on the remaining service life and risk level of the bridge.

[0007] In one possible implementation, the acceleration sensor, strain sensor, displacement sensor, and temperature and humidity sensor are installed together through pre-embedding in the bridge structure and external deployment.

[0008] In one possible implementation, the effective multi-source heterogeneous data obtained by filtering the real-time multi-source heterogeneous data on the edge computing device at the bridge site using an improved sliding window algorithm includes: The improved sliding window algorithm is obtained by setting the window length of the sliding window algorithm based on the real-time vibration frequency at the bridge site. The effective multi-source heterogeneous data is obtained by calculating the real-time multi-source heterogeneous data using the improved sliding window algorithm and linear interpolation replacement.

[0009] In one possible implementation, feature packages and damage sensitivity indices are obtained by performing feature extraction and damage sensitivity index calculation on the effective multi-source heterogeneous data using a damage sensitivity index formula on an edge computing device at the bridge site. These include: The time-domain and frequency-domain features in the effective multi-source heterogeneous data are packaged with the damage sensitivity index to obtain a feature package.

[0010] In one possible implementation, the bridge condition assessment model constructed on the cloud server based on multi-scale feature fusion according to historical valid multi-source heterogeneous data, the feature package, and the damage sensitivity index includes: A training dataset is constructed using real bridge calibration data, the feature package, and the historical valid multi-source heterogeneous data. Labels are constructed based on static load test results, standard evaluation levels, expert scores, and the aforementioned damage sensitivity indicators; The bridge state assessment model based on multi-scale feature fusion is constructed using the training dataset, labels, mean squared error loss function, and Adam optimizer.

[0011] In one possible implementation, obtaining the overall bridge condition score based on the bridge condition assessment model based on multi-scale feature fusion and the effective multi-source heterogeneous data includes: The bridge status assessment model based on multi-scale feature fusion uploads the effective multi-source heterogeneous data to the cloud server; Local features of the effective multi-source heterogeneous data are extracted using multi-scale CNN; Global temporal features of the effective multi-source heterogeneous data are extracted using Transformer; The overall bridge status score is obtained by weighted fusion of the local features and the global temporal features using the damage sensitivity index.

[0012] In one possible implementation, the remaining service life and risk level of the bridge are predicted by the cloud server based on historical degradation data, design service life, and the overall bridge condition score, including: A correlation prediction formula is constructed based on the overall bridge condition score as the independent variable. The remaining service life of the bridge is calculated using the aforementioned correlation prediction formula, design service life, and historical degradation data. The risk level is set and assessed based on the remaining service life of the bridge and the overall condition score of the bridge.

[0013] On the other hand, embodiments of the present invention provide a bridge status monitoring system based on edge-cloud collaboration, including: The bridge site data acquisition module is used to collect real-time multi-source heterogeneous data of the bridge site through acceleration sensors, strain sensors, displacement sensors, temperature and humidity sensors and camera equipment. The bridge site data filtering module is used to filter the real-time multi-source heterogeneous data on the edge computing device at the bridge site using an improved sliding window algorithm to obtain effective multi-source heterogeneous data. The bridge site data processing module is used to perform feature extraction and damage sensitivity index calculation on the effective multi-source heterogeneous data using a damage sensitivity index formula on the edge computing device at the bridge site, thereby obtaining a feature package and a damage sensitivity index; the damage sensitivity index formula is as follows: ; in, It is a damage sensitivity indicator, with a value of 0-1; It is the real-time strain value. It is the maximum allowable strain designed according to the parameters of the bridge design specifications. This is the real-time displacement deviation value. It is the allowable displacement value specified in the standard. It is real-time vibration acceleration. It is the safe vibration acceleration threshold. These are weighting coefficients determined based on the bridge structure type. ; The edge-cloud dynamic data transmission module is used to dynamically allocate bandwidth according to the damage sensitivity index; and to transmit the feature packet and the effective multi-source heterogeneous data to the cloud server through the dynamic bandwidth allocation. The cloud-based data fusion processing module is used to construct a bridge condition assessment model based on multi-scale feature fusion on the cloud server based on historical valid multi-source heterogeneous data, the feature package, and the damage sensitivity index; and to obtain an overall bridge condition score based on the bridge condition assessment model based on multi-scale feature fusion and the valid multi-source heterogeneous data. The input of the bridge condition assessment model based on multi-scale feature fusion is the valid multi-source heterogeneous data, the feature package, and the damage sensitivity index, and the output of the bridge condition assessment model based on multi-scale feature fusion is an overall bridge condition score of 0 to 100. The cloud-integrated data assessment module is used to predict the remaining service life and risk level of the bridge based on historical degradation data, design service life and the overall condition score of the bridge on the cloud server. The bridge data reporting module is used to send multi-level notifications and alarms to the management platform, regulatory agencies, and edge computing devices at the bridge site based on the remaining service life and risk level of the bridge on the cloud server.

