Stability monitoring method and system for bridge swivel construction

Abstract

The invention provides a stability monitoring method and system for bridge swivel construction, belongs to the field of bridge construction monitoring, and is used for solving the problems of long adaptation period, poor real-time working condition response and inaccurate risk early warning of a single bridge swivel project monitoring model in related technologies. The scheme comprises four core support systems of data acquisition, local calculation, cloud collaboration and communication transmission, a global collaborative knowledge model is constructed through multi-project parameter weighted aggregation, model parameters are optimized through a self-organizing evolutionary optimization algorithm, and dynamic weight and Bayesian parameter collaborative calculation is performed, so that efficient processing of monitoring data and dynamic adaptation of the model are realized. The model adaptation period can be shortened, the monitoring real-time performance and the risk pre-judgment precision are improved, and the bridge swivel construction stability is guaranteed.

Description

Technical Field

[0001] This application relates to the field of bridge construction monitoring, and more particularly to a stability monitoring method and system for bridge rotation construction. Background Technology

[0002] Bridge rotation construction is widely used in complex scenarios such as crossing railways and rivers due to its strong spanning capacity and minimal disruption to existing traffic. The stability of the construction process directly determines the safety of the bridge structure, thus requiring real-time monitoring of the rotation status through professional monitoring technology. Currently, bridge rotation stability monitoring mostly relies on local data modeling for single projects. The monitoring model needs to be retrained from the initial stage, resulting in a long adaptation period and difficulty in quickly meeting the monitoring needs of the basic working conditions in the early stages of construction.

[0003] In existing technologies, the construction of monitoring models often ignores the value of reusing knowledge across projects. Monitoring data from each project are stored in isolation, making it impossible to improve model accuracy through experience from multiple projects. At the same time, monitoring data often focuses on explicit parameters of rotating structures (such as rotation angle and stress), without fully integrating implicit data such as equipment health and precursors to environmental changes. This results in insufficient adaptability of the model to complex working conditions, which can easily lead to delayed risk warnings or misjudgments.

[0004] In addition, the existing monitoring system has poor coordination between algorithms and hardware environment. Local computing devices have difficulty executing model optimization algorithms efficiently. Data interaction between the cloud and local is easily affected by signals from the construction scene, resulting in delays in monitoring data processing and untimely updates of global knowledge. It cannot meet the high requirements of real-time and accuracy for bridge rotation construction. There is an urgent need for a stability monitoring solution that can integrate cross-project knowledge, adapt to complex working conditions, and has efficient hardware coordination. Summary of the Invention

[0005] This application provides a stability monitoring method and system for bridge rotation construction. It can solve the problems of slow adaptation of single-project monitoring models and poor response to complex working conditions by cross-project knowledge aggregation and local algorithm optimization, thereby improving the real-time performance and accuracy of stability monitoring during bridge rotation construction.

[0006] Firstly, this application provides a stability monitoring method for bridge rotation construction. It includes the following steps: acquiring initial monitoring data for the target bridge rotation construction project, including explicit monitoring data reflecting the posture and stress of the rotating structure and implicit correlation data affecting monitoring accuracy; fine-tuning the initial monitoring data locally based on a global collaborative knowledge model formed by weighted aggregation of multi-project monitoring model parameters to obtain an initial monitoring model adapted to the basic working conditions of the target project, wherein the aggregation uses the proportion of multi-project data and the representativeness of the working conditions to dynamically allocate weight values; continuously acquiring real-time monitoring data during the target bridge rotation construction process, and dynamically optimizing the initial monitoring model through a self-organizing evolutionary optimization algorithm, wherein the self-organizing evolutionary optimization algorithm includes new working condition identification and correction of dynamic weight parameters and Bayesian early warning parameters to obtain a dynamic monitoring model adapted to the real-time working conditions; and outputting the stability monitoring results of the target bridge rotation construction based on the collaborative calculation of the dynamic weight parameters and Bayesian early warning parameters.

[0007] By adopting the above technical solution, a global collaborative knowledge model is constructed by aggregating knowledge from multiple projects. This eliminates the need to train the monitoring model from scratch for the target project, significantly shortening the model adaptation cycle. At the same time, the aggregation weights are dynamically allocated based on the amount of data and the representativeness of the working conditions, ensuring that the global model can reflect the experience of multiple projects and adapt to the basic characteristics of the target project, laying the foundation for improving the accuracy of subsequent monitoring.

[0008] Furthermore, the algorithm for constructing the global collaborative knowledge model includes: collecting historical monitoring parameters from multiple bridge rotation projects and unifying the parameter dimensions through feature normalization; calculating the working condition matching coefficients of each project using the analytic hierarchy process and generating comprehensive weights based on the proportion of data volume; and fusing parameters from multiple projects based on a weighted average algorithm to form a dynamically invoked global knowledge model.

