A rigid frame bridge deflection real-time monitoring and early warning device

Abstract

The application belongs to the technical field of measurement, and discloses a rigid frame bridge downward deflection real-time monitoring and early warning device, a high-stability measurement system is built through a laser vision module, a rigid anchor seat is adopted for connecting a reference host computer with a bridge foundation, and a constant-temperature constant-damping vibration isolation cavity is matched, so that reference interference caused by bridge structure deformation, environmental temperature change and vibration is eliminated; vertical displacement data is directly collected through a laser emission array, binocular vision cameras and the laser array work cooperatively to correct various measurement errors, and can also adaptively adjust measurement parameters according to monitoring requirements, and adapt to rigid frame bridges with different spans; a decoupling and purification module adopts a fusion algorithm combined with a four-dimensional frequency characteristic library, can accurately separate instantaneous elastic deformation, temperature reversible deformation and irreversible downward deflection caused by concrete creep, filters various interference data through a triple abnormal value elimination algorithm, and can also predict the downward deflection development trend based on the algorithm.

Description

Technical Field

[0001] This invention belongs to the field of measurement technology, specifically a real-time monitoring and early warning device for deflection under rigid frame bridges. Background Technology

[0002] Prestressed concrete rigid frame bridges, with their core advantages such as strong span capacity, high structural stiffness, good driving comfort, and controllable cost, have become one of the mainstream bridge types for long-span bridge engineering in my country. However, during long-term service, due to the coupled influence of multiple factors such as concrete creep, prestress loss, vehicle cyclic loads, temperature gradient effects, and changes in geological conditions, the main beam of the rigid frame bridge will experience irreversible vertical deflection, and the deflection will continue to increase with the service life. When the cumulative deflection exceeds the design limit, it will directly cause problems such as deterioration of the beam alignment, redistribution of internal forces, beam cracking, and reduction of structural bearing capacity. In severe cases, it may even lead to bridge collapse. Current automated monitoring technologies for the deflection deformation of rigid frame bridges mainly face the following technical challenges: Strain data is collected by deploying fiber optic grating sensors, and the deflection is obtained through mechanical calculation. However, there are problems such as high installation difficulty, fiber optics being susceptible to aging and breakage due to vibration / temperature and humidity, poor long-term service stability, and high maintenance costs. Moreover, the strain-deflection conversion process will generate serious error accumulation, and the long-term monitoring accuracy cannot be guaranteed, making it difficult to meet the needs of millimeter-level precision monitoring of deflection of rigid frame bridges and long-term operation throughout the entire life cycle.

[0003] The reference points of existing devices are mostly set on the main beam or pier, which are easily affected by factors such as bridge structural deformation, temperature changes, and vibration, resulting in reference drift, which directly leads to distortion of measurement data and cannot guarantee the reliability of measurement results. It is impossible to accurately separate irreversible structural deflection caused by concrete creep / prestress loss from instantaneous elastic deformation caused by vehicle load / wind vibration and reversible deformation caused by temperature effect. The monitoring data has many interference items, and the false alarm and missed alarm rates of early warning remain high, making it difficult to achieve accurate measurement and data purification. During long-term service, sensors are prone to parameter drift and accuracy decay, and cannot achieve in-situ self-calibration and fault self-diagnosis, resulting in no guarantee of measurement accuracy throughout the entire life cycle. Existing laser measurement or vision measurement technologies are mostly single-mode and have not achieved deep integration of multi-dimensional measurement data. Furthermore, they have not optimized the measurement logic for the structural characteristics of rigid frame bridges, making it difficult to meet the high-precision measurement requirements of the entire cross section of long-span rigid frame bridges. Summary of the Invention

[0004] The purpose of this invention is to provide a real-time monitoring and early warning device for deflection of rigid frame bridges, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a real-time monitoring and early warning device for deflection of a rigid frame bridge, comprising the following modules: Preferably, the laser vision module is as follows: The reference host adopts an integrated structure consisting of a rigid anchor and a constant temperature and constant damping vibration isolation cavity. The rigid anchor is rigidly connected to the bridge foundation. The constant temperature and constant damping vibration isolation cavity has a built-in temperature control system and a multi-layer damping buffer structure to eliminate reference drift caused by changes in ambient temperature and ground vibration. The laser emission array fully covers all monitoring sections along the bridge direction, and adopts adjustable focus laser emission technology. The vertical displacement of the laser spot is collected by a position-sensitive detector. The binocular vision camera adopts a synchronous trigger acquisition mode and works in conjunction with the laser emission array to synchronously collect the lateral offset and torsional angle data of the section. Through the laser spot and visual feature point collaborative matching algorithm, the laser measurement value is corrected in real time to eliminate the measurement error caused by the lateral sway and torsion of the beam. The multi-section synchronous calibration acquisition unit and each measurement terminal are synchronously networked through LoRa local area network. It has the function of adaptive adjustment of measurement distance and can automatically optimize laser power and visual focal length according to the distance between the monitoring section and the reference host. The measurement terminal is an integrated laser receiving and visual acquisition terminal deployed at each monitoring section of the bridge.