[0014] The bridge status monitoring method and system based on edge-cloud collaboration in this invention have the following advantages: (1) By using edge-cloud collaboration and hierarchical processing, the accuracy of data preprocessing and damage identification is greatly improved, and transmission latency and bandwidth usage are effectively reduced.

[0015] (2) By adopting a multi-scale feature fusion model and a special calculation formula, the bridge condition can be accurately assessed and the remaining life can be reliably predicted, making the decision more scientific.

[0016] (3) A closed-loop early warning mechanism is used to achieve full-process linkage, improve the real-time performance and emergency response efficiency of the monitoring system, and is more suitable for the long-term safe operation and maintenance of large bridges. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating the bridge condition monitoring method based on edge-cloud collaboration provided in this application embodiment; Figure 2 This is a schematic diagram of the structure of a bridge condition monitoring system based on edge-cloud collaboration provided in an embodiment of this application. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Figure 1 This is a flowchart illustrating the bridge condition monitoring method based on edge-cloud collaboration provided in an embodiment of the present invention. The embodiment of the present invention provides a bridge condition monitoring method based on edge-cloud collaboration, comprising: S1. Real-time multi-source heterogeneous data of the bridge site is collected at the bridge site using acceleration sensors, strain sensors, displacement sensors, temperature and humidity sensors and camera equipment; S2. On the edge computing device at the bridge site, the real-time multi-source heterogeneous data is filtered using an improved sliding window algorithm to obtain effective multi-source heterogeneous data; S3. On the edge computing device at the bridge site, feature extraction and damage sensitivity index calculation are performed on the effective multi-source heterogeneous data using the damage sensitivity index formula to obtain feature packets and damage sensitivity indices; the damage sensitivity index formula is as follows: ； in, It is a damage sensitivity indicator, with a value of 0-1; It is the real-time strain value. It is the maximum allowable strain designed according to the parameters of the bridge design specifications. This is the real-time displacement deviation value. It is the allowable displacement value specified in the standard. It is real-time vibration acceleration. It is the safe vibration acceleration threshold. These are weighting coefficients determined based on the bridge structure type. ; S4. Perform dynamic bandwidth allocation based on the damage sensitivity index; transmit the feature packet and the effective multi-source heterogeneous data to the cloud server through the dynamic bandwidth allocation; S5. On the cloud server, a bridge condition assessment model based on multi-scale feature fusion is constructed based on historical valid multi-source heterogeneous data, the feature package, and the damage sensitivity index; the overall bridge condition score is obtained based on the bridge condition assessment model based on multi-scale feature fusion and the valid multi-source heterogeneous data; the input of the bridge condition assessment model based on multi-scale feature fusion is the valid multi-source heterogeneous data, the feature package, and the damage sensitivity index, and the output of the bridge condition assessment model based on multi-scale feature fusion is the overall bridge condition score from 0 to 100; S6. The remaining service life and risk level of the bridge are predicted by the cloud server based on historical degradation data, design service life and the overall condition score of the bridge. S7. The cloud server sends multi-level notifications and alarms to the management platform, regulatory agencies, and edge computing devices at the bridge site based on the remaining service life and risk level of the bridge.

[0021] The acceleration sensor, strain sensor, displacement sensor, and temperature and humidity sensor are installed together through pre-embedded parts in the bridge structure and external deployment.

[0022] The effective multi-source heterogeneous data obtained by filtering the real-time multi-source heterogeneous data using an improved sliding window algorithm on the edge computing device at the bridge site includes: The improved sliding window algorithm is obtained by setting the window length of the sliding window algorithm based on the real-time vibration frequency at the bridge site. The effective multi-source heterogeneous data is obtained by calculating the real-time multi-source heterogeneous data using the improved sliding window algorithm and linear interpolation replacement.

[0023] On the edge computing device at the bridge site, feature extraction and damage sensitivity index calculation are performed on the effective multi-source heterogeneous data using a damage sensitivity index formula to obtain a feature package and damage sensitivity index, including: The time-domain and frequency-domain features in the effective multi-source heterogeneous data are packaged with the damage sensitivity index to obtain a feature package.

