Crack monitoring system for fixed end of steel-concrete composite beam

Abstract

The invention relates to the field of bridge monitoring, and particularly discloses a crack monitoring system for a fixed end of a steel-concrete composite beam. The method is used for solving the problems that multi-source information fusion, hot spot stress on-line accurate calibration, fatigue life real-time accurate evaluation, S-N curve dynamic correction and remote visual early warning are difficult to realize in crack monitoring of a fixed end of an existing steel-concrete composite beam under complex load and environmental interference. Comprising a multi-sensor data acquisition module, an edge preprocessing module, a data fusion and damage identification module, a hot spot stress calibration module, an early warning decision module, a visual remote communication module and a digital twinborn calibration module. According to the invention, through integration of multi-sensor synchronous acquisition, edge intelligent preprocessing, multi-source data fusion, online hot spot stress calibration and digital twinborn calibration, Monte life prediction and early warning and hybrid energy management, high sensitivity, low power consumption and accurate early warning of steel-concrete composite beam fixed end crack monitoring are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge monitoring, and in particular to a crack monitoring system for a fixed end of a steel-concrete composite beam. Background Art

[0002] The vertical joints between the concrete deck and beams of steel-concrete composite bridges are typically connected using shear connectors. Under prolonged vehicle loads, the steel-concrete interface (SCI) is prone to slippage, resulting in interfacial cracks. These cracks allow corrosive media such as acid rain to penetrate the SCI along these cracks, corroding components such as the steel beams, shear connectors, and rebar at the bottom of the concrete. This accelerates crack propagation, creating a vicious cycle. The wet joints between ultra-high performance concrete (UHPC) and conventional concrete (NC) are a key component in precast bridge decks and are most susceptible to cracking. Wet joints located at the end sections of the beams are subject to large negative bending moments, exacerbating cracking and water seepage. Especially in the following design situations: the lower deck at the side main beam is designed as a steel-concrete composite deck, the structure is designed to set node crossbeams at the main truss nodes, several inter-segment crossbeams are set in the intersegment, no longitudinal beams are set, the crossbeams are directly welded to the bottom chord, the concrete slab is only combined with the crossbeams, not with the bottom chord, the concrete panel design adopts precast panels, the vertical joint section of the precast concrete panel and the crossbeam is connected by shear nails, and the transverse joint section design uses not only shear connectors but also UHPC cast-in-place joints to strengthen the connection between the precast concrete slab and the steel bridge deck, and its transverse joint section is as follows: Figure 1 、 2 As shown, there is an urgent need to conduct static and fatigue tests on the local structure of the wet joints of UHPC prefabricated panels to explore the stress laws of the steel-UHPU interface area and the UHPC-NC wet joints under static and dynamic loads, and to find the structural scheme with minimum stress in this area under load. How to monitor the cracks at the fixed end of the steel-concrete composite beam in real time based on the test data and crack width development in different wet joint structural schemes and steel-UHPC joint section schemes, and evaluate the structural form of the wet joint structural scheme and the steel-UHPC joint section scheme, has become a technical problem that needs to be solved urgently. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a crack monitoring system for the fixed end of a steel-concrete composite beam. By integrating multi-sensor synchronous acquisition, edge intelligent preprocessing, multi-source data fusion, online hot spot stress calibration and digital twin calibration, Monte life prediction and early warning, and hybrid energy management, high sensitivity, low power consumption and accurate early warning of crack monitoring at the fixed end of a steel-concrete composite beam can be achieved.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] A crack monitoring system for the fixed end of a steel-concrete composite beam includes a multi-sensor data acquisition module, an edge preprocessing module, a data fusion and damage identification module, a hotspot stress calibration module, an early warning decision module, a visual remote communication module, and a digital twin calibration module. The multi-sensor data acquisition module synchronously collects displacement, strain, acoustic emission, and crack images through LVDT, FBG, acoustic emission, DIC, and strain gauges and is electrically connected to the edge preprocessing module. The edge preprocessing module transmits the preprocessed data to the data fusion and damage identification module. The data fusion and damage identification module uses extended Kalman filtering and wavelet packet decomposition to fuse multi-source data, extract crack initiation / propagation features, and is electrically connected to the hotspot stress Calibration module and early warning decision module. The hot spot stress calibration module reads the finite element node stress at 0.4 times the plate thickness and 1.0 times the plate thickness from the weld toe, calculates the interface hot spot stress according to the linear proportion of first taking the high stress and then subtracting the low stress, and corrects the fatigue S-N curve in combination with the strain inversion results measured on site. The calibration information is then electrically connected to the digital twin calibration module. The early warning decision module issues a three-level alarm based on the dual thresholds of 0.2mm crack width and 0.5mm lateral slip and Miner cumulative damage, and is electrically connected to the visual remote communication module. The digital twin calibration module uses the least squares method to iteratively update the model material parameters and contact friction coefficient online and adaptively correct the finite element model.

[0006] It should be noted that the crack monitoring system described in the present invention is designed to target the complex stress and crack evolution characteristics of the fixed end of the steel-concrete composite beam, and realize full-process online monitoring and calibration through modular design: the multi-sensor data acquisition module arranges LVDT displacement sensors with a range of 50mm, 150mm extended-range LVDTs, metal strain gauges and concrete strain gauges attached to the longitudinal tensile steel bars and the top, side and bottom surfaces of the UHPC, FBG fiber optic sensors and three-way acoustic emission sensors at the beam end support section and mid-span, and combines the DIC support sub-module to draw a 50mm×50mm white grid on the surface of the structure to perform multi-spectral real-time grayscale difference enhancement, and output sub-pixel crack contours after frame-level FPGA preprocessing; the edge preprocessing module executes sixth-order bandpass filtering and GPS in parallel based on the CORDIC core on the FPGA+ARM platform. / IEEE-1588 clock synchronization is used to suppress 50Hz grid noise and steel beam vibration interference, and the time-domain windowed signal and sub-pixel grayscale image are electrically connected to the data fusion and damage identification module; the edge preprocessing module uses a double-layer adaptive extended Kalman filter to fuse LVDT / FBG displacement and strain, and combines improved wavelet packet decomposition to perform five-level sub-band decomposition of the acoustic emission waveform to extract envelope energy, spectral entropy, and instantaneous frequency characteristics, and uses an online incremental SVM to distinguish crack initiation, high-frequency tearing noise, and friction slip events. After the redundant sensor fusion fallback strategy is automatically activated in the event of abnormal mutations, the crack initiation and propagation characteristics are electrically connected to the hot spot stress calibration and early warning decision module; the hot spot stress calibration module reads the ABAQUS finite element node stress at the positions 0.4t and 1.0t from the weld toe after each load step, where t is the mainboard thickness and is calculated based on σ hs =1.67σ 0.4t -0.67σ 1.0t Linear extrapolation calculation of interface hot spot stress, σ hs is the interface hot spot stress, σ 0.4t , σ 1.0t The ABAQUS finite element node stresses at positions 0.4t and 1.0t from the weld toe are respectively calculated, and the fatigue S-N curve is least squares corrected in combination with the field strain inversion data, and the calibration coefficient is electrically connected to the digital twin calibration module; the early warning decision module dynamically predicts fatigue life based on the dual thresholds of crack width 0.2mm and lateral slip 0.5mm, combined with Miner cumulative damage and Monte Carlo random sampling, triggers a three-level SMS / email alarm and is electrically connected to the visual remote communication module; the digital twin calibration module uses the least squares method or Bayesian-least squares hybrid parameter identification to iteratively update the elastic modulus and contact friction coefficient online, adaptively correct the finite element model, and push the uncertainty information to the WebGL three-dimensional visualization interface and mobile app.