[0006] Preferably, the decoupling and purification module uses a fusion algorithm combining variational mode decomposition and wavelet packet multi-band decomposition to perform two-layer time-frequency domain decomposition on the original measurement data based on the original measurement data collected by the laser vision module. A four-dimensional feature library is pre-established for the natural frequency of the rigid frame bridge, vehicle load frequency, temperature gradient effect frequency, concrete creep and prestress loss frequency. The separation of multi-frequency components is achieved through feature matching, which can separate instantaneous vibration elastic deformation and temperature-induced reversible deformation from irreversible cumulative deflection caused by concrete creep and prestress loss. The irreversible deflection trend prediction submodule, based on purified irreversible deflection data and combined with a deep learning LSTM model, predicts short-term and medium-to-long-term deflection development trends in real time and outputs deflection rate change curves to provide advance warning. It has a built-in triple outlier removal algorithm consisting of the 3σ criterion, deformation continuity constraint, and trend consistency verification to remove abnormal data caused by measurement noise and sudden interference, and simultaneously outputs three core measurement parameters: cumulative deflection, deflection rate, and deflection trend.

[0007] Preferably, the calibration and diagnostic module triggers a full-link in-situ self-calibration mechanism based on the metrological parameters output by the decoupling and purification module and the measurement equipment characteristics of the laser vision module. The calibration module automatically performs calibration during low-load periods each day. The laser array of the reference host emits a calibration signal with a preset standard displacement. Each measurement terminal synchronously collects and automatically corrects the nonlinear error of the position-sensitive detector, the drift of the intrinsic and extrinsic parameters of the binocular camera, and the signal attenuation error of the data transmission link.

[0008] The calibration accuracy feedback adjustment submodule compares the self-calibration data with the preset standard value. If the deviation exceeds the preset accuracy threshold, the measurement parameters are automatically adjusted. The adjacent mutual calibration mechanism achieves bidirectional mutual calibration through laser mutual emission verification and deformation correlation analysis of adjacent measurement terminals, diagnoses terminal fault status in real time, and verifies the rationality of measurement data through the deformation law of adjacent sections. Terminals with abnormal data are automatically marked and fault warnings are triggered. The cross-section mutual calibration library integrates historical calibration data and normal measurement data. When the terminal calibration is abnormal, compensation calibration is achieved through the sample library, forming a two-dimensional calibration system that combines self-calibration closed loop and adjacent mutual calibration compensation.

[0009] Preferably, the alignment reconstruction module establishes a cubic B-spline curve reconstruction model with structural mechanical constraints based on the deflection and rotation data calibrated by the calibration and diagnosis module, combined with the assumption of the rigid frame bridge plane section and the constitutive relationship between bending moment and curvature. The section weight allocation algorithm allocates different weights according to the stress characteristics of different sections to achieve continuous reconstruction of the vertical alignment of the entire bridge.

[0010] The finite element boundary conditions are updated in real time. Based on key environmental parameters such as the bridge's service life, temperature, humidity, and vibration frequency, the constraint parameters of the reconstructed model are dynamically adjusted to correct the reconstruction error. The automatic alignment deviation correction function automatically optimizes the model parameters when the deviation between the reconstructed alignment and the actual measurement data exceeds a threshold.

[0011] Preferably, the early warning triggering module establishes a dynamic early warning model based on the full bridge vertical alignment data obtained by the alignment reconstruction module and the core measurement parameters of cumulative deflection, deflection rate, and deflection trend output by the decoupling and purification module. This model combines bridge design specification limits, historical measurement data, and real-time service status, and updates early warning thresholds at each level using an adaptive threshold algorithm. A four-level monitoring and early warning system is formed by setting a normal monitoring status and three levels of early warning, with different levels corresponding to different early warning push methods. When an early warning is triggered, it synchronously links the original measurement data, calibration data, purified deflection data, full-bridge alignment reconstruction results, and deflection trend curve to achieve full-process traceability of the early warning; it automatically generates a standardized early warning report, which clearly defines the abnormal cross section, abnormal data, deviation value, early warning level, and maintenance recommendations; The self-optimization function for early warning thresholds dynamically adjusts the early warning thresholds at all levels based on historical early warning data and bridge maintenance results.

[0012] Preferably, the power transmission module adopts a triple complementary power supply mode combining solar energy, bridge vibration energy harvesting and backup lithium battery. Combined with the low power sleep and wake-up mechanism of the measurement terminal, the terminal enters sleep state when there is no vehicle load, and only the reference host and wake-up circuit are kept working. It adopts a composite transmission architecture with dual-mode redundancy of LoRa local area network, 5G wide area network and Beidou short message. Each measurement terminal achieves close-range synchronous networking and data transmission through LoRa. The aggregation node transmits data to the cloud platform through 5G with encryption. When there is no 5G signal, it automatically switches to Beidou short message transmission. The transmitted data has built-in timestamp, cyclic check code and AES encryption mechanism to ensure the integrity and authenticity of the measurement data. The data transmission status monitoring function monitors the data transmission rate and packet loss rate in real time, and automatically switches the transmission mode and triggers an alarm when transmission is abnormal.

