Bridge construction intelligent monitoring method based on AI

Abstract

The invention discloses an AI-based bridge construction intelligent monitoring method, particularly relates to the field of data analysis, and comprises the steps of multi-modal sensing network construction, cross-domain characteristic parameter extraction, comprehensive state evaluation modeling and intelligent decision and regulation. According to the method, deformation, corrosion, vibration and other multi-physical field data are synchronously collected through a multi-modal sensing network, a cross-scale feature extraction algorithm is adopted to construct creep-corrosion collaborative factors and other composite indexes, the limitation of traditional single-dimensional parameter analysis is broken through, a dynamic evaluation model is fused with weighted integral and a nonlinear function, and the dynamic evaluation accuracy is improved. Precise early warning of risks such as bridge pier instability and cantilever cracking is achieved, the multi-field coupling effect quantification capacity is remarkably improved, an intelligent joint control mechanism automatically triggers instructions such as jacking reinforcement and damping tuning through a three-level emergency response system, a monitoring-analysis-regulation closed-loop system is constructed, the hysteresis of manual decision is effectively eliminated, and the safety and reliability of the system are improved. And all-weather self-adaptive safety guarantee is provided for bridge construction.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis. More specifically, the present invention relates to an intelligent monitoring method for bridge construction based on AI. Background Art

[0002] Bridge construction monitoring is the core link to ensure project safety and quality. In specific implementation, the construction unit usually arranges static level gauges at the pier foundation to monitor settlement, installs vibrating wire strain gauges at the cantilever section to measure concrete stress, configures fiber Bragg grating sensors in the cable system to detect cable force changes, and records basic parameters such as environmental temperature, humidity, and wind speed through daily manual inspections. The operation process of the existing system is divided into three stages: first, collect the original sensor data through a wired transmission network, then calculate the single-parameter deviation degree using a finite element model, and finally trigger an audible and visual alarm device according to a preset threshold. This mode has supported the construction of more than 75% of the extra-large bridges in China in the past decade, but gradually exposes the defect of insufficient adaptability in the era of intelligent construction.

[0003] The existing technical system has three major structural defects: firstly, the ability to process multi-source heterogeneous data is weak. The millimeter-level GNSS data used for deformation monitoring and the microampere-level electrochemical signals used for material corrosion monitoring are difficult to effectively fuse due to differences in time-frequency characteristics, resulting in the inability to establish a correlation model between pier settlement and steel bar corrosion; secondly, the dimension of feature parameter extraction is single. Traditional methods only use linear indicators such as strain extreme values and displacement cumulative amounts, lacking quantification means for non-linear phenomena such as creep-corrosion synergy effect and multi-field coupling deformation of wind-rain-temperature; thirdly, the decision-making mechanism is rigid. The existing system only realizes the primary response from over-standard alarm to manual handling, lacking the adaptive regulation ability based on real-time state evaluation. Summary of the Invention

[0004] To overcome the above-mentioned defects of the prior art, the present invention provides an intelligent monitoring method for bridge construction based on AI. Through the following solutions, it solves the defects of insufficient data fusion, single feature parameter, and decision-making lag proposed in the above background art, which are difficult to quantify the multi-field coupling effect and realize dynamic regulation, and restricts the safety guarantee ability of complex bridge construction.

[0005] To achieve the above object, the present invention provides the following technical solution: an intelligent monitoring method for bridge construction based on AI, including:

[0006] S1: Construction of a multimodal sensing network: Deploy an intelligent sensing array at the pier, cantilever, and cable system to respectively collect multi-physical field data of structural deformation, environmental erosion, and dynamic response, forming three groups of heterogeneous monitoring data streams;

[0007] S2: Cross - domain Feature Parameter Extraction: Conduct settlement dynamics analysis on the pier group collected in S1, perform unsteady hygrothermal coupling calculation on the cantilever group, and carry out thermal gradient tensor analysis on the cable group to extract three key feature parameters characterizing structural deterioration.

[0008] S3: Comprehensive State Assessment Modeling: Integrate multiple groups of feature parameters extracted in S2 to construct a pier dynamic instability warning index, a risk accumulation index for cantilever casting, and a critical state index for the cable system. Use weighted integration and non - linear functions to achieve state dimensionality reduction and form a quantitative evaluation index system.

[0009] S4: Intelligent Decision - making and Regulation: Establish a three - level joint control mechanism based on the three evaluation indexes calculated in S3, including the foundation construction control layer, the accumulation index, and the critical state index of the cable system. Achieve early warning classification response and equipment linkage control through a logical judgment tree.

[0010] The technical effects and advantages of the present invention:

[0011] 1. By constructing a multi - modal sensing network, the present invention realizes the deep integration of multi - physical - field data in the whole process of bridge construction. Intelligent sensing arrays are respectively deployed on the piers, cantilevers, and cable systems to synchronously collect data on structural deformation, environmental erosion, and dynamic response, breaking through the limitations of traditional single - parameter monitoring. Using cross - domain feature analysis technology, key indicators such as creep - corrosion synergy factors and hygrothermal coupling effect indexes are extracted from heterogeneous data streams to accurately quantify material deterioration and multi - field coupling effects, significantly improving the systematicness and relevance of feature parameters.