[0007] As a further solution of the present invention, the multi-sensor data acquisition module also includes a DIC support submodule and a deep learning terminal. The DIC support submodule utilizes the difference in multi-spectral reflection intensity on the surface of the structure to enhance the contrast of the crack contour, performs real-time frame-level image preprocessing through FPGA, eliminates interference from direct sunlight and shadow diffusion, and outputs a grayscale difference map for sub-pixel measurement of cracks in real time. The deep learning front-end quickly screens the crack ROI based on the distilled YOLOv4-Tiny model and generates an edge suggestion box. The data fusion and damage identification module then fuses the DIC and AI detection results.

[0008] It should be noted that the multi-sensor data acquisition module also integrates a DIC support sub-module and a deep learning terminal. The DIC support sub-module whitewashes the specimen surface before the test and uses ink lines to pop up a 50mm×50mm positioning grid. It collects reflection images in different bands through a multi-spectral light source and an industrial camera, and uses FPGA to perform real-time frame-level denoising, illumination balancing, and shadow separation, and outputs a sub-pixel grayscale difference map of the cracks. The deep learning front-end is based on the distilled YOLOv4-Tiny model to quickly screen the crack ROI in the grayscale difference map and generate an edge suggestion box. The data fusion and damage identification module then fuses the DIC and AI detection results to achieve high-precision positioning and quantitative analysis of structural cracks.

[0009] As a further solution of the present invention, the data fusion and damage identification module uses a two-layer adaptive Kalman filter framework. The first layer uses an extended Kalman filter to fuse the displacement / strain data of the LVDT and FBG. The second layer uses an improved wavelet packet decomposition based on the Kalman filter output to perform more than 5 sub-band sub-band decomposition on the acoustic emission time domain signal, extracting three types of features: envelope energy, spectral entropy, and instantaneous frequency. Combined with an online incremental support vector machine, real-time classification is performed to distinguish between crack initiation, high-frequency tearing noise, and friction slip events. When abnormal data points such as strain mutation and LVDT distortion occur, the redundant sensor fusion fallback strategy is automatically activated.

[0010] It should be noted that the data fusion and damage identification module adopts a two-layer adaptive Kalman filter framework: the first layer uses the TST3826E static signal test and analysis system to collect real-time displacement and strain data from LVDT displacement sensors installed on the left and right support sections and mid-span positions, as well as BF120-5AA and BMB120-80AA strain gauges attached to the web of the steel beam, the UHPC top plate and the top, side and bottom surfaces of the ordinary concrete slab. The extended Kalman filter algorithm is used to perform spatiotemporal coupling fusion and high-frequency noise filtering on the multi-source data; the second layer uses the above filter Based on the wave output, the time domain signal collected by the acoustic emission sensor is decomposed into sub-bands using wavelet packets of level five or above, and the envelope energy, spectral entropy, instantaneous frequency and multi-scale statistical features are extracted. Dynamic classification and identification are then performed in combination with an online incremental support vector machine to distinguish between crack initiation, high-frequency tearing noise and friction and slip events at the steel-UHPC interface. At the same time, when abnormal measurement points such as strain mutation, LVDT jump or strain gauge distortion are detected, the module can automatically activate the fusion fallback strategy of the built-in FBG sensor and the backup YWC strain gauge displacement meter to ensure the continuity of monitoring data and the robustness of algorithm recognition.

[0011] As a further solution of the present invention, the hot spot stress calibration module automatically generates a microscopic sub-model that matches the in-situ crack morphology for online reconstruction after each load step. Based on a predefined secondary mesh refinement strategy, this sub-model refines the mesh at a ratio of 1:10 within a set range around the crack tip, retaining the coarse mesh of the overall structural model. Then, a single-step dynamic explicit simulation is used to accurately regress the stress field at the crack tip. The simulation results are then inverted with the crack deformation measured by DIC using the least squares error method, and the extrapolation coefficient is updated.

[0012] It should be noted that after each load step is completed, the hotspot stress calibration module first automatically splices and maps the crack morphology data captured by the DIC sub-pixel grayscale difference map arranged on the top and side surfaces of the junction segment and the high-density strain rosette array to a predefined microscopic sub-model based on the overall finite element coarse mesh. Then, a secondary mesh refinement strategy of 1:10 is used at the crack tip and its periphery to locally refine only this area, while retaining the overall coarse mesh of the structure. ABAQUS single-step dynamic explicit simulation is then used to calculate the crack tip stress field distribution of this sub-model under the current load step to obtain a high-precision hotspot stress value. This value is then fitted with the crack opening displacement and local strain field obtained by DIC inversion using the least squares method to obtain the extrapolation coefficient with the minimum inversion error. Finally, the updated extrapolation coefficient and the stress results of the reconstructed sub-model are fed back to the subsequent load step calibration algorithm, realizing the online iteration and continuous calibration of the fine interface hotspot stress required for fatigue S–N curve correction.