[0024] The bridge condition assessment model constructed on the cloud server based on historical valid multi-source heterogeneous data, the feature package, and the damage sensitivity index includes: A training dataset is constructed using real bridge calibration data, the feature package, and the historical valid multi-source heterogeneous data. Labels are constructed based on static load test results, standard evaluation levels, expert scores, and the aforementioned damage sensitivity indicators; The bridge state assessment model based on multi-scale feature fusion is constructed using the training dataset, labels, mean squared error loss function, and Adam optimizer.

[0025] The overall bridge condition score obtained based on the bridge condition assessment model based on multi-scale feature fusion and the effective multi-source heterogeneous data includes: The bridge status assessment model based on multi-scale feature fusion uploads the effective multi-source heterogeneous data to the cloud server; Local features of the effective multi-source heterogeneous data are extracted using multi-scale CNN; Global temporal features of the effective multi-source heterogeneous data are extracted using Transformer; The overall bridge status score is obtained by weighted fusion of the local features and the global temporal features using the damage sensitivity index.

[0026] The remaining service life and risk level of the bridge are predicted by the cloud server based on historical degradation data, design service life, and the overall condition score of the bridge, including: A correlation prediction formula is constructed based on the overall bridge condition score as the independent variable. The remaining service life of the bridge is calculated using the aforementioned correlation prediction formula, design service life, and historical degradation data. The risk level is set and assessed based on the remaining service life of the bridge and the overall condition score of the bridge.

[0027] For example, data from the bridge is first collected to form multi-source heterogeneous data. This multi-source heterogeneous data includes acceleration, strain, displacement, temperature and humidity data, and image data. This data is collected by a sensor array deployed at predetermined locations and densities at key stress-bearing components such as the main beam, piers, cap beams, expansion joints, and supports of the bridge. To ensure that all data are time-synchronized, aligned, and fused, the entire sensor system employs local high-precision crystal oscillator compensation to achieve a unified timestamp for all sensors.

[0028] Various sensors are configured with different sampling rates according to structural monitoring requirements: accelerometers are set at 500Hz for vibration, modal, and damage identification; strain sensors are set at 200Hz for stress, fatigue, and crack monitoring; temperature and humidity sensors are set at 1Hz for environmental compensation; and video acquisition equipment is set at 25 frames per second for auxiliary judgment of apparent defects.

[0029] Edge computing gateways aggregate raw data from all sensors in real time via wired / wireless connections using technologies such as Ethernet, LoRa, and 4G / 5G, employing standard industrial protocols like Modbus / TCP and MQTT. The gateway does not perform any filtering, noise reduction, or feature extraction; it simply organizes the data in a fixed format: timestamp + sensor number + data type + raw physical quantity value, storing it in a local high-speed cache to ensure the raw data is not tampered with or lost.

[0030] After collecting the raw data, outliers, jumps, and missing values ​​are found due to electromagnetic interference, momentary sensor malfunctions, lightning strikes, and power supply fluctuations. Traditional fixed windows cannot adapt to changes in bridge vibration and are prone to misjudgment. This application uses an adaptive improved sliding window algorithm for data filtering.

[0031] The adaptive window length of the improved sliding window algorithm is no longer fixed, but is determined by the real-time vibration frequency, as shown in the following equation: ； in, This is the length of the sliding window in seconds. k This is a proportionality coefficient, determined according to the type of bridge, and typically ranges from 3 to 5. f The real-time vibration frequency of the bridge is calculated from the acceleration data in S1 using Fast Fourier Transform (FFT). As the bridge load, vehicle speed, and traffic flow change, the window automatically lengthens or shortens to ensure that each window contains a complete vibration cycle.

[0032] The following formula is used to calculate all data points within the current window: mean ; Standard deviation ; Anomaly detection is performed through a dual-assessment process, specifically by applying the 3σ principle and the upper limit of the measurement range, to avoid misjudgments based on a single condition. The formula is as follows: ; in, It is the first one in the window i One set of raw data, It is the window mean. It is the window standard deviation. It is the theoretical maximum range of the sensor hardware. n This is the total number of data points contained within the current sliding window. Data that satisfies the above condition is considered abnormal.

[0033] Linear interpolation replacement is used to process outlier data, as shown in the following formula: ; in, This is the corrected data. and These are the valid data adjacent to the outlier.