[0012] 2. By establishing a dynamic instability warning index, a risk accumulation index, and a critical state index, the present invention realizes the quantitative evaluation and trend prediction of the construction state. Integrate weighted integration and non - linear functions to construct a comprehensive evaluation model, solving the defect of poor adaptability of the traditional threshold method to complex working conditions. Potential risks such as pier instability, cantilever cracking, and cable flutter can be identified in advance. Combining the three - level joint control mechanism and the logical judgment tree, a closed - loop control from data collection to equipment regulation is realized, greatly improving the response speed and disposal accuracy.

[0013] 3. The present invention innovatively embeds AI algorithms into the whole process of construction monitoring. Through the intelligent decision - making module, control instructions such as jacking reinforcement, casting speed adjustment, and cable force adjustment are automatically triggered. Use industrial Internet of Things protocols to achieve multi - device collaborative linkage, build a monitoring - analysis - regulation integrated platform, effectively overcoming the lag and subjectivity of manual intervention, and providing all - weather and adaptive intelligent safety guarantee for bridge construction. Brief Description of the Drawings

[0014] Figure 1 It is a schematic diagram of the overall structure of the present invention. Detailed Embodiment

[0015] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0016] Refer to Figure 1 An AI-based intelligent monitoring method for bridge construction shown as follows includes:

[0017] S1: Construction of a multi-modal sensing network: Deploy intelligent sensing arrays on bridge piers, cantilevers, and cable systems to collect multi-physical field data of structural deformation, environmental erosion, and dynamic response respectively, forming three groups of heterogeneous monitoring data streams.

[0018] The heterogeneous monitoring data streams include a pier group, a cantilever group, and a cable group. The pier group includes the settlement difference between adjacent piers, concrete resistivity, chloride ion concentration, and micro-vibration acceleration; the cantilever group includes formwork strain, GPS displacement, acoustic emission event energy, and environmental temperature and humidity; the cable group includes cable force, temperature gradient, three-dimensional wind speed, and magnetic flux.

[0019] The acquisition method of the settlement difference between adjacent piers is as follows: Install dual-frequency GNSS receivers on the tops of adjacent piers, adopt the RTK measurement mode, set the sampling frequency to 2Hz, and obtain millimeter-level settlement data through post-processing differential. Synchronously install tilt sensors to compensate for the influence of attitude changes, with a range of ±15° and an accuracy of 0.001°; the acquisition method of concrete resistivity is as follows: Use a four-electrode resistivity meter, arrange stainless steel probes on the pier surface in a 20×20 cm grid, apply 100Hz alternating current, and measure to a depth of 8 cm. Automatically scan and record the surface resistivity distribution every 30 minutes; the acquisition method of chloride ion concentration is as follows: Embed concrete durability monitoring sensors, arrange Ag / AgCl electrode pairs in the steel bar protection layer, apply a voltage of -0.2V through a potentiostat, measure the diffusion current and convert it to the Cl- concentration, with a sampling interval of 4 hours; the acquisition method of micro-vibration acceleration is as follows: Install a three-axis MEMS accelerometer, with an XYZ-axis vector range of ±5g and a bandwidth of 0.5 - 2000Hz. Fix it on the side wall of the pier at 1 / 3 height using a magnetic base, and the continuous sampling rate is 2048Hz.

[0020] The acquisition method of template strain is as follows: Fiber Bragg grating strain gauges are pasted in the stress concentration area of the cantilever template. The grating area length is 10 mm, and the arrangement spacing is 50 cm in a cross array. The demodulator samples at 250 Hz, and the temperature compensation accuracy is ±0.5 με. The acquisition method of GPS displacement is as follows: A high-precision GPS dynamic measurement system is adopted. The receiver is fixed at the end of the cantilever. L1+L5 dual-frequency signals are used, and a short baseline is formed in cooperation with the local reference station. The output frequency is 20 Hz, and the three-dimensional positioning accuracy is ±2 mm + 0.5 ppm. The acquisition method of acoustic emission event energy is as follows: Wideband acoustic emission sensors are installed, with a frequency range of 50 - 400 kHz and a threshold set at 40 dB. The event energy is calculated through waveform feature extraction technology, and the sampling rate is 10 MHz. The acquisition method of ambient temperature and humidity is as follows: Wireless sensor nodes are deployed, integrating SHT35 temperature and humidity sensors. The measurement range is -40~125℃ / 0 - 100%RH, and the accuracy is ±0.2℃ / ±2%RH. Data is uploaded every 5 minutes.

