Intelligent monitoring system for bridge rotation

Abstract

The invention provides a bridge swivel intelligent monitoring system, belongs to the field of bridge engineering monitoring, and is used for solving the problems of poor multi-form adaptation, low calculation precision, insufficient extreme pre-judgment, weak fault tolerance and no long-term evolution ability of a traditional bridge swivel monitoring system in related technologies. The system comprises a multi-form identification module, a multi-field coupling calculation module, a digital twinborn mapping module, a fault-tolerant cooperation module and a full-cycle evolution module, all the modules cooperate to achieve accurate identification of a rotation form, nonlinear calculation of core parameters, generation of extreme scene samples, multi-fault fault tolerance and full-life-cycle parameter optimization, and horizontal rotation, vertical rotation and combined rotation scenes can be covered. The monitoring precision and the system robustness are improved, the bridge swivel construction safety is guaranteed, and the intelligent level is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of bridge engineering monitoring, in particular to a bridge rotation intelligent monitoring system. BACKGROUND

[0002] In bridge engineering construction, rotation construction can reduce the influence on existing traffic and surrounding environment, and is widely used in complex scenes such as crossing railways and rivers. Rotation forms mainly include horizontal rotation, vertical rotation and combined rotation. The parameter monitoring precision and real-time performance in the construction process are extremely high to ensure the safety of the rotation structure and the construction efficiency.

[0003] At present, the bridge rotation monitoring system is mostly designed for a single rotation form, such as only adapting to horizontal rotation or vertical rotation, and lacks effective coverage of combined rotation forms. The function modules in the system mostly work independently, the correlation calculation of multi-field parameters (temperature, wind load, etc.) and core monitoring parameters (spherical hinge stress, main tower eccentricity, etc.) mostly uses linear models, and there is no dynamic constraint adjustment mechanism, which is difficult to adapt to complex working conditions in the rotation process. At the same time, the existing system lacks fault-tolerant design for multiple concurrent faults, and does not integrate construction and operation data to realize whole-cycle parameter optimization.

[0004] The existing technology has obvious defects: first, the multi-form adaptation ability is poor, which cannot meet the monitoring needs of combined rotation scenes; second, the multi-field coupling calculation precision is low, the linear model does not match the dynamic working condition, and the fixed constraint boundary leads to large early warning deviation; third, the system robustness is insufficient, the single fault handling capacity is limited, and the monitoring is easily interrupted when multiple faults occur; fourth, there is no whole-cycle evolution ability, the construction and operation data are separated, and the monitoring model cannot be continuously optimized, which seriously affects the safety and economy of bridge rotation construction. SUMMARY

[0005] The present application provides a bridge rotation intelligent monitoring system, which can solve the problems of insufficient multi-form coverage, low calculation precision, weak fault tolerance and no whole-cycle evolution of the existing monitoring system through multi-module cooperation and multi-form adaptation, and improve the safety and intelligent level of bridge rotation monitoring.

[0006] In a first aspect, the application provides a bridge swivel intelligent monitoring system. It comprises a multi-form identification module, a multi-field coupling calculation module, a digital twin mapping module and an early warning module, which are sequentially connected in signal. The multi-form identification module is used to identify the form of the bridge swivel, which includes horizontal rotation, vertical rotation and combined rotation. The digital twin mapping module is used to map the entity state of the bridge swivel. The multi-field coupling calculation module is used to calculate the core parameters of the bridge swivel based on the entity state data and multi-field parameters output by the digital twin mapping module. The multi-field parameters include temperature, wind load, vibration and stress, and the core parameters include spherical hinge stress, main tower eccentricity and arch rib cable force. The early warning module is used to send an early warning signal based on the core parameters output by the multi-field coupling calculation module.

[0007] By adopting the above technical solution, the multi-form identification module realizes full coverage identification of horizontal rotation, vertical rotation and combined rotation. Combined with the entity state data provided by the digital twin mapping module, the multi-field coupling calculation module can accurately associate multi-field parameters with core parameters. The early warning module responds to core parameter abnormalities in real time, forming a basic monitoring closed loop of "identification-calculation-mapping-early warning". This effectively solves the problems of insufficient multi-form coverage and broken basic monitoring link in existing systems, and improves the parameter monitoring integrity and real-time performance of the swivel process.

[0008] Further, the multi-form identification module comprises a feature extraction unit and a form matching unit, which are connected in signal. The feature extraction unit is used to process sensor data of the bridge swivel, including horizontal rotation angle data, arch rib inclination angle data and bridge type parameter data. The form matching unit is used to match the corresponding swivel form physical model according to the feature data output by the feature extraction unit.

[0009] By adopting the above technical solution, the multi-form identification module is divided into feature extraction and form matching units. The feature extraction unit processes key sensor data such as horizontal rotation angle and arch rib inclination angle. The form matching unit accurately matches the physical model based on the feature data, avoiding form misjudgment caused by single identification logic, improving the accuracy of multi-form identification, and providing a basis for subsequent form-specific calculation of modules.

[0010] Further, the feature extraction unit further comprises an attention enhancement unit. The attention enhancement unit is used to assign a preset weight to the form-specific features in the feature data, including spherical hinge rotation angle features corresponding to horizontal rotation and arch rib inclination angle features corresponding to vertical rotation.

[0011] By adopting the technical scheme, the attention enhancement unit gives high weight to morphological exclusive features such as spherical hinge angle and arch rib inclination, strengthens the influence of key features on the recognition result, and weakens the interference of irrelevant data such as environmental noise, which can further improve the morphological recognition accuracy, especially in a small sample or feature overlap scene, and solve the recognition deviation problem caused by equal feature weight in the existing recognition technology.

[0012] Further, the multi-field coupling calculation module includes a nonlinear modeling unit and a dynamic constraint generation unit, the nonlinear modeling unit is signal connected with the dynamic constraint generation unit; the nonlinear modeling unit is used for fitting the correlation between the multi-field parameters and the core parameters; the dynamic constraint generation unit is used for adjusting the constraint boundary of the core parameter based on real-time multi-field parameters.

[0013] By adopting the technical scheme, the nonlinear modeling unit replaces the traditional linear model, which can accurately fit the complex correlation between the multi-field parameters and the core parameters, and the dynamic constraint generation unit adjusts the constraint boundary according to the real-time working condition, avoiding the pre-warning lag or false alarm caused by fixed constraints, significantly improving the multi-field coupling calculation accuracy, and adapting to the dynamic working condition change in the rotation process.

[0014] Further, the dynamic constraint generation unit further includes a physical attribution unit; the physical attribution unit is used for quantifying the contribution degree of each multi-field parameter to the constraint boundary.

[0015] By adopting the technical scheme, the physical attribution unit quantifies the influence degree of temperature, wind load and other multi-field parameters on the constraint boundary, makes the constraint adjustment logic transparent, facilitates the construction personnel to understand the constraint source and rationality, reduces the trust concern for the black box model, at the same time provides an interpretable adjustment basis for subsequent parameter optimization, and improves the engineering practicability of the system.

[0016] Further, the digital twin mapping module includes an extreme scene sample generation unit and a consistency verification unit, the extreme scene sample generation unit is signal connected with the consistency verification unit; the extreme scene sample generation unit is used for generating the extreme scene sample of the bridge rotation based on the constraint boundary output by the dynamic constraint generation unit; the consistency verification unit is used for verifying the consistency of the extreme scene sample and the entity state data.

