Large-span bridge body structure health monitoring system based on deformation analysis

Abstract

The invention discloses a large-span bridge body structure health monitoring system based on deformation analysis, and relates to the technical field of large-span bridge body structure health monitoring, the system comprises a multi-modal sensor fusion module, a data processing module, an adaptive algorithm module and a deep learning and predictive maintenance module; and the multi-modal sensor fusion module is used for acquiring data of a plurality of sensors and carrying out cooperative processing. According to the large-span bridge body structure health monitoring system based on deformation analysis, through intelligent fusion of multi-modal sensors, the system can provide more accurate angle measurement under the complex conditions of dynamic loads, environment changes, sensor aging and the like, the measurement precision is improved through sensor redundancy design and a self-adaptive algorithm, and the system is suitable for popularization and application. And meanwhile, the influence of environmental interference and sensor faults on a measurement result is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of long-span bridge beam structure health monitoring, and in particular to a long-span bridge beam structure health monitoring system based on deformation analysis. Background Art

[0002] In the health monitoring of the beam structure of a long-span bridge, angle measurement is one of the important indicators for evaluating the deformation and health status of the structure. However, the existing angle measurement technology faces a series of technical difficulties in practical applications, especially in terms of dynamic loads, environmental changes and sensor accuracy, which restricts its effective application in long-span bridges. First, since long-span bridges are often subjected to complex dynamic loads, including traffic flow, wind loads, and temperature changes, traditional angle measurement equipment, such as tilt sensors and accelerometers, have low measurement accuracy under dynamic loads. It is usually difficult to accurately distinguish the angle changes caused by external dynamic loads from the static deformation of the structure, which makes the measurement results easily interfered by vibration noise. In addition, due to the particularity of the long-span bridge structure, the angle changes of the beam at different positions and at different time points are very important. There may be significant differences in the shapes of the bridge, and a single sensor cannot fully monitor the angle changes of the entire bridge. Secondly, the changes in temperature, humidity, and wind speed in environmental factors cannot be ignored in the impact of angle measurement sensors. In addition, angle sensors in the prior art, such as inclinometers and fiber optic sensors, often experience sensor aging, drift, or damage after long-term operation, further affecting the accuracy and reliability of the measurement. Especially in the long-term monitoring process, due to the drift of the sensor, the measurement data may gradually deviate from the actual value, affecting the accurate assessment of the health status of the bridge structure. Therefore, how to accurately and stably perform angle measurements in the complex environment and dynamic conditions of large-span bridges to avoid accuracy problems and environmental interference in the prior art has become a key technical problem that needs to be urgently solved in the field of bridge health monitoring. Summary of the invention

[0003] 1. Technical issues to be resolved In view of the shortcomings of the prior art, the present invention provides a long-span bridge beam structure health monitoring system based on deformation analysis, which solves the problem of how to accurately and stably perform angle measurement under the complex environment and dynamic conditions of the long-span bridge, avoiding the accuracy problems and environmental interference problems in the prior art.

[0004] (II) Technical solution To achieve the above objectives, the present invention is implemented through the following technical solutions: a long-span bridge beam structure health monitoring system based on deformation analysis, the system includes a multimodal sensor fusion module, a data processing module and an adaptive algorithm module and a deep learning and predictive maintenance module; The multimodal sensor fusion module is used to acquire multiple sensor data and perform collaborative processing; The data processing module is used to perform denoising, feature extraction, pattern recognition, data fusion and error correction on the sensor data; The adaptive algorithm module dynamically adjusts the working mode and weight of the multimodal sensor according to real-time monitoring data to ensure accurate and stable angle measurement under dynamic load or environmental changes.

[0005] The deep learning and predictive maintenance module is used to analyze the health status of the bridge based on the accumulated monitoring data, identify potential faults, and predict future structural deformation trends, so as to provide early warning and maintenance suggestions to the bridge maintenance team; The multimodal sensor fusion module includes a sensor network composed of a tilt sensor, an accelerometer, a fiber optic sensor and a laser ranging technology, which are fused through a redundant design to improve the reliability and anti-interference ability of the system; The redundant design adopts a collaborative backup mechanism between different sensors to ensure that when a single sensor fails or data drifts, other sensors can take over the work in time, thereby ensuring the long-term stability and accuracy of the system; The data processing module performs refined processing on the sensor data in stages through a three-cycle processing flow to obtain fully optimized high-quality data.

[0006] Preferably, the data processing module includes a data preprocessing unit, a feature extraction unit, a pattern recognition unit and a data fusion unit; The data preprocessing unit uses low-pass filtering, Kalman filtering or adaptive noise estimation filtering algorithm to denoise the raw data collected by the sensor, and unifies the data of different sensors through standardization and normalization methods to eliminate the influence of different dimensions; The feature extraction unit conducts in-depth mining of the denoised sensor data through time domain analysis and frequency domain analysis methods to extract key feature values ​​of the bridge angle deformation, including the maximum value, mean value, change rate and vibration frequency of the deformation; The pattern recognition unit classifies and recognizes patterns of the extracted feature values ​​through a machine learning algorithm based on a support vector machine (SVM) or a convolutional neural network (CNN), thereby accurately determining the health status of the bridge; The data fusion unit dynamically fuses the data from different sensors based on a weighted average algorithm, a Kalman filter algorithm or a multimodal fusion technology of deep learning, so as to obtain a measurement result with higher comprehensive accuracy.

[0007] The data processing module responds to environmental changes (such as temperature, humidity and dynamic load changes) in real time during each stage of processing, and improves the stability and reliability of the processing results by adaptively adjusting the parameters of each module.