[0013] As a further solution of the present invention, the early warning decision module constructs a dual closed-loop fatigue life prediction framework based on Monte Carlo random sampling on the basis of the original dual thresholds. First, the real-time measured crack widening rate and lateral slip rate are called as the initial input, and the corresponding Miner damage evolution curve is generated through more than 1,000 Monte Carlo simulations. Then, the fatigue crack propagation curve is dynamically calculated on each curve according to the Paris law, and the propagation exponent m and coefficient are corrected in real time based on the acoustic emission count. When the crack width or slip rate exceeds the set threshold, an early warning is triggered, and a moderate warning is automatically issued when the 75th percentile of the cumulative damage D distribution of all simulated curves exceeds 0.7. When the 90th percentile D ≥ 1, it switches to a high-level red alarm.

[0014] It should be noted that the early warning decision module, while retaining the dual threshold trigger based on crack width 0.2mm and lateral slip 0.5mm, integrates Monte Carlo fatigue failure criterion in accordance with GB / T50082-2009. Carlo random sampling double closed-loop life prediction framework: This framework uses the crack widening rate of each step measured by edge preprocessing and DIC and the lateral slip rate inverted by LVDT / FBG as initial input. According to the statistical distribution of the first crack load and ultimate failure load recorded in the static test, the corresponding Miner cumulative damage D evolution curve is generated through more than 1000 sampling times. On each curve, the crack propagation index m and coefficient C are adaptively corrected in real time according to the Paris law combined with the DH3816 acoustic emission counting. The fatigue crack propagation curve is dynamically calculated and a first-level warning is immediately triggered when the crack width or slip rate of any curve exceeds the set threshold. When the 75th percentile of the D distribution of all simulated curves exceeds 0.7, it switches to a moderate yellow warning. When the 90th percentile D ≥ 1, it is upgraded to a high-level red alarm, and SMS / email notifications are pushed to bridge operation and maintenance personnel through the visual remote communication module, realizing a full-process digital warning closed loop from static load testing to online monitoring.

[0015] As a further solution of the present invention, the digital twin calibration module iteratively updates the model material parameters and contact friction coefficient online and adaptively corrects the finite element model based on the Bayesian-least squares hybrid parameter identification method. The process includes:

[0016] Step 1: Least squares initial identification: Use least squares to fit the deviation between the on-site hot spot stress and the FEM predicted stress to obtain an initial estimate of the material elastic modulus and friction coefficient;

[0017] Step 2, Bayesian posterior inference: Using this estimate as a Bayesian prior, combined with the joint likelihood function of each field point, the Markov chain Monte Carlo method is used to generate the posterior distribution and automatically calculate the 95% confidence interval;

[0018] Step 3: Digital twin parameter update: Update the digital twin model parameters with the a posteriori mean and push the uncertainty information to the visual remote communication module.

[0019] It should be noted that after completing the preliminary correction of the fatigue S-N curve based on the hot spot stress method, the digital twin calibration module also integrates a Bayesian-least squares hybrid parameter identification process to iteratively update the material parameters and contact surface friction coefficient of the finite element model online: First, at the end of each load step, the interface hot spot stress obtained by inverting the on-site crack tip strain collected by the edge preprocessing module and the TST3826E static signal test and analysis system is fitted with the FEM predicted stress extracted at the nodes 0.4 times and 1.0 times the plate thickness from the weld toe in the software ABAQUS by least squares fitting, and the initial estimates of the material elastic modulus E and the interface friction coefficient μ are quickly obtained. Then, this estimate is used as a Bayesian prior to introduce the joint likelihood function of multi-directional strain sensors deployed on-site on the top, side, and bottom surfaces of the cross-section. The Markov Chain Monte Carlo (MCMC) algorithm is used to generate a posterior distribution in a cloud or local digital twin environment, and a 95% confidence interval is automatically calculated to quantify parameter uncertainty. Finally, E and μ in the digital twin model are updated with the mean of the posterior distribution, and the parameter update records and their credible intervals are synchronously pushed to the visual remote communication module through the web dashboard and mobile app, thereby realizing adaptive correction of the finite element model of the fixed end of the steel-concrete composite beam and online digital twin closed-loop maintenance.

[0020] As a further solution of the present invention, the visual remote communication module displays the load-deflection, crack evolution, and hot spot stress field in real time through a web dashboard and mobile app, and pushes SMS / email. The edge preprocessing module performs bandpass filtering and GPS clock synchronization on the signal and then electrically connects it to the data fusion and damage identification module. The energy health management module is based on a hybrid power supply of piezoelectric vibration and solar energy and performs periodic self-inspection and fault switching on each node.

[0021] It should be noted that the visualization remote communication module uses the load-deflection curves, LVDT displacements, acoustic emission counts, DIC grayscale difference maps, and ABAQUS finite element hotspot stress fields collected by the DH3816 static signal test system and TST3826E data. It performs real-time three-dimensional rendering through a microservices-based web dashboard and mobile app, and pushes SMS / email when a set threshold is triggered. The edge preprocessing module implements CORDIC sixth-order bandpass filtering on the FBG strain gauge and LVDT signals on an FPGA+ARM heterogeneous platform, combines it with GPS-PTP precision clock synchronization, and then electrically connects it to the data fusion and damage identification module. The energy health management module integrates micro photovoltaic panels and bridge deck vibration piezoelectric energy harvesting circuits, uses a bidirectional DC-DC converter to dynamically switch power between supercapacitors and lithium batteries, and performs periodic self-tests and fault switching on each node to ensure long-term continuous and stable operation of the system.

[0022] As a further solution of the present invention, the visual remote communication module adopts containerized deployment based on microservice architecture and WebSocket two-way push technology, and performs three-dimensional rendering and thermal map superposition of the real-time collected load-deflection data, crack evolution curve and FEM hot spot stress field through WebGL. It supports multi-perspective rotation, scaling and cross-sectional cutting, and sets dynamic query and threshold valves on the page end, freely configures alarm display rules, and automatically switches to local IndexedDB cache mode when the network is interrupted. After the network is restored, all unsent key data is supplemented in batch increments.

[0023] In this solution, the visual remote communication module is deployed through a containerized microservices architecture. It implements bidirectional real-time push of multi-source sensor data and FEA interface hotspot stress fields based on WebSocket. It also uses WebGL to perform three-dimensional visualization of load-deflection curves, crack evolution images, and hotspot stress cloud maps. It supports multi-perspective rotation, zooming, and cross-sectional cutting. The front-end interface provides dynamic query and threshold gating functions. Users can customize alarm rules. When the network is interrupted, it automatically switches to the IndexedDB local cache to store all key monitoring data. After the link is restored, it retransmits the unsent data in batch increments to ensure the integrity and real-time performance of the crack monitoring data at the fixed end of the composite deck beam.