[0017] By adopting the technical scheme, the extreme scene sample generation unit can generate monitoring samples under extreme working conditions based on the dynamic constraint, provide data support for extreme risk prediction, and the consistency verification unit ensures the matching of the sample and the entity state, avoids the prediction deviation caused by invalid samples, solves the problem that the existing system lacks extreme scene coping ability, and improves the risk prevention and control level of the rotation construction.

[0018] Further, the extreme scene sample generation unit comprises a morphology-specific branch unit and a step adjustment unit, the morphology-specific branch unit is in signal connection with the step adjustment unit; the morphology-specific branch unit is configured to generate corresponding extreme scene sample branches for different rotation morphologies; and the step adjustment unit is configured to adjust the step of sample generation according to the generation stage of the extreme scene sample.

[0019] By adopting the above technical solutions, the morphology-specific branch unit generates exclusive sample branches for different rotation morphologies, ensuring the adaptability of the sample to the physical law of the morphology; the step adjustment unit dynamically adjusts the step according to the generation stage, balancing the efficiency and accuracy of sample generation, avoiding the problems of long sample generation time or poor consistency caused by a single branch or fixed step, and improving the timeliness and reliability of extreme scene prediction.

[0020] Further, it further comprises a fault-tolerant cooperation module, the fault-tolerant cooperation module is in signal connection with the multi-field coupling calculation module and the early warning module respectively; the fault-tolerant cooperation module comprises a fault detection unit, a redundant backup unit and a priority decision unit, the fault detection unit, the redundant backup unit and the priority decision unit are sequentially in signal connection; the fault detection unit is configured to monitor the running state of each module; the redundant backup unit is configured to deploy backup devices of each module; and the priority decision unit is configured to determine the priority of fault handling according to the detection result of the fault detection unit.

[0021] By adopting the above technical solutions, the fault-tolerant cooperation module monitors the module state in real time through the fault detection unit, the redundant backup unit provides device-level backup, and the priority decision unit determines the processing order according to the fault influence degree, avoiding the interruption of overall monitoring caused by single module failure, solving the problem of insufficient robustness of the existing system, and ensuring the continuity of the rotation monitoring, especially suitable for time-sensitive scenes such as adjacent iron.

[0022] Further, the priority decision unit further comprises a coupling risk calculation unit; the coupling risk calculation unit is configured to quantify the joint risk when multiple faults occur concurrently.

[0023] By adopting the above technical solutions, the coupling risk calculation unit can quantify the joint risk when multiple faults occur concurrently, avoiding the delay of processing high-risk faults caused by single fault priority determination, ensuring that the problem with the greatest impact on the rotation safety is solved first in the multi-fault scene, and further improving the fault tolerance and safety guarantee level of the system in complex fault working conditions.

[0024] Further, a full-cycle evolution module is further included, and the full-cycle evolution module is in signal connection with the digital twin mapping module and the multi-field coupling calculation module respectively; the full-cycle evolution module includes a data fusion unit and a parameter optimization unit, the data fusion unit is in signal connection with the parameter optimization unit; the data fusion unit is used for fusing construction stage data and operation and maintenance stage data of the bridge turntable; and the parameter optimization unit is used for optimizing parameters of the multi-field coupling calculation module and the digital twin mapping module based on fusion data output by the data fusion unit.

[0025] By adopting the technical solution, the full-cycle evolution module fuses construction and operation and maintenance data, the parameter optimization unit continuously optimizes parameters of the calculation and mapping modules based on the fusion data, the full life cycle iteration upgrade of the monitoring model is realized, the problem of data fragmentation and no evolution capability of the existing system is solved, the system continuously maintains high monitoring accuracy in the full life cycle of the bridge, and long-term operation and maintenance cost is reduced.

[0026] In summary, the present application at least has the following beneficial effects:

[0027] 1. A multi-form adaptive bridge turntable intelligent monitoring system is provided, and full-link intelligent monitoring is realized.

[0028] 2. By attention enhancement and nonlinear modeling, multi-form recognition and multi-field coupling calculation accuracy are improved.

[0029] 3. By fault tolerance cooperation and full-cycle evolution, system robustness and long-term applicability are strengthened.

[0030] It should be understood that the content described in the summary section is not intended to limit the key or important features of the embodiments of the present application, nor to limit the scope of the present application. Other features of the present application will become apparent through the following description. BRIEF DESCRIPTION OF DRAWINGS

[0031] The above and other features, advantages, and aspects of the embodiments of the present application will become more apparent by describing in detail the following detailed description in conjunction with the accompanying drawings. In the drawings, the same or similar reference numerals indicate the same or similar elements, in which:

[0032] Figure 1 A principle diagram of a bridge turntable intelligent monitoring system in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0033] To make the objects, technical solutions and advantages of the embodiments of the present application clearer, the following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0034] In addition, the term "and / or" in this paper is only a description of the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which can represent the three cases of A alone, A and B together, and B alone. In addition, the character " / " in this paper generally represents that the front and rear associated objects are in an "or" relationship.

[0035] The present application provides a bridge swivel intelligent monitoring system, which can cover horizontal rotation, vertical rotation and combined rotation multi-form, improve multi-field coupling calculation precision, have multi-fault tolerance ability and full-cycle evolution characteristics, can guarantee swivel construction safety, adapt to complex scenes, and effectively improve the intelligent level of monitoring.

[0036] The embodiments of the present application disclose a bridge swivel intelligent monitoring system.

[0037] Figure 1 The principle diagram of a bridge swivel intelligent monitoring system in the embodiments of the present application is shown.

[0038] Reference Figure 1 The system comprises a multi-form identification module, a multi-field coupling calculation module, a digital twin mapping module, a warning module, a fault tolerance cooperation module and a full-cycle evolution module. The multi-form identification module, the multi-field coupling calculation module, the digital twin mapping module and the warning module are sequentially signal connected.

[0039] The multi-form identification module is the "form perception core" of the bridge swivel intelligent monitoring system, and connects the perception layer sensor and the subsequent multi-field coupling calculation module. Its core function is to accurately identify the swivel form to adapt to the exclusive physical model, and specifically used for identifying the form of the bridge swivel, the form comprising horizontal rotation, vertical rotation and horizontal-vertical combined rotation. In actual construction, the module receives the multi-dimensional sensor data transmitted by the perception layer in real time, first performs data validity check (such as eliminating abnormal values), and then starts the feature extraction and form matching process, to provide form basis for the subsequent dynamic constraint generation and extreme scene prediction of the system.

[0040] The multi-form recognition module comprises a feature extraction unit and a form matching unit, the feature extraction unit is signal connected with the form matching unit; the feature extraction unit is used for processing sensor data of bridge rotation, and the sensor data comprises horizontal rotation angle data, arch rib inclination angle data and bridge type parameter data; specifically, the unit first pre-processes the collected original sensor data, including data normalization (mapping the horizontal rotation angle, arch rib inclination angle and the like to the [0, 1] interval to eliminate the dimensional difference) and time sequence smoothing (filtering high-frequency noise by using a sliding average method, and the window size is set to 5 sampling points), and then inputs the pre-processed data into a lightweight MobileNet-V3 network for deep feature extraction, the network reduces the calculation amount by using a deep separable convolution kernel (reducing 70% of the power consumption compared with a traditional convolution), adapts to the real-time requirement of rotation construction, and the extracted feature data covers key distinguishing information of the rotation form (such as the continuous rotation angle change feature of horizontal rotation and the arch rib inclination angle increment feature of vertical rotation).