[0008] Preferably, the data preprocessing unit includes a filtering algorithm based on adaptive noise estimation, which is used to adjust the filtering parameters in real time according to environmental changes, so as to adapt to the data noise conditions in different environments, wherein the environmental changes include factors that may affect the sensor data, such as temperature changes, humidity changes, and load changes; The filtering algorithm dynamically adjusts the filter's cutoff frequency, bandwidth or gain by real-time monitoring of environmental parameters, such as temperature and humidity, thereby effectively suppressing data noise caused by environmental fluctuations; in addition, the filtering algorithm further includes an outlier detection mechanism for identifying and eliminating sudden noise or abnormal signals in the original data; the identification of outliers is based on the historical change pattern and current real-time change trend of sensor data, and is achieved by comparing the data fluctuation amplitude, frequency and other characteristic values ​​within a specific time window; after eliminating outliers, the system will automatically interpolate missing data according to the historical data compensation algorithm to ensure the continuity and reliability of the pre-processed data, thereby providing high-quality input for subsequent feature extraction and pattern recognition.

[0009] Preferably, the feature extraction unit includes a signal processing method of wavelet transform and Fourier transform, performs a joint analysis of the sensor data in the time domain and the frequency domain, and extracts the maximum value, mean value, and rate of change of the bridge angle change; the wavelet transform is used to analyze the local feature changes in the sensor data, and by selecting appropriate wavelet basis functions, including Haar wavelet and Daubechies wavelet, the signal is decomposed into multiple scale components to extract the angular deformation characteristics of the bridge structure at different frequencies; the Fourier transform is used to identify the frequency domain characteristics of the sensor data, and by analyzing the frequency spectrum, the main frequency components and amplitude changes in the signal are extracted to determine whether the bridge is affected by periodic vibration or abnormal loads; these extracted feature values ​​will be used as input to the pattern recognition unit to provide basic data support for the classification of the bridge health status and abnormality detection.

[0010] Preferably, the pattern recognition unit adopts a deep learning algorithm to perform pattern learning on the bridge angle change through a convolutional neural network (CNN) or a recurrent neural network (RNN) to identify the structural deformation mode under different working conditions; The CNN algorithm is used to extract features from the input two-dimensional sensor data matrix, and extract local features such as the rate of change and mutation point of the bridge angle deformation through the operation of multi-layer convolution kernels; The RNN algorithm is used to model time series data, capture the time dependence of bridge deformation characteristics, and identify the angle change pattern under dynamic loads.

[0011] Preferably, the adaptive algorithm module includes a dynamic weighting algorithm, a temperature compensation mechanism and a self-calibration mechanism; The dynamic weighting algorithm automatically adjusts the working modes and weights of multiple sensors based on the real-time dynamic state; that is, the dynamic weighting algorithm assigns weights to the output signals of the sensors according to the vibration frequency, vibration amplitude, and sensor signal stability, and dynamically reduces the weight of the sensor when the sensor has a large signal noise under dynamic load; The temperature compensation mechanism dynamically adjusts the working mode of the sensor by real-time monitoring of temperature and humidity changes to cope with the impact of temperature changes on measurement accuracy; that is, the temperature compensation mechanism corrects the measured value of the sensor through the environmental parameter correction model based on the real-time monitored temperature and humidity changes; the correction model is dynamically updated according to the different characteristics of the sensor type and the working environment, including the material expansion coefficient and the temperature and humidity coupling effect parameters; The self-calibration mechanism automatically calibrates the sensor by comparing historical data with real-time measurements to ensure long-term stability; that is, the self-calibration mechanism automatically detects and corrects sensor drift, data deviation or error accumulation by comparing real-time measurement data with historical data or design parameters of the bridge; Among them, during drift detection, the self-calibration mechanism uses known reference points or redundant sensor data within the system for comparison, automatically calculates the offset and corrects the system output, thereby ensuring accuracy and stability in long-term operation.

[0012] Preferably, the dynamic weighting algorithm automatically selects the sensor that best suits the current environment and dynamic state according to the vibration frequency and amplitude changes in the real-time state of the sensor to optimize the measurement accuracy; The dynamic weighting algorithm adopts a method based on multi-objective optimization, and dynamically adjusts the weight of each sensor through a preset objective function combined with real-time environmental data; the objective function includes but is not limited to comprehensive indicators of measurement accuracy, data stability, response speed and sensor health status; during the dynamic weighting process, the system first evaluates the data quality of all sensors in real time to obtain evaluation results, including evaluation of key parameters such as signal fluctuations, noise amplitude, and measurement errors; based on the evaluation results, the system ensures the accuracy of the fusion results by assigning high weights to high-quality data.

[0013] Preferably, the self-calibration mechanism automatically detects and corrects sensor drift or error by comparing known standard data with real-time measurement data to ensure the accuracy of the measurement results; the self-calibration mechanism is implemented by the following steps: Step 1: Establish a benchmark database based on the historical health data of the bridge, and record the angle change range, dynamic response mode and environmental parameter influence rules of the bridge under normal conditions; Step 2: Regularly obtain real-time measurement data and compare it with historical data in the benchmark database to identify potential drift or deviation; Step 3: When it is detected that the sensor data deviates from the preset threshold, the calibration mode is started to correct the current measurement value through redundant sensor data or known environmental parameters; Step 4: The system records all calibration operations and updates the historical database to provide a more accurate reference for subsequent calibration and prediction.

