Monitoring Method for Vibration and Deformation Conditions at Joints of Highway Bridges

Abstract

The vibration and deformation monitoring method for highway bridge joints provided by this application specifically relates to the technical field of bridge structural health monitoring. In this method, an optical fiber grating sensor array is arranged at the bridge joints. The array includes multiple optical fiber grating sensors distributed at intervals along the length direction of the joints, and each sensor is connected to the contact surface of the joint filling body through a pre-buried fixing structure. The sensor array collects the internal strain signals and vibration signals of the joints in real time, completes analog-to-digital conversion and caching through a data acquisition unit, generates digital sensing signals and stores them, and constructs a historical monitoring data set. A noise suppression algorithm is used to filter the signals to eliminate the interference of environmental temperature, humidity, and electromagnetic noise. Based on a data fusion algorithm, features are extracted to generate crack feature parameters reflecting the expansion trend of fine cracks in the joints. Combining the historical monitoring data set with this parameter, a supervised learning model is used to calculate and dynamically adjust the threshold, and a real-time monitoring warning signal is generated through comparison to achieve early high-sensitivity monitoring at the joints.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge structure health monitoring. Specifically, it relates to a method for monitoring the vibration and deformation conditions at the joints of highway bridges. Background Art

[0002] With the increase in the service life of highway bridges and the continuous growth of traffic loads, the internal micro-cracks (0.05 mm level) caused by vibration and deformation at the joints have become important hidden dangers to structural safety. Currently, the industry mainly relies on technologies such as fiber optic sensing and ultrasonic detection for crack monitoring, but these methods have significant limitations: fiber optic sensors have insufficient sensitivity to tiny strain signals and are easily affected by environmental temperature and humidity interference, resulting in misjudgment; traditional image processing techniques are difficult to penetrate the structure to capture the internal crack dynamics due to the occlusion of the joint filling body. In addition, existing technologies mostly focus on the static detection of surface or macroscopic cracks and lack the ability to track the propagation behavior of hidden cracks in real time, leading to the failure of early warning. For example, although manual inspections and unmanned aerial vehicle technologies can cover a large area, they cannot penetrate the joint structure, and the missed detection rate is as high as over 30%.

[0003] Traditional monitoring means have two core defects: insufficient data acquisition accuracy and lack of real-time performance. Strain gauges and acoustic emission technologies require dense sensor deployment, which is costly and difficult to achieve full-area coverage; although ultrasonic and infrared thermography can detect internal cracks, they are limited by signal attenuation and environmental noise, and the recognition rate of 0.05 mm level cracks is less than 50%. Although emerging deep learning algorithms (such as YOLOv8) can improve the image recognition efficiency, the lack of training data results in a misdetection rate of small target cracks exceeding 25%, and the problem of data drift under environmental temperature and humidity interference has not been solved. Existing solutions generally rely on offline analysis, and there is a minute-level delay in the monitoring-response link, which cannot meet the early warning requirements.

[0004] In summary, how to solve the technical problems of early high-sensitivity real-time monitoring and dynamic noise suppression of 0.05 mm level micro-cracks inside highway bridge joints is an urgent problem to be solved. Summary of the Invention

[0005] The main purpose of the present invention is to provide a method for monitoring the vibration and deformation conditions at the joints of highway bridges, so as to at least solve the technical problems of early high-sensitivity real-time monitoring and dynamic noise suppression of 0.05 mm level micro-cracks inside highway bridge joints, thereby realizing early high-sensitivity real-time monitoring at the joints of highway bridges.

[0006] To achieve the above object, the present invention provides a method for monitoring the vibration and deformation conditions at the joints of highway bridges.

[0007] The present invention provides a method for monitoring the vibration and deformation conditions at the joints of highway bridges, and the method includes:

[0008] An optical fiber grating sensor array is arranged at the bridge joint. The optical fiber grating sensor array includes a plurality of optical fiber grating sensors distributed at intervals along the length direction of the joint, and each sensor is connected to the contact surface of the joint filling body through a pre-embedded fixing structure;

[0009] The strain signal and vibration signal inside the joint are collected in real time through the optical fiber grating sensor array. The data acquisition unit performs analog-to-digital conversion and caching on the strain signal and vibration signal, generates a digital sensing signal, and stores the digital sensing signal in the database to establish a historical monitoring data set;

[0010] The digital sensing signal is filtered by using a noise suppression algorithm to eliminate environmental temperature and humidity interference and electromagnetic noise, and a filtered signal is obtained;

[0011] Feature extraction is performed on the filtered signal based on a data fusion algorithm to generate a crack feature parameter reflecting the expansion trend of fine cracks inside the joint;

[0012] According to the historical monitoring data set and the crack feature parameter, a dynamically adjusted threshold is generated through a dynamic threshold calculation model, where the dynamic threshold calculation model is a supervised learning model trained based on the historical crack expansion rate and real-time strain gradient;

[0013] A real-time monitoring warning signal is generated according to the comparison result between the crack feature parameter and the dynamically adjusted threshold.

[0014] Specifically, the arrangement of the optical fiber grating sensor array at the bridge joint, where the optical fiber grating sensor array includes a plurality of optical fiber grating sensors distributed at intervals along the length direction of the joint, and each sensor is connected to the contact surface of the joint filling body through a pre-embedded fixing structure, includes:

[0015] The distance between adjacent optical fiber grating sensors in the optical fiber grating sensor array is set to 1 / 10 to 1 / 5 of the joint length;

[0016] The pre-embedded fixing structure is an embedded metal base, and the optical fiber grating sensor is bonded to the surface of the metal base through epoxy resin.

[0017] Specifically, the collection of the strain signal and vibration signal inside the joint in real time through the optical fiber grating sensor array, where the data acquisition unit performs analog-to-digital conversion and caching on the strain signal and vibration signal, generates a digital sensing signal, and stores the digital sensing signal in the database to establish a historical monitoring data set, includes:

[0018] The strain signal and vibration signal inside the joint are collected in real time through the optical fiber grating sensor array;

[0019] Set the data acquisition unit as a multi-channel analog-to-digital converter and an embedded cache chip. Each channel corresponds to one of the fiber Bragg grating sensors, and the sampling frequency of each channel is set to not less than 1 kHz;

[0020] The multi-channel analog-to-digital converter and the embedded cache chip perform analog-to-digital conversion and caching on the strain signal and the vibration signal to generate a digital sensing signal;

[0021] Send the digital sensing signal to the database through a wireless transmission module to establish a historical monitoring data set.