[0009] By adopting the above technical solution, the historical monitoring parameters are first normalized to eliminate the interference of differences in the dimensions of parameters from different projects on model aggregation; then, the working condition matching coefficient is accurately calculated using the analytic hierarchy process, and a comprehensive weight is generated by combining the data volume ratio to ensure the rationality of the fusion of parameters from multiple projects, so that the global knowledge model has the value of cross-project reuse.

[0010] Furthermore, explicit monitoring data includes rotation angle data, stress data, wind speed data, and vibration frequency data of the rotating structure; implicit correlation data includes health data of monitoring equipment and precursor data of environmental changes; the synchronous fusion processing of the two types of data is achieved through a multi-source data spatiotemporal alignment algorithm.

[0011] By adopting the above technical solutions, explicit structural parameters are combined with implicit influencing factors to comprehensively cover the key monitoring dimensions of the stability of the rotation construction; at the same time, the spatiotemporal alignment algorithm is used to achieve synchronous data fusion, avoiding monitoring errors caused by data time differences and spatial deviations, and improving the integrity and accuracy of the model input data.

[0012] Furthermore, the local fine-tuning algorithm for the initial monitoring model includes: calculating the characteristic difference between the target project's basic working conditions and the global typical working conditions; adaptively adjusting the initial values ​​of the global model's parameters based on the difference; and correcting the error of the adjusted model using the least squares method to keep the model's output error within a preset range.

[0013] By adopting the above technical solution, parameters are adaptively adjusted to address the differences between the target project and the overall operating conditions. Then, the error is corrected using the least squares method, ensuring that the initial monitoring model can accurately adapt to the basic characteristics of the target project, avoiding monitoring deviations caused by differences in operating conditions, and improving the monitoring accuracy in the initial stage of the model.

[0014] Furthermore, the self-organizing evolutionary optimization algorithm includes: constructing a feature matrix containing real-time monitoring data and implicit correlation data; using a density clustering algorithm to identify new operating conditions not included in the basic operating conditions; triggering a parameter optimization process for the new operating conditions, and iteratively updating the dynamic weight parameters and Bayesian early warning parameters through a particle swarm optimization algorithm.

[0015] By adopting the above technical solutions, new working conditions that occur during the construction of the target project can be automatically identified, and parameter optimization can be triggered without manual intervention. Combined with the iterative advantages of the particle swarm optimization algorithm, weights and early warning parameters can be updated quickly, enabling the model to continuously adapt to dynamically changing construction conditions and improve the monitoring and response capabilities under complex working conditions.

[0016] Furthermore, the dynamic weight parameter optimization algorithm includes: establishing a multi-dimensional evaluation function with the goals of corner control accuracy and risk warning timeliness; using reinforcement learning algorithm to dynamically adjust the weight ratio of each monitoring parameter; and setting operational robustness constraints to make the weight adjustment results adapt to the dynamic characteristics of the rotation construction.

[0017] By adopting the above technical solution, taking the core needs of the rotation construction as the evaluation target, and using reinforcement learning to achieve dynamic adjustment of weights, we can ensure that high-impact monitoring parameters receive more attention. At the same time, we add operational robustness constraints so that the weight adjustment can adapt to the dynamic changes in the construction process and avoid monitoring instability caused by weight fluctuations.

[0018] Furthermore, the Bayesian early warning parameter correction algorithm includes: introducing environmental abrupt change precursor data as new nodes into the Bayesian network; updating the conditional probabilities between nodes based on the maximum likelihood estimation method; and implementing probabilistic reasoning under complex working conditions through the Markov chain Monte Carlo algorithm to improve the accuracy of risk prediction.

[0019] By adopting the above technical solutions, environmental aberration precursor data are integrated into Bayesian networks to capture risk factors in advance; combined with maximum likelihood estimation and Markov chain Monte Carlo algorithm, the accuracy of probabilistic reasoning under complex working conditions is improved, effectively reducing the probability of delayed or misjudged risk warnings and ensuring construction safety.

[0020] Furthermore, the collaborative computing algorithm includes: constructing a priority ranking rule for monitoring data based on dynamic weight parameters; dynamically allocating computing resources according to priority; inputting high-priority data into the corrected Bayesian model for risk calculation; and integrating multi-dimensional calculation results through a weighted fusion algorithm to generate a comprehensive stability assessment value representing the stability monitoring results.