[0013] Preferably, the digital twin module establishes a bridge deflection deformation database based on the full monitoring data output by the laser vision module, decoupling and purification module, calibration and diagnosis module, alignment reconstruction module, early warning triggering module, and power transmission module. This database stores original measurement data, calibration data, purified data, alignment reconstruction data, early warning data, and deflection trend data. A digital twin model of the rigid frame bridge is constructed, and the real-time reconstructed bridge alignment, deflection measurement data, and calibration data are synchronously presented to the digital twin model, achieving a three-dimensional visualization of the bridge's deflection deformation. Based on historical measurement data and deep learning algorithms, the development trend of downward deflection is predicted; the linkage optimization function of digital twin and laser vision module optimizes the acquisition parameters of measurement module and the constraints of reconstructed model through the simulation results of digital twin model. The anomaly simulation function uses a digital twin model to simulate different downward anomaly scenarios and verify the effectiveness of the early warning mechanism.

[0014] The beneficial effects of this invention are as follows: 1. This invention establishes a highly stable measurement system using a laser vision module. The reference host is connected to the bridge foundation using a rigid anchor, and is equipped with a constant temperature and constant damping vibration isolation cavity to eliminate reference interference caused by bridge structural deformation, environmental temperature changes, and vibration. Vertical displacement data is directly acquired through a laser emission array. The binocular vision camera and laser array work together to correct various measurement errors and can also adaptively adjust measurement parameters according to monitoring needs, adapting to rigid frame bridges of different spans. The measurement terminal has an integrated design, and the magnetic assembly does not damage the bridge structure, improving the stability and measurement accuracy of the device during long-term service.

[0015] 2. The decoupling and purification module of this invention adopts a fusion algorithm combined with a four-dimensional frequency feature library, which can accurately separate instantaneous elastic deformation, temperature reversible deformation and irreversible deflection caused by concrete creep. It filters various interference data through a triple outlier removal algorithm and can also predict the deflection development trend based on the algorithm. The early warning triggering module establishes a multi-dimensional dynamic early warning model, sets four early warning levels and is equipped with differentiated push methods. When an early warning is issued, full-process data traceability can be realized. It can also self-optimize the early warning threshold based on the bridge maintenance effect. Combined with a two-dimensional calibration system, the accuracy of the entire process from data purification to early warning release is guaranteed, reducing the probability of false alarms and missed alarms.

[0016] 3. The calibration and diagnosis module of this invention achieves daily automatic full-link in-situ self-calibration, forming a two-dimensional calibration closed loop with the adjacent mutual calibration mechanism to continuously ensure measurement accuracy; the power supply and transmission module adopts a triple complementary power supply mode with a low-power mechanism, and the dual-mode redundant transmission architecture ensures long-term uninterrupted operation in the field without external power supply, while also achieving secure data transmission; the digital twin module integrates all monitoring data to build a model, realizing a three-dimensional visualization of bridge deflection, and can also predict deformation trends, reverse optimize measurement parameters, and combine precise modeling of linear reconstruction to provide a basis for daily bridge maintenance, repair and reinforcement, realizing intelligent full-process from data collection to maintenance recommendations. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the entire process of real-time monitoring and early warning of deflection in rigid frame bridges according to the present invention. Figure 2 This is a flowchart of the multi-source data acquisition and preprocessing sub-process of the present invention; Figure 3 This is a flowchart of the intelligent early warning and response closed-loop process of the present invention. Detailed Implementation

[0018] 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.

[0019] like Figures 1 to 3 As shown in the embodiment of the present invention, a real-time monitoring and early warning device for deflection of a rigid frame bridge is provided, comprising the following modules, the specific structure and working principle of each module being as follows: The laser vision module is specifically as follows: The reference host adopts an integrated structure consisting of a rigid anchor seat and a constant temperature and constant damping vibration isolation cavity. The rigid anchor seat is rigidly connected to the bridge foundation (non-main beam, non-pier) to avoid the influence of the bridge structure's own deformation on the reference. The constant temperature and constant damping vibration isolation cavity has a built-in temperature control system (temperature control accuracy ±0.1℃) and a multi-layer damping buffer structure to eliminate reference drift caused by changes in ambient temperature (-40℃~60℃) and ground vibration (frequency 0.1Hz~100Hz). The multi-layer damping buffer structure consists of three layers: the first layer is a rubber elastic damping layer used to absorb low-frequency vibration interference; the middle layer is a metal honeycomb damping layer used to attenuate medium-frequency vibration energy; and the last layer is a silicone viscoelastic damping layer used to suppress high-frequency vibration disturbances. The three damping layers are stacked and fixed inside the vibration isolation cavity and rigidly connected to the reference host. The temperature control system adopts a semiconductor refrigeration and heating integrated module, which collects temperature data in real time through the temperature sensor built into the cavity to achieve closed-loop feedback control of the temperature.

[0020] The laser emission array fully covers all monitoring sections along the bridge direction. It adopts adjustable-focus laser emission technology with a focal length adjustment range of 5 meters to 500 meters. The vertical displacement of the laser spot is collected by a position-sensitive detector to achieve absolute deflection measurement with a resolution of 0.01 mm, without the need for strain calculation. The sampling frequency of the position-sensitive detector is synchronized with the laser emission array, and the adjustable range is 1Hz~100Hz. It supports manual / automatic switching according to monitoring needs, and the sampling accuracy of the light spot positioning is 0.005mm, providing redundant accuracy guarantee for displacement measurement.