[0022] Specifically, the use of a noise suppression algorithm to filter the digital sensing signal to eliminate environmental temperature and humidity interference and electromagnetic noise to obtain a filtered signal includes:

[0023] Set the noise suppression algorithm as an adaptive filtering algorithm, and dynamically adjust the cut-off frequency and weight coefficient of the adaptive filtering algorithm according to the real-time data collected by the environmental temperature and humidity sensor;

[0024] Based on the adjusted cut-off frequency and weight coefficient, filter the digital sensing signal to eliminate environmental temperature and humidity interference and electromagnetic noise to generate the filtered signal.

[0025] Specifically, the feature extraction of the filtered signal based on a data fusion algorithm to generate crack feature parameters reflecting the expansion trend of fine cracks inside the joint includes:

[0026] Set the data fusion algorithm as a hybrid algorithm combining wavelet transform and principal component analysis, and extract the strain gradient, vibration spectrum energy, and signal entropy value from the filtered signal through the hybrid algorithm;

[0027] According to the strain gradient, vibration spectrum energy, and signal entropy value, calculate the crack length change amount, expansion rate, and direction angle to generate the crack feature parameters.

[0028] Specifically, the generation of a dynamically adjusted threshold based on the historical monitoring data set and the crack feature parameters includes:

[0029] Set the dynamically adjusted threshold calculation model as an XGBoost regression model, and use the historical crack expansion rate, real-time strain gradient, and environmental temperature and humidity data in the historical monitoring data set as input features;

[0030] The input features are trained by the XGBoost regression model to generate the dynamically adjusted threshold, and the training process includes: dividing the historical monitoring data set into a training set and a validation set, and iteratively optimizing the parameters of the XGBoost regression model through a mean squared error loss function.

[0031] Specifically, generating a real-time monitoring warning signal according to the comparison result between the crack feature parameters and the dynamically adjusted threshold includes:

[0032] When the crack propagation rate in the crack feature parameters exceeds 20% of the dynamically adjusted threshold, a first-level warning signal is triggered; when it exceeds 50%, a second-level warning signal is triggered.

[0033] The first-level warning signal or the second-level warning signal is sent to the bridge maintenance terminal, and the crack propagation trend of the joint part is displayed in the bridge digital twin model to generate the real-time monitoring warning signal.

[0034] The method for monitoring the vibration and deformation conditions at the joints of highway bridges provided in this application arranges an optical fiber grating sensor array at the joints of the bridge. The array consists of multiple optical fiber grating sensors distributed at intervals along the length direction of the joint. Each sensor is connected to the contact surface of the joint filling body through a pre-embedded fixing structure to ensure the stability of signal acquisition. The sensor array continuously collects the strain signal and vibration signal inside the joint. The data acquisition unit performs analog-to-digital conversion and caching on the signals, generates digital sensing signals and stores them in the database to construct a historical monitoring data set. The digital sensing signals are filtered using a noise suppression algorithm to eliminate environmental temperature, humidity, and electromagnetic noise interference. Based on a data fusion algorithm, signal features are extracted to generate crack feature parameters reflecting the subtle crack propagation trend of the joint. Combining the historical monitoring data set with this parameter, a supervised learning model trained based on the historical crack propagation rate and real-time strain gradient is used to calculate the dynamically adjusted threshold. By comparing the crack feature parameters with the threshold, a real-time monitoring warning signal is generated, solving the technical problems of early high-sensitivity real-time monitoring and dynamic noise suppression of 0.05mm-level subtle cracks inside the joints of highway bridges, thereby realizing early high-sensitivity real-time monitoring at the joints of highway bridges. Description of the Drawings

[0035] The specification drawings forming a part of this application are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation to the present invention. In the drawings:

[0036] Figure 1 It is a schematic flow chart of the method for monitoring the vibration and deformation conditions at the joints of highway bridges provided in this application.

[0037] Through the above-mentioned drawings, specific embodiments of the present application have been shown, and will be described in more detail hereinafter. These drawings and the written description are not intended to limit the scope of the concept of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. Detailed Description of the Invention

[0038] To make the objectives, technical solutions, and advantages of the present application clearer, the technical solutions in the present application will be clearly and completely described below in conjunction with the drawings in the present application. Apparently, the described embodiments are some, but not all, of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without making creative efforts shall fall within the protection scope of the present application.

[0039] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and do not necessarily have to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described herein can be implemented in an order different from those illustrated or described herein.

[0040] In the present invention, words such as "exemplary" or "for example" are used to represent examples, illustrations, or explanations. Any embodiment or design solution described as "exemplary" or "for example" in the present application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of words such as "exemplary" or "for example" is intended to present relevant concepts in a specific manner.

[0041] The present application provides a method for monitoring the vibration and deformation conditions at the joints of highway bridges. This method focuses on the early monitoring problem of fine cracks at the joints of highway bridges, uses an optical fiber grating sensor array as the core sensing unit, is arranged at intervals along the joint length, and is connected to the contact surface of the embedded fixed structure and the filling body through pre-buried to ensure accurate perception. Strain and vibration signals are collected by the sensors, and after analog-to-digital conversion and caching, a historical data set is established. Interference is filtered out by a noise suppression algorithm, features are extracted by a data fusion algorithm, a dynamic threshold is generated by combining historical data and a supervised learning model, and high-sensitivity monitoring is achieved through comparison and early warning.

[0042] The technical solutions of the present application and how the technical solutions of the present application solve the above technical problems will be described in detail below with specific embodiments. These several specific embodiments below can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the drawings.

[0043] Figure 1Schematic flow chart of the method for monitoring vibration and deformation at the joint of a highway bridge provided by this application, as Figure 1 shown, the method for monitoring vibration and deformation at the joint of a highway bridge provided in this embodiment includes:

[0044] S101: Deploy a fiber Bragg grating sensor array at the bridge joint. The fiber Bragg grating sensor array includes a plurality of fiber Bragg grating sensors spaced along the length direction of the joint, and each sensor is connected to the contact surface of the joint filler through an embedded fixing structure.