[0021] The acquisition method of cable force is as follows: A magnetic flux cable force meter is adopted. The annular sensor is sleeved behind the anchor. The excitation frequency is 10 kHz. The cable force is measured through the magneto-elastic effect. The measurement range is 200 - 20000 kN, and the accuracy is ±1%FS. The sampling rate is 100 Hz. The acquisition method of temperature gradient is as follows: A distributed optical fiber temperature measurement system is arranged along the cable length direction. The spatial resolution is 0.5 m, and the temperature measurement accuracy is ±0.5℃. Temperature-sensing gratings are arranged every meter to generate three-dimensional temperature field data in real time. The acquisition method of three-dimensional wind speed is as follows: An ultrasonic anemometer is installed. The measurement range is 0 - 60 m / s, and the resolution is 0.01 m / s. The three-dimensional vector synthesis frequency is 32 Hz. It is installed 5 m above the mid-span of the cable. The acquisition method of magnetic flux is as follows: Array Hall sensors are adopted. Six groups of probes are arranged equidistantly along the circumferential direction of the cable, with a spacing of 50 cm, to measure the change in the background magnetic field intensity. 16-bit ADC conversion is performed, and the sampling rate is 1 kHz.

[0022] S2: Cross-domain feature parameter extraction: Perform settlement dynamics analysis on the pier group collected in S1, execute unsteady hygrothermal coupling calculation on the cantilever group, and carry out thermal gradient tensor analysis on the cable group to extract three key feature parameters characterizing structural deterioration.

[0023] The key characteristic parameters include the monitoring data of pier stability, the monitoring data of the dynamic response of cantilever casting, and the monitoring data of the multi-field coupling of the cable system. The monitoring data of pier stability includes the uneven settlement trend coefficient K_s, the creep-corrosion synergistic factor Q_c of materials, the micro-vibration energy entropy E_v, and the electrochemical impedance phase angle Φ_z. The monitoring data of the dynamic response of cantilever casting includes the time-varying stiffness reduction coefficient η_t, the hygrothermal coupling effect index H_w, the chaotic characteristic quantity D_d of dynamic displacement, and the fractal-entropy composite index F_e of acoustic emission. The monitoring data of the multi-field coupling of the cable system includes the cable force-temperature gradient coupling coefficient K_T, the vortex-induced vibration energy capture rate E_v, the magnetoelastic hysteresis loss factor H_m, and the multi-order frequency weight ratio R_f.

[0024] The uneven settlement trend coefficient is specifically expressed as: ΔS_p(t) represents the time series of the settlement difference between adjacent piers, t represents the monitoring time, m represents the number of pier pairs, and S_ref represents the reference settlement amount, specifically taking the design allowable settlement value. The formula first uses the logistic function 1 + e^{-0.1t} to perform time-varying attenuation correction on the settlement difference between adjacent piers, and then takes its time derivative to reflect the trend change rate. The logarithmic term quantifies the overall offset degree by accumulating the ratio of the settlement difference between adjacent piers to the reference value. Finally, the dynamic change rate is combined with the total dimension. The creep-corrosion synergistic factor of materials is specifically expressed as: ε_c(τ) represents the function of creep strain varying with time, C_{ion}(τ) represents the measured value of chloride ion concentration, C_th represents the corrosion threshold concentration, specifically taken according to the concrete code, and α represents the corrosion acceleration index. The formula characterizes the time-varying characteristics of materials through the creep strain rate dε_c / dlnτ, uses the C_{ion} / C_th exponential term to quantify the environmental corrosion effect, and the integral operation reflects the cumulative synergistic effect of the two within the monitoring period [t_1, t_2]. The micro-vibration energy entropy is specifically expressed as: where N_b represents the number of frequency band divisions, taking values from 2^4 to 2^6, A(f) represents the Fourier amplitude spectrum of the vibration acceleration signal, and E_total represents the total vibration energy integrated from 0 to f_max. The formula divides the vibration spectrum into N_b sub-frequency bands, calculates the energy proportion of each frequency band, and uses the Shannon entropy formula to quantify the chaos degree of energy distribution. An increase in the entropy value indicates that the vibration energy distribution tends to be dispersed, reflecting the development of structural damage. The electrochemical impedance phase angle is specifically expressed as: Z'(ω) and Z”(ω): the real and imaginary parts of the AC impedance, ρ_c(t): the measured value of the concrete resistivity, ρ_ref: the initial resistivity reference value. The formula characterizes the dielectric properties of materials through the original value of the impedance phase angle arctan(Z” / Z'), and multiplies it by the resistivity change correction term 1 + (ρ_ref - ρ_c) / ρ_ref to reflect the influence of the degradation of the electrical conductivity on the phase measurement.