[0021] The laser emitting array has an adjustable emission power range of 5mW to 500mW, which can be adaptively adjusted according to the measurement distance. The effective diameter of the laser spot within the measurement range of 5 meters to 500 meters is controlled within 1mm to 5mm, ensuring accurate capture by the position-sensitive detector.

[0022] The binocular vision camera adopts a synchronous trigger acquisition mode and works in conjunction with the laser emission array to synchronously acquire data on the lateral offset and torsional angle of the cross section. Through a laser spot and visual feature point matching algorithm, the laser measurement value is corrected in real time to eliminate measurement errors caused by the lateral sway and torsion of the beam. The binocular vision camera's acquisition frame rate is synchronized with the laser emission array, and the shooting resolution is 2560×1440. The recognition accuracy of the laser spot and bridge cross-section feature points is 0.01mm, which can accurately capture the minute deformations of the cross-section's lateral offset and torsional angle.

[0023] The multi-section synchronous calibration acquisition unit and each measurement terminal are synchronously networked via LoRa local area network. The sampling frequency is adjustable from 1Hz to 100Hz, taking into account the needs of real-time monitoring, ease of deployment, and anti-aging capabilities. It has an adaptive adjustment function for measurement distance, which can automatically optimize the laser power and visual focal length according to the distance between the monitoring section and the reference host, adapting to the measurement needs of rigid frame bridges with different spans. The measurement terminal is an integrated laser receiving and visual acquisition terminal deployed at each monitoring section of the bridge.

[0024] The decoupling and purification module uses a fusion algorithm combining variational mode decomposition and wavelet packet multi-band decomposition to perform two-layer time-frequency domain decomposition on the original measurement data collected by the laser vision module. The fusion algorithm first decomposes the original measurement data into multiple intrinsic modal components through variational mode decomposition to achieve preliminary separation of signals of different frequencies. Then, it performs wavelet packet multi-band decomposition on each intrinsic modal component to refine the frequency bands corresponding to concrete creep, temperature effect, and vehicle load. The effective components after decomposition are reconstructed to obtain the frequency feature data to be matched, and the remaining noise components are directly eliminated.

[0025] A four-dimensional feature library is pre-established for the natural frequency, vehicle load frequency, temperature gradient effect frequency, concrete creep and prestress loss frequency of rigid frame bridge. Through feature matching, the separation of multiple frequency components is achieved. It can separate instantaneous vibration elastic deformation and temperature-induced reversible deformation from irreversible cumulative deflection caused by concrete creep and prestress loss, with a separation accuracy better than 0.01mm. The specific frequency range of the four-dimensional frequency feature library is as follows: the natural frequency of rigid frame bridge is 0.1Hz~5Hz, the vehicle load frequency is 2Hz~20Hz, the temperature gradient effect frequency is 0.0001Hz~0.01Hz, and the frequency of concrete creep and prestress loss is 0.00001Hz~0.001Hz. There is no overlap between the frequency bands, so as to achieve accurate feature matching.

[0026] After frequency feature matching is completed, time-domain signal reconstruction is performed on the effective frequency components corresponding to each type of deformation. The frequency components of the same type of deformation are integrated by linear superposition to generate independent time-domain data curves for instantaneous vibration elastic deformation, temperature reversible deformation, and irreversible cumulative deflection. The time axis of each curve is completely synchronized with the original acquired data.

[0027] The irreversible downward trend prediction submodule, based on purified irreversible downward data and combined with a deep learning LSTM model, predicts the downward development trend in the next 30 days and 90 days in real time, and outputs the downward rate change curve to provide advance warning. The input to the deep learning LSTM model is purified irreversible deflection time series data, including five core feature data: cumulative deflection, deflection rate, bridge service life, environmental temperature and humidity, and vehicle load frequency. The time step is set to 1 hour, and the input sequence length is 720 steps (30 days). The model output is the predicted deflection value for the next 30 days (short-term) and 90 days (medium-to-long-term) and the deflection rate change curve. The prediction results are divided into dimensions according to the monitoring section, and each monitoring section corresponds to an independent model output result. The resolution of the model output data is consistent with the acquisition resolution of the laser vision module, which is 0.01 mm.

[0028] It incorporates a triple outlier removal algorithm consisting of the 3σ criterion, deformation continuity constraint, and trend consistency verification. This algorithm is used to remove abnormal data caused by measurement noise and sudden interference (such as strong winds and rainstorms), and simultaneously outputs three core measurement parameters: cumulative deflection, deflection rate, and deflection trend.

[0029] The triple outlier removal algorithm is executed in the following order: First, the original purified data is coarsely screened using the 3σ criterion to remove extreme outliers that deviate from the mean by three times the standard deviation; then, the deformation continuity constraint is checked to remove outliers whose deflection changes beyond a reasonable range at adjacent time steps; finally, the trend consistency is checked by comparing the single-section data with the overall deflection trend of the entire bridge to remove outliers with conflicting trends. The dataset after triple removal is the valid analysis data.