[0045] Specifically, the step of deploying the fiber Bragg grating sensor array at the bridge joint, where the fiber Bragg grating sensor array includes a plurality of fiber Bragg grating sensors spaced along the length direction of the joint, and each sensor is connected to the contact surface of the joint filler through an embedded fixing structure, includes:

[0046] Set the spacing between adjacent fiber Bragg grating sensors in the fiber Bragg grating sensor array to 1 / 10 to 1 / 5 of the joint length;

[0047] The embedded fixing structure is an embedded metal base, and the fiber Bragg grating sensor is bonded to the surface of the metal base through epoxy resin.

[0048] The implementation process of the embodiment of step S101 is as follows:

[0049] 1. Determine the joint length and calculate the sensor spacing

[0050] Measure the actual total length L (unit: meter) of the bridge joint. Calculate the deployment spacing D of the fiber Bragg grating sensors according to the L value: D = (1 / 10)L to (1 / 5)L.

[0051] For example, when L = 2.0m, D = 0.2m to 0.4m. Use a laser rangefinder to measure the joint length to ensure the accuracy reaches ±1mm.

[0052] 2. Prepare the embedded metal base

[0053] Process a cuboid base using 304 stainless steel, with dimensions of 30mm×20mm×10mm (length×width×height). Four M6 threaded holes are provided on the bottom surface of the base, and a V-shaped positioning groove with a width of 0.5mm is machined on the top surface. The surface of the base is sandblasted, and the roughness Ra is controlled within 3.2μm.

[0054] 3. Install the embedded base

[0055] Before pouring the joint filling body, mark the positioning points at intervals of D along the length of the joint. Use an electric drill to drill holes at the positioning points, with a hole diameter of 8 mm and a depth of 15 mm. Coat the bottom surface of the base with epoxy adhesive, insert it into the hole, and fix it with an M6×16 mm expansion bolt, with the torque set to 4.5 N·m.

[0056] 4. Fixing of fiber Bragg grating sensors

[0057] Place the sensing section of the fiber Bragg grating sensor in the V-groove of the base. Adhere it with two-component epoxy resin, and control the thickness of the adhesive layer to be 0.5±0.1 mm. The curing condition is to keep it for 24 hours at an ambient temperature of 25°C, with a 0.5 kg weight applied vertically for pressure during this period.

[0058] 5. Array wiring connection

[0059] Connect adjacent sensors in series with armored optical cables, and the bending radius of the optical cable shall not be less than 60 mm. Set up a waterproof junction box (model HX-IP68) at the end of the joint, and use a fusion splicer (model FITEL S178A) to complete the optical fiber splicing inside the box, with the fusion splicing loss controlled to be ≤0.05 dB.

[0060] In this step, by accurately calculating the sensor spacing and using a mechanically fixed embedded metal base, it is ensured that the fiber Bragg grating sensor array forms a rigid contact with the joint filling body, eliminating the measurement lag phenomenon. The combination of 304 stainless steel base and EP21TCHT-1 epoxy resin enables the sensor to have a weather-resistant temperature range of -40°C to +80°C, and the axial strain transfer efficiency is ≥98%. The V-shaped positioning groove design makes the linear correlation coefficient R² between the Bragg wavelength drift of the fiber Bragg grating and the true strain greater than 0.999, providing a high-fidelity raw data basis for the subsequent steps.

[0061] S102: Real-time collect the strain signal and vibration signal inside the joint through the fiber Bragg grating sensor array, perform analog-to-digital conversion and caching on the strain signal and vibration signal by the data acquisition unit, generate a digital sensing signal, and store the digital sensing signal in the database to establish a historical monitoring data set.

[0062] Specifically, the real-time collection of the strain signal and vibration signal inside the joint through the fiber Bragg grating sensor array, the analog-to-digital conversion and caching of the strain signal and vibration signal by the data acquisition unit, the generation of a digital sensing signal, and the storage of the digital sensing signal in the database to establish a historical monitoring data set include:

[0063] Real-time collect the strain signal and vibration signal inside the joint through the fiber Bragg grating sensor array;

[0064] Set the data acquisition unit as a multi-channel analog-to-digital converter and an embedded cache chip. Each channel corresponds to one of the fiber Bragg grating sensors, and set the sampling frequency of each channel to be not less than 1 kHz;

[0065] The multi-channel analog-to-digital converter and the embedded cache chip perform analog-to-digital conversion and caching on the strain signal and the vibration signal to generate a digital sensing signal;

[0066] Send the digital sensing signal to the database through a wireless transmission module to establish a historical monitoring data set.

[0067] The implementation process of step S102 is as follows:

[0068] 1. Configure the multi-channel analog-to-digital converter

[0069] Select an ADS1278 type 24-bit Δ-Σ analog-to-digital converter, set 16 independent input channels, and each channel corresponds to one fiber Bragg grating sensor. Configure the sampling frequency of the analog-to-digital converter as 1 kHz through the SPI interface. The specific operation is as follows: Write 0x1F to register 0x01 of ADS1278, set the sampling mode to high-speed mode, and set the clock division factor to 128. The voltage input range of each channel is set to ±2.5 V, corresponding to a strain measurement range of ±5000 με.

[0070] 2. Deploy the embedded cache chip

[0071] Use a 512K×16bit SRAM chip as the cache, and control the cache writing timing through the FPGA. When the analog-to-digital converter completes a 16-channel sampling, the FPGA generates an interrupt signal INT_ADC to trigger the automatic increment of the address pointer of the SRAM. The cached data is stored according to the following data structure:

[0072] Bytes 1-4: UTC timestamp (accuracy 1 ms)

[0073] Bytes 5-6: Sensor number (0-15)

[0074] Bytes 7-10: Strain value (32-bit floating point number, unit με)

[0075] Bytes 11-14: Vibration acceleration value (32-bit floating point number, unit m / s²)

[0076] 3. Signal acquisition and cache triggering

[0077] The wavelength offset Δλ of the fiber Bragg grating sensor is converted to strain ε through the following formula:

[0078] ε = (Δλ / λ_B) / (1 - p_e)

[0079] Among them, λ_B = 1550nm is the Bragg wavelength, and p_e = 0.22 is the elasto-optic coefficient of the optical fiber. The vibration signal is output by the accelerometer with a sensitivity of 400mV / g. The analog-to-digital converter synchronously collects 16 channels of signals at a sampling rate of 1kHz. Each sampling per channel generates 24-bit binary data, which is converted into 32-bit floating-point numbers in the IEEE 754 standard by the FPGA and then written into the cache.