[0034] After replacing the abnormal data, the cleaned and valid data are subjected to feature extraction and damage-sensitive feature index calculation.

[0035] The formula for calculating the damage sensitivity index is as follows: ； in, It is a damage sensitivity indicator, ranging from 0 to 1, with a higher value indicating a higher risk of damage. It is the real-time strain value. It is the maximum allowable strain designed according to the parameters of the bridge design specifications. This is the real-time displacement deviation value. It is the allowable displacement value specified in the standard. It is real-time vibration acceleration. It is the safe vibration acceleration threshold. These are weighting coefficients, determined according to the bridge structure type. Simultaneously extract time-domain features (mean, variance, peak value) and frequency-domain features (fundamental frequency, main frequency amplitude), and combine all features with... DSI Pack them into lightweight feature packages to reduce data size.

[0036] Then according to In other words, risk levels are classified based on damage sensitivity indicators, specifically: Level 1 (High Priority): DSI >0.7, high risk of damage, feature packet + raw data full transmission; Level 2 (Medium Priority): 0.3 DSI ≤0.7, normal state, only feature packets are transmitted; Level 3 (low priority): DSI ≤0.3, safe state, transmit the feature packet once every 10 cycles.

[0037] The bandwidth allocation formula is as follows: ； in, It is the first i Level data allocation bandwidth, It is the total transmission bandwidth. It is the first i Level Priority Weight (Level 1) P =5, Level 2 P =2, Level 3 P=1). The edge gateway automatically compresses and transmits data according to this strategy. After receiving the data in the cloud, it sorts it by timestamp to ensure that the data is consistent and without loss.

[0038] Combining CNN and Transformer for extracting multi-scale damage features of bridge structures; Multi-scale convolutional layers extract local micro-features of the S4 input, such as micro-cracks and local strain concentrations, through three different convolutional kernels: 3×3, 5×5, and 6×6.

[0039] The Transformer encoding layer extracts global macroscopic features from time-series data, such as overall structural settlement and long-term fatigue trends.

[0040] Then, the micro and macro features are concatenated and introduced into the DSI in S3 as attention weights to output the overall state score of the bridge, as shown in the following formula: ; in, It is the overall condition score of the bridge (0~100). It is a local feature score. It is a global feature score. It is the fusion coefficient, with a value of 0.4. The scoring result is directly related to the structural health level and provides a core dependent variable for subsequent predictions.

[0041] Based on the overall condition score of the bridge S The remaining life of a bridge is calculated using the design life and material fatigue parameters, as shown in the following formula: ; in, This is the remaining useful life, measured in years. It refers to the bridge's design service life, a parameter inherent in the specifications. It is a degradation index, determined by the fatigue characteristics of the material; for concrete bridges, θ = 1.2. According to... RUL and S Jointly determine the risk level.

[0042] High risk is defined as: S <60 or RUL <10 years; Medium risk: 60≤ S <85 or 10≤ RUL <30 years; Low risk: S ≥85 and RUL ≥30 years; Instructions are generated based on risk levels and issued in tiers. Details are as follows: Low risk: The cloud only stores data, generates daily reports periodically, and pushes them to the management terminal.

[0043] Medium risk: The cloud sends inspection instructions to the edge gateway, which triggers an on-site audible and visual alarm, notifying maintenance personnel to conduct fixed-point inspections.

[0044] High risk: The cloud immediately issues an emergency command, the edge gateway cuts off unnecessary power consumption of the associated monitoring devices, and simultaneously sends multi-level alarms to the management platform, on-site personnel, and regulatory authorities, and feeds back the handling results to the cloud, forming a closed loop of collection-analysis-decision-early warning-feedback.