[0024] As a further solution of the present invention, an edge preprocessing module is built on an FPGA+ARM heterogeneous platform. It uses a sixth-order bandpass filter configured based on the CORDIC algorithm to parallel suppress 50Hz power grid noise and structural vibration interference. In combination with a multi-frequency GPS receiver and the IEEE 1588 precision clock synchronization protocol, it performs nanosecond timestamp alignment, runs adaptive threshold event detection and dynamic sampling rate control logic, and automatically wakes up the corresponding sensor node when a transient burst pulse or a sudden change in DIC image edge contrast is detected, and completes data packaging and transmission within 5ms.

[0025] In this solution, the edge preprocessing module is deployed on an FPGA+ARM heterogeneous platform to perform frame-level preprocessing on LVDT, strain gauge, acoustic emission, and DIC image signals. A sixth-order bandpass filter is constructed in parallel based on the CORDIC algorithm to effectively suppress 50Hz power frequency interference and structural vibration noise. Multi-frequency GPS and the IEEE 1588 protocol are combined to achieve nanosecond clock synchronization. Digital compensation and adaptive balancing logic are used to correct drift in real time and eliminate abnormal measurement points. Adaptive threshold event detection and dynamic sampling rate control are implemented. When a sudden change in strain or crack grayscale is detected, the corresponding sensor node is automatically awakened within 5ms and data is packaged and sent, effectively ensuring the accuracy and timeliness of signals throughout the entire crack monitoring process.

[0026] As a further solution of the present invention, the system also includes an energy health management module, which integrates an MPPT photovoltaic power management chip and a piezoelectric resonant energy harvesting circuit. It dynamically switches between the vanadium battery supercapacitor and the small lithium-ion battery through a bidirectional DC-DC converter. It prioritizes solar power supply under sunshine conditions and piezoelectric power generation at night and when the vehicle vibrates. It has an internal embedded SOC and SOH estimation algorithm, monitors voltage, current, and ambient temperature and humidity in real time, and sends energy health reports to the cloud via LoRaWAN. When the node energy is lower than 20% or the output power is detected to be continuously decreasing, it automatically switches to the ultra-low power sleep wake-up mode.

[0027] In this solution, the energy health management module integrates a high-efficiency MPPT photovoltaic power management chip and a high-Q piezoelectric resonant energy harvesting circuit. It balances the energy flow between the vanadium battery + supercapacitor and the small lithium-ion battery in real time through a bidirectional DC-DC converter. It prioritizes solar power supply when there is sufficient sunshine, and automatically switches to piezoelectric power generation at night and when the vehicle vibrates. It embeds an SOC / SOH estimation algorithm based on the fusion of Kalman filtering and neural networks, continuously collects and analyzes multi-dimensional environmental and power status data such as voltage, current, temperature, and humidity, and periodically reports energy health reports to the cloud through the LoRaWAN node. When it detects that the remaining power of the node is less than 20% or the output power continues to decline, it automatically triggers the ultra-low power sleep wake-up strategy to ensure that the system can still maintain key monitoring and communication functions under extreme conditions.

[0028] The technical effects of the crack monitoring system for the fixed end of a steel-concrete composite beam of the present invention are as follows:

[0029] The present invention integrates a multi-sensor data acquisition module of LVDT, FBG, acoustic emission, DIC and strain gauge, and relies on FPGA real-time frame-level filtering and distilled YOLOv4-Tiny model for rapid ROI identification to achieve synchronous high-precision measurement of millimeter-level displacement, microstrain and sub-pixel crack width in complex lighting and vibration environments; edge preprocessing and double-layer adaptive Kalman filtering + improved wavelet packet decomposition fuse multi-source signals and online incremental SVM classification to distinguish crack initiation, high-frequency tearing noise and friction slip in real time, automatically start sensor redundancy fallback, and reduce false alarm and missed alarm rate; based on the linear extrapolation of stress at nodes 0.4 times / 1.0 times the plate thickness from the weld toe, supplemented by DIC crack deformation least squares inversion and secondary grid refinement sub-model dynamic explicit simulation, the error between FEM prediction and actual stress on site is reduced to within 5%, and the fatigue S-N curve is corrected in a targeted manner; the early warning decision module uses more than 1000 Monte Carlo random sampling and double closed-loop Paris method dynamic crack propagation simulation, and real-time correction of the propagation index based on acoustic emission, triggering moderate and advanced red alarms at 75% / 90% cumulative damage quantile thresholds to achieve dynamic quantification of remaining life; the digital twin calibration module uses least squares initial identification and Bayesian / MCMC posterior inference to iteratively update the material elastic modulus and contact friction coefficient online, adaptively correct the finite element model and continuously optimize the simulation; the visual remote communication module provides multi-perspective sectioning and threshold customized alarms based on microservice architecture and WebGL 3D rendering; the energy health management module uses solar energy + piezoelectric hybrid energy extraction, MPPT power management and SOC / SOH estimation to ensure all-weather ultra-low power autonomous operation, improving the accuracy, reliability and intelligence level of crack monitoring at the fixed end of steel-concrete composite beams. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1This is a schematic diagram of the transverse bonding section in the prior art of the present invention;

[0031] Figure 2 This is a structural diagram of UHPC cast-in-situ joints in the prior art of the present invention;

[0032] Figure 3 This is a distribution diagram of the steel bar measurement points of the present invention;

[0033] Figure 4 The distribution diagram of the measuring points on the side and top of the concrete cross section of the specimen of the present invention;

[0034] Figure 5 The distribution of measuring points on the bottom surface of the test piece and the arrangement of strain gauges on the steel web are shown in the figure;

[0035] Figure 6 This is the cross-sectional strain arrangement diagram of the specimen of the present invention;

[0036] Figure 7 This is the distribution diagram of concrete cracks in the transverse section of the test specimen of the present invention;

[0037] Figure 8 This is a diagram showing the side strain data recorded at the transverse joint section of the steel-concrete composite bridge deck according to the present invention;

[0038] Figure 9 This is a graph recording strain data on the top surface of the transverse joint section of the steel-concrete composite bridge deck of the present invention;

[0039] Figure 10 This is a record of strain data on the bottom surface of the transverse joint section of the steel-concrete composite bridge deck of the present invention. DETAILED DESCRIPTION