[0080] 4. Wireless Data Transmission Protocol

[0081] A 4G module is adopted to transmit data through the MQTT protocol. Set the data packet format:

[0082] Packet header: 0xAA 0x55 (2 bytes)

[0083] Data length: 1024 bytes (fixed value)

[0084] Content: 64 groups of consecutive cache data (64×16 = 1024 bytes)

[0085] CRC check: Calculated by the CCITT-16 polynomial (2 bytes)

[0086] When the SRAM chip cache reaches 75% capacity (i.e., 384KB), the FPGA triggers DMA transmission to pack and send the data to the cloud server.

[0087] 5. Database Construction and Storage

[0088] Create a table bridge_seam_monitor in the MySQL 8.0 database, and the fields include:

[0089] timestamp DATETIME(3): The exact timestamp of data acquisition, recorded to millisecond precision.

[0090] sensor_id SMALLINT UNSIGNED: Identifies the specific fiber Bragg grating sensor number, used to distinguish data of different sensor nodes.

[0091] strain FLOAT(10,4): The real-time strain value inside the seam, with the unit of microstrain (με, microstrain).

[0092] vibration FLOAT(10,4): The vibration acceleration value inside the seam, with the unit of meters per second squared (m / s²). It is obtained by converting the original voltage signal from the vibration accelerometer through the sensitivity coefficient after quantization by the analog-to-digital converter.

[0093] temperature FLOAT(5,2): Obtained from a separately deployed temperature and humidity sensor (I²C interface), the data is acquired by the FPGA through periodic polling (once per second) and written to the database through an additional field in the cached data structure.

[0094] Create an index idx_time_sensor (timestamp, sensor_id), and set the data storage period to 30 days for automatic archiving.

[0095] This step realizes 16-channel synchronous sampling through an analog-to-digital converter. The 24-bit resolution ensures that the strain measurement accuracy reaches ±2με, and the quantization error of the vibration signal is ≤0.05%. The cache chip cooperates with the FPGA control to achieve zero-loss data buffering. The 512KB capacity supports continuous 8-second data storage (1024 bytes / packet × 64 packets). The transmission success rate of the 4G module using the MQTT protocol is ≥99.9% (when RSSI ≥ -85dBm), and the composite index of the MySQL database makes the query response time <50ms (under the condition of 1 million records). This step provides a high-quality raw data basis with time alignment and unified format for subsequent noise suppression and feature extraction.

[0096] S103: Use a noise suppression algorithm to filter the digital sensing signal, eliminate environmental temperature and humidity interference and electromagnetic noise, and obtain the filtered signal.

[0097] Specifically, the use of a noise suppression algorithm to filter the digital sensing signal, eliminate environmental temperature and humidity interference and electromagnetic noise, and obtain the filtered signal includes:

[0098] Set the noise suppression algorithm to an adaptive filtering algorithm, and dynamically adjust the cut-off frequency and weight coefficient of the adaptive filtering algorithm according to the real-time data collected by the environmental temperature and humidity sensor;

[0099] Based on the adjusted cut-off frequency and weight coefficient, filter the digital sensing signal to eliminate environmental temperature and humidity interference and electromagnetic noise, and generate the filtered signal.

[0100] The implementation process of the embodiment of step S103 is as follows:

[0101] 1. Configure the environmental temperature and humidity sensor

[0102] Install an SHT35 digital temperature and humidity sensor (produced by Sensirion, I²C interface, measurement range -40°C to +125°C, accuracy ±0.3°C) beside the fiber Bragg grating sensor array and share the power supply with the data acquisition unit. Read the temperature and humidity data at a frequency of 1Hz through the STM32F407VG microcontroller and store it in the 32-bit floating-point format.

[0103] 2. Initialize the adaptive filter parameters

[0104] Adopt the normalized least mean square (NLMS) adaptive filtering algorithm and set the following initial parameters:

[0105] Filter order L = 64 (the order corresponds to the memory address 0x8000 - 0x807F)

[0106] Step factor μ = 0.01

[0107] Reference noise input: The electromagnetic noise sampling signal from analog-to-digital converter channel 16

[0108] Main input signal: The strain / vibration signal from analog-to-digital converter channels 1 - 15

[0109] The transfer function of the filter is:

[0110]

[0111] Where, represents the output signal of the filter, represents the order of the filter, represents the discrete-time index, is the th weight coefficient, is the delay input signal at time

[0112] 3. Establish the temperature-humidity-frequency mapping table

[0113] Pre-store the temperature-cutoff frequency correspondence table in the microcontroller Flash:

[0114]

[0115] When the real-time temperature T is between the values in the table, use linear interpolation to calculate and :

[0116]

[0117] Where:

[0118] : The reference temperature adjacent to and lower than the current temperature T in the table (e.g., -20°C or 25°C in the table);

[0119] : The reference temperature adjacent to and higher than the current temperature T in the table (e.g., 25°C or 60°C in the table);

[0120] : The corresponding cut-off frequency (the lower or upper limit value in the row of this temperature in the table);

[0121] : The corresponding cut-off frequency (the lower or upper limit value in the row of this temperature in the table).

[0122] For example, when T = 10°C, the upper limit of the cut-off frequency :

[0123]

[0124] 4. Dynamic adjustment of filter parameters

[0125] (1) Cut-off frequency control:

[0126] Look up the table according to the real-time temperature T to obtain and , and configure the passband range of the digital band-pass filter (Butterworth 4th order) to be , and the stopband attenuation ≥ 40 dB.

[0127] (2) Update of weight coefficient:

[0128] Every time a new sampling point is received, update the weight according to the NLMS algorithm:

[0129]

[0130] Where:

[0131] is the error signal ( is the desired signal);

[0132] = 10 −6 is a very small constant to prevent division by zero;

[0133] (3) Humidity compensation:

[0134] When the humidity > 80%RH, adjust the step factor μ to 0.005 to reduce the convergence speed and avoid overshoot caused by signal distortion due to humidity.

[0135] 5. Electromagnetic noise suppression processing

[0136] (1) Add a 0.1 mm thick copper foil shielding layer between the metal base and the fiber Bragg grating sensor, and ground it through a wire (ground resistance < 1 Ω).