[0025] The time-varying stiffness reduction coefficient is specifically expressed as: ε(t) represents the template strain time series, EWMA_β represents the exponentially weighted moving average of the β parameter, β takes values from 0.05 to 0.2, ε_y represents the material yield strain threshold, t_char represents the characteristic time constant, taking one-third of the age. The formula quantifies the influence of sudden loads through the difference between the strain extreme value and the moving average. The tanh function introduces a time-related stiffness degradation mechanism. When t approaches t_char, tanh approaches 1, reflecting the age effect. The hydrothermal coupling effect index is specifically expressed as: T_a(t) represents the ambient temperature, T_g represents the characteristic temperature of cement hydration, k_h = 0.03 °C⁻¹ represents the humidity influence coefficient, t_d represents the diffusion time constant, specifically the square of the curing layer thickness divided by the diffusion coefficient, ΔT_max represents the critical temperature difference threshold of material thermal stress, RH represents the real-time relative humidity, RH_0 represents the required reference humidity value for concrete curing. The temperature term (T_a - T_g) / ΔT_max in the formula quantifies the thermal stress, and the exponential term e^{k_h(RH - RH_0)} characterizes the non-linear influence of humidity on material properties. The error function erf() describes the characteristics of the hydrothermal coupling effect diffusing over time. The dynamic displacement chaos characteristic quantity is specifically expressed as: where δ_i and δ_j represent different sampling points of the GPS displacement sequence, N represents the total number of sampling points, ε_d represents the phase space distance threshold, Θ() represents the Heaviside step function. The formula maps the displacement time series to an m-dimensional space through the phase space reconstruction technique, calculates the correlation integral C(ε_d), and uses the correlation dimension D_d to quantify the chaos degree of the system. A decrease in the D_d value indicates that the displacement pattern tends to be orderly. The acoustic emission fractal-entropy composite index is specifically expressed as: D_b represents the box-counting method fractal dimension, H_q represents the width of the multifractal spectrum, E_k represents the energy of the k-th acoustic emission event, γ = 1.2 represents the energy weighting coefficient, N_AE represents the total number of acoustic emission events. The numerator term D_b / √H_q in the formula comprehensively characterizes the spatial complexity and multi-scale characteristics of damage evolution. The logarithmic term highlights the contribution of high-energy events through the γ-th power weighted summation, reflecting the degree of damage accumulation.

[0026] The cable force-temperature gradient coupling coefficient is specifically expressed as: F_s(t) represents the measured value of the cable force, represents the cable body temperature gradient vector, θ_{TF} represents the angle between the temperature gradient and the cable force change direction, It represents the standard value of the critical temperature gradient. The formula performs a vector dot product operation on the cable force change rate and the normalized temperature gradient. The cosθTF term reflects the direction correlation between the two. A positive value indicates that the thermo-mechanical coupling intensifies the cable force fluctuation. The vortex-induced vibration energy capture rate is specifically expressed as: Vn(t) represents the normal wind speed component, \ddot{x}(t) represents the lateral acceleration of the cable, fv represents the vortex shedding frequency, fst represents the Strouhal characteristic frequency, λ = 0.3 represents the Reynolds number correction exponent, ρa represents the air density, Cvm = 0.8 - 1.6 represents the vortex-induced vibration mode coupling coefficient, Δt represents the integration time window. The formula calculates the energy exchange rate through the time-domain integration of the vibration acceleration and the wind speed, and introduces the (fv / fst)-λ term to correct the influence of non-ideal flow states, accurately characterizing the intensity of vortex-induced resonance under the actual wind field. The magneto-elastic hysteresis loss factor is specifically expressed as: Ψs represents the magnetic flux, Hm represents the magnetization field strength, σF(t) represents the standard deviation of the cable force fluctuation, Β = 0.7 represents the stress-magnetization coupling coefficient, cycle represents the loop integral of the complete closed path of the hysteresis loop, σF0 represents the reference standard deviation of the cable force fluctuation. The numerator of the formula calculates the area enclosed by the hysteresis loop to characterize the iron loss, and the denominator performs the maximum normalization process. The exponential term (1 + σF / σF0)β introduces the modulation effect of the cable force fluctuation on the magneto-elastic effect. The multi-order frequency weight ratio is specifically expressed as: wm represents the low-order mode weight, wn represents the high-order mode weight, taking n-1, fd represents the dominant frequency, fc represents the critical flutter frequency, M represents the total number of low-order modes, N represents the upper limit of the total number of high-order modes, N0 represents the starting serial number of the high-order modes, Am 2 represents the square of the vibration amplitude of the m-th mode, An2 represents the square of the vibration amplitude of the n-th mode. The formula highlights the contribution of the low-order mode through the weighted sum of squares ratio. The tangent function tan(π / 2·fd / fc) increases when fd approaches fc, which is used to warn of the risk of aerodynamic instability.

[0027] S3: Comprehensive state assessment modeling: Integrate multiple groups of characteristic parameters extracted in S2 to construct the dynamic instability warning index of the bridge pier, the risk accumulation index of cantilever casting, and the critical state index of the cable system. Use weighted integration and non-linear functions to achieve state dimension reduction, forming a quantitative evaluation index system.

[0028] The dynamic instability warning index of the bridge pier is specifically expressed as: DVI represents the dynamic instability warning index of the bridge pier, Ev0 represents the reference value of the micro-vibration energy entropy, taking the entropy value in the healthy state, Φcrit represents the critical phase angle.

[0029] The dynamic instability warning index of piers characterizes the driving force of structural degradation by multiplying the settlement trend coefficient and the creep corrosion factor, and normalizes the energy state by dividing it by the benchmark entropy value. The index of 0.8 is used to weaken the influence of extreme values. The sinusoidal function term maps the phase angle change nonlinearly to the interval [0, 1]. When Φ_z is close to Φ_crit, the sensitivity is enhanced, and the index of 1.5 improves the sensitivity of abnormal state recognition.