[0030] The calibration and diagnostic module is based on the metrological parameters output by the decoupling and purification module and the measurement equipment characteristics of the laser vision module. The full-link in-situ self-calibration mechanism is automatically triggered during the low-load period of each day. The laser array of the reference host emits a calibration signal with a preset standard displacement (0.1mm to 10mm). Each measurement terminal synchronously collects and automatically corrects the nonlinear error of the position sensitive detector, the drift of the intrinsic and extrinsic parameters of the binocular camera, and the signal attenuation error of the data transmission link, thereby eliminating the accuracy attenuation caused by sensor aging and temperature and humidity changes. The daily low-load period is set from 0:00 to 4:00 AM, during which no vehicles pass over the bridge or the load is extremely low. During calibration, the reference host continuously transmits 300 standard displacement calibration signals at a frequency of 10Hz. Each measurement terminal collects the data synchronously. The maximum and minimum values ​​of the 300 sets of data are removed, and the arithmetic mean is taken as the reference data for calibration correction.

[0031] The calibration accuracy feedback adjustment submodule compares the self-calibration data with the preset standard value. If the deviation exceeds the threshold (0.05mm), it automatically adjusts measurement parameters such as laser power, position-sensitive detector sensitivity, and visual exposure time to ensure stable measurement accuracy throughout the entire life cycle. The adjacent mutual calibration mechanism achieves bidirectional mutual calibration through laser mutual refraction verification and deformation correlation analysis between adjacent measurement terminals, diagnoses terminal fault status in real time, and verifies the rationality of measurement data through the deformation law of adjacent sections. Terminals with abnormal data are automatically marked and fault warnings are triggered to prevent invalid data from entering subsequent processes. The deformation correlation analysis uses the Pearson correlation coefficient as the criterion. The correlation coefficient threshold for deflection data of adjacent sections is set to 0.85. When the correlation coefficient is lower than this threshold, the corresponding measurement terminal data is judged to be abnormal and marked.

[0032] The cross-section mutual calibration library integrates historical calibration data and normal measurement data. When a terminal calibration is abnormal, compensation calibration is achieved through the sample library, reducing maintenance costs and forming a two-dimensional calibration system that combines self-calibration closed loop with adjacent mutual calibration compensation.

[0033] The triggering condition for calling the cross-section mutual calibration library is: when the deviation of the self-calibration data of a single measurement terminal exceeds the 0.05mm threshold for three consecutive times, and the adjacent mutual calibration cannot complete the compensation; the sample matching rule of the calibration library is to accurately match according to the bridge section type (mid-span, support, ordinary section), service environment, and monitoring duration, and select the recent normal calibration data of the same type of section as the compensation sample, with a sample number of no less than 50 groups.

[0034] The alignment reconstruction module, based on the deflection and rotation data calibrated by the calibration and diagnosis module, and combined with the rigid frame bridge plane section assumption and the constitutive relationship between bending moment and curvature, establishes a cubic B-spline curve reconstruction model with structural mechanical constraints. The section weight allocation algorithm assigns different weights according to the stress characteristics of different sections (key sections such as the mid-span of the main beam and the support), realizing the continuous reconstruction of the vertical alignment of the entire bridge. The reconstruction alignment accuracy is better than 0.1 mm / m, which can help verify the accuracy of the deflection data and further improve the measurement accuracy.

[0035] The cross-section weight allocation algorithm divides the weights according to the stress characteristics of the bridge: the mid-span section of the main beam is the key monitoring section, with a weight of 0.4; the support section of the main beam is the secondary key monitoring section, with a weight of 0.3; the weights of the remaining ordinary sections are set to 0.3, and the sum of the weights of all sections of the entire bridge is 1. The weight allocation results are recalibrated every 5 years as the bridge's service life increases.

[0036] The finite element boundary conditions are updated in real time, and the constraint parameters of the reconstructed model are dynamically adjusted based on key environmental parameters such as the bridge's service life, temperature, humidity, and vibration frequency to correct reconstruction errors. The automatic alignment deviation correction function automatically optimizes the model parameters when the deviation between the reconstructed alignment and the actual measured data exceeds a threshold.

[0037] The threshold for determining the linear deviation is set to 0.1 mm / m. When the deviation between the reconstructed linear shape and the actual measured data exceeds this threshold for 10 consecutive time steps, the model parameter optimization is automatically triggered. The optimized parameters include the node vector, curvature correction coefficient, and cross-sectional stiffness coefficient of the cubic B-spline curve. After optimization, the reconstructed linear shape is regenerated.

[0038] The update cycle of the finite element boundary conditions is linked to the acquisition frequency of the laser vision module, with an update cycle of 10 minutes. The constraint parameters are dynamically adjusted based on the average environmental parameters within the cycle, balancing the timeliness of updates and computational efficiency.

[0039] The early warning triggering module establishes a dynamic early warning model based on the full bridge vertical alignment data obtained by the alignment reconstruction module and the core measurement parameters of cumulative deflection, deflection rate, and deflection trend output by the decoupling and purification module. It combines the bridge design specification limits, historical measurement data, and real-time service status, and updates the early warning thresholds at all levels through an adaptive threshold algorithm. The adaptive threshold algorithm takes six types of parameters as input: cumulative measured deflection, predicted deflection rate, bridge design specification limits, bridge service life, historical early warning data, and maintenance and treatment effect evaluation. The algorithm processes each input parameter through weighted normalization, with the bridge design specification limits having a weight of 0.4, the predicted deflection rate having a weight of 0.3, the maintenance and treatment effect evaluation having a weight of 0.2, and the remaining parameters having a combined weight of 0.1. The algorithm outputs dynamic threshold values ​​for Level 1, Level 2, and Level 3 early warnings. The threshold values ​​are differentiated according to the bridge monitoring sections, with the threshold accuracy for key stress sections such as mid-span and supports set at 0.01 mm, and the threshold accuracy for ordinary sections set at 0.1 mm. The threshold update cycle is synchronized with the self-calibration cycle of the calibration and diagnosis module, which is once a day.