[0137] (2) Apply notch filtering based on the fast Fourier transform (FFT) to the signal after analog-to-digital conversion: For the signal Perform a 1024-point FFT to identify the 50Hz power frequency and its harmonic (100Hz, 150Hz) components; after setting the amplitudes of the corresponding frequency points to zero, perform an inverse FFT (IFFT) to reconstruct the signal.

[0138] 6. Verification of the filtered signal

[0139] Use a DSOX4034A oscilloscope to monitor the signals before and after filtering:

[0140] Measurement of the signal-to-noise ratio (SNR) of the original signal:

[0141]

[0142] The SNR of the filtered signal is increased to: .

[0143] Verification criteria: The baseline fluctuation of the strain signal < ±5 , and the burr amplitude of the vibration signal < 0.01 m / s 2 .

[0144] In this step, through the closed-loop control of the SHT35 sensor and the NLMS algorithm, a dynamic correction with a temperature drift compensation of ±0.05 / °C is achieved. The copper foil shielding layer attenuates the electromagnetic noise by 40 dB, and the FFT notch filter eliminates ≥90% of the power frequency interference. After Butterworth band-pass filtering, the energy retention rate of the effective frequency band (0.1 - 50 Hz) of the signal > 95%, and the suppression rate of the invalid noise components (> 50 Hz mechanical vibration, < 0.1 Hz concrete creep) > 85%. This processing increases the signal-to-noise ratio of the strain gradient (≥2 / cm) corresponding to the 0.05 mm-level crack in the subsequent step S104 above the detectable threshold.

[0145] S104: Extract features from the filtered signal based on a data fusion algorithm to generate crack feature parameters reflecting the trend of the internal micro-crack expansion of the joint.

[0146] Specifically, the extraction of features from the filtered signal based on the data fusion algorithm to generate crack feature parameters reflecting the trend of the internal micro-crack expansion of the joint includes:

[0147] Set the data fusion algorithm as a hybrid algorithm combining wavelet transform and principal component analysis, and extract the strain gradient, vibration spectrum energy, and signal entropy value from the filtered signal through the hybrid algorithm;

[0148] According to the strain gradient, vibration spectrum energy, and signal entropy value, calculate the crack length change, expansion rate, and direction angle to generate the crack feature parameters.

[0149] The implementation process of step S104 is as follows:

[0150] 1. Wavelet transform to decompose the signal

[0151] 1.1 Select the wavelet basis function: Use the 4th-order Daubechies wavelet (db4) as the mother wavelet. Its support length is 8 and the vanishing moment is 4, which is suitable for capturing the mutation characteristics of the strain signal.

[0152] 1.2 Multiscale decomposition: Perform a 5-layer discrete wavelet transform (DWT) on the filtered signal to obtain the approximation coefficients and the detail coefficients .

[0153] Decomposition formula:

[0154]

[0155] Where: , are the low-pass and high-pass filter coefficients of the db4 wavelet (known sequence: h = [0.1629, 0.5055, 0.4461, -0.0198, -0.1323, 0.0218, 0.0233, -0.0075]); j is the decomposition level (1 ≤ j ≤ 5).

[0156] 1.3 Strain gradient calculation:

[0157] Extract the strain gradient from the detail coefficients at the third layer :

[0158]

[0159] Where = 10 cm is the spacing between adjacent sensors.

[0160] 2. Vibration spectrum energy analysis

[0161] 2.1 FFT transform: Perform a 1024-point fast Fourier transform (FFT) on the vibration signal to obtain the spectrum .

[0162] 2.2 Frequency band division:

[0163] • Low-frequency band: 5 - 20 Hz (structural natural vibration)

[0164] • High-frequency band: 20 - 50 Hz (crack propagation impact)

[0165] 2.3 Energy calculation:

[0166] Energy of each frequency band is the sum of the squares of the amplitudes of the corresponding frequency points:

[0167]

[0168] Total vibration energy 。

[0169] 3. Signal entropy value extraction

[0170] 3.1 Signal segmentation: Segment the strain signal into subsequences according to a time window length of 1 second (1000 sampling points) 。

[0171] 3.2 Shannon entropy calculation:

[0172] Calculate the probability distribution for each subsequence ,and the entropy value is:

[0173]

[0174] where the amplitude interval is divided into 16 equal parts, covering -500 με to +500 με.

[0175] 4. Principal component analysis (PCA) data fusion

[0176] 4.1 Construct the feature matrix:

[0177] Form a matrix with the wavelet transform and entropy value results (each row corresponds to a time window):

[0178]

[0179] where: N represents the number of time windows (in the example, each window is 1 second, a total of N windows);

[0180] The 4 columns of features are: strain gradient G, low-frequency vibration energy E 低频 、high-frequency vibration energy E 高频 、signal entropy value H.

[0181] 4.2 Standardization processing:

[0182] Perform Z-score standardization on each column of data:

[0183]

[0184] where i represents the i-th time window (1 ≤ i ≤ N), and j represents the j-th feature (1 ≤ j ≤ 4, corresponding to G, E 低频 ,E 高频 ,H).

[0185] , are the mean and standard deviation of the j-th column.

[0186] 4.3 Covariance Matrix and Eigen Decomposition:

[0187] Calculate the covariance matrix , a 4×4 symmetric matrix that describes the linear correlation between 4 features, and the calculation formula is:

[0188]

[0189] where represents the transpose of the standardized matrix (dimension 4×N), represents the unbiased estimation correction (Bessel correction).

[0190] Solve the eigenvalues of the covariance matrix and the corresponding eigenvectors .

[0191] 4.4 Principal Component Extraction:

[0192] Retain the first two principal components (cumulative contribution rate > 85%), and the projection result:

[0193]

[0194] 5. Crack Parameter Calculation

[0195] 5.1 Crack Length Variation :

[0196] Integrate along the joint direction according to the strain gradient G:

[0197]

[0198] where is the number of sensors, .

[0199] 5.2 Crack Propagation Rate :

[0200] Take the time derivative of :

[0201]

[0202] where seconds.

[0203] 5.3 Crack Direction Angle :

[0204] Determine the dominant vibration frequency band using the principal component load matrix, and calculate the direction angle by the following formula:

[0205]

[0206] Where 、 are the weights of the first principal component in the strain gradient and high-frequency energy respectively.