[0030] The cantilever casting risk accumulation index is specifically expressed as: CRI represents the cumulative risk index of cantilever casting, and F_e0 represents the reference value of acoustic emission, which is the typical value in the initial stage of casting.

[0031] The cantilever casting risk accumulation index accumulates the construction risk in the form of time integration. The first term η_t^1.3 in the square brackets amplifies the nonlinear effect of stiffness degradation. The second term constructs the inverse relationship between the hygrothermal effect and the displacement order through H_w / D_d^0.4. The exponential attenuation term exp(-F_e / F_e0) realizes the modulation effect of acoustic emission energy on the overall risk. When F_e exceeds F_e0, the attenuation effect is significantly enhanced.

[0032] The critical state index of the cable system is specifically expressed as: CSI stands for Cable System Criticality Index.

[0033] The numerator of the critical state index of the cable system K_TG·E_vw represents the external energy input intensity, and the denominator H_m^0.7 reflects the internal loss capacity of the material. The ratio of the two constructs the energy input-dissipation balance relationship. The hyperbolic tangent function tanh(10(R_f-0.9)) is used as the switching function of frequency instability. When R_f<0.9, the output is close to -1, 0.9 <R_f<1.1时呈陡峭线性变化,R_f> Saturates to 1 at 1.1.

[0034] S4: Intelligent decision-making and regulation: A three-level cascade control mechanism is established based on the three evaluation indexes calculated in S3, including the basic construction control layer, the cumulative index and the critical state index of the cable system. The early warning hierarchical response and equipment linkage control are realized through the logic judgment tree.

[0035] The foundation construction control layer activates a yellow warning when 0.5≤DVI<0.75, and reduces the pier construction speed to 50% of the design value. The settlement and corrosion data are remeasured every 2 hours. If the DVI drops for 3 consecutive times, the warning is lifted. A red warning is triggered when 0.75≤DVI<1.2, and the foundation construction is suspended. The emergency reinforcement plan is activated until DVI<0.5 and stabilizes for 6 hours before resuming work. When DVI≥1.2, people within a radius of 200m are urgently evacuated, and the multi-point synchronous jacking system of the pier is activated, and a decision is made after the expert group's on-site evaluation.

[0036] When the daily increase in CRI ∈ [1.0, 2.5), the pouring process control layer automatically adjusts the pouring speed V_new = V_design×(2.5 - CRI) / 1.5, where design represents the preset value, and increases the sampling frequency of the temperature and humidity sensors to 5 Hz; when the daily increase in CRI ≥ 2.5, the pouring is immediately suspended, and the template stress release device is started to conduct three-dimensional laser scanning detection on the completed section. The resumption of work can only be carried out when the following conditions are met: CRI < 1.0 for 2 consecutive hours, F_e < 12.0, and D_d > 1.8; when the cumulative CRI > 15, a comprehensive assessment of the structural health is started, and fiber Bragg grating sensors are implanted to strengthen the monitoring.

[0037] When |CSI| ∈ [0.5, 0.8), the cable force dynamic adjustment layer starts the cable force fine-tuning mode ΔF = 0.03F_design×sign(CSI), where sign() is the sign function, and activates the active control of the damper, updating the wind speed prediction model every 10 minutes; when |CSI| ≥ 0.8, the wind vibration suppression system is triggered, activating the full-bridge tuned mass damper, and the cable force adjustment amplitude is increased to ΔF = 0.1F_design×sign(CSI), closing the bridge deck traffic until |CSI| < 0.5 for 1 hour; when CSI×d(CSI) / dt > 0.62, the aerodynamic measures are started, deploying the bridge deck spoiler to 45°, and injecting active airflows to interfere with the vortex shedding frequency.

[0038] S4 constructs a three-level PLC + SCADA control system. The basic construction control layer connects the hydraulic jacking equipment through the Modbus TCP protocol. The pouring control layer controls the pump truck flow valve using the OPC UA interface. The cable force adjustment is connected to the electric servo tensioner through the industrial Internet of Things platform. The control instruction generation module integrates the MATLAB / Simulink real-time simulator, and the logical judgment tree is implemented using the Drools rule engine, deployed in the Kubernetes cluster to ensure high availability.