[0040] The system is configured with four warning levels: routine monitoring, Level 1 warning (abnormal deflection rate or sudden trend change), Level 2 warning (cumulative deflection approaching the limit), and Level 3 warning (cumulative deflection exceeding the limit). Different warning levels correspond to different warning push methods, including local audible and visual alarms, cloud platform push, and SMS reminders from maintenance personnel. The push rules for each warning level are as follows: Level 1 warning only triggers local audible and visual alarms and cloud platform push notifications; Level 2 warning triggers local audible and visual alarms, cloud platform push notifications, and SMS reminders from the maintenance team leader; Level 3 warning triggers local audible and visual alarms, cloud platform push notifications, and SMS reminders from the maintenance team leader and the bridge management unit's supervisor. The SMS reminders include the abnormal section, warning level, deviation value, and emergency response suggestions.

[0041] When an early warning is triggered, it synchronously links the original measurement data, calibration data, purified deflection data, full-bridge alignment reconstruction results, and deflection trend curve to achieve full-process traceability of the early warning and avoid unfounded false alarms; it automatically generates a standardized early warning report, which clearly defines the abnormal cross section, abnormal data, deviation value, early warning level, and maintenance recommendations. The self-optimization function for early warning thresholds dynamically adjusts the early warning thresholds at all levels based on historical early warning data and bridge maintenance results, thereby improving the accuracy and practicality of early warnings.

[0042] The standardized early warning report includes six core chapters: basic bridge information, monitoring data tracing, abnormal data analysis, early warning level determination criteria, targeted maintenance recommendations, and key points for subsequent monitoring. The report uses a unified format combining tables and text and can be directly exported as a PDF for bridge maintenance filing.

[0043] The power transmission module adopts a triple complementary power supply mode that combines solar energy, bridge vibration energy harvesting and backup lithium battery. Combined with the low power sleep and wake-up mechanism of the measurement terminal, each measurement terminal enters sleep state when there is no vehicle load, and only the reference host and wake-up circuit are kept working, so as to realize long-term uninterrupted service in the field without external power supply and ensure the continuity of measurement data. The energy dispatch priority of the triple complementary power supply mode is as follows: solar power supply is the first priority, which prioritizes powering the device and charging the lithium battery, accounting for about 70% of the power; bridge vibration energy harvesting is the second priority, which supplements power supply during periods without sunlight, accounting for about 20% of the power; and the backup lithium battery is the third priority, which provides emergency power supply when the first two power supply methods fail, accounting for about 10% of the power. When the lithium battery is fully charged, it can ensure that the device can operate at full load for no less than 72 hours.

[0044] The wake-up mechanism is triggered by the vibration sensor built into the measurement terminal. When the bridge vibration amplitude is detected to be greater than 0.01 mm and the vibration frequency matches the characteristics of vehicle traffic, the terminal wakes up from the sleep state within 0.5 seconds and resumes full-load acquisition. In the sleep state, the power consumption of the measurement terminal is ≤5mW, and the total power consumption of the reference host and the wake-up circuit is ≤20mW.

[0045] It adopts a composite transmission architecture with dual-mode redundancy of LoRa local area network, 5G wide area network and Beidou short message. Each measurement terminal achieves close-range synchronous networking and data transmission through LoRa. The aggregation node transmits data to the cloud platform with encryption via 5G. When there is no 5G signal on bridges in remote mountainous areas, it automatically switches to Beidou short message transmission. The transmitted data has built-in timestamp, cyclic check code and AES encryption mechanism to prevent data tampering and loss and ensure the integrity and authenticity of the measurement data. The AES encryption uses a 256-bit encryption algorithm, the cyclic check code uses CRC32, and the timestamp is marked in milliseconds. This triple verification mechanism ensures that the data is secure and tamper-proof throughout the entire process from the collection end to the cloud platform.

[0046] The LoRa local area network has an effective outdoor communication distance of up to 5km, a maximum number of nodes in a single network of 200, supports the distributed deployment of multiple measurement terminals, and the time synchronization error of each terminal after networking is ≤1ms, ensuring the time consistency of the collected data.

[0047] The data transmission status monitoring function monitors the data transmission rate and packet loss rate in real time. When the transmission is abnormal, it automatically switches the transmission mode and triggers an alarm. Each measurement terminal of the device adopts a magnetic detachable assembly design, which has no rigid connection with the bridge monitoring section. This optimizes the layout structure and does not require damage to the bridge structure. It further makes up for the defects of fiber optic sensing, such as difficulty in deployment and susceptibility to environmental influences, and ensures the long-term stable operation of the measurement device.

[0048] The magnetic detachable assembly and placement design of the measurement terminal adopts a strong permanent magnet adsorption structure. The magnet is wrapped with a stainless steel protective layer. The contact surface with the bridge monitoring section is equipped with an anti-slip and wear-resistant rubber pad. The static friction force of a single measurement terminal is not less than 500N, which can adapt to the working conditions of daily bridge vibration. There is no rigid connection between the terminal and the bridge section such as drilling or welding, and there is no structural residue when disassembling.