[0014] Preferably, the data processing module includes a fault detection and diagnosis unit for identifying and processing abnormal states of sensors; the fault detection and diagnosis unit determines whether the sensor has faults or performance degradation by analyzing the changing trends and signal characteristics of real-time monitoring data; when an abnormality is detected in the sensor signal, the system will issue an alarm; at the same time, the fault detection unit will automatically enable the backup sensor in the redundant design to ensure uninterrupted monitoring tasks; the fault detection and diagnosis unit uses a multi-parameter analysis method to diagnose the cause of the fault, including environmental interference, electromagnetic interference, and sensor aging; after the diagnosis is completed, the system will automatically record the fault information and generate a report, suggesting that maintenance personnel take corresponding maintenance or replacement measures; through this fault detection and diagnosis mechanism, the system can quickly respond to various emergencies and ensure the stability and continuity of the bridge monitoring system.

[0015] Preferably, the data processing module processes the sensor data through a three-cycle processing flow, and the three-cycle processing includes the following stages: The first cycle: basic data preprocessing and noise filtering, removing noise from the original data and smoothing the signal; that is, through low-pass filtering or Kalman filtering in the filtering algorithm, remove high-frequency noise in the original data, smooth the signal, remove outliers and interpolate missing data; The second cycle: feature extraction and pattern recognition, extracting the bridge angle change features from the denoised data and performing pattern recognition through machine learning algorithms; that is, extracting key feature values ​​from the denoised data, including maximum value, mean value, change rate and frequency domain characteristics, and performing pattern classification on the feature values ​​through machine learning algorithms; The third cycle: data fusion and error correction, through weighted data fusion and error correction algorithm, optimize the final angle measurement result; that is, through dynamic weighting algorithm and Kalman filter algorithm, weighted fusion of different sensor data, and combined with temperature compensation and multi-dimensional error compensation method to optimize the measurement result; The data fusion and error correction algorithm includes a multi-dimensional error compensation algorithm, which comprehensively corrects the sensor data according to environmental changes and load changes.

[0016] The system optimizes the feature extraction and pattern recognition steps in the data processing process by introducing environmental monitoring sensor data, wherein the environmental monitoring sensor data includes temperature, humidity, and wind speed, so as to improve the measurement accuracy of the system in complex environments.

[0017] (III) Beneficial effects The present invention provides a long-span bridge beam structure health monitoring system based on deformation analysis. It has the following beneficial effects: (I) This large-span bridge beam structure health monitoring system based on deformation analysis, through the intelligent fusion of multi-modal sensors, can provide more accurate angle measurement under complex conditions such as dynamic loads, environmental changes and sensor aging. The sensor redundancy design and adaptive algorithm improve the measurement accuracy, while reducing the impact of environmental interference and sensor failure on the measurement results; the system can automatically adjust the working mode according to real-time data and select the sensor that best suits the current environment and dynamic state, greatly improving the intelligence level of the system. At the same time, the intelligent self-calibration mechanism and dynamic error compensation ensure the long-term stable operation of the system; it can flexibly respond to different bridge types, environmental changes and load conditions, has strong scalability, and is suitable for bridge health monitoring tasks in a variety of complex environments.

[0018] (II) The large-span bridge beam structure health monitoring system based on deformation analysis not only ensures the accuracy and stability of the data through three-cycle processing, but also enables the system to perform angle measurement more accurately when facing complex environments and dynamic loads. That is, under complex conditions such as environmental changes and dynamic loads, the system can ensure the accuracy of the data through adaptive adjustments; the system also uses adaptive algorithms to perform intelligent adjustments between different sensors, and can dynamically adapt to factors such as environmental changes and sensor state changes, thereby avoiding errors caused by environmental changes; and with the help of the introduced machine learning and deep learning algorithms, the system can learn and adapt to the characteristic changes of the bridge structure, thereby realizing predictive maintenance and early fault warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of the framework of the long-span bridge beam structure health monitoring system based on deformation analysis of the present invention; Figure 2 It is a schematic diagram of the flow of the self-calibration mechanism of the present invention. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] See also Figure 1 and Figure 2 ,The present invention provides a technical solution: a long-span bridge beam structure health monitoring system based on deformation analysis, the system includes a multi-modal sensor fusion module, a data processing module and an adaptive algorithm module as well as a deep learning and predictive maintenance module; The multimodal sensor fusion module is used to acquire multiple sensor data and perform collaborative processing; The data processing module is used to perform denoising, feature extraction, pattern recognition, data fusion and error correction on sensor data; The adaptive algorithm module dynamically adjusts the working mode and weight of the multimodal sensor according to real-time monitoring data to ensure accurate and stable angle measurement under dynamic loads or environmental changes.