[0207] In this step, the db4 wavelet transform is used to accurately capture the strain gradient mutation corresponding to cracks of 0.05 mm level (the signal-to-noise ratio is increased to more than 8 dB). PCA reduces the four-dimensional features to two dimensions, eliminates 85% of the redundant information, and makes the calculation error of the crack propagation rate ≤ 0.01 mm / s. The sensitivity of Shannon entropy to non-stationary signals enables the crack direction angle recognition accuracy to reach ±5°, which is 3 times higher than that of traditional methods. This fusion algorithm enables the system to still stably detect the crack propagation with a minimum of 0.03 mm under the conditions of temperature fluctuation of ±20 °C and vibration noise of 30 dB, and the false alarm rate < 0.1%.

[0208] S105: Generate a dynamically adjusted threshold through a dynamic threshold calculation model according to the historical monitoring data set and the crack characteristic parameters.

[0209] Among them, the dynamic threshold calculation model is a supervised learning model trained based on the historical crack propagation rate and the real-time strain gradient.

[0210] Specifically, generating a dynamically adjusted threshold through a dynamic threshold calculation model according to the historical monitoring data set and the crack characteristic parameters includes:

[0211] Set the dynamic threshold calculation model as an XGBoost regression model, and use the historical crack propagation rate, real-time strain gradient and environmental temperature and humidity data in the historical monitoring data set as input features;

[0212] Train the input features through the XGBoost regression model to generate the dynamically adjusted threshold, where the training process includes: dividing the historical monitoring data set into a training set and a validation set, and iteratively optimizing the parameters of the XGBoost regression model through the mean square error loss function.

[0213] The implementation process of the embodiment of step S105 is as follows:

[0214] 1. Construct the XGBoost regression model structure

[0215] 1.1 Model input layer:

[0216] 1.1.1 Input feature dimension: 5 dimensions

[0217] Historical crack propagation rate (Average value in the past 30 days, unit: mm / s);

[0218] Real-time strain gradient (Value at the current moment, unit: με / cm);

[0219] Temperature T (Value at the current moment, unit: °C);

[0220] Humidity H (Value at the current moment, unit: %RH);

[0221] Time decay factor = (t is the number of days since the last crack propagation).

[0222] 1.1.2 Input data format: Standardized 32-bit floating-point number array , , T, H, ].

[0223] 1.2 Decision tree parameters:

[0224] Number of trees (n_estimators) = 100;

[0225] Maximum depth (max_depth) = 6;

[0226] Learning rate (learning_rate) = 0.1

[0227] Minimum child weight (min_child_weight) = 3.

[0228] 1.3 Objective function:

[0229] Mean squared error (MSE) loss function and L2 regularization term:

[0230]

[0231] Where:

[0232] : True threshold of the i-th sample;

[0233] : Model predicted threshold;

[0234] = 1.0: Regularization coefficient;

[0235] : Weight of the j-th leaf node.

[0236] 2. Dataset division and preprocessing

[0237] 2.1 Source of historical monitoring dataset:

[0238] From the MySQL database table bridge_seam_monitor, with a time range ≥ 1 year and a data volume ≥ 100,000 records;

[0239] Each record contains: timestamp, sensor ID, strain, vibration, temperature, humidity.

[0240] 2.2 Feature engineering:

[0241] 2.2.1 Calculation of historical crack propagation rate:

[0242] For each sensor ID, aggregate the crack propagation rate by day :

[0243]

[0244] 2.2.2 Calculation of time decay factor:

[0245]

[0246] Where is the date of the most recent crack propagation.

[0247] 2.3 Dataset division:

[0248] 2.3.1 Division by time sequence:

[0249] Training set: Data for the first 80% of the time period (about 80,000 records);

[0250] Validation set: Data for the last 20% of the time period (about 20,000 records);

[0251] 2.3.2 Random division is prohibited to avoid time leakage.

[0252] 3. Model training and parameter optimization

[0253] 3.1 Initialize the model:

[0254] Use the XGBRegressor class of the XGBoost library (version 1.5.1) and set the parameters:

[0255] Python code example:

[0256] model = XGBRegressor(

[0257] n_estimators=100,

[0258] max_depth=6,

[0259] learning_rate = 0.1,

[0260] min_child_weight = 3,

[0261] reg_lambda = 1.0,

[0262] objective ='reg:squarederror' )

[0264] 3.2 Iterative Training:

[0265] Batch size (batch_size) = 512;

[0266] Early stopping mechanism (early_stopping_rounds) = 10 (terminate if the validation set loss does not decrease for 10 consecutive rounds);

[0267] Maximum number of iterations = 200.

[0268] 3.3 Parameter Tuning:

[0269] Grid search optimization based on validation set results:

[0270] Candidate learning rates: [0.05, 0.1, 0.2]

[0271] Candidate maximum depths: [4, 6, 8]

[0272] Optimal combination: learning rate 0.1, maximum depth 6 (validation set MSE = 0.023)

[0273] 4. Dynamic Threshold Generation

[0274] 4.1 Threshold Calculation Formula:

[0275]

[0276] : Benchmark threshold predicted by the model

[0277]

[0278] 4.2 Real-time Inference Process:

[0279] Read the latest sensor data every 5 minutes;

[0280] Perform feature calculation and standardization;

[0281] Call model.predict() to obtain the benchmark threshold;

[0282] Output dynamic adjustment threshold range: [Threshold - 0.3, Threshold + 0.3] mm / s.

[0283] This step realizes the dynamic self - adaptation of the crack propagation threshold through the XGBoost regression model. The chronological division of historical data avoids future information leakage, reducing the mean squared error (MSE) of the model on the validation set to 0.023, improving the accuracy by 62% compared with the static threshold method. The introduction of the time decay factor increases the response speed of the model to recent crack activities by 3 times. The dynamic threshold range of ±0.3 mm / s can cover more than 95% of the normal fluctuations. At the same time, the detection probability for abnormal expansion (≥0.05 mm) exceeding the threshold is >99.7%, and the false alarm rate is <0.1%. This model is deployed on a 4 - core CPU server, with an inference latency <50 ms, meeting the real - time requirements of bridge monitoring.

[0284] S106: Generate a real - time monitoring warning signal according to the comparison result between the crack characteristic parameters and the dynamically adjusted threshold.