[0049] The digital twin module establishes a bridge deflection deformation database based on the full monitoring data output by the laser vision module, decoupling and purification module, calibration and diagnosis module, linear reconstruction module, early warning triggering module and power transmission module. It stores original measurement data, calibration data, purification data, linear reconstruction data, early warning data and deflection trend data, and realizes full-process traceability of monitoring data. The storage rules for the deflection deformation database are as follows: original measurement data and calibration data are stored permanently; purified data and linear reconstruction data are stored for a period not less than the bridge's design service life; early warning data and deflection trend data are stored permanently; historical data cleaning only targets temporarily cached intermediate data, with a caching period of 7 days, which is automatically cleaned upon expiration. No manual intervention is required before cleaning, and it does not affect the retention of core data.

[0050] A digital twin model of a rigid frame bridge is constructed, and the real-time reconstructed full bridge alignment, deflection measurement data, and calibration data are synchronously presented to the digital twin model to realize a three-dimensional visualization of the bridge's deflection deformation and intuitively present the deformation state of the entire bridge. Based on historical measurement data and deep learning algorithms, the development trend of deflection is predicted, providing accurate measurement basis for the daily maintenance, repair and reinforcement of bridges; The deep learning algorithm reuses the LSTM model structure of the decoupled purification module. The model input is the full life cycle deflection deformation database data of all monitoring sections of the entire bridge, including four types of core data: original measurement data, calibration data, purification data, and linear reconstruction data. The input sequence length is 365 steps (1 year). The model output is the deflection trend prediction result of the overall vertical linearity of the entire bridge and the annual predicted value of the deflection of each monitoring section. The prediction results are synchronized to the digital twin model in real time and presented in a three-dimensional visualization form. The predicted data output by the model is directly used as the basis for the digital twin module to back-optimize the acquisition parameters of the laser vision module.

[0051] The digital twin and measurement module linkage optimization function uses the simulation results of the digital twin model to reverse optimize the sampling frequency, laser power and other acquisition parameters of the measurement module and the constraints of the reconstructed model, thereby further improving the measurement accuracy. The optimization rules for the acquisition parameters are as follows: when the cross-sectional deformation fluctuation is large in the simulation display of the digital twin model, the sampling frequency is automatically increased by 5~20Hz and the laser power is increased by 10%~30%; when the deformation tends to stabilize, the sampling frequency and laser power are automatically reduced back to the reference value, and the reference value is executed according to the initial monitoring parameters of the bridge.

[0052] The anomaly simulation function uses a digital twin model to simulate different downward anomaly scenarios, verifying the effectiveness of the early warning mechanism. At the same time, it strengthens the integration of measurement data with actual engineering applications, improving the practicality of the technical solution.

[0053] The simulation of abnormal deflection scenarios includes three typical scenarios: irreversible deflection exceeding the limit at mid-span of the main beam, simultaneous deformation change of multiple sections, and sudden increase in deflection rate of local sections. The core indicators for verifying the early warning mechanism include early warning trigger response time, early warning accuracy, missed alarm rate, and false alarm rate. The requirements are that the early warning trigger response time is ≤10 seconds, the early warning accuracy is ≥99%, and the missed alarm rate and false alarm rate are both ≤1%.

[0054] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0055] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A real-time monitoring and early warning device for deflection under a rigid frame bridge, characterized in that, It consists of the following modules: Laser vision module: The reference host adopts an integrated structure of rigid anchor seat and constant temperature and constant damping vibration isolation cavity. The anchor seat is rigidly connected to the bridge foundation. The laser emission array combined with the position sensitive detector collects vertical displacement data. The binocular vision camera works together to correct measurement errors. With multi-section synchronous calibration and distance adaptive adjustment function, the original data of the bridge deflection can be collected. Decoupling and purification module: The original data is decomposed in the time and frequency domain using a fusion decomposition algorithm, and different types of deformation are separated based on a preset four-dimensional frequency feature library; through the irreversible deflection trend prediction and outlier removal algorithm, three core measurement parameters are selected and output: cumulative deflection, deflection rate and deflection trend. Calibration and Diagnosis Module: Ensures measurement accuracy through a full-link in-situ self-calibration and feedback adjustment mechanism that is automatically triggered during daily low-load periods; Realizes measurement terminal fault diagnosis and data rationality verification based on adjacent mutual calibration mechanism and cross-section mutual calibration library; Linear Reconstruction Module: Based on the calibrated deflection and rotation data, and combined with the structural characteristics of the rigid frame bridge, a reconstruction model with structural mechanical constraints is established. Through section weight allocation and dynamic parameter adjustment, the vertical linear reconstruction of the entire bridge is realized. Early warning trigger module: Establish a dynamic early warning model, set hierarchical early warning standards and differentiated push methods, realize traceability of the entire early warning process, and automatically generate standardized maintenance suggestions; Power transmission module: It adopts a triple complementary power supply mode of solar energy, bridge vibration energy harvesting and backup lithium battery, and realizes secure data transmission through dual-mode redundant transmission architecture, with real-time transmission status monitoring function. Digital twin module: Establish a database of bridge deflection, construct a digital twin model to achieve deformation visualization and trend prediction, and improve the measurement accuracy of the device through linkage and optimization with other modules.