[0022] Deep learning and predictive maintenance module, which is used to analyze the health status of bridges based on accumulated monitoring data, identify potential failures, and predict future structural deformation trends, so as to provide early warning and maintenance suggestions to the bridge maintenance team; The multimodal sensor fusion module includes a sensor network composed of tilt sensors, accelerometers, fiber optic sensors, and laser ranging technology, which are integrated through redundant design to improve the reliability and anti-interference ability of the system. The redundant design uses a collaborative backup mechanism between different sensors to ensure that when a single sensor fails or data drifts, other sensors can take over the work in time, thereby ensuring the long-term stability and accuracy of the system; It should be further explained that in the health monitoring system of long-span bridges, sensors may fail or data drift due to environmental changes, aging or external interference. To this end, the redundant design establishes a backup relationship between different sensors through a collaborative backup mechanism to ensure the stability and measurement accuracy of the system; The collaborative backup mechanism of the redundant design is based on the grouping and collaborative work of tilt sensors, accelerometers, fiber optic sensors and laser rangefinders. The tilt sensors are used to monitor the static component of the bridge angle change, which is suitable for long-term static monitoring. The accelerometers are used to monitor the dynamic response and high-frequency vibration of the bridge, which is suitable for capturing the angle changes under dynamic loads. The fiber optic sensors are used to monitor the subtle deformation of the bridge with high precision, and have anti-electromagnetic interference characteristics, which is suitable for long-term monitoring of key points. The laser rangefinder is used to measure the distance changes between key points of the bridge, which is suitable for real-time high-precision monitoring. The above sensors are divided into main sensor and backup sensor groups according to the monitoring tasks, and work in coordination through the redundancy mechanism. The redundancy mechanism is designed as follows: First, the system is initialized. When it starts, the tilt sensor is selected as the default main sensor, and the accelerometer is selected as the auxiliary sensor for dynamic monitoring; the fiber optic sensor and laser rangefinder are used as backup sensors to handle abnormal situations on standby; Then the main sensor is used for operation and real-time monitoring. The main sensor tilt sensor continuously collects static angle data of the bridge, and filters and smoothes it through the data processing module to obtain real-time angle information. At the same time, the auxiliary sensor accelerometer collects dynamic angle changes in real time to supplement the shortcomings of the tilt sensor in high-frequency response. The system then uses the status monitoring unit to conduct real-time evaluation of the working status of each sensor, including data output stability, signal strength and environmental adaptability. It monitors the fluctuation range and drift trend of the data, detects the signal noise level and sampling frequency, and evaluates the impact of temperature and humidity on the sensor. When the signal output of the main sensor: the tilt sensor is unstable or fails (such as signal interruption, drift beyond the threshold), the system triggers the following steps: Step a1: Activate the fiber optic sensor as the new main sensor to take over the monitoring task of the static angle of the bridge; Step a2: Record the failure time and affected area of ​​the sensor, and notify the maintenance team at the same time; Step a3: Switch the tilt sensor to the standby mode for status diagnosis and repair; In the event of a fault, the fiber optic sensor takes over the monitoring task of the tilt sensor, and the accelerometer continues to provide high-frequency vibration data as a dynamic auxiliary sensor; at the same time, the laser rangefinder can be used to verify the static deformation between key points of the bridge, further improving data reliability; After the sensor is switched, the system will perform data fusion and error correction according to the following steps: Step b1: Initialize the data of the replacement sensor (fiber optic sensor) and compare it with the historical data to ensure data continuity; Step b2: Use a dynamic weighting algorithm to weight all sensor data, giving priority to signals from high-precision sensors; Step b3: Apply the Kalman filter algorithm to correct the differences between different sensor data and eliminate the jump effect during the switching process.

[0023] In actual application scenarios, for scenario 1: dynamic load monitoring: the bridge is affected by wind load and traffic flow at the same time, resulting in a significant increase in vibration frequency, and the tilt sensor cannot output stably due to increased noise; the system automatically enables the accelerometer and fiber optic sensor to work together, the accelerometer provides high-frequency vibration angle changes in real time, and the fiber optic sensor is responsible for providing high-precision angle measurement; For scenario 2: Environmental interference processing: During the bridge construction, the tilt sensor fails due to electromagnetic interference; the system detects the loss of the tilt sensor signal and automatically enables the fiber optic sensor and laser rangefinder to supplement the monitoring data to ensure the continuity of bridge deformation monitoring; Among them, the redundancy switching time: less than 10 milliseconds, the data fusion accuracy: higher than 99.9%, the sensor fault detection threshold: signal drift exceeds 3σ or there is no output for 3 consecutive seconds, and the calibration frequency: automatic calibration once after each switch; Through the redundant design of the above-mentioned collaborative backup mechanism, the system realizes continuous, stable and high-precision monitoring of the health status of the bridge. It can operate effectively even in complex environments and sensor failure conditions, ensuring the reliability and robustness of the monitoring system.

[0024] The data processing module performs phased and refined processing on the sensor data through a three-cycle processing flow to obtain fully optimized high-quality data.

[0025] The data processing module includes a data preprocessing unit, a feature extraction unit, a pattern recognition unit, and a data fusion unit: The data preprocessing unit uses a filtering algorithm to denoise and smooth the sensor data; the feature extraction unit extracts the characteristic values ​​of the bridge angle deformation from the denoised data through time domain analysis and frequency domain analysis; the pattern recognition unit classifies the extracted characteristic values ​​through a machine learning algorithm to identify the health status of the bridge; the data fusion unit uses a weighted average algorithm or a Kalman filter algorithm to fuse the data from different sensors; The data preprocessing unit uses low-pass filtering, Kalman filtering or adaptive noise estimation filtering algorithm to denoise the raw data collected by the sensor, and unifies the data of different sensors through standardization and normalization methods to eliminate the influence of different dimensions; The feature extraction unit conducts in-depth mining of the denoised sensor data through time domain analysis and frequency domain analysis methods to extract key feature values ​​of the bridge angle deformation, including the maximum value, mean value, change rate and vibration frequency of the deformation; The pattern recognition unit classifies and recognizes the extracted feature values ​​through machine learning algorithms based on support vector machines (SVM) or convolutional neural networks (CNN), thereby accurately determining the health status of the bridge; The data fusion unit dynamically fuses data from different sensors based on weighted average algorithm, Kalman filter algorithm or multimodal fusion technology of deep learning to obtain measurement results with higher comprehensive accuracy.

[0026] The data processing module responds to environmental changes (such as temperature, humidity and dynamic load changes) in real time during each stage of processing, and improves the stability and reliability of the processing results by adaptively adjusting the parameters of each module.