[0039] First, in the multi-modal sensing network construction stage of the present invention, intelligent sensing arrays are respectively deployed on the bridge pier, cantilever, and cable system. Professional devices such as dual-frequency GNSS receivers, fiber Bragg grating strain gauges, and magnetic flux cable force meters are used to collect three types of heterogeneous data streams of structural deformation, environmental erosion, and dynamic response in real time at different sampling frequencies from 2 Hz to 10 MHz. Specifically, it includes 21 core parameters such as the millimeter-level settlement difference and micro-vibration acceleration of the bridge pier group, the GPS displacement and acoustic emission energy of the cantilever group, and the three-dimensional wind speed and temperature gradient of the cable group. Subsequently, it enters the cross-domain feature parameter extraction stage. Algorithms such as settlement dynamics analysis, unsteady hygrothermal coupling calculation, and thermal gradient tensor analysis are used to extract 12 key feature indicators from the original data, such as the uneven settlement trend coefficient, time-varying stiffness reduction coefficient, and vortex-induced vibration energy capture rate. Among them, the stability monitoring of the bridge pier uses the creep-corrosion synergistic factor to quantify the degree of material deterioration, the analysis of cantilever casting evaluates the displacement orderliness through chaotic characteristic quantities, and the cable system uses the magnetoelastic hysteresis loss factor to characterize the energy dissipation characteristics. In the comprehensive state assessment and modeling stage, multi-source feature parameters are fused through weighted integration and non-linear functions to construct three assessment models: the dynamic instability warning index of the bridge pier, the risk accumulation index of cantilever casting, and the critical state index of the cable system. Among them, the DVI index combines the micro-vibration energy entropy and phase angle change to achieve instability warning, the CRI index evaluates the construction risk through the non-linear coupling of stiffness reduction and acoustic emission energy, and the CSI index uses the frequency weight ratio and energy balance relationship to determine the critical state. Finally, it enters the intelligent decision-making and control stage, establishing a three-level joint control mechanism: when DVI≥0.75, the multi-point synchronous jacking system of the bridge pier is activated; when the daily increase of CRI≥2.5, the template stress release device is triggered; when |CSI|≥0.8, the full-bridge tuned mass damper is activated. Real-time linkage with hydraulic jacking equipment, pump truck flow valves, and electric servo tensioners is achieved through industrial protocols such as Modbus TCP and OPC UA, forming a closed-loop intelligent control system of monitoring - analysis - warning - control.

[0040] Second, in the attached drawings of the disclosed embodiments of the present invention, only the structures related to the disclosed embodiments are involved. For other structures, reference can be made to the general design. Without conflict, the same embodiment and different embodiments of the present invention can be combined with each other;

[0041] Finally, the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An AI-based intelligent monitoring method for bridge construction, characterized in that, include: S1: Construction of multimodal sensor network: deploy intelligent sensor arrays on bridge piers, cantilevers, and cable systems to collect multi-physical field data of structural deformation, environmental erosion, and dynamic response, forming three sets of heterogeneous monitoring data streams; S2: Extraction of cross-domain characteristic parameters: The pier group collected in S1 is subjected to settlement dynamics analysis, the cantilever group performs non-steady-state hygrothermal coupling calculation, and the cable group performs thermal gradient tensor analysis to extract three groups of key characteristic parameters that characterize structural degradation; S3: Comprehensive state assessment modeling: Integrate multiple sets of characteristic parameters extracted in S2 to construct the pier dynamic instability warning index, cantilever casting risk accumulation index and cable system critical state index, use weighted integration and nonlinear functions to achieve state dimension reduction, and form a quantitative assessment index system; S4: Intelligent decision-making and regulation: A three-level cascade control mechanism is established based on the three evaluation indexes calculated in S3, including the basic construction control layer, the cumulative index and the critical state index of the cable system. The early warning hierarchical response and equipment linkage control are realized through the logic judgment tree.

2. The intelligent monitoring method for bridge construction based on AI according to claim 1, characterized in that: The heterogeneous monitoring data stream includes pier group, cantilever group and cable group. The pier group includes the settlement difference between adjacent piers, concrete resistivity, chloride ion concentration and micro-vibration acceleration; the cantilever group includes template strain, GPS displacement, acoustic emission event energy and ambient temperature and humidity; the cable group includes cable force, temperature gradient, three-dimensional wind speed and magnetic flux.

3. The intelligent monitoring method for bridge construction based on AI according to claim 1, characterized in that: The key characteristic parameters include pier stability monitoring data, cantilever casting dynamic response monitoring data, and cable system multi-field coupling monitoring data. The pier stability monitoring data include uneven settlement trend coefficient K_s, material creep-corrosion synergy factor Q_c, micro-vibration energy entropy E_v, and electrochemical impedance phase angle Φ_z; the cantilever casting dynamic response monitoring data include time-varying stiffness reduction coefficient η_t, moisture-heat coupling effect index H_w, dynamic displacement chaos characteristic D_d, and acoustic emission fractal-entropy composite index F_e; the cable system multi-field coupling monitoring data include cable force-temperature gradient coupling coefficient K_T, vortex vibration energy capture rate E_v, magnetoelastic hysteresis loss factor H_m, and multi-order frequency weight ratio R_f.