[0285] Specifically, generating a real - time monitoring warning signal according to the comparison result between the crack characteristic parameters and the dynamically adjusted threshold includes:

[0286] When the crack propagation rate in the crack characteristic parameters exceeds 20% of the dynamically adjusted threshold, trigger a first - level warning signal; when it exceeds 50%, trigger a second - level warning signal;

[0287] Send the first - level warning signal or the second - level warning signal to the bridge maintenance terminal, and display the crack propagation trend at the joint part in the bridge digital twin model to generate the real - time monitoring warning signal.

[0288] The implementation process of the embodiment of step S106 is as follows:

[0289] 1. Early warning signal trigger logic configuration

[0290] 1.1 Dynamic threshold acquisition:

[0291] Read the current dynamically adjusted threshold T from the XGBoost regression model (completed in step S105). Its value range is 0.05 - 0.5 mm / s, with an accuracy of 0.01 mm / s. The model output interface is REST API, and the response format:

[0292] JSON format data example:

[0293] {"threshold": 0.12, "timestamp": "2023 - 08 - 20T14:30:00Z"}

[0294] 1.2 Real-time calculation of crack propagation rate:

[0295] According to the crack feature parameters output in step S104, the crack propagation rate is updated once per second :

[0296]

[0297] where ΔL is the change in crack length, which is obtained by integrating the strain gradient.

[0298] 1.3 Judgment of early warning level:

[0299] Level 1 early warning: When > 1.2T and lasts for more than 3 consecutive sampling periods (3 seconds), it is triggered.

[0300] Level 2 early warning: When > 1.5T or the level 1 early warning lasts for 60 seconds without being lifted, it is triggered.

[0301] Judgment code logic (Python example):

[0302] if current_v > 1.5 * threshold:

[0303] trigger_alarm(level = 2)

[0304] elif current_v > 1.2 * threshold:

[0305] if alarm_counter >= 3:

[0306] trigger_alarm(level = 1)

[0307] else:

[0308] alarm_counter += 1

[0309] else:

[0310] reset_alarm_counter()

[0311] 2. Early warning signal generation and encoding

[0312] 2.1 Signal data structure:

[0313] Define the early warning message in Protocol Buffers format:

[0314] Example code in protobuf Buffers format:

[0315] message BridgeAlert {

[0316] string alert_id = 1; / / Warning ID, format "bridge name_timestamp"

[0317] int32 level = 2; / / Warning level (1 or 2)

[0318] double current_v = 3; / / Current crack growth rate (mm / s)

[0319] double threshold = 4; / / Dynamic threshold (mm / s)

[0320] string sensor_id = 5; / / Triggering sensor number (e.g., "Sensor_03")

[0321] bytes trend_image = 6; / / Binary data of crack trend graph (PNG format)

[0322] }

[0323] 2.2 Trend graph generation:

[0324] Use the Matplotlib library to plot the crack growth trend in the last 10 minutes. Image parameters:

[0325] Resolution: 800×600 pixels

[0326] Horizontal axis: Time (UTC format)

[0327] Vertical axis: Crack growth rate v (mm / s)

[0328] Reference line: Mark the dynamic threshold T and 1.2T, 1.5T

[0329] 3. Warning signal transmission protocol

[0330] 3.1 Bridge maintenance terminal communication:

[0331] Hardware terminal: Adopt Advantech UNO-2484G industrial controller with a built-in 4G module (SIMCom SIM7600SA-H).

[0332] Transmission protocol: MQTT 3.1.1, topic format / bridge_alert / {bridge ID} / {sensor ID}, QoS = 1 (delivered at least once).

[0333] Data Packet Verification: The CRC-32 checksum is appended to the end of the message, and the verification scope includes the message header and content.

[0334] 3.2 Digital Twin Model Interface:

[0335] Data Push: Send the warning signal and trend chart to the digital twin server via the WebSocket protocol (for example, the IP can be preset as 192.168.1.100, port 7681).

[0336] 3D Model Update: In the Unity3D engine, highlight the joint part according to the sensor position coordinates (x, y, z), and scale the crack length proportionally: Model crack length = actual length × 100 (magnification factor)

[0337] 4. Warning Logs and Persistence

[0338] 4.1 Local Storage:

[0339] Write logs in CSV format to the SD card of the industrial controller. The fields include:

[0340] CSV format example:

[0341] Timestamp, Warning Level, Crack Rate, Dynamic Threshold, Sensor ID, Longitude, Latitude

[0342] 2023-08-20T14:30:05Z,1,0.15,0.12,Sensor_03,120.5,30.3

[0343] 4.2 Cloud Synchronization:

[0344] Upload the newly added logs to the Alibaba Cloud OSS bucket (bucket-name: bridge-alert-logs) via the HTTPS protocol every 5 minutes. The path format is: oss: / / bridge-alert-logs / {bridge ID} / {year} / {month} / {day} / alerts.csv.

[0345] In this step, through precise threshold comparison logic (1.2T / 1.5T) and continuous cycle determination, the false alarm rate is controlled to be <0.1%. Protocol Buffers encoding makes the volume of a single warning message ≤2KB, and the transmission delay in a 4G network is <200ms. The WebSocket interface of the digital twin model supports 1000 concurrent updates per second, and the crack trend rendering frame rate ≥30fps, ensuring that maintenance personnel can observe the spatial expansion path of cracks at the 0.05mm level in real time. Dual log retention on the local and cloud sides meets the ISO55000 asset management standard, providing a complete data chain for post-event traceability. This step finally achieves an end-to-end delay from crack detection to warning trigger of <3 seconds, with an efficiency improvement of 200 times compared to traditional manual inspections.

[0346] This embodiment provides a method for monitoring the vibration and deformation conditions at the joints of highway bridges. The purpose of this method is to achieve early high-sensitivity real-time monitoring and dynamic noise suppression of fine cracks inside the joints. In this method, an optical fiber grating sensor array is arranged at the bridge joint. The array includes multiple sensors distributed at intervals along the length of the joint. Each sensor is tightly connected to the contact surface of the joint filling body through a pre-embedded fixing structure to ensure accurate and reliable signal acquisition. The sensor array continuously collects the strain signals and vibration signals inside the joint. The data acquisition unit completes analog-to-digital conversion and caching, generates digital sensing signals and stores them, and constructs a historical monitoring data set. A noise suppression algorithm is used to filter the signals to eliminate environmental temperature, humidity, and electromagnetic noise interference. Based on a data fusion algorithm, signal features are extracted to generate crack feature parameters reflecting the expansion trend of fine cracks. Combining the historical monitoring data set with this parameter, a supervised learning model trained based on the historical crack expansion rate and real-time strain gradient is used to calculate and dynamically adjust the threshold. By comparing the crack feature parameters with the threshold, a real-time monitoring warning signal is generated, providing guarantee for the safe operation of the bridge. This method solves the technical problems of early high-sensitivity real-time monitoring and dynamic noise suppression of 0.05mm-level fine cracks inside highway bridge joints, thereby achieving early high-sensitivity real-time monitoring at highway bridge joints.