[0027] The data preprocessing unit includes a filtering algorithm based on adaptive noise estimation, which is used to adjust the filtering parameters in real time according to environmental changes, so as to adapt to the data noise conditions in different environments, wherein the environmental changes include factors that may affect the sensor data, such as temperature changes, humidity changes, and load changes; The filtering algorithm monitors environmental parameters such as temperature and humidity in real time, and dynamically adjusts the filter's cutoff frequency, bandwidth or gain, thereby effectively suppressing data noise caused by environmental fluctuations. In addition, the filtering algorithm further includes an outlier detection mechanism for identifying and eliminating sudden noise or abnormal signals in the original data. The identification of outliers is based on the historical change pattern and current real-time change trend of sensor data, and is achieved by comparing the data fluctuation amplitude, frequency and other characteristic values ​​within a specific time window. After eliminating outliers, the system will automatically interpolate missing data based on the historical data compensation algorithm to ensure the continuity and reliability of the pre-processed data, thereby providing high-quality input for subsequent feature extraction and pattern recognition.

[0028] The feature extraction unit includes the signal processing methods of wavelet transform and Fourier transform, which jointly analyzes the sensor data in the time domain and frequency domain to extract the maximum value, mean value and change rate of the bridge angle change; Wavelet transform is used to analyze local feature changes in sensor data. By selecting appropriate wavelet basis functions, including Haar wavelet and Daubechies wavelet, the signal is decomposed into multiple scale components to extract the angular deformation characteristics of the bridge structure at different frequencies. Fourier transform is used to identify the frequency domain characteristics of sensor data. By analyzing the spectrum, the main frequency components and amplitude changes in the signal are extracted to determine whether the bridge is affected by periodic vibration or abnormal load. These extracted feature values ​​will be used as input to the pattern recognition unit to provide basic data support for the classification of bridge health status and anomaly detection.

[0029] The pattern recognition unit adopts a deep learning algorithm to learn the pattern of bridge angle changes through convolutional neural network (CNN) or recurrent neural network (RNN) to identify the structural deformation patterns under different working conditions; the CNN algorithm is used to extract features from the input two-dimensional sensor data matrix, and extract local features such as the change rate and mutation points of the bridge angle deformation through multi-layer convolution kernel operations; the RNN algorithm is used to model time series data, capture the time dependence of bridge deformation characteristics, and identify the angle change law under dynamic loads.

[0030] Through the comprehensive application of the above algorithms, the system can distinguish the normal deformation mode and potential abnormal state of the bridge, and realize early fault warning and health assessment.

[0031] The adaptive algorithm module includes a dynamic weighting algorithm, a temperature compensation mechanism, and a self-calibration mechanism; The dynamic weighting algorithm automatically adjusts the working modes and weights of multiple sensors based on real-time dynamic status; the dynamic weighting algorithm assigns weights to the output signals of the sensors according to the vibration frequency, vibration amplitude, and sensor signal stability. When the sensor has large signal noise under dynamic load, the weight of the sensor is dynamically reduced; The temperature compensation mechanism monitors temperature and humidity changes in real time and dynamically adjusts the sensor's operating mode to cope with the impact of temperature changes on measurement accuracy; The self-calibration mechanism automatically calibrates the sensor by comparing historical data with real-time measurements to ensure long-term stability. The temperature compensation mechanism corrects the sensor's measured value through the environmental parameter correction model based on the real-time monitored temperature and humidity changes; the correction model is dynamically updated according to the different characteristics of the sensor type and working environment, including material expansion coefficient and temperature and humidity coupling parameters; The self-calibration mechanism automatically detects and corrects sensor drift, data deviation or error accumulation by comparing real-time measurement data with historical data or design parameters of the bridge; During drift detection, the self-calibration mechanism uses known reference points or redundant sensor data within the system for comparison, automatically calculates the offset and corrects the system output, thereby ensuring accuracy and stability in long-term operation.

[0032] The dynamic weighting algorithm automatically selects the sensor that best suits the current environment and dynamic state according to the vibration frequency and amplitude changes in the real-time state of the sensor to optimize the measurement accuracy; The dynamic weighting algorithm adopts a multi-objective optimization-based method to dynamically adjust the weight of each sensor through a preset objective function combined with real-time environmental data; Objective functions include but are not limited to comprehensive indicators of measurement accuracy, data stability, response speed, and sensor health status; During the dynamic weighting process, the system first evaluates the data quality of all sensors in real time and obtains the evaluation results, including key parameters such as signal fluctuation, noise amplitude, and measurement error; Based on the evaluation results, the system ensures the accuracy of the fusion results by giving high weights to high-quality data. For example, when the bridge is in a state of severe vibration, the dynamic weighting algorithm will give priority to the accelerometer data that is sensitive to high-frequency vibration, while when the bridge is static or slowly deforming, it will rely more on the high-precision fiber optic sensor signal. Through this real-time optimized sensor weight distribution method, the dynamic weighting algorithm significantly improves the measurement accuracy of the system in complex environments and under various load conditions.