[0041] The form matching unit is used for matching a corresponding rotation form physical model according to the feature data output by the feature extraction unit; the matching process takes “feature similarity + physical law verification” as the double criteria, first calculates the cosine similarity of the feature vector output by the feature extraction unit and the preset form label vector (horizontal rotation, vertical rotation and combined rotation each correspond to a group of label vectors), and the similarity calculation formula is wherein is the extracted feature vector, is the preset label vector, represents the L2 norm of the vector, and when the similarity of a certain form is , the form is preliminarily determined as the form; secondly, the physical law verification is combined, such as “the vertical rotation angle change rate ” needs to be met in the horizontal rotation scene, “the horizontal rotation angle change rate ” needs to be met in the vertical rotation scene, and “both the horizontal rotation angle and the vertical rotation angle exist effective changes” needs to be met in the combined rotation scene, so that the double criteria ensure that the form matching is unambiguous, and the misjudgment caused by a single feature similarity (such as misjudging the horizontal rotation of slight vertical rotation as combined rotation) is avoided.

[0042] The feature extraction unit further comprises an attention enhancement unit; the attention enhancement unit is used for giving a preset weight to form exclusive features in the feature data, and the form exclusive features comprise a spherical hinge rotation angle feature corresponding to horizontal rotation and an arch rib inclination angle feature corresponding to vertical rotation; the unit realizes feature enhancement by using a convolution block attention module (CBAM), and specifically divides into two branches of channel attention and spatial attention: the channel attention branch is used for highlighting the channel weight where the form exclusive features are located, and the calculation process is: calculating the channel attention map of the feature map output by the feature extraction unit (C is the number of channels, H is the height of the feature map, and W is the width of the feature map), first global average pooling is performed to obtain statistical features in the channel dimension , then through two fully connected networks (the first layer reduces the dimension of the coefficient , and the second layer increases the dimension to the original channel number) and the Sigmoid activation function, the channel attention weight is generated , where is the weight of the fully connected layer, is the bias, and this weight gives higher values (usually ) to the horizontal rotation ball hinge angle feature channel and the vertical rotation arch rib inclination angle feature channel, and gives lower values (usually ) to the channel where the environmental noise is located; the spatial attention branch is used to focus on the spatial location of the shape-specific features on the feature map, and the calculation process is: taking the average of the feature map ( represents element-level multiplication) along the channel dimension to obtain spatial features , then through convolution layer (padding is 1 to maintain the feature map size) and Sigmoid activation function, the spatial attention weight is generated ( is the bias of the convolution layer), which gives high values to the spatial positions corresponding to the ball hinge angle data and the arch rib inclination angle data, further suppressing the noise interference in irrelevant areas; through the synergistic effect of channel and spatial attention, the signal-to-noise ratio of the shape-specific features is improved by more than 3 times, providing more accurate feature data for the subsequent shape matching unit.

[0043] In addition, to solve the problem of insufficient model generalization ability caused by the scarcity of rotation body shape samples (especially the flat and rigid combined rotation samples, with less than 50 actual cases in engineering), the attention enhancement unit also combines the transfer learning strategy, taking the CBAM-MobileNet model parameters pre-trained on the general building structure monitoring dataset (containing more than 100,000 samples) as the initial weights, only fine-tuning the parameters of the model's convolution layer and attention module, and introducing virtual samples (1000+ groups, covering extreme working conditions of different rotation body shapes) generated by the digital twin mapping module to supplement the training data. The loss function in the fine-tuning process uses a weighted cross-entropy loss, the formula is CrossEntropy CrossEntropy , where is the number of real samples, is the number of virtual samples, and 0.3 is the weight coefficient of virtual samples (to avoid overfitting caused by too many virtual samples), is the real shape label, To predict the label of the model, the morphological recognition accuracy in the small sample scene is improved from the traditional method of 70% to 95% through this training strategy to The above, and the recognition delay is controlled within 100 ms, fully meeting the real-time monitoring needs of the body construction.

[0044] The digital twin mapping module, as the "entity digitization core" of the system, takes in the morphological labels output by the multi-morphological recognition module and the real-time monitoring data from the perception layer. By constructing a multi-degree-of-freedom dynamic model that is 1:1 with the entity bridge, it realizes accurate mapping of the entity state, specifically for mapping the entity state of the bridge body. The construction of this model is based on the bridge design drawings (including structural dimensions and material parameters) and dynamically corrects the model parameters by combining with the real-time collected entity data (such as the initial position of the spherical hinge and the installation accuracy of the arch rib) during the construction period. The freedom configuration of the model adapts to different body morphologies: in the horizontal rotation scene, it focuses on configuring "spherical hinge rotation freedom + main beam translation freedom" (a total of 2 main freedoms), in the vertical rotation scene, it focuses on configuring "arch rib rotation freedom around hinge + cable stretching freedom" (a total of 2 main freedoms), and in the combined rotation scene, it integrates both types of freedom and adds "horizontal-vertical coupling freedom" (a total of 3 main freedoms), ensuring that the model can truly reflect the mechanical behavior of the entity body and the mapping error is controlled within 2%.

[0045] The digital twin mapping module includes an extreme scene sample generation unit and a consistency verification unit. The extreme scene sample generation unit is signal connected with the consistency verification unit. The extreme scene sample generation unit is used to generate extreme scene samples of the bridge body based on the constraint boundaries output by the dynamic constraint generation unit. In actual application, this unit needs to first receive the real-time constraint vectors transmitted by the dynamic constraint generation unit (such as the upper limit of the spherical hinge stress , the upper limit of the main tower eccentricity , the upper limit of the arch rib cable force , and the upper limit of the tower inclination . Taking these constraints as "sample generation boundaries" can avoid generating invalid samples that violate the laws of engineering physics. The generation process uses a condition diffusion model specific to the morphology. This model gradually generates time series samples under extreme scenarios (such as a 10-minute parameter change sequence under "strong wind + adjacent iron vibration") through a reverse denoising process. The loss function of the denoising process needs to embed the morphology-specific physical constraints, with the formula being where is the mean square error loss (to ensure that the sample distribution matches), is the physical constraint weight, is the morphology-specific physical penalty term - in the horizontal rotation scene , is the predicted spherical hinge stress of the sample, (For the sample prediction of main tower eccentricity), in the vertical rotation scenario ( For the predicted buckle force of the sample, The tower tilt angle predicted for the sample is used to ensure that the generated extreme samples conform to the characteristics of extreme working conditions without exceeding the safety constraints of the physical structure.