[0040] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0041] Example 1

[0042] The present invention proposes a crack monitoring system for the fixed end of a steel-concrete composite beam, comprising a multi-sensor data acquisition module, an edge preprocessing module, a data fusion and damage identification module, a hot spot stress calibration module, an early warning decision module, a visual remote communication module and a digital twin calibration module. The multi-sensor data acquisition module synchronously collects displacement, strain, acoustic emission and crack images through LVDT, FBG, acoustic emission, DIC and strain gauges and is electrically connected to the edge preprocessing module. The edge preprocessing module transmits the preprocessed data to the data fusion and damage identification module. The data fusion and damage identification module uses extended Kalman filtering and wavelet packet decomposition to fuse multi-source data, extract crack initiation / propagation features and is electrically connected to the hot spot. Point stress calibration module and early warning decision module. The hot spot stress calibration module reads the finite element node stress at 0.4 times the plate thickness and 1.0 times the plate thickness from the weld toe, calculates the interface hot spot stress according to the linear proportion of first taking the high stress and then subtracting the low stress, and corrects the fatigue S-N curve in combination with the strain inversion results measured on site. The calibration information is then electrically connected to the digital twin calibration module. The early warning decision module issues a three-level alarm based on the dual thresholds of 0.2mm crack width and 0.5mm lateral slip and Miner cumulative damage, and is electrically connected to the visual remote communication module. The digital twin calibration module uses the least squares method to iteratively update the model material parameters and contact friction coefficient online and adaptively correct the finite element model.

[0043] Among them, the multi-sensor data acquisition module also includes a DIC support sub-module and a deep learning terminal. The DIC support sub-module uses the difference in multi-spectral reflection intensity on the structural surface to enhance the contrast of the crack contour, performs real-time frame-level image preprocessing through FPGA, eliminates interference from direct sunlight and shadow diffusion, and outputs grayscale difference maps for sub-pixel measurement of cracks in real time. The deep learning front-end quickly screens the crack ROI based on the distilled YOLOv4-Tiny model and generates edge suggestion boxes. The data fusion and damage identification module then fuses the DIC and AI detection results.

[0044] Thirdly, the edge preprocessing module is built on an FPGA+ARM heterogeneous platform. It uses a sixth-order bandpass filter based on the CORDIC algorithm to suppress 50Hz power grid noise and structural vibration interference in parallel. It combines a multi-frequency GPS receiver and the IEEE 1588 precision clock synchronization protocol to perform nanosecond timestamp alignment, run adaptive threshold event detection and dynamic sampling rate control logic, and automatically wake up the corresponding sensor node when a transient burst pulse or a sudden change in the edge contrast of the DIC image is detected, and completes data packaging and transmission within 5ms.

[0045] Thirdly, the data fusion and damage identification module uses a two-layer adaptive Kalman filter framework. The first layer uses extended Kalman filtering to fuse the displacement / strain data of LVDT and FBG. The second layer uses improved wavelet packet decomposition based on the Kalman filter output to perform more than 5 sub-band divisions on the acoustic emission time domain signal, extract three types of features: envelope energy, spectral entropy and instantaneous frequency, and combine it with an online incremental support vector machine for real-time classification to distinguish crack initiation, high-frequency tearing noise and friction slip events. When abnormal data points such as strain mutation and LVDT distortion occur, the redundant sensor fusion fallback strategy is automatically activated.

[0046] In addition, after each load step, the hotspot stress calibration module automatically generates a microscopic sub-model that matches the in-situ crack morphology for online reconstruction. Based on a predefined secondary mesh refinement strategy, this sub-model refines the mesh at a ratio of 1:10 within a set range around the crack tip, retaining the coarse mesh of the overall structural model. The stress field at the crack tip is then accurately regressed through a single-step dynamic explicit simulation. The simulation results are then inverted with the crack deformation measured by DIC using the least squares error method, and the extrapolation coefficient is updated. Based on the original dual thresholds, the early warning decision module constructed a dual closed-loop fatigue life prediction framework based on Monte Carlo random sampling. First, the real-time measured crack widening rate and lateral slip rate were used as initial inputs, and the corresponding Miner damage evolution curve was generated through more than 1,000 Monte Carlo simulations. Then, the fatigue crack propagation curve was dynamically calculated on each curve according to the Paris law, and the propagation exponent m and coefficient were corrected in real time based on the acoustic emission counts. An early warning was triggered when the crack width or slip rate exceeded the set threshold, and a moderate warning was automatically issued when the 75th percentile of the cumulative damage D distribution of all simulated curves exceeded 0.7. When the 90th percentile D ≥ 1, it switched to a high-level red alarm.

[0047] Specifically, the digital twin calibration module uses the Bayesian-least squares hybrid parameter identification method to iteratively update the model material parameters and contact friction coefficient online and adaptively correct the finite element model. The process includes:

[0048] Step 1: Least squares initial identification: Use least squares to fit the deviation between the on-site hot spot stress and the FEM predicted stress to obtain an initial estimate of the material elastic modulus and friction coefficient;

[0049] Step 2, Bayesian posterior inference: Using this estimate as a Bayesian prior, combined with the joint likelihood function of each field point, the Markov chain Monte Carlo method is used to generate the posterior distribution and automatically calculate the 95% confidence interval;

[0050] Step 3: Digital twin parameter update: Update the digital twin model parameters with the a posteriori mean and push the uncertainty information to the visual remote communication module.

[0051] The visual remote communication module then displays load-deflection, crack evolution, and hotspot stress fields in real time via a web dashboard and mobile app, and pushes SMS / email. The edge preprocessing module performs bandpass filtering and GPS clock synchronization on the signals before electrically connecting them to the data fusion and damage identification module. The energy health management module, powered by a hybrid of piezoelectric vibration and solar power, performs periodic self-tests and fault switching on each node. The visual remote communication module utilizes containerized deployment based on a microservices architecture and WebSocket bidirectional push technology. It uses WebGL to render real-time load-deflection data, crack evolution curves, and FEM hotspot stress fields in 3D and overlay them with thermal maps. It supports multi-view rotation, zooming, and cross-sectioning. Dynamic queries and thresholds can be configured on the page, and alarm display rules can be freely configured. In the event of a network interruption, it automatically switches to local IndexedDB caching mode and retransmits all unsent critical data in batches and incrementally after the network is restored.