[0033] It is necessary to further explain that the dynamic weighting algorithm automatically adjusts the working mode and weight of multiple sensors based on the real-time dynamic state, and is used in the health monitoring of the beam structure of a large-span bridge: In the health monitoring system of the beam structure of a large-span bridge, the quality and stability of the data collected by the sensor may change because the environment in which the bridge is located is complex and is often affected by different types of dynamic loads (such as traffic loads, wind loads, etc.) and environmental changes (such as temperature, humidity, etc.); in order to ensure the reliability of the health monitoring system in a complex environment, a dynamic weighting algorithm is used to automatically adjust the working mode and weight of multiple sensors, thereby optimizing the data fusion process and ensuring that the final monitoring results are accurate and stable; In this embodiment, the system includes four main sensors, including a tilt sensor, an accelerometer, a fiber optic sensor, and a laser rangefinder. The tilt sensor is used for static angle measurement, the accelerometer is used for dynamic vibration measurement, the fiber optic sensor is used for high-precision deformation monitoring, and the laser rangefinder is used for distance change monitoring between key points. The data of all sensors will be fused and analyzed by the data processing module. The dynamic weighting algorithm will automatically adjust the working mode and weight of each sensor based on the real-time dynamic state to ensure the accuracy and stability of the sensor data under complex loads or environmental changes. The goal of the dynamic weighting algorithm is to optimize data fusion and the final angle measurement results by dynamically assigning weights to each sensor based on measurement quality, environmental changes, and dynamic load conditions by evaluating the sensor status in real time. The system will regularly monitor the stability and quality of each sensor output signal, including evaluating the reliability of each sensor data by calculating the noise amplitude, fluctuation range, and change rate of the signal. The system also monitors the temperature, humidity, and wind speed of environmental conditions in real time, as these factors may affect the measurement accuracy of the sensor. For example, temperature changes can affect the accuracy of fiber optic sensors, while strong winds may affect the stability of laser rangefinders. Weighted adjustment strategies include: According to real-time monitoring data, the dynamic weighting algorithm automatically adjusts the weight of each sensor to optimize the data fusion result; the main adjustment strategies include signal stability weight adjustment, environmental adaptability adjustment, and dynamic load impact adjustment; During signal stability weight adjustment, when a sensor signal is noisy or unstable, the algorithm will reduce the weight of that sensor and increase the weight of other signal stable sensors; for example, if the accelerometer is disturbed by short-term high-amplitude noise, the algorithm will reduce the weight of the accelerometer and increase the weight of the fiber optic sensor; During environmental adaptation, when environmental parameters change significantly, the algorithm adjusts the weights of sensors accordingly; for example, in hot weather, the weight of temperature-sensitive fiber optic sensors may be reduced, while the weight of more stable tilt sensors may be increased. During the dynamic load impact adjustment period, under the action of dynamic loads (such as when the bridge carries traffic flow or wind loads), the system will adjust according to the response sensitivity of each sensor to different loads; at this time, the weight of the accelerometer will increase because it can sensitively capture high-frequency dynamic vibrations, while the weight of the tilt sensor may be appropriately reduced; on the basis of the above, the algorithm has an adaptive mechanism; that is, as the bridge health monitoring data accumulates, the system will continuously optimize the weight allocation strategy and gradually improve the system's adaptability to complex environments and different load conditions.

[0034] The implementation is as follows in combination with specific application scenarios: Scenario 3: The bridge is under dynamic load, that is, the combined effect of traffic flow and wind load; assuming that a long-span bridge is under the influence of a large number of vehicles passing and strong winds, the system will be affected by dynamic loads; at this time, the system uses a dynamic weighting algorithm to automatically adjust the weights of each sensor; due to the high-frequency vibrations generated by the bridge under the action of strong winds and vehicle traffic, the accelerometer will respond to these changes first and provide high-precision dynamic angle change data, and the accelerometer is responsible for monitoring this signal; although the tilt sensor has a good response to static angle changes, its accuracy may decrease under dynamic vibration conditions, so the system will automatically reduce its weight; The process for weighted adjustments includes: Increased weight of accelerometer: Based on the high-frequency vibration information captured by the accelerometer, the system automatically increases its weight to ensure the accuracy of dynamic data; Fiber optic sensors and laser rangefinders: Since wind loads may cause slight deformations on the surface of the structure, fiber optic sensors and laser rangefinders will enhance monitoring and ensure that the system can accurately perceive the slight deformations of the bridge; Environmental parameter monitoring: When the system detects a sudden increase in wind speed and a large temperature change, it will adjust the weight of the sensor in a timely manner according to the change in wind speed, reduce the weight of the sensor that is more affected by wind speed, and increase the weight of the more stable accelerometer and fiber optic sensor; Scenario 4: The bridge is under static load, i.e., maintenance or no-load conditions. When the bridge is under maintenance or there is no traffic flow, the bridge will be in a static load state with a low vibration frequency. The system will also automatically adjust the weight of the sensor. At this time, the tilt sensor can more accurately monitor the angular deformation of the bridge, and the system will increase the weight of the tilt sensor. Since there is no dynamic load, the system will appropriately reduce the weight of the accelerometer because it has a weak response to static angle changes. The process for weighted adjustments includes: The weight of the tilt sensor is increased: In the absence of dynamic load, the signal of the tilt sensor is relatively stable, and the system increases its weight to ensure accuracy; Accelerometer and fiber optic sensor weight reduction: The measurement accuracy of these two sensors may be low in static state, so the system will reduce their weight to improve static angle measurement accuracy.

[0035] As bridge health monitoring data continues to accumulate, the system will optimize the weighting algorithm through machine learning so that the performance of each sensor under different environmental conditions can be fully learned and utilized; through the combination of deep learning and adaptive algorithms, the system can predict the optimal weight configuration under different environmental conditions, thereby improving the performance and accuracy of the system; Through the dynamic weighting algorithm, the system can automatically adjust the working modes and weights of different sensors in real time, ensuring that the health monitoring system can continue to provide high-precision data under various environmental changes and dynamic load conditions; this mechanism can effectively cope with complex environments, dynamic loads and sensor failures in the health monitoring of large-span bridge beam structures, and ensure the reliability and stability of the monitoring results.