[0046] The consistency verification unit is used to verify the consistency between the extreme scenario sample and the entity state data. The verification process uses the multi-degree-of-freedom dynamic model of the digital twin mapping module as a "benchmark." First, key parameters from the extreme scenario sample (such as horizontal rotation angle, arch rib inclination angle, and spherical hinge stress) are input into the model. The theoretical response of the entity is calculated by solving the dynamic equations. The general form of the dynamic equations is... ,in The mass matrix (calculated from the density and dimensions of the structural materials) The damping matrix is ​​(based on the fitting of the measured vibration attenuation curve). The stiffness matrix (which changes dynamically with the rotation angle, such as the linear increase in the contact stiffness of a ball joint as the rotation angle increases during horizontal rotation). This is the external load vector (including wind load, vibration, self-weight, etc.). Let be the displacement / rotation vector; then calculate the relative deviation between the sample predicted response and the model theoretical response, using the deviation formula: ,in The response values ​​are for extreme samples. The theoretical response value calculated for the model; when If the samples are consistent, it is considered valid. Then, the "local denoising correction" process is initiated—the time step data with excessive bias in the sample is fine-tuned using the gradient descent method, and the correction formula is as follows: ( To adjust the step size, until the deviation meets the requirements, ensuring that extreme samples can truly reflect the behavior of the entity under extreme working conditions.

[0047] The extreme scene sample generation unit includes a morphology-specific branch unit and a step size adjustment unit. The morphology-specific branch unit is signal-connected to the step size adjustment unit. The morphology-specific branch unit is used to generate corresponding extreme scene sample branches for different rotation morphologies. This unit is configured with independent diffusion branch links for horizontal rotation, vertical rotation, and combined rotation, and each branch embeds the core physical law of the corresponding morphology: the horizontal rotation branch embeds the ball-joint friction law (…). , The coefficient of friction, (for the normal force of the ball joint) and the force balance formula of the support leg ( (G is the self-weight of the main beam) Off-center from the main tower This is the offset of the support leg. (where the radius is the ball joint radius), ensuring that the ball joint stress and support reaction force in the sample conform to mechanical equilibrium; the vertical rotation branch is embedded in the arch rib bending equation ( E is the elastic modulus. Let the moment of inertia of the cross section be... The relationship between the arch rib deflection and the cable force ( L is the lever arm of the cable. (For the self-weight arm of the arch rib), ensuring that the bending moment of the arch rib and the cable force in the sample do not exceed the structural limits; the combined rotation branch integrates two types of physical laws and adds an additional "horizontal-vertical coupling constraint" (such as the influence coefficient of the change in horizontal rotation traction speed on the inclination angle of the vertical rotation arch rib). This avoids contradictory cases where "excessive horizontal rotation leads to instability of the arch rib".

[0048] The step size adjustment unit is used to adjust the step size of sample generation according to the generation stage of extreme scenario samples; this unit optimizes the step size through a two-dimensional adjustment strategy of "stage division + morphological adaptation": firstly, the diffusion generation process is divided into an initial stage ( ), medium term ( ), later stage ( The process consists of three stages, with T being the base step size (horizontal rotation). Vertical rotation Combination Transformation The step size coefficients for different stages are calculated according to the formula. The initial setup uses a large step size (coefficient 5) to quickly reduce noise (completing 80% of the denoising process); the intermediate setup uses a medium step size (coefficient 2) to balance efficiency and accuracy; and the final setup uses a small step size (coefficient 1) to finely optimize sample details (ensuring the accuracy of key parameter deviations). Secondly, for combined transformation scenarios, a "branch parallel diffusion" mechanism is enabled, where horizontal and vertical transformation branches are generated simultaneously, with a parallel coefficient. Total generation time Single branch time For example, generating a single branch from a combined branch takes 9.6 seconds, but after parallelization, it only takes 4.8 seconds, perfectly adapting to adjacent railway window points. The unit also monitors the consistency deviation of samples in real time during the step size adjustment process (by calling the preliminary verification results of the consistency verification unit). If the deviation suddenly increases at a certain stage (e.g., ...), the unit will detect the deviation. If the deviation returns to a reasonable range, the step size coefficient will be automatically reduced by 1 level (e.g., from medium step size 2 to small step size 1) until the deviation returns to a reasonable range.

[0049] The multi-field coupling calculation module is the "parameter calculation core" of the bridge swing intelligent monitoring system, which is connected to the entity state data (such as real-time rotation angle and displacement) output by the digital twin mapping module and connected to the warning module and the fault tolerance cooperation module. The core function is to solve the problem of "nonlinear coupling of temperature, wind load, vibration and other multi-field parameters on the core structure parameters", and accurately calculate the key indicators to ensure the safety of the swing. Specifically, the core parameters of the bridge swing are calculated based on the entity state data and multi-field parameters output by the digital twin mapping module. The multi-field parameters include temperature, wind load, vibration and stress. The core parameters include spherical hinge stress, main tower eccentricity and arch rib cable force. Wherein, the multi-field parameters are all from the special sensors deployed in the perception layer: the temperature parameter is collected by the fiber bragg grating sensor embedded in the spherical hinge and the main tower (sampling frequency 1 Hz, accuracy ±0.1℃), the wind load parameter is obtained by the three-dimensional ultrasonic anemometer installed on the top of the main tower (sampling frequency 10 Hz, accuracy ±0.1m / s), the vibration parameter is monitored by the acceleration sensor fixed on the main beam (sampling frequency 50 Hz, accuracy ±0.001m / s²), and the initial stress parameter is collected by the strain gauge pasted on the spherical hinge and the arch rib (sampling frequency 10 Hz, accuracy ±1με). The calculation results of the core parameters need to be fed back to the digital twin mapping module in real time to correct the dynamic model, form a closed loop of "data acquisition-calculation-model correction", and ensure that the calculation accuracy is continuously optimized with the construction process.

[0050] The multi-field coupling calculation module includes a nonlinear modeling unit and a dynamic constraint generation unit, which are signal connected. Because the correlation between the multi-field parameters and the core parameters in the bridge swing has significant nonlinearity (such as when the temperature is greater than 30℃, the rate of spherical hinge stress increase with temperature is 2 times faster than when the temperature is lower; when the wind speed is greater than 10m / s, the main tower eccentricity increases with the square of the wind speed), the traditional linear regression model error is more than 8%, which cannot meet the accuracy requirement. Therefore, the nonlinear modeling unit is used to fit the correlation between the multi-field parameters and the core parameters. Specifically, a radial basis function (RBF) neural network is used to realize nonlinear fitting. The input layer of the network is the standardized multi-field parameter vector ( for temperature deviation, for wind speed, for vibration acceleration, for initial stress), the hidden layer uses Gaussian kernel activation function is the center vector of the jth hidden layer neuron, is the width parameter, which is determined by K-means clustering algorithm), and the output layer is the core parameter vector ( is the spherical hinge stress, is the main tower eccentricity, is the arch rib cable force), and the output formula is The number of hidden layer neurons, The weight, The bias, minimize the mean square error loss by gradient descent method Training, The number of training samples is 1000+ virtual extreme samples generated by the digital twin module and The number of conventional samples collected in the entity construction, ensuring that the model can maintain high fitting accuracy (error ) under both conventional and extreme working conditions.