4. The intelligent monitoring method for bridge construction based on AI according to claim 3, characterized in that: The non-uniform settlement trend coefficient is specifically expressed as: ΔS_p(t) represents the time series of the settlement difference between adjacent piers, t represents the monitoring time, m represents the number of pier pairs, and S_ref represents the reference settlement amount, specifically taking the design allowable settlement value. First, the logistic function 1 + e^{-0.1t} is used to correct the time-varying attenuation of the settlement difference between adjacent piers, and then its time derivative is calculated to reflect the trend change rate. The logarithmic term quantifies the overall deviation degree by accumulating the ratio of the settlement difference between adjacent piers to the reference value. Finally, the dynamic change rate is combined with the total dimension. The material creep-corrosion synergy factor is specifically expressed as: ε_c(τ) represents the function of creep strain varying with time, C_{ion}(τ) represents the measured value of chloride ion concentration, C_th represents the corrosion threshold concentration, which is specifically taken according to the concrete code, and α represents the corrosion acceleration index. The formula characterizes the time-varying characteristics of the material through the creep strain rate dε_c / dlnτ, quantifies the environmental corrosion effect using the C_{ion} / C_th exponential term, and the integral operation reflects the cumulative synergy effect of the two during the monitoring period [t_1, t_2]. The micro-vibration energy entropy is specifically expressed as: where N_b represents the number of frequency band divisions, taking values from 2^4 - 2^6, A(f) represents the Fourier amplitude spectrum of the vibration acceleration signal, and E_total represents the total vibration energy integrated from 0 to f_max. The formula divides the vibration spectrum into N_b sub-frequency bands, calculates the energy proportion of each frequency band, and uses the Shannon entropy formula to quantify the chaos degree of energy distribution. An increase in the entropy value indicates that the vibration energy distribution tends to be dispersed, reflecting the development of structural damage. The electrochemical impedance phase angle is specifically expressed as: Z'(ω) and Z”(ω): the real and imaginary parts of the AC impedance, ρ_c(t): the measured value of the concrete resistivity, ρ_ref: the reference value of the initial resistivity. The formula characterizes the dielectric properties of the material through the original value of the impedance phase angle arctan(Z” / Z'), and multiplies it by the resistivity change correction term 1 + (ρ_ref - ρ_c) / ρ_ref to reflect the influence of the degradation of the electrical conductivity on the phase measurement.

5. The intelligent monitoring method for bridge construction based on AI according to claim 3, characterized in that: The time-varying stiffness reduction coefficient is specifically expressed as: ε(t) represents the template strain time series, EWMA_β represents the exponential weighted moving average of the β parameter, β takes values from 0.05 to 0.2, ε_y represents the material yield strain threshold, t_char represents the characteristic time constant, taking one-third of the age, the formula quantifies the influence of sudden loads through the difference between the strain extreme value and the moving average, and the tanh function introduces a time-related stiffness degradation mechanism. When t approaches t_char, tanh approaches 1, reflecting the age effect; the hydrothermal coupling effect index is specifically expressed as: T_a(t) represents the ambient temperature, T_g represents the cement hydration characteristic temperature, k_h = 0.03 °C^-1 represents the humidity influence coefficient, t_d represents the diffusion time constant, specifically the square of the curing layer thickness divided by the diffusion coefficient, ΔT_max represents the critical temperature difference threshold of the material thermal stress, RH represents the real-time relative humidity, RH_0 represents the concrete curing reference humidity requirement value, the temperature term (T_a - T_g) / ΔT_max in the formula quantifies the thermal stress, the exponential term e^{k_h(RH - RH_0)} characterizes the non-linear influence of humidity on the material properties, and the error function erf() describes the characteristics of the hydrothermal coupling effect diffusing with time; the dynamic displacement chaos characteristic quantity is specifically expressed as: Where δ_i and δ_j represent different GPS displacement sequence sampling points, N represents the total number of sampling points, ε_d represents the phase space distance threshold, Θ() represents the Heaviside step function, the formula maps the displacement time series to an m-dimensional space through the phase space reconstruction technology, calculates the correlation integral C(ε_d), and uses the correlation dimension D_d to quantify the chaos degree of the system. A decrease in the D_d value indicates that the displacement pattern tends to be orderly; the acoustic emission fractal-entropy composite index is specifically expressed as: D_b represents the box counting method fractal dimension, H_q represents the multi-fractal spectrum width, E_k represents the energy of the kth acoustic emission event, γ = 1.2 represents the energy weighting coefficient, N_AE represents the total number of acoustic emission events, the numerator term D_b / √H_q in the formula comprehensively characterizes the spatial complexity and multi-scale characteristics of damage evolution, and the logarithmic term highlights the contribution of high-energy events through the γ-th power weighted summation, reflecting the degree of damage accumulation.