[0036] The self-calibration mechanism automatically detects and corrects sensor drift or errors by comparing known standard data with real-time measurement data to ensure the accuracy of measurement results; The self-calibration mechanism is implemented through the following steps: Step 1: Establish a benchmark database based on the historical health data of the bridge, and record the angle change range, dynamic response mode and environmental parameter influence rules of the bridge under normal conditions; Step 2: Regularly obtain real-time measurement data and compare it with historical data in the benchmark database to identify potential drift or deviation; Step 3: When it is detected that the sensor data deviates from the preset threshold, the calibration mode is started to correct the current measurement value through redundant sensor data or known environmental parameters; Step 4: The system records all calibration operations and updates the historical database to provide a more accurate reference for subsequent calibration and prediction.

[0037] The data processing module also includes a fault detection and diagnosis unit, which is used to automatically detect sensor failures or performance degradation through intelligent algorithms, and automatically switch to backup sensors when a fault is detected, and trigger an alarm; the fault detection and diagnosis unit is used to identify and process abnormal states of sensors; the fault detection and diagnosis unit determines whether the sensor has a fault or performance degradation by analyzing the changing trend and signal characteristics of real-time monitoring data; when an abnormality is detected in the sensor signal, such as no signal output for a long time, abnormal fluctuations, or drift exceeding a preset threshold, the system will issue an alarm prompt; at the same time, the fault detection unit will automatically enable the backup sensor in the redundant design to ensure uninterrupted monitoring tasks; the unit uses a multi-parameter analysis method to diagnose the cause of the fault, including environmental interference, electromagnetic interference, and sensor aging; after the diagnosis is completed, the system will automatically record the fault information and generate a report, suggesting that maintenance personnel take corresponding maintenance or replacement measures; through this fault detection and diagnosis mechanism, the system can quickly respond to various emergencies and ensure the stability and continuity of the bridge monitoring system.

[0038] The data processing module processes the sensor data through a three-cycle processing flow, which includes the following stages: the first cycle: basic data preprocessing and noise filtering, removing noise from the original data and smoothing the signal; that is, through low-pass filtering or Kalman filtering in the filtering algorithm, high-frequency noise in the original data is removed, and signal smoothing is performed, while outliers are eliminated and missing data is interpolated; the second cycle: feature extraction and pattern recognition, extracting the bridge angle change characteristics from the denoised data, and performing pattern recognition through a machine learning algorithm; that is, extracting key eigenvalues ​​from the denoised data, including maximum value, mean value, change rate and frequency domain characteristics, and classifying the eigenvalues ​​through a machine learning algorithm; the third cycle: data fusion and error correction, optimizing the final angle measurement result through a weighted data fusion and error correction algorithm; that is, weighted fusion of different sensor data is performed through a dynamic weighting algorithm and a Kalman filtering algorithm, and optimizing the measurement result in combination with temperature compensation and multi-dimensional error compensation methods; among them, the data fusion and error correction algorithm includes a multi-dimensional error compensation algorithm, which comprehensively corrects the sensor data according to environmental changes and load changes.

[0039] The system optimizes the feature extraction and pattern recognition steps in the data processing process by introducing environmental monitoring sensor data, where the environmental monitoring sensor data includes temperature, humidity, and wind speed, in order to improve the measurement accuracy of the system in complex environments. Through this multi-stage cyclic processing method, the system can achieve high-precision and high-stability angle measurement results, and significantly improve the measurement robustness in complex environments.

[0040] Through the intelligent fusion of multimodal sensors, the system can provide more accurate angle measurements under complex conditions such as dynamic loads, environmental changes, and sensor aging. Sensor redundancy design and adaptive algorithms improve measurement accuracy while reducing the impact of environmental interference and sensor failures on measurement results; the system can automatically adjust the working mode according to real-time data and select the sensor that best suits the current environment and dynamic state, greatly improving the intelligence level of the system. At the same time, the intelligent self-calibration mechanism and dynamic error compensation ensure the long-term stable operation of the system; it can flexibly respond to different bridge types, environmental changes, and load conditions, has strong scalability, and is suitable for bridge health monitoring tasks in a variety of complex environments; The three-cycle processing not only ensures the accuracy and stability of the data, but also enables the system to measure angles more accurately when facing complex environments and dynamic loads. That is, under complex conditions such as environmental changes and dynamic loads, the system can ensure the accuracy of the data through adaptive adjustments. The system also uses adaptive algorithms to perform intelligent adjustments between different sensors, and can dynamically adapt to factors such as environmental changes and sensor state changes, thereby avoiding errors caused by environmental changes. The introduced machine learning and deep learning algorithms enable the system to learn and adapt to the characteristic changes of the bridge structure, thereby achieving predictive maintenance and early fault warning.

[0041] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0042] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A long-span bridge beam structure health monitoring system based on deformation analysis, characterized in that: The system includes a multimodal sensor fusion module, a data processing module, an adaptive algorithm module, and a deep learning and predictive maintenance module; The multimodal sensor fusion module is used to acquire multiple sensor data and perform collaborative processing; The data processing module is used to perform denoising, feature extraction, pattern recognition, data fusion and error correction on the sensor data; the data processing module processes the sensor data through a three-cycle processing flow, wherein the first cycle is basic data preprocessing and noise filtering, the second cycle is feature extraction and pattern recognition, and the third cycle is data fusion and error correction; The adaptive algorithm module dynamically adjusts the working mode and weight of the multimodal sensor according to the real-time monitoring data; the adaptive algorithm module includes a dynamic weighting algorithm, a temperature compensation mechanism and a self-calibration mechanism; the dynamic weighting algorithm automatically adjusts the working mode and weight of each sensor according to the real-time state of the sensor, wherein the real-time state includes vibration frequency, amplitude change and signal stability; The deep learning and predictive maintenance module is used to analyze the health status of the bridge based on the accumulated monitoring data, identify potential faults, and predict future structural deformation trends, so as to provide early warning and maintenance suggestions to the bridge maintenance team; The multimodal sensor fusion module includes a sensor network composed of a tilt sensor, an accelerometer, a fiber optic sensor and a laser rangefinder, which are fused through a redundant design to improve the reliability and anti-interference ability of the system; The redundant design adopts a collaborative backup mechanism between different sensors; The data processing module performs refined processing on the sensor data in stages through a three-cycle processing flow.