[0051] The dynamic constraint generation unit is used to adjust the constraint boundary of the core parameter based on real-time multi-field parameters. Its original intention is to solve the problem of "over-conservative" or "insufficient constraint" of traditional static constraints (such as fixed spherical hinge stress ) under dynamic working conditions - for example, in severe cold weather , the yield strength of steel increases compared to the baseline temperature , and static constraints will cause unnecessary construction adjustments; while in strong wind , the anti-overturning capacity of the main tower decreases, and static constraints are easy to cause over-warning risk. This unit realizes dynamic adjustment by establishing a quantitative correlation model of "multi-field parameters-constraint boundary". The constraint formulas for different core parameters are as follows: spherical hinge stress constraint (The baseline constraint is , the temperature influence coefficient is , the real-time temperature is , the vibration influence coefficient is ), main tower eccentricity constraint (The baseline constraint is , the wind speed influence coefficient is (The temperature difference influence coefficient is ), arch rib cable force constraint (The baseline constraint is , the wind speed influence coefficient is , the real-time vertical rotation angle is , the vertical rotation angle influence coefficient is ).

[0052] The update frequency of the constraint boundary is synchronized with the acquisition frequency of the multi-field parameters (1Hz), ensuring that the constraint at each moment is accurately matched with the current working condition. The dynamic constraint generation unit also includes a physical attribution unit. The physical attribution unit is used to quantify the contribution of each multi-field parameter to the constraint boundary. Its core value lies in solving the "black box problem" of the nonlinear constraint model. Construction personnel do not need to understand the complex principles of neural networks. They can intuitively understand "why the constraint boundary is adjusted" through the attribution results, such as "the current spherical hinge stress constraint is relaxed from 101.5MPa to 124.5MPa, and the temperature reduction contributes , vibration reduction contribution , so as to improve the confidence and execution efficiency of constraint adjustment. The contribution quantification is realized by calculating the sensitivity of the constraint boundary to each multi-field parameter: first, the partial derivative of the dynamic constraint formula to each parameter is solved (such as , reflecting the change of the constraint boundary when the temperature changes ), then the absolute influence of each parameter is calculated combined with the real-time parameter value (such as , finally the contribution ratio is calculated according to the formula Cont , where is the absolute influence of the i-th parameter, and is the number of parameters); the attribution result is output in the form of numerical value and heat map, where the deeper the color of the parameter in the heat map, the higher its contribution, for example, the contribution ratio of wind load is corresponds to dark red, the contribution ratio of temperature is corresponds to light red, which is intuitive and easy to understand.

[0053] In addition, there is a data interaction closed loop between the nonlinear modeling unit and the dynamic constraint generation unit: the real-time constraint boundary output by the dynamic constraint generation unit will be used as the "parameter rationality check threshold" of the nonlinear modeling unit, if the core parameter calculated by the model exceeds the constraint boundary (such as ), the model parameter fine-tuning will be triggered - the weights of the RBF neural network are updated through Bayesian estimation ( where is the measured spherical hinge stress, and is the model calculation value), to ensure that the model calculation result is always within the safety constraint range; at the same time, the trend of the core parameter output by the nonlinear modeling unit will be fed back to the dynamic constraint generation unit to predict the adjustment direction of the constraint boundary (such as predicting that the temperature will continue to decrease in 5 minutes, and reserving the constraint relaxation space in advance), further improving the dynamic adaptability of the system.

[0054] The early warning module, as the "safety response core" of the intelligent monitoring system of the bridge tower, connects the core parameters (spherical hinge stress, main tower eccentricity, arch rib cable force, etc.) output by the multi-field coupling calculation module on the top and the fault-tolerant collaboration module and the construction control terminal on the bottom, its core function is to accurately determine the abnormal state of the core parameter, generate and transmit the early warning signal in time, guide the construction personnel or the system to start the corresponding emergency measures, and avoid the risk of structural safety. Specifically, the early warning signal is issued based on the core parameters output by the multi-field coupling calculation module; this module is not simply compared with the fixed threshold, but constructs the early warning logic of "shape adaptation + dynamic grading + collaborative response", to ensure that the early warning neither misses extreme risks nor produces false alarms due to transient fluctuations; the receiving frequency of the core parameter is synchronized with the output frequency of the multi-field coupling calculation module (1 Hz), to ensure real-time performance and avoid the risk of risk expansion due to delay.

[0055] The core of the early warning module is the "dynamic hierarchical early warning threshold system", which needs to adapt to the exclusive safety needs of different rotating forms, and is linked with the constraint boundary output by the dynamic constraint generation unit. In the horizontal rotation scene, the threshold setting of the spherical hinge stress and the main tower eccentricity is focused on, and the arch rib cable force threshold is only used as an auxiliary (because the arch rib stress is smaller in horizontal rotation); In the vertical rotation scene, the arch rib cable force and the tower inclination angle (converted from the main tower eccentricity, the conversion formula is , H is the height of the main tower) are the core early warning parameters; In the combined rotation scene, the threshold values of both types of parameters need to be activated, and "parameter coupling early warning" (such as when the spherical hinge stress exceeds the threshold value in horizontal rotation, the arch rib cable force also exceeds the threshold value in vertical rotation, a higher level of early warning needs to be triggered) is added. The setting of each level of threshold is based on the dynamic constraint boundary, combined with the structural safety factor and engineering experience calibration, and the specific calculation formula is: the first level of early warning threshold , the second level of early warning threshold , and the third level of early warning threshold , where is the real-time constraint boundary (such as spherical hinge stress constraint, cable force constraint) output by the dynamic constraint generation unit, is the safety factor gradient - the first level of early warning corresponds to "the parameter is close to the warning range, which needs to be monitored", the second level corresponds to "the parameter enters the warning range, which needs to prepare emergency measures", and the third level corresponds to "the parameter breaks through the safety boundary, which needs to stop rotating immediately and start fault tolerance", which can avoid "response lag" or "overreaction" caused by a single threshold.

[0056] To reduce false positives caused by instantaneous fluctuations (such as sudden wind gusts causing the main tower eccentricity to rise instantaneously), the early warning module introduces "time judgment conditions", that is, the core parameters need to continuously meet "exceed the corresponding level threshold" and the duration reaches the set time length, to trigger the early warning signal. The time length is determined according to the adjustment window of the rotating construction, for example, in the horizontal rotation scene, the adjustment window is usually 3min (180s), so the setting "the parameter exceeds the threshold for 3 consecutive sampling periods (1s per sampling period, coupled with the calculation frequency of multiple fields)" is the trigger condition, and the judgment formula is count , where is the core parameter value of the ith sampling period, is the kth threshold , is the indicator function, which takes 1 when the condition in the parentheses is true, and 0 otherwise), when count=3, the parameter is confirmed to be abnormal and the corresponding level of early warning is triggered. This design can control the false positive rate to below (based on group historical construction data verification).

[0057] The generation of early warning signals needs to include "key information elements" to ensure that construction personnel or systems can quickly locate the problem and respond. The specific elements include: parameter identification (such as "horizontal rotation-sphere hinge stress" and "vertical rotation-arch rib cable force"), current parameter value (accurate to one decimal place, such as 124.5 MPa), corresponding level threshold value (such as a three-level threshold value of 124.5 MPa), early warning level (distinguished by words and colors, with level one being blue, level two being yellow, and level three being red), and recommended response measures (such as level one warning "increase monitoring frequency to 0.5s / second", level two warning "suspend rotation and check the traction system", and level three warning "start the backup traction sensor and apply for adjacent iron window extension"). These information is transmitted through two channels: one is the visual interface of the construction control terminal (real-time pop-up display), and the other is the signal interface of the fault-tolerant cooperation module (level three warning automatically triggers the backup device start instruction of the fault-tolerant module), ensuring "human-machine" cooperative response.