[0052] In addition, the system also includes an energy health management module, which integrates an MPPT photovoltaic power management chip and a piezoelectric resonant energy harvesting circuit. It dynamically switches between vanadium battery supercapacitors and small lithium-ion batteries through a bidirectional DC-DC converter. It prioritizes solar power supply under sunny conditions and piezoelectric power generation at night and when the vehicle vibrates. It has an internal embedded SOC and SOH estimation algorithm, monitors voltage, current, and ambient temperature and humidity in real time, and sends energy health reports to the cloud via LoRaWAN. When the node energy is lower than 20% or a continuous decline in output power is detected, it automatically switches to an ultra-low power sleep wake-up mode.

[0053] In order to clearly illustrate the technical effects of the above technical solutions proposed by the present invention, the following laboratory static and fatigue test arrangements were carried out:

[0054] Model components: 1:2 scale steel-UHPC composite beam specimens (see original Figures 3 to 6 ), the transverse joint section is in the middle of the model span, the UHPC material strength is C60, the thickness is 50mm, and the ordinary concrete is C30; the longitudinal tensile steel bar has a diameter of 12mm and a spacing of 150mm;

[0055] Loading device: Both static and fatigue loading use a 500kN hydraulic jack, acting on the mid-span position through a geometric range extension system, increasing in steps of 100kN to the ultimate load or 2 million fatigue cycles, with a loading rate controlled at 0.5mm / min;

[0056] Surface preparation: The specimen surface was painted with white latex paint and positioned using a 50mm×50mm grid ( Figure 7 ), which facilitates the observation of crack distribution with naked eyes and DIC.

[0057] The sensors are arranged as follows:

[0058] Displacement measurement: LVDTs (range 50mm, 150mm) are arranged at the left and right supports and in the mid-span to record the overall deflection;

[0059] Strain measurement points: (1) Steel beam: BF120-5AA(11)-P150-D metal strain gauge, grid length 5 mm × 3 mm; (2) Concrete slab: BMB120-80AA-150-D, strain gauge grid length 80 mm × 3 mm (see Table 1); (3) DIC: dual-camera stereo vision system, resolution 5 μm, capable of measuring crack width at the sub-pixel level; (4) Acoustic emission: 8 sensors are placed to collect 100 kHz–1 MHz signals; (5) Multi-sensor data acquisition module: TST3826E / DH3816 static signal test and analysis system, multi-channel synchronous sampling, GPS / IEEE1588 calibrated clock, sampling rate up to 20 kHz.

[0060] The test equipment layout and result statistics are shown in Table 1:

[0061] Table 1 Test equipment layout and result statistics

[0062]

[0063] The key monitoring results are as follows:

[0064] Crack initiation and propagation: Figure 8 The curves of the side strain measurement points J-1-CC-1 to C-15 with load change are shown. The maximum side strain of 339.6με appears at 420kN (measurement point J-1-CC-6), corresponding to the first visible crack width of 0.05mm. After 400kN, the crack quickly expands to the entire span ( Figure 7 );

[0065] Top and bottom surface strains: Figure 9 and Figure 10 In the middle, the strain on the top surface increases to 46.5 με (J-1-CS-5), and the peak value on the bottom surface is 28.3 με (J-1-CX-4), which is consistent with the hot spot stress of 185 MPa predicted by FEM, proving that the error of the S–N curve after calibration of the calibration module is <4%;

[0066] Digital twin calibration: Based on the least squares fitting in step one, the material elastic modulus was adjusted from the original 30 GPa to 31.2 GPa, and the friction coefficient μ was optimized from 0.35 to 0.28. In step two, the 95% confidence interval of the Bayesian posterior distribution was [30.8, 31.6] GPa. After the automatic update, the maximum residual difference between the FEM load-deformation curve and the test curve was only 3.8%.

[0067] Fatigue life prediction verification: 1000 Monte Carlo simulations combined with the Paris law extended model corrected by acoustic emission showed a deviation of less than 8% from the crack width and length measurements after 2 million fatigue cycles. The 75% cumulative damage was D = 0.71 and the 90% cumulative damage was D = 1.02, achieving a level 3 warning accuracy of 95%.

[0068] The technical solution proposed by the present invention has significant technical effects in the following aspects:

[0069] High-sensitivity crack detection: Based on DIC sub-pixel measurement and fusion of EKF+incremental SVM, the crack initiation threshold is lowered from the traditional 480kN to 400kN, providing 17% early warning.

[0070] Improved stress calibration accuracy: Dynamic refinement of the microscopic sub-model (1:10) and least squares inversion increased the consistency between FEM and measured hotspot stresses to 96%, and increased the reliability of the corrected S–N curve from 85% to 95%.

[0071] Model adaptive update: The digital twin module iterates material parameters and friction coefficients in real time, ensuring that the residual error of the laboratory model under different loading conditions is ≤5%, significantly better than the 12% residual error of the uncalibrated model.

[0072] Example 2

[0073] The difference between Example 2 of the present invention and Example 1 is that this example introduces a long-term online monitoring application of a crack monitoring system for the fixed end of a steel-concrete composite beam, focusing on the on-site deployment and operation effects of the remote visualization remote communication module, edge preprocessing hardware platform and energy health management module.

[0074] For the application scenario proposed in the background technology, eight sensor nodes were deployed at each mid-span and support. Each group includes displacement (LVDT), strain (FBG, metal strain gauge), acoustic emission (AE), DIC (binocular camera), and crack propagation camera. An edge preprocessing unit (FPGA + ARM) and an energy health management unit were integrated into the steel anchor box. LoRaWAN and 4G dual channels were used for parallel security, and the edge module pushed raw time series and compressed feature data to the cloud. The node layout information is shown in Table 2:

[0075] Table 2 Node layout information table

[0076]

[0077] Among them, the hardware platform of the edge pre-processing module includes Xilinx Zynq-7000 (ARM Cortex-A9+FPGA), CORDIC-accelerated sixth-order band-pass filtering and parallel suppression of 50Hz power frequency noise and structural vibration, GPS multi-frequency and IEEE1588 phase-locked to achieve nanosecond timestamps, adaptive threshold event detection (displacement mutation, acoustic emission pulse, DIC edge contrast mutation), dynamic sampling rate control (1Hz when there is no event, 20kHz when there is an event). The backend is deployed in a Docker container and uses WebSocket for two-way push. Front-end WebGL rendering: load-deflection curve, crack evolution thermal map, hot spot stress field three-dimensional overlay, support multi-view and dynamic sectioning. Network disconnection cache: based on IndexedDB, 7 days of data are stored locally; after the network is restored, the unreached data is supplemented in batch increments. The supplementary transmission time data corresponding to the network disconnection time is shown in Table 3 below:

[0078] Table 3

[0079] Internet disconnection duration Cache data volume Re-pass duration 3h 1.2GB 5min 12h 4.8GB 18min

[0080] The energy health management module enables multi-source energy data collection (solar energy: a 10W peak monocrystalline silicon MPPT chip can charge 150Wh after 6 hours of sunshine; piezoelectric energy: bridge vehicle vibration power generation, an average of 20Wh per day). Vanadium battery supercapacitors store short-term impact energy, and small Li-ion batteries are used for nighttime and cloudy conditions. SOC / SOH estimation is based on an extended Kalman filter and is updated every 15 minutes. The low-battery strategy is to enter an ultra-low-power sleep-wake-up mode when the SOC is <20% or the output power is <5W for 3 consecutive hours. The specific power supply time periods and node status are shown in Table 4:

[0081] Table 4 Energy health management module node status table

[0082] Time period Solar power supply Piezoelectric power supply Node Status Daytime (6–18 hours) 90％ 10％ Full module online Nighttime (18–6 hours) 0％ 5％ Some sensors wake up Rainy / Cloudy (all day) 30％ 15％ Periodic self-test

[0083] In the above test setting environment, the system operated continuously for 180 days, with no data loss during network disconnection and retransmission. A total of 12 three-level warnings were issued, including 7 moderate and 5 high warnings. All maintenance was carried out within less than 2 hours after the joint indication of acoustic emission and DIC, and the energy self-sufficiency rate reached 85%.

[0084] In this embodiment, the system of the present invention deployed a total of 8 groups of sensor nodes in the main span and support of the test, and ran continuously for 180 days. During this period, the FPGA+ARM heterogeneous edge platform was used to achieve real-time preprocessing and feature extraction of LVDT, FBG, acoustic emission, DIC and strain gauge multi-source signals at nanosecond synchronization. Combined with the microservice architecture and WebGL 3D visualization (multi-view rotation, zoom and cross-section) and dynamic threshold valve configuration, a 99.2% data integrity rate and zero manual intervention were achieved. IndexedDB offline caching and batch incremental retransmission were used to ensure that key data was not lost. Monte Carlo random sampling and Miner cumulative damage model triggered 12 level 3 alarms (7 moderate and 5 high) and accurately located crack initiation and expansion 2.3 hours in advance on average; the Bayesian-least squares digital twin method was used to control the residual between the finite element predicted stress and the on-site measured hot spot stress to within 5%; the energy health management module achieved an 85% system self-sufficiency rate under mixed power supply of sunlight and vibration, and automatically switched to ultra-low power sleep wake-up mode when the battery was low. At the same time, the operation and maintenance cost was reduced by about 60% compared with the traditional solution. Ultimately, it was verified that this system has the comprehensive technical effects of high reliability, high sensitivity, low energy consumption, long life, accurate early warning and easy maintenance in a complex real bridge environment.

[0085] In summary, the present invention integrates a multi-sensor data acquisition module of LVDT, FBG, acoustic emission, DIC and strain gauge, and relies on FPGA real-time frame-level filtering and distilled YOLOv4-Tiny model for rapid ROI identification to achieve synchronous high-precision measurement of millimeter-level displacement, microstrain and sub-pixel crack width in complex lighting and vibration environments; edge preprocessing and double-layer adaptive Kalman filtering + improved wavelet packet decomposition fuse multi-source signals and online incremental SVM classification, real-time distinction between crack initiation, high-frequency tearing noise and friction slip, automatic activation of sensor redundancy fallback, and reduction of false alarm and missed alarm rate; based on linear extrapolation of stress at nodes 0.4 times / 1.0 times the plate thickness from the weld toe, supplemented by DIC crack deformation least squares inversion and secondary grid refinement sub-model dynamic explicit simulation, the error between FEM prediction and actual stress on site is reduced to within 5%, and fatigue S-N curve is corrected in a targeted manner; the early warning decision module uses more than 1000 Monte Carlo random sampling and double closed-loop Paris method dynamic crack propagation simulation, and real-time correction of the propagation index based on acoustic emission, triggering moderate and advanced red alarms at 75% / 90% cumulative damage quantile thresholds to achieve dynamic quantification of remaining life; the digital twin calibration module uses least squares initial identification and Bayesian / MCMC posterior inference to iteratively update the material elastic modulus and contact friction coefficient online, adaptively correct the finite element model and continuously optimize the simulation; the visual remote communication module provides multi-perspective sectioning and threshold customized alarms based on microservice architecture and WebGL 3D rendering; the energy health management module uses solar energy + piezoelectric hybrid energy extraction, MPPT power management and SOC / SOH estimation to ensure all-weather ultra-low power autonomous operation, improving the accuracy, reliability and intelligence level of crack monitoring at the fixed end of steel-concrete composite beams.

[0086] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0087] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A crack monitoring system for the fixed end of a steel-concrete composite beam, characterized in that: It includes a multi-sensor data acquisition module, an edge preprocessing module, a data fusion and damage identification module, a hot spot stress calibration module, an early warning decision module, a visual remote communication module and a digital twin calibration module. The multi-sensor data acquisition module synchronously collects displacement, strain, acoustic emission and crack images through LVDT, FBG, acoustic emission, DIC and strain gauges and is electrically connected to the edge preprocessing module. The edge preprocessing module transmits the preprocessed data to the data fusion and damage identification module. The data fusion and damage identification module uses extended Kalman filtering and wavelet packet decomposition to fuse multi-source data, extract crack initiation / extension features and is electrically connected to the hot spot stress calibration module and the early warning decision module. The hot spot stress calibration module reads the finite element node stress at 0.4 times the plate thickness and 1.0 times the plate thickness from the weld toe, calculates the interface hot spot stress in a linear ratio of first taking the high stress and then subtracting the low stress, and corrects the fatigue S-N curve in combination with the strain inversion results measured on site. The calibration information is then electrically connected to the digital twin calibration module. The early warning decision module issues a three-level alarm based on the dual thresholds of 0.2 mm crack width and 0.5 mm lateral slip and Miner cumulative damage, and is electrically connected to the visual remote communication module. The digital twin calibration module uses the least squares method to iteratively update the model material parameters and contact friction coefficient online and adaptively correct the finite element model.