2. According to claim 1, a long-span bridge beam structure health monitoring system based on deformation analysis is characterized in that: The data processing module includes a data preprocessing unit, a feature extraction unit, a pattern recognition unit and a data fusion unit; the data preprocessing unit uses a filtering algorithm to denoise and smooth the sensor data; the feature extraction unit extracts the characteristic value of the bridge angle deformation from the denoised data through time domain analysis and frequency domain analysis; the pattern recognition unit classifies the extracted characteristic values ​​through a machine learning algorithm to identify the health status of the bridge; the data fusion unit uses a weighted average algorithm or a Kalman filter algorithm to fuse data from different sensors.

3. The long-span bridge beam structure health monitoring system based on deformation analysis according to claim 2 is characterized by: The data preprocessing unit includes a filtering algorithm based on adaptive noise estimation, which is used to adjust filtering parameters in real time according to environmental changes to adapt to data noise conditions in different environments, wherein environmental changes include temperature changes, humidity changes, and load changes.

4. The long-span bridge beam structure health monitoring system based on deformation analysis according to claim 3 is characterized by: The feature extraction unit includes a signal processing method of wavelet transform and Fourier transform, which performs a joint analysis of the sensor data in the time domain and the frequency domain to extract the maximum value, mean value and change rate of the bridge angle change; the wavelet transform is used to analyze the local feature changes in the sensor data, and by selecting a suitable wavelet basis function, including Haar wavelet and Daubechies wavelet, the signal is decomposed into multiple scale components to extract the angle deformation characteristics of the bridge structure at different frequencies; the Fourier transform is used to identify the frequency domain characteristics of the sensor data, and by analyzing the frequency spectrum, the main frequency components and amplitude changes in the signal are extracted to determine whether the bridge is affected by periodic vibration or abnormal load.

5. The long-span bridge beam structure health monitoring system based on deformation analysis according to claim 4 is characterized in that: The pattern recognition unit adopts a deep learning algorithm to perform pattern learning on the angle change of the bridge through a convolutional neural network (CNN) or a recurrent neural network (RNN) to identify the structural deformation patterns under different working conditions; wherein the convolutional neural network algorithm is used to extract features from the input two-dimensional sensor data matrix and extract local features through multi-layer convolution kernel operations; the recurrent neural network algorithm is used to model time series data, capture the time dependence of the bridge deformation characteristics, and identify the angle change law under the action of dynamic loads.

6. The long-span bridge beam structure health monitoring system based on deformation analysis according to claim 5 is characterized by: The dynamic weighting algorithm assigns weights to the output signals of the sensor according to the vibration frequency, vibration amplitude, and sensor signal stability. When the sensor has large signal noise under dynamic load, the weight of the sensor is dynamically reduced. The temperature compensation mechanism dynamically adjusts the working mode of the sensor by real-time monitoring of temperature and humidity changes to cope with the impact of temperature changes on measurement accuracy. The self-calibration mechanism automatically calibrates the sensor by comparing historical data with real-time measurement values.

7. The long-span bridge beam structure health monitoring system based on deformation analysis according to claim 6 is characterized by: The dynamic weighting algorithm automatically selects the sensor that best suits the current environment and dynamic state according to the vibration frequency and amplitude changes in the real-time state of the sensor; the dynamic weighting algorithm adopts a method based on multi-objective optimization, and dynamically adjusts the weight of each sensor through a preset objective function combined with real-time environmental data; the objective function includes comprehensive indicators of measurement accuracy, data stability, response speed and sensor health status; during the dynamic weighting process, the system first evaluates the data quality of all sensors in real time to obtain evaluation results, including evaluation of signal fluctuations, noise amplitude, and measurement errors; based on the evaluation results, the system ensures the accuracy of the fusion results by assigning high weights to high-quality data.

8. The long-span bridge beam structure health monitoring system based on deformation analysis according to claim 7 is characterized by: The self-calibration mechanism automatically detects and corrects sensor drift or error by comparing known standard data with real-time measurement data; the self-calibration mechanism is implemented by the following steps: Step 1: Establish a benchmark database based on the historical health data of the bridge, and record the angle change range, dynamic response mode and environmental parameter influence rules of the bridge under normal conditions; Step 2: Regularly obtain real-time measurement data and compare it with historical data in the benchmark database to identify potential drift or deviation; Step 3: When it is detected that the sensor data deviates from the preset threshold, the calibration mode is started to correct the current measurement value through redundant sensor data or known environmental parameters; Step 4: The system records all calibration operations and updates the historical database to provide a more accurate reference for subsequent calibration and prediction.

9. The long-span bridge beam structure health monitoring system based on deformation analysis according to claim 8 is characterized by: The data processing module also includes a fault detection and diagnosis unit, which is used to automatically detect the failure or performance degradation of the sensor through an intelligent algorithm, and automatically switch to a backup sensor when a failure is detected, and trigger an alarm.

10. The long-span bridge beam structure health monitoring system based on deformation analysis according to claim 9 is characterized in that: The three-cycle process includes the following stages: First cycle: remove noise from raw data and smooth the signal; Second cycle: extract the bridge angle change features from the denoised data and perform pattern recognition through machine learning algorithms; The third cycle: Optimize the final angle measurement result through weighted data fusion and error correction algorithm.

Images