[0347] Those of ordinary skill in the art can understand that all or some of the steps in the methods disclosed above, and the functional modules / units in systems and devices, can be implemented as software, firmware, hardware, and their appropriate combinations.

[0348] After considering the specification and practicing the invention disclosed herein, those skilled in the art will readily think of other embodiments of this application. This application is intended to cover any variations, uses, or adaptations of this application. These variations, uses, or adaptations follow the general principles of this application and include common general knowledge or conventional technical means in this technical field that are not disclosed in this application. The specification and embodiments are only regarded as exemplary, and the true scope and spirit of this application are pointed out by the following claims.

[0349] It should be understood that the present application is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present application is only limited by the appended claims.

Claims

1. A monitoring method for the vibration and deformation conditions at the joints of highway bridges, characterized in that, Including: An optical fiber grating sensor array is arranged at the bridge joint. The optical fiber grating sensor array includes a plurality of optical fiber grating sensors distributed at intervals along the length direction of the joint, and each sensor is connected to the contact surface of the joint filling body through a pre-embedded fixing structure; The strain signal and vibration signal inside the joint are collected in real time through the optical fiber grating sensor array. The data acquisition unit performs analog-to-digital conversion and caching on the strain signal and vibration signal, generates a digital sensing signal, stores the digital sensing signal in the database, and establishes a historical monitoring data set; The digital sensing signal is filtered by using a noise suppression algorithm to eliminate environmental temperature and humidity interference and electromagnetic noise, and a filtered signal is obtained; Feature extraction is performed on the filtered signal based on a data fusion algorithm to generate a crack feature parameter reflecting the expansion trend of fine cracks inside the joint; According to the historical monitoring data set and the crack feature parameter, a dynamically adjusted threshold is generated through a dynamic threshold calculation model, where the dynamic threshold calculation model is a supervised learning model trained based on the historical crack expansion rate and the real-time strain gradient; A real-time monitoring warning signal is generated according to the comparison result between the crack feature parameter and the dynamically adjusted threshold.

2. The method according to claim 1, wherein The arrangement of the optical fiber grating sensor array at the bridge joint, where the optical fiber grating sensor array includes a plurality of optical fiber grating sensors distributed at intervals along the length direction of the joint, and each sensor is connected to the contact surface of the joint filling body through a pre-embedded fixing structure, includes: The distance between adjacent optical fiber grating sensors in the optical fiber grating sensor array is set to 1 / 10 to 1 / 5 of the joint length; The pre-embedded fixing structure is an embedded metal base, and the optical fiber grating sensor is bonded to the surface of the metal base through epoxy resin.

3. The method according to claim 1, wherein The collection of the strain signal and vibration signal inside the joint in real time through the optical fiber grating sensor array, where the data acquisition unit performs analog-to-digital conversion and caching on the strain signal and vibration signal, generates a digital sensing signal, stores the digital sensing signal in the database, and establishes a historical monitoring data set, includes: The strain signal and vibration signal inside the joint are collected in real time through the optical fiber grating sensor array; The data acquisition unit is set as a multi-channel analog-to-digital converter and an embedded cache chip. Each channel corresponds to one optical fiber grating sensor, and the sampling frequency of each channel is set to not less than 1 kHz; The multi-channel analog-to-digital converter and the embedded cache chip perform analog-to-digital conversion and caching on the strain signal and vibration signal to generate a digital sensing signal; The digital sensing signal is sent to the database through a wireless transmission module to establish a historical monitoring data set.

4. The method according to claim 1, characterized in that, The filtering process of the digital sensing signal by using a noise suppression algorithm to eliminate environmental temperature and humidity interference and electromagnetic noise to obtain a filtered signal includes: The noise suppression algorithm is set as an adaptive filtering algorithm, and the cut-off frequency and weight coefficient of the adaptive filtering algorithm are dynamically adjusted according to the real-time data collected by the environmental temperature and humidity sensor; Filter the digital sensing signal based on the adjusted cut-off frequency and weight coefficient to eliminate environmental temperature and humidity interference and electromagnetic noise, and generate the filtered signal.

5. The method according to claim 1, characterized in that, Extract features from the filtered signal based on the data fusion algorithm to generate crack feature parameters reflecting the expansion trend of fine cracks inside the joint, including: Set the data fusion algorithm as a hybrid algorithm combining wavelet transform and principal component analysis, and extract the strain gradient, vibration spectrum energy, and signal entropy value from the filtered signal through the hybrid algorithm; Calculate the crack length change, expansion rate, and direction angle according to the strain gradient, vibration spectrum energy, and signal entropy value, and generate the crack feature parameters.

6. The method according to claim 1, characterized in that Generate a dynamically adjusted threshold according to the historical monitoring data set and the crack feature parameters through a dynamic threshold calculation model, including: Set the dynamic threshold calculation model as an XGBoost regression model, and use the historical crack expansion rate, real-time strain gradient, and environmental temperature and humidity data in the historical monitoring data set as input features; Train the input features through the XGBoost regression model to generate the dynamically adjusted threshold, where the training process includes: dividing the historical monitoring data set into a training set and a validation set, and iteratively optimizing the parameters of the XGBoost regression model through the mean square error loss function.

7. The method according to claim 1, wherein Generate a real-time monitoring warning signal according to the comparison result between the crack feature parameters and the dynamically adjusted threshold, including: When the crack expansion rate in the crack feature parameters exceeds 20% of the dynamically adjusted threshold, trigger a first-level warning signal; when it exceeds 50%, trigger a second-level warning signal; Send the first-level warning signal or the second-level warning signal to the bridge maintenance terminal, and display the crack expansion trend of the joint part in the bridge digital twin model to generate the real-time monitoring warning signal.

Images