[0045] Figure 2 This is a schematic diagram of the structure of a bridge condition monitoring system based on edge-cloud collaboration provided in an embodiment of this application; the embodiment of this invention provides a bridge condition monitoring system based on edge-cloud collaboration, including: The bridge site data acquisition module is used to collect real-time multi-source heterogeneous data of the bridge site through acceleration sensors, strain sensors, displacement sensors, temperature and humidity sensors and camera equipment. The bridge site data filtering module is used to filter the real-time multi-source heterogeneous data on the edge computing device at the bridge site using an improved sliding window algorithm to obtain effective multi-source heterogeneous data. The bridge site data processing module is used to perform feature extraction and damage sensitivity index calculation on the effective multi-source heterogeneous data using a damage sensitivity index formula on the edge computing device at the bridge site, thereby obtaining a feature package and a damage sensitivity index; the damage sensitivity index formula is as follows: ； in, It is a damage sensitivity indicator, with a value of 0-1; It is the real-time strain value. It is the maximum allowable strain designed according to the parameters of the bridge design specifications. This is the real-time displacement deviation value. It is the allowable displacement value specified in the standard. It is real-time vibration acceleration. It is the safe vibration acceleration threshold. These are weighting coefficients determined based on the bridge structure type. ; The edge-cloud dynamic data transmission module is used to dynamically allocate bandwidth according to the damage sensitivity index; and to transmit the feature packet and the effective multi-source heterogeneous data to the cloud server through the dynamic bandwidth allocation. The cloud-based data fusion processing module is used to construct a bridge condition assessment model based on multi-scale feature fusion on the cloud server based on historical valid multi-source heterogeneous data, the feature package, and the damage sensitivity index; and to obtain an overall bridge condition score based on the bridge condition assessment model based on multi-scale feature fusion and the valid multi-source heterogeneous data. The input of the bridge condition assessment model based on multi-scale feature fusion is the valid multi-source heterogeneous data, the feature package, and the damage sensitivity index, and the output of the bridge condition assessment model based on multi-scale feature fusion is an overall bridge condition score of 0 to 100. The cloud-integrated data assessment module is used to predict the remaining service life and risk level of the bridge based on historical degradation data, design service life and the overall condition score of the bridge on the cloud server. The bridge data reporting module is used to send multi-level notifications and alarms to the management platform, regulatory agencies, and edge computing devices at the bridge site based on the remaining service life and risk level of the bridge on the cloud server.

[0046] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0047] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A bridge condition monitoring method based on edge-cloud collaboration, characterized in that, include: S1. Real-time multi-source heterogeneous data of the bridge site is collected at the bridge site using acceleration sensors, strain sensors, displacement sensors, temperature and humidity sensors and camera equipment; S2. On the edge computing device at the bridge site, the real-time multi-source heterogeneous data is filtered using an improved sliding window algorithm to obtain effective multi-source heterogeneous data; S3. On the edge computing device at the bridge site, feature extraction and damage sensitivity index calculation are performed on the effective multi-source heterogeneous data using the damage sensitivity index formula to obtain feature packets and damage sensitivity indices; the damage sensitivity index formula is as follows: ； in, It is a damage sensitivity indicator, with a value of 0-1; It is the real-time strain value. It is the maximum allowable strain designed according to the parameters of the bridge design specifications. This is the real-time displacement deviation value. It is the allowable displacement value specified in the standard. It is real-time vibration acceleration. It is the safe vibration acceleration threshold. These are weighting coefficients determined based on the bridge structure type. ; S4. Perform dynamic bandwidth allocation based on the damage sensitivity index; transmit the feature packet and the effective multi-source heterogeneous data to the cloud server through the dynamic bandwidth allocation; S5. On the cloud server, a bridge condition assessment model based on multi-scale feature fusion is constructed based on historical valid multi-source heterogeneous data, the feature package, and the damage sensitivity index; the overall bridge condition score is obtained based on the bridge condition assessment model based on multi-scale feature fusion and the valid multi-source heterogeneous data; the input of the bridge condition assessment model based on multi-scale feature fusion is the valid multi-source heterogeneous data, the feature package, and the damage sensitivity index, and the output of the bridge condition assessment model based on multi-scale feature fusion is the overall bridge condition score from 0 to 100; S6. The remaining service life and risk level of the bridge are predicted by the cloud server based on historical degradation data, design service life and the overall condition score of the bridge. S7. The cloud server sends multi-level notifications and alarms to the management platform, regulatory agencies, and edge computing devices at the bridge site based on the remaining service life and risk level of the bridge.

2. The bridge condition monitoring method based on edge-cloud collaboration according to claim 1, characterized in that, The acceleration sensor, strain sensor, displacement sensor, and temperature and humidity sensor are installed together through pre-embedded parts in the bridge structure and external deployment.

3. The bridge condition monitoring method based on edge-cloud collaboration according to claim 1, characterized in that, The effective multi-source heterogeneous data obtained by filtering the real-time multi-source heterogeneous data on the edge computing device at the bridge site using an improved sliding window algorithm includes: The improved sliding window algorithm is obtained by setting the window length of the sliding window algorithm based on the real-time vibration frequency at the bridge site. The effective multi-source heterogeneous data is obtained by calculating the real-time multi-source heterogeneous data using the improved sliding window algorithm and linear interpolation replacement.