6. The intelligent monitoring method for bridge construction based on AI according to claim 3, characterized in that: The coupling coefficient of cable force - temperature gradient is specifically expressed as: F_s(t) represents the measured value of the cable force, represents the cable body temperature gradient vector, and θ_{TF} represents the angle between the temperature gradient and the direction of cable force change. represents the standard value of the critical temperature gradient. The formula performs a vector dot product operation on the cable force change rate and the normalized temperature gradient. The cosθ_{TF} term reflects the direction correlation between the two. A positive value indicates that the thermo - mechanical coupling intensifies the cable force fluctuation. The vortex - induced vibration energy capture rate is specifically expressed as: V_n(t) represents the normal wind speed component, \ddot{x}(t) represents the lateral acceleration of the cable body, f_v represents the vortex shedding frequency, f_st represents the Strouhal characteristic frequency, λ = 0.3 represents the Reynolds number correction exponent, ρ_a represents the air density, C_vm = 0.8 - 1.6 represents the vortex - induced vibration mode coupling coefficient, Δt represents the integration time window. The formula calculates the energy exchange rate through the time - domain integration of the vibration acceleration and the wind speed, and introduces the term (f_v / f_st)^{-λ} to correct the influence of non - ideal flow states, accurately characterizing the intensity of vortex - induced resonance under the actual wind field. The magneto - elastic hysteresis loss factor is specifically expressed as: Ψ_s represents the magnetic flux, H_m represents the magnetization field strength, σ_F(t) represents the standard deviation of the cable force fluctuation, Β = 0.7 represents the stress - magnetization coupling coefficient, cycle represents the loop integral along the complete closed path of the hysteresis loop, σ_F0 represents the reference standard deviation of the cable force fluctuation. The numerator of the formula calculates the area enclosed by the hysteresis loop to characterize the iron loss, and the denominator performs a maximum normalization process. The exponential term (1 + σ_F / σ_F0)^β introduces the modulation effect of the cable force fluctuation on the magneto - elastic effect. The multi - order frequency weight ratio is specifically expressed as: w_m represents the low - order mode weight, w_n represents the high - order mode weight, taking n^{-1}, f_d represents the dominant frequency, f_c represents the critical flutter frequency, M represents the total number of low - order modes, N represents the upper limit of the total number of high - order modes, N0 represents the starting serial number of the high - order modes, A_m 2 represents the square of the vibration amplitude of the m - th mode, A_n2 represents the square of the vibration amplitude of the n - th mode. The formula highlights the contribution of the low - order mode through the weighted square - sum ratio. The tangent function tan(π / 2·f_d / f_c) increases when f_d approaches f_c, which is used to warn of the risk of aerodynamic instability.

7. The intelligent monitoring method for bridge construction based on AI according to claim 1, characterized in that: The dynamic instability warning index of the pier is specifically expressed as follows: DVI represents the dynamic instability warning index of the pier, E_v0 represents the reference value of the micro-vibration energy entropy, which is the entropy value in the healthy state, and Φ_crit represents the critical phase angle.

8. The intelligent monitoring method for bridge construction based on AI according to claim 1, characterized in that: The risk accumulation index of cantilever casting is specifically expressed as: CRI represents the risk accumulation index of cantilever casting, and F_e0 represents the acoustic emission reference value, taking the typical value in the initial stage of casting.

9. The intelligent monitoring method for bridge construction based on AI according to claim 1, characterized in that: The critical state index of the cable system is specifically expressed as: CSI represents the critical state index of the cable system.

10. The AI-based intelligent monitoring method for bridge construction according to claim 1 is characterized in that: The foundation construction control layer activates a yellow warning when 0.5≤DVI<0.75, reduces the pier construction speed to 50% of the design value, re-measures settlement and corrosion data every 2 hours, and lifts the warning if DVI decreases for 3 consecutive times; a red warning is triggered when 0.75≤DVI<1.2, suspends foundation construction, and activates an emergency reinforcement plan until DVI<0.5 and stabilizes for 6 hours before resuming work; When DVI is ≥1.2, people within a radius of 200m will be evacuated urgently, and the multi-point synchronous lifting system of the bridge pier will be activated, and the decision will be made after the expert group's on-site evaluation; When the daily increase in CRI ∈ [1.0, 2.5), the pouring process control layer automatically adjusts the pouring speed V_new = V_design × (2.5 - CRI) / 1.5, where design represents the preset value, and increases the sampling frequency of the temperature and humidity sensor to 5 Hz; when the daily increase in CRI ≥ 2.5, the pouring is immediately suspended, and the template stress release device is started to conduct three-dimensional laser scanning detection on the completed section. The resumption of work can only be carried out when the following conditions are met: CRI is continuously < 1.0 for 2 hours, and F_e < 12.0, and D_d > 1.8; when the cumulative CRI > 15, a comprehensive assessment of the structural health is started, and fiber Bragg grating sensors are implanted to strengthen the monitoring; When |CSI| ∈ [0.5, 0.8), the cable force dynamic adjustment layer starts the cable force fine-tuning mode ΔF = 0.03F_design × sign(CSI), where sign() is the sign function, and activates the active control of the damper, and updates the wind speed prediction model every 10 minutes; when |CSI| ≥ 0.8, the wind vibration suppression system is triggered, the full-bridge tuned mass damper is activated, and the cable force adjustment amplitude is increased to ΔF = 0.1F_design × sign(CSI), and the bridge deck traffic is closed until |CSI| < 0.5 for 1 hour; when CSI × d(CSI) / dt > 0.62, the aerodynamic measures are started, the bridge deck spoiler is deployed to 45°, and the active air flow is sprayed to interfere with the vortex shedding frequency.

Images