[0058] In addition, the early warning module also has the function of "historical early warning data recording and feedback". After each early warning trigger, the early warning time, trigger parameter, threshold value, duration, and processing result (such as "2025-XX-XX 14:30, horizontal rotation main tower eccentricity reaches 126mm (two-level threshold value 126.2mm), lasts for 3s, processing measures: reduce traction speed 0.02° / min") are automatically recorded. These data are archived by day and then fed back to the whole cycle evolution module regularly (such as every week) for optimizing the early warning threshold and the core parameter calculation model. For example, if the level one warning of a certain parameter is frequently triggered but does not develop into a higher level, the whole cycle evolution module will adjust the level one threshold safety factor of that parameter based on the feedback data (such as from 0.8 to 0.85) to reduce invalid early warnings. If a parameter appears "skips level two and directly triggers level three" multiple times, the fitting accuracy of the multi-field coupling calculation model will be calibrated to ensure that the parameter change trend can be predicted in advance, forming a "early warning-processing-recording-optimization" whole cycle closed loop, further improving the safety response reliability of the system.

[0059] The fault-tolerant collaborative module is the "robust core" of the bridge rotation intelligent monitoring system in response to equipment failures and parameter anomalies. Its design, which connects to the multi-field coupling calculation module and the early warning module respectively, aims to simultaneously receive two types of key information: first, the real-time status of core parameters output by the multi-field coupling calculation module (such as the stress of the ball joint and whether the eccentricity of the main tower is approaching the constraint boundary); and second, the early warning signals issued by the early warning module at various levels (especially the emergency state of the third-level early warning). Through the collaborative judgment of these two types of information, the fault tolerance delay caused by relying solely on parameters or early warnings is avoided. For example, when the multi-field coupling module detects "abnormal traction sensor data" but the early warning module does not trigger an early warning, the sensor fault tolerance still needs to be activated to prevent potential data deviations from causing subsequent misjudgments. However, when the early warning module triggers a third-level early warning and the multi-field coupling module reports "core parameters exceeding the limit," the fault tolerance of the execution equipment (such as the traction system and the cable pump) should be activated first to ensure rapid loss mitigation.

[0060] The fault-tolerant collaborative module includes a fault detection unit, a redundant backup unit, and a priority decision unit, which are sequentially signal-connected. This "detection-decision-backup" link design ensures a seamless fault handling process, with the output of each unit directly serving as the input of the next, avoiding information loss or delay. The fault detection unit monitors the operating status of each module, and its monitoring logic is divided into two categories: "communication status detection" and "data consistency verification." Communication status detection is implemented through a "heartbeat mechanism." The fault detection unit sends heartbeat packets every second to key modules such as the multi-field coupling calculation module, early warning module, and traction execution system. If no heartbeat response is received from the target module for three consecutive times (within 3 seconds), it is considered a "communication failure." Data consistency verification calculates the deviation rate of the same parameter between the sensor and execution device data at the primary and backup nodes, using the formula: ( Master node data, (For backup node data), when Time (e.g., the speed feedback from the main traction sensor) Backup sensor feedback If a fault is detected, it is classified as a "data fault". Both types of detection results are transmitted to the priority decision unit in real time to ensure that no fault is missed.

[0061] The redundant backup unit is used to deploy backup equipment for each module. Its deployment strategy needs to adapt to the multi-form characteristics of bridge rotation to avoid poor adaptability caused by "general backup": In horizontal rotation scenarios, the focus is on deploying "traction sensor backup units" (identical to the main sensor model, installed symmetrically at the traction cylinder position) and "railway dispatch backup links" (a 4G private network independent of the main communication link); in vertical rotation scenarios, the focus is on deploying "cable hydraulic pump backup units" (matching the main pump's flow and pressure parameters, connected in parallel to the hydraulic pipeline) and "tower tilt sensor backup units" (installed on the other side of the main tower to avoid blind spots caused by unilateral obstruction); in combined rotation scenarios, both types of backup equipment need to be deployed simultaneously, and an additional "primary and backup data synchronization server" needs to be added to ensure real-time consistency of primary and backup node data (synchronization frequency 1Hz). The synchronization formula is as follows: ( This is the backup node data at time t. The master node data at time t, For communication status coefficients, This indicates a communication interruption, and the backup data from the previous moment will be used. This indicates that communication is normal and the data is updated to the master node. Through this deployment and synchronization mechanism, the preparation time for switching of backup devices is controlled within 0.5 seconds, laying the foundation for a 1-second rapid switchover.

[0062] The priority decision-making unit is used to determine the priority of fault handling based on the detection results of the fault detection unit. Its decision-making logic does not rely on fixed rules (such as "always prioritize handling traction faults"), but is based on a fuzzy comprehensive evaluation method of "rotation progress, parameter deviation, and risk level" to ensure that the priority matches the real-time working conditions: First, the three major evaluation indicators and their weights are determined—rotation completion degree (weight 0.3, such as horizontal rotation completion degree). At the same time, the weight of progress indicators is implicitly increased to avoid rework due to failures near the end of the process, and the deviation rate of core parameters (weight 0.4, i.e.) The higher the value, the higher the priority; warning levels (weight 0.3, with level 3 warning having the highest weight and level 1 warning having the lowest); then, each indicator is fuzzyened (e.g., rotation completion rate). Corresponding to "high" membership degree. (For "medium" and <60% corresponds to "low"), construct a fuzzy evaluation matrix. ( (where i is the membership degree of the i-th index to the j-th type of fault); finally, through the weight vector... With matrix Product calculation priority ,Pick The fault type with the highest value is prioritized for handling, such as horizontal transition completion. At that time, the traction sensor malfunctioned. , the buckle cable pump fault , the traction backup sensor is switched preferentially.

[0063] The priority decision unit also includes a coupling risk calculation unit; the coupling risk calculation unit is used to quantify the joint risk when multiple faults occur, and is designed to solve the problem that "single fault priority cannot cope with multiple fault superposition" - for example, when "traction sensor fault + buckle cable pump fault" occurs, the priority of both is determined to be 0.7 individually, but after superposition, the synergistic risk of "rotation angle out of control + arch rib instability" may occur, and the processing priority needs to be raised. This unit quantifies the coupling risk through a "joint probability + risk superposition" model, and the formula is , where is the probability of two faults occurring simultaneously (based on historical construction data statistics, such as the probability of traction and buckle cable faults occurring simultaneously ), is the single fault risk value (converted from the value output by the priority decision unit, ), is the coupling amplification coefficient (reflecting the non-linear risk growth of fault superposition); the calculated coupling risk value needs to be compared with a preset threshold (such as 0.8), and when , "emergency synergistic fault tolerance" is triggered - that is, the backup devices of both types of faults are started simultaneously, and a "pause rotation" instruction is sent to the construction control terminal to avoid risk expansion. In addition, the results of the coupling risk calculation unit are also synchronized to the whole cycle evolution module, which is used to optimize the value (for example, if the actual risk after the occurrence of a certain type of multiple faults is lower than the calculated value, then the value is lowered), forming an iterative optimization closed loop for risk quantification.