4. The bridge condition monitoring method based on edge-cloud collaboration according to claim 1, characterized in that, On the edge computing device at the bridge site, feature extraction and damage sensitivity index calculation are performed on the effective multi-source heterogeneous data using a damage sensitivity index formula to obtain a feature package and damage sensitivity index, including: The time-domain and frequency-domain features in the effective multi-source heterogeneous data are packaged with the damage sensitivity index to obtain a feature package.

5. The bridge condition monitoring method based on edge-cloud collaboration according to claim 1, characterized in that, The bridge condition assessment model constructed on the cloud server based on historical valid multi-source heterogeneous data, the feature package, and the damage sensitivity index includes: A training dataset is constructed using real bridge calibration data, the feature package, and the historical valid multi-source heterogeneous data. Labels are constructed based on static load test results, standard evaluation levels, expert scores, and the aforementioned damage sensitivity indicators; The bridge state assessment model based on multi-scale feature fusion is constructed using the training dataset, labels, mean squared error loss function, and Adam optimizer.

6. The bridge condition monitoring method based on edge-cloud collaboration according to claim 1, characterized in that, The overall bridge condition score obtained based on the bridge condition assessment model based on multi-scale feature fusion and the effective multi-source heterogeneous data includes: The bridge status assessment model based on multi-scale feature fusion uploads the effective multi-source heterogeneous data to the cloud server; Local features of the effective multi-source heterogeneous data are extracted using multi-scale CNN; Global temporal features of the effective multi-source heterogeneous data are extracted using Transformer; The overall bridge status score is obtained by weighted fusion of the local features and the global temporal features using the damage sensitivity index.

7. The bridge condition monitoring method based on edge-cloud collaboration according to claim 1, characterized in that, The remaining service life and risk level of the bridge are predicted by the cloud server based on historical degradation data, design service life, and the overall condition score of the bridge, including: A correlation prediction formula is constructed based on the overall bridge condition score as the independent variable. The remaining service life of the bridge is calculated using the aforementioned correlation prediction formula, design service life, and historical degradation data. Risk levels are set and assessed based on the remaining service life of the bridge and the overall condition score of the bridge.

8. A bridge condition monitoring system based on edge-cloud collaboration, characterized in that, include: The bridge site data acquisition module is used to collect real-time multi-source heterogeneous data of the bridge site through acceleration sensors, strain sensors, displacement sensors, temperature and humidity sensors and camera equipment. The bridge site data filtering module is used to filter the real-time multi-source heterogeneous data on the edge computing device at the bridge site using an improved sliding window algorithm to obtain effective multi-source heterogeneous data. The bridge site data processing module is used to perform feature extraction and damage sensitivity index calculation on the effective multi-source heterogeneous data using a damage sensitivity index formula on the edge computing device at the bridge site, thereby obtaining a feature package and a damage sensitivity index; the damage sensitivity index formula is as follows: ； in, It is a damage sensitivity indicator, with a value of 0-1; It is the real-time strain value. It is the maximum allowable strain designed according to the parameters of the bridge design specifications. This is the real-time displacement deviation value. It is the allowable displacement value specified in the standard. It is real-time vibration acceleration. It is the safe vibration acceleration threshold. These are weighting coefficients determined based on the bridge structure type. ; The edge-cloud dynamic data transmission module is used to dynamically allocate bandwidth according to the damage sensitivity index; and to transmit the feature packet and the effective multi-source heterogeneous data to the cloud server through the dynamic bandwidth allocation. The cloud-based data fusion processing module is used to construct a bridge condition assessment model based on multi-scale feature fusion on the cloud server based on historical valid multi-source heterogeneous data, the feature package, and the damage sensitivity index; and to obtain an overall bridge condition score based on the bridge condition assessment model based on multi-scale feature fusion and the valid multi-source heterogeneous data. The input of the bridge condition assessment model based on multi-scale feature fusion is the valid multi-source heterogeneous data, the feature package, and the damage sensitivity index, and the output of the bridge condition assessment model based on multi-scale feature fusion is an overall bridge condition score of 0 to 100. The cloud-integrated data assessment module is used to predict the remaining service life and risk level of the bridge based on historical degradation data, design service life and the overall condition score of the bridge on the cloud server. The bridge data reporting module is used to send multi-level notifications and alarms to the management platform, regulatory agencies, and edge computing devices at the bridge site based on the remaining service life and risk level of the bridge on the cloud server.

Images