2. The real-time monitoring and early warning device for deflection of rigid frame bridges according to claim 1, characterized in that, In the laser vision module, the constant temperature and constant damping vibration isolation cavity has a built-in temperature control and multi-layer damping buffer structure to eliminate reference drift caused by changes in ambient temperature and ground vibration interference; the laser emission array covers all monitoring sections along the bridge direction and adopts adjustable focus technology, while the position-sensitive detector collects the vertical displacement of the laser spot; the binocular vision camera adopts a synchronous trigger mode to collect data on the lateral offset and torsional angle of the section, and corrects the laser measurement error through a collaborative matching algorithm; the multi-section synchronous calibration acquisition unit is networked and linked with each measurement terminal to adaptively adjust parameters including laser power and visual focal length related to the measurement distance.

3. The real-time monitoring and early warning device for deflection of rigid frame bridges according to claim 2, characterized in that, The decoupling and purification module employs a fusion algorithm of variational mode decomposition and wavelet packet multi-band decomposition to perform a two-layer time-frequency domain decomposition on the original data. The four-dimensional frequency feature library includes the natural frequencies of the rigid frame bridge, vehicle load frequencies, temperature gradient effect frequencies, and frequencies of concrete creep and prestress loss, achieving the separation of instantaneous vibration elastic deformation, temperature reversible deformation, and irreversible cumulative deflection. The trend prediction submodule combines a deep learning LSTM model to predict short-term and medium-to-long-term deflection trends and uses a triple outlier removal algorithm to eliminate abnormal data caused by measurement noise and sudden interference.

4. The real-time monitoring and early warning device for deflection of rigid frame bridges according to claim 3, characterized in that, The calibration and diagnostic module's end-to-end in-situ self-calibration mechanism transmits a standard displacement calibration signal from the reference host. Each measurement terminal synchronously collects and corrects the nonlinear error of the position-sensitive detector, the parameter drift of the binocular vision camera, and the signal attenuation error of the data transmission link. The calibration accuracy feedback adjustment submodule compares the self-calibration data with the preset standard value. When the deviation exceeds the preset accuracy threshold, it automatically adjusts the measurement parameters, including laser power and visual exposure time. The adjacent mutual calibration mechanism realizes bidirectional mutual calibration and terminal fault diagnosis between measurement terminals through laser mutual emission verification and deformation correlation analysis. The cross-section mutual calibration library is used for compensation calibration when the terminal calibration is abnormal, forming a two-dimensional calibration system that combines self-calibration and mutual calibration.

5. The real-time monitoring and early warning device for deflection of rigid frame bridges according to claim 4, characterized in that, The linear reconstruction module establishes a cubic B-spline curve reconstruction model with structural mechanical constraints based on the assumption of plane section of rigid frame bridge and the constitutive relationship between bending moment and curvature. The section weight allocation algorithm allocates weights according to the stress characteristics of each section, updates the finite element boundary conditions in real time based on key environmental parameters, and dynamically adjusts the model constraint parameters to achieve automatic correction of the vertical linear deviation of the entire bridge.

6. The real-time monitoring and early warning device for deflection of rigid frame bridges according to claim 5, characterized in that, The dynamic early warning model of the early warning triggering module is based on three core dimensions: cumulative deflection, deflection rate, and deflection trend. It combines bridge design specification limits, historical measurement data, and real-time service status, and dynamically updates the early warning thresholds at each level through an adaptive threshold algorithm. A four-level monitoring and early warning system is adopted, including routine monitoring and level 1, 2, and 3 early warnings, with different early warning levels corresponding to different early warning push methods. When an early warning is triggered, the entire early warning process is traceable by linking monitoring data across the entire chain. A standardized early warning report containing abnormal information and maintenance suggestions is automatically generated. It also has a self-optimization function for early warning thresholds based on historical early warning data and bridge maintenance effects.

7. The real-time monitoring and early warning device for deflection of rigid frame bridges according to claim 6, characterized in that, The power supply and transmission module combines the low-power sleep and on-demand wake-up mechanisms of the measurement terminal to reduce the overall energy consumption of the device. The transmission architecture adopts LoRa local area network local networking, combined with 5G wide area network and Beidou short message dual-mode redundant transmission. When there is no 5G wide area network signal, it automatically switches to Beidou short message transmission mode. The transmitted data adopts a multi-encryption and verification mechanism of AES encryption, cyclic check code and timestamp to ensure data integrity and authenticity. The real-time transmission status monitoring function can automatically trigger alarm and switch transmission mode when data transmission is abnormal.

8. The real-time monitoring and early warning device for deflection of rigid frame bridges according to claim 7, characterized in that, The digital twin module stores all monitoring data in its deflection deformation database; the digital twin model synchronously presents real-time alignment, measurement data, and calibration data, enabling real-time three-dimensional visualization of bridge deflection deformation; based on historical monitoring data and deep learning LSTM algorithm, it predicts the development trend of bridge deflection; through the linkage optimization function with the laser vision module and alignment reconstruction module, it reversely optimizes the measurement acquisition parameters and reconstruction model constraint parameters, and has the function of simulating abnormal bridge deflection scenarios, thereby verifying the practical application effectiveness of the early warning mechanism.

Images