[0064] The whole cycle evolution module is the core of the bridge rotation intelligent monitoring system to achieve "long-term high-precision adaptation", and its signal connection with the digital twin mapping module and the multi-field coupling calculation module builds a "data acquisition - parameter optimization - effect feedback" two-way closed loop: on the one hand, the module obtains the entity state mapping data during the construction period (such as the displacement response of the multi-degree-of-freedom model, the verification results of extreme scenario samples) and the structural aging data during the operation period (such as the change of friction coefficient caused by ball joint wear, the attenuation of elastic modulus caused by arch rib concrete carbonization) from the digital twin mapping module, and obtains the whole cycle core parameter calculation data (such as the ball joint stress fitting results under different working conditions, the dynamic constraint boundary adjustment record) from the multi-field coupling calculation module; on the other hand, the module feeds back the optimized parameters (such as the dynamic coefficients of the stiffness matrix of the digital twin, the weights of the RBF neural network of the multi-field coupling) to the two modules in real time, ensuring that the system can adapt to the changes in the structure state during the bridge construction period, operation period and even the whole life cycle, and avoiding the precision decay caused by the traditional "fixed parameters after calibration during the construction period".

[0065] The full-cycle evolution module includes a data fusion unit and a parameter optimization unit, and the data fusion unit is signal connected with the parameter optimization unit. The link design of "fusion first and then optimization" aims to solve the "heterogeneous contradiction" of construction period and operation period data. The construction period data has the characteristics of high frequency (1 Hz), real-time and short-term (such as 10 hours of continuous stress monitoring in the process of rotation), and the operation period data has the characteristics of low frequency (once a month), long-term and trend (such as the spherical hinge settlement data in 5 years). Direct splicing will lead to unbalanced data weight (such as high-frequency construction data covering low-frequency operation trend). Therefore, when the data fusion unit is used to fuse the construction stage data and operation stage data of the bridge rotation, "data standardization processing" and "credibility weighted fusion" need to be carried out first. Data standardization maps different dimensional data to the [0, 1] interval through min-max normalization, and the formula is is the original data, is the minimum value and the maximum value of the data respectively), eliminating the dimensional difference of parameters such as stress (MPa), rotation angle and displacement (mm); the credibility weighting is based on the timeliness and accuracy of the data, and the initial credibility of the construction period new project data , and the initial credibility of the old project data in the operation period , and the credibility is updated in real time through the Bayes formula: , wherein is a certain type of data (construction / operation), is the current measured data, is the prediction probability of the measured value (subject to normal distribution, ), and the final fusion data , ensuring that the fusion result not only retains the real-time accuracy of the construction period, but also includes the long-term trend of the operation period.

[0066] The parameter optimization unit is used to optimize the parameters of the multi-field coupling calculation module and the digital twin mapping module based on the fusion data output by the data fusion unit. The optimization logic needs to be designed respectively according to the core parameter characteristics of the two types of modules: for the multi-field coupling calculation module, the core optimization objects are the weights and the hidden layer centers of the RBF neural network, and the optimization goal is to minimize the error between the "measured core parameters" and the "model calculated core parameters" in the fusion data set, and the loss function is is the number of fusion data samples, is the measured core parameter of the i-th sample, ​​​The gradient descent method is used to update the parameters: , , and ensure that the error of the multi-field coupling calculation after optimization is reduced from the initial to below; for the digital twin mapping module, the core optimization object is the stiffness matrix and the damping matrix of the dynamic model, where the stiffness matrix changes dynamically with the shape of the rotor (such as the contact stiffness of the spherical hinge increasing with the increase of the rotation angle), and the operation and maintenance period structure aging parameters (such as the spherical hinge wear amount ) in the fusion data are introduced during optimization to establish , where is the initial stiffness matrix, and is the wear influence coefficient), the least squares method is used to fit the deviation between the displacement response in the fusion data and the model calculation response , and the coefficients of and are adjusted to ensure that the mapping deviation of the digital twin is always controlled within .

[0067] In addition, the parameter optimization unit also needs to consider the "reusability of cross-shape data", for example, when migrating the wind load coefficient optimization results of the horizontal rotation construction period to the vertical rotation scene, physical prior constraints need to be embedded to avoid parameter out-of-bounds - based on the common law that "the wind load stress of horizontal rotation and vertical rotation in bridge mechanics follows Bernoulli's equation", the relationship between the vertical rotation wind load coefficient and the horizontal rotation wind load coefficient is set as , and a regularization term is added to the optimization loss function to ensure that the parameters migrated across shapes comply with the laws of mechanics; after optimization, the module will automatically record the precision changes before and after parameter adjustment (such as the multi-field coupling calculation error from to ), and feed this record back to the data fusion unit as "effect verification data" for the next round of credibility weighting, for example, if a certain type of operation and maintenance data significantly improves the accuracy after optimization, its credibility will be increased from 0.6 to 0.7, forming a "fusion-optimization-verification-re-fusion" full-cycle iterative closed loop, so that the monitoring accuracy of the system continues to be optimized as the bridge service time increases, rather than gradually decreasing.

[0068] ​It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations, but those skilled in the art should know that the present application is not limited by the order of the described actions, because according to the embodiments of the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily required by the present application.

[0069] The multi-morphology recognition module can effectively weaken the interference of environmental noise on feature recognition, avoid morphological misjudgment caused by single feature similarity, and simultaneously adapt to the real-time demand of the rotating body construction, by means of the technical means of "attention augmented cell (CBAM) focusing on morphological specific features (horizontal rotating hinge rotation angle, vertical rotating arch rib inclination angle) + MobileNet-V3 network realizing lightweight feature extraction + morphological matching double criteria (feature similarity calculation + physical law verification)", thereby deducing the effect of "improving the recognition accuracy of multi-morphology (horizontal rotation, vertical rotation, horizontal and vertical combined rotation), and quickly responding to the morphological switching of the rotating body to provide a reliable morphological basis for subsequent modules".

[0070] The multi-field coupling calculation module can break through the limitation of traditional linear models that cannot adapt to complex coupling relationships, avoid "excessive conservatism" or "insufficient constraints" of fixed constraint boundaries under dynamic working conditions, and solve the "black box problem" of nonlinear models, by means of the technical means of "RBF neural network fitting nonlinear correlation between multi-field parameters (temperature, wind load, vibration) and core parameters (hinge stress, main tower eccentricity) + dynamic constraint generation unit linkage real-time multi-field parameter adjustment constraint boundary + physical attribution unit quantifying the contribution of parameters to constraints", thereby deducing the effect of "improving the calculation accuracy of core parameters, accurately matching the constraint boundary with real-time working conditions, and enabling construction personnel to intuitively understand the constraint adjustment logic and enhance their trust in the constraint results".