2. A crack monitoring system for fixed ends of steel-concrete composite beams according to claim 1, characterized in that: The multi-sensor data acquisition module also includes a DIC support sub-module and a deep learning terminal. The DIC support sub-module uses the difference in multi-spectral reflection intensity on the structural surface to enhance the contrast of the crack contour, performs real-time frame-level image preprocessing through FPGA, eliminates interference from direct sunlight and shadow diffusion, and outputs grayscale difference maps for sub-pixel measurement of cracks in real time. The deep learning front-end quickly screens the crack ROI and generates edge suggestion boxes based on the distilled YOLOv4-Tiny model. The data fusion and damage identification module then fuses the DIC and AI detection results.

3. The crack monitoring system for the fixed end of a steel-concrete composite beam according to claim 1, characterized in that: The data fusion and damage identification module uses a two-layer adaptive Kalman filter framework. The first layer uses an extended Kalman filter to fuse the displacement / strain data of the LVDT and FBG. The second layer uses an improved wavelet packet decomposition based on the Kalman filter output to perform more than five sub-band sub-band divisions on the acoustic emission time domain signal, extracting three types of features: envelope energy, spectral entropy, and instantaneous frequency. Combined with an online incremental support vector machine, it performs real-time classification to distinguish between crack initiation, high-frequency tearing noise, and friction slip events. When abnormal data points such as strain mutations and LVDT distortion occur, the redundant sensor fusion fallback strategy is automatically activated.

4. The crack monitoring system for the fixed end of a steel-concrete composite beam according to claim 1, characterized in that: After each load step, the hotspot stress calibration module automatically generates a microscopic sub-model that matches the in-situ crack morphology for online reconstruction. Based on a predefined secondary mesh refinement strategy, this sub-model refines the mesh at a ratio of 1:10 within a set range around the crack tip, retaining the coarse mesh of the overall structural model. The stress field at the crack tip is then accurately regressed through a single-step dynamic explicit simulation. The simulation results are then inverted with the crack deformation measured by DIC using the least squares error method, and the extrapolation coefficient is updated.

5. The crack monitoring system for the fixed end of a steel-concrete composite beam according to claim 1, characterized in that: Based on the original dual thresholds, the early warning decision module constructed a dual closed-loop fatigue life prediction framework based on Monte Carlo random sampling. First, the real-time measured crack widening rate and lateral slip rate were used as initial inputs, and the corresponding Miner damage evolution curve was generated through more than 1,000 Monte Carlo simulations. Then, the fatigue crack propagation curve was dynamically calculated on each curve according to the Paris law, and the propagation exponent m and coefficient were corrected in real time based on the acoustic emission counts. An early warning was triggered when the crack width or slip rate exceeded the set threshold, and a moderate warning was automatically issued when the 75th percentile of the cumulative damage D distribution of all simulated curves exceeded 0.

7. When the 90th percentile D ≥ 1, it switched to a high-level red alarm.

6. The crack monitoring system for the fixed end of a steel-concrete composite beam according to claim 1, characterized in that: The digital twin calibration module uses the Bayesian-least squares hybrid parameter identification method to iteratively update the model material parameters and contact friction coefficient online and adaptively correct the finite element model. The process includes: Step 1: Least squares initial identification: Use least squares to fit the deviation between the on-site hot spot stress and the FEM predicted stress to obtain an initial estimate of the material elastic modulus and friction coefficient; Step 2, Bayesian posterior inference: Using this estimate as a Bayesian prior, combined with the joint likelihood function of each field point, the Markov chain Monte Carlo method is used to generate the posterior distribution and automatically calculate the 95% confidence interval; Step 3: Digital twin parameter update: Update the digital twin model parameters with the a posteriori mean and push the uncertainty information to the visual remote communication module.

7. The crack monitoring system for the fixed end of a steel-concrete composite beam according to claim 1, characterized in that: The visual remote communication module displays load-deflection, crack evolution, and hotspot stress fields in real time through a web dashboard and mobile app, and pushes SMS / email. The edge preprocessing module performs bandpass filtering and GPS clock synchronization on the signal before electrically connecting it to the data fusion and damage identification module. The energy health management module uses a hybrid power supply of piezoelectric vibration and solar energy to perform periodic self-tests and fault switching on each node.

8. The crack monitoring system for the fixed end of a steel-concrete composite beam according to claim 1, characterized in that: The visual remote communication module adopts containerized deployment based on microservice architecture and WebSocket two-way push technology. It uses WebGL to perform three-dimensional rendering and thermal map overlay on the real-time collected load-deflection data, crack evolution curves and FEM hot spot stress fields. It supports multi-view rotation, zooming and cross-section cutting, and sets dynamic queries and threshold valves on the page end, freely configures alarm display rules, and automatically switches to local IndexedDB cache mode when the network is interrupted. After the network is restored, all unsent key data will be retransmitted in batch increments.

9. The crack monitoring system for the fixed end of a steel-concrete composite beam according to claim 1, characterized in that: The edge preprocessing module is built on an FPGA+ARM heterogeneous platform. It uses a sixth-order bandpass filter based on the CORDIC algorithm to parallel suppress 50Hz grid noise and structural vibration interference. It combines a multi-frequency GPS receiver and the IEEE 1588 precision clock synchronization protocol to perform nanosecond timestamp alignment, run adaptive threshold event detection and dynamic sampling rate control logic, and automatically wake up the corresponding sensor node when a transient burst pulse or a sudden change in DIC image edge contrast is detected, and complete data packaging and transmission within 5ms.

10. The crack monitoring system for the fixed end of a steel-concrete composite beam according to claim 1, characterized in that: The system also includes an energy health management module, which integrates an MPPT photovoltaic power management chip and a piezoelectric resonant energy harvesting circuit. It dynamically switches between vanadium battery supercapacitors and small lithium-ion batteries through a bidirectional DC-DC converter. It prioritizes solar power supply under sunny conditions and piezoelectric power generation at night and when the vehicle vibrates. It has an internal embedded SOC and SOH estimation algorithm, monitors voltage, current, and ambient temperature and humidity in real time, and sends energy health reports to the cloud via LoRaWAN. When the node energy is lower than 20% or a continuous decline in output power is detected, it automatically switches to an ultra-low power sleep wake-up mode.

Images