[0071] The digital twin mapping module can ensure that the digital twin model is consistent with the mechanical behavior of the physical rotating body, exclude invalid extreme samples that violate physical laws, and balance the efficiency and accuracy of sample generation, by means of the technical means of "morphology-adapted multi-degree-of-freedom dynamic model (horizontal rotation configuration rotating / translation degree of freedom, vertical rotation configuration bending / extension degree of freedom) + conditional diffusion model generating extreme scenario samples (embedding morphology-specific physical constraints) + consistency verification unit verifying sample effectiveness based on dynamic equations + step adjustment unit dynamically optimizing step size (large step size for denoising in the early stage, small step size for accuracy preservation in the later stage) according to the generation stage", thereby deducing the effect of "providing reliable data support for extreme scenario prediction, reducing the time-consuming of sample generation, and adapting to the risk prevention and control needs of time-sensitive construction scenarios such as adjacent iron windows".

[0072] The fault-tolerant cooperative module can realize fault-free detection, lay a hardware foundation for rapid fault switching, avoid response disorder caused by single fault priority or multiple fault superposition, and derive the effect of "single fault scenario can quickly start standby equipment to restore monitoring, multiple fault concurrent scenario has no cooperative interruption, system robustness is significantly improved, and structural safety risk caused by fault delay is effectively avoided".

[0073] The full-cycle evolution module can integrate heterogeneous data of the whole life cycle of the bridge through the technical means of "data fusion unit (min-max standardization to eliminate dimension difference + Bayesian credibility weighted to distinguish construction / operation data effectiveness) + parameter optimization unit (gradient descent method to correct multi-field coupling / RBF network parameters + physical prior constraint to ensure the rationality of cross-mode parameter migration)", avoid long-term precision decay caused by fixed parameters, ensure that cross-mode data reuse conforms to the laws of engineering mechanics, and derive the effect of "system monitoring precision continuously optimizes with the progress of bridge construction and operation period, rather than gradually decays, and can adapt to the dynamic monitoring needs of long-term service of the bridge".

[0074] In summary, the five modules of multi-mode recognition, multi-field coupling calculation, digital twin mapping, fault-tolerant cooperation, and full-cycle evolution achieve the overall technical effect of "intelligent monitoring closed loop of bridge rotation in full mode, full working condition, and full life cycle, effectively solving the core problems of traditional monitoring systems such as poor multi-mode adaptation, low calculation accuracy, insufficient extreme prediction, weak fault tolerance, and no long-term evolution ability, and ensuring the safety of rotation construction and the improvement of intelligent level" through layer-by-layer cooperation of technical means.

[0075] Those skilled in the art should understand that the disclosure range involved in the present application is not limited to the technical solutions formed by the specific combination of the above technical features, and also covers other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the aforementioned disclosed concept. For example, the above features can be replaced with similar functional technical features disclosed in the present application (but not limited to) to form technical solutions.

Claims

1. A bridge rotation intelligent monitoring system, characterized in that, It includes a multi-morphological recognition module, a multi-field coupling calculation module, a digital twin mapping module, and an early warning module, which are sequentially connected by signals. The multi-morphology recognition module is used to identify the shape of the bridge rotation, including horizontal rotation, vertical rotation, and a combination of horizontal and vertical rotation. The digital twin mapping module is used to map the physical state of the bridge during rotation; The multi-field coupling calculation module is used to calculate the core parameters of bridge rotation based on the entity state data and multi-field parameters output by the digital twin mapping module. The multi-field parameters include temperature, wind load, vibration and stress, and the core parameters include ball joint stress, main tower eccentricity and arch rib cable force. The early warning module is used to issue early warning signals based on the core parameters output by the multi-field coupling calculation module.

2. The intelligent monitoring system for bridge rotation according to claim 1, characterized in that, The multimorphic recognition module includes a feature extraction unit and a morphology matching unit, and the feature extraction unit is signal-connected to the morphology matching unit. The feature extraction unit is used to process the sensor data of the bridge rotation, which includes horizontal rotation angle data, arch rib inclination angle data and bridge type parameter data. The morphology matching unit is used to match the corresponding rotational morphology physical model based on the feature data output by the feature extraction unit.

3. The intelligent monitoring system for bridge rotation according to claim 2, characterized in that, The feature extraction unit also includes an attention enhancement unit; The attention enhancement unit is used to assign preset weights to the shape-specific features in the feature data. The shape-specific features include the ball joint angle feature corresponding to horizontal rotation and the arch rib tilt angle feature corresponding to vertical rotation.

4. The intelligent monitoring system for bridge rotation according to claim 1, characterized in that, The multi-field coupling calculation module includes a nonlinear modeling unit and a dynamic constraint generation unit, and the nonlinear modeling unit is signal-connected to the dynamic constraint generation unit. The nonlinear modeling unit is used to fit the correlation between the multi-field parameters and the core parameters; The dynamic constraint generation unit is used to adjust the constraint boundaries of the core parameters based on real-time multi-field parameters.

5. The intelligent monitoring system for bridge rotation according to claim 4, characterized in that, The dynamic constraint generation unit also includes a physical attribution unit; The physical attribution unit is used to quantify the contribution of each of the multi-field parameters to the constraint boundary.

6. The intelligent monitoring system for bridge rotation according to claim 5, characterized in that, The digital twin mapping module includes an extreme scenario sample generation unit and a consistency verification unit, wherein the extreme scenario sample generation unit is signal-connected to the consistency verification unit; The extreme scenario sample generation unit is used to generate extreme scenario samples of bridge rotation based on the constraint boundaries output by the dynamic constraint generation unit. The consistency verification unit is used to verify the consistency between the extreme scenario sample and the entity state data.

7. The intelligent monitoring system for bridge rotation according to claim 6, characterized in that, The extreme scenario sample generation unit includes a morphology-specific branch unit and a step size adjustment unit, wherein the morphology-specific branch unit is signal-connected to the step size adjustment unit. The morphology-specific branch unit is used to generate corresponding extreme scenario sample branches for different rotational morphologies; The step size adjustment unit is used to adjust the step size of sample generation according to the generation stage of extreme scenario samples.

8. The intelligent monitoring system for bridge rotation according to claim 1, characterized in that, It also includes a fault-tolerant coordination module, which is signal-connected to the multi-field coupling calculation module and the early warning module, respectively. The fault-tolerant coordination module includes a fault detection unit, a redundancy backup unit, and a priority decision unit, which are sequentially connected by signals. The fault detection unit is used to monitor the operating status of each module; The redundant backup unit is used to deploy backup equipment for each module; The priority decision unit is used to determine the priority of fault handling based on the detection results of the fault detection unit.

9. The intelligent monitoring system for bridge rotation according to claim 8, characterized in that, The priority decision-making unit also includes a coupled risk calculation unit; The coupled risk calculation unit is used to quantify the joint risk when multiple faults occur concurrently.

10. The intelligent monitoring system for bridge rotation according to claim 1, characterized in that, It also includes a full-cycle evolution module, which is signal-connected to the digital twin mapping module and the multi-field coupling calculation module, respectively; The full-cycle evolution module includes a data fusion unit and a parameter optimization unit, and the data fusion unit and the parameter optimization unit are signal-connected. The data fusion unit is used to fuse construction phase data and operation and maintenance phase data of the bridge rotation. The parameter optimization unit is used to optimize the parameters of the multi-field coupling calculation module and the digital twin mapping module based on the fused data output by the data fusion unit.

Images