[0021] By adopting the above technical solutions, computing resources are allocated according to weight and priority to ensure that high-impact monitoring data is processed first and improve the real-time performance of risk calculation. Furthermore, by weighted fusion and integration of multi-dimensional results, the one-sidedness of the assessment caused by a single data dimension is avoided, and the stability assessment value more comprehensively reflects the rotation construction status.

[0022] Furthermore, the global knowledge model update algorithm includes: receiving the optimized model parameters of the target project and marking the parameters to generate timestamps; calculating the timeliness coefficient of the parameters based on the time decay function; and using an incremental learning algorithm to integrate the weighted new parameters into the global model to achieve incremental evolution of the model.

[0023] By adopting the above technical solutions, a time decay function is introduced to distinguish the timeliness of parameters, avoiding outdated knowledge from interfering with global model updates. At the same time, incremental learning enables model evolution without the need to retrain the global model, significantly reducing update costs and allowing the global knowledge model to continuously absorb optimization experience from various projects, thereby enhancing its cross-project reusability.

[0024] Secondly, this application provides a stability monitoring system for bridge rotation construction. It includes: a data acquisition layer comprising a monitoring network of various sensors for collecting various monitoring data during the rotation construction; an edge computing layer including local servers deployed at the construction site for executing the methods described in any one of the first aspects above, enabling real-time processing of monitoring data and model optimization; a cloud collaboration layer including a remote server cluster for storing a global collaborative knowledge model and providing cross-project data collaboration services; and a communication transmission layer employing multi-mode communication technology to achieve data interaction between layers, ensuring data transmission requirements during algorithm execution.

[0025] By adopting the above technical solutions, a four-layer collaborative hardware architecture of "acquisition-computation-storage-transmission" is constructed. The data acquisition layer ensures comprehensive coverage of monitoring data, the edge computing layer enables efficient execution of local algorithms, the cloud collaboration layer supports cross-project knowledge sharing, and the communication transmission layer ensures stable data interaction. The four-layer collaboration ensures that the monitoring system can efficiently execute algorithm logic and meet the real-time monitoring needs of rotating construction.

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

[0027] 1. It provides a monitoring solution that integrates cross-project knowledge and local optimization, significantly shortening the model adaptation cycle and improving monitoring accuracy;

[0028] 2. Enhance the model's adaptability to complex working conditions through multi-source data fusion and self-identification of new working conditions;

[0029] 3. Construct a four-layer hardware architecture to ensure efficient algorithm execution and stable data interaction, supporting real-time monitoring.

[0030] It should be understood that the description in the Summary Section is not intended to limit the key or essential features of the embodiments of this application, nor is it intended to restrict the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0031] The above and other features, advantages, and aspects of the embodiments of this application will become more apparent from the accompanying drawings and the following detailed description. In the drawings, the same or similar reference numerals denote the same or similar elements, wherein:

[0032] Figure 1 A schematic diagram of a stability monitoring system for bridge rotation construction is shown in an embodiment of this application.

[0033] Figure 2 A flowchart of a stability monitoring method for bridge rotation construction is shown in an embodiment of this application. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0035] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0036] This application provides a stability monitoring system and method for bridge rotation construction. It improves monitoring accuracy and reliability through multi-dimensional data coupling analysis, dynamic data preprocessing and clear hardware function division, and effectively solves the problems of one-sided monitoring, poor data processing adaptability and confusing hardware functions in the existing technology.

[0037] In a first aspect, embodiments of this application disclose a stability monitoring system for bridge rotation construction.

[0038] Figure 1 A schematic diagram of a stability monitoring system for bridge rotation construction is shown in an embodiment of this application.

[0039] Reference Figure 1 The system includes a macro-level hardware environment that supports the execution of the "stability monitoring method for bridge rotation construction". The hardware environment is divided into four core support systems: data acquisition, local computing, cloud collaboration, and communication transmission. Each system is connected through standardized interfaces to form a complete hardware link of "data acquisition - real-time processing - knowledge sharing - stable transmission", ensuring that the algorithm steps such as global knowledge aggregation, local model optimization, and collaborative computing in the monitoring method are implemented efficiently.

[0040] The data acquisition system, serving as the source support for monitoring data, consists of various specialized sensors and supporting acquisition modules. Among them, the rotating structure parameter monitoring sensors include a high-precision GNSS receiver, MEMS stress sensors, ultrasonic anemometers, and vibration sensors, used to collect rotation angle data, structural stress data, construction environment wind speed data, and vibration frequency data, respectively. The implicit correlation data acquisition module includes an equipment health monitoring unit and an environmental precursor monitoring unit. The equipment health monitoring unit calculates equipment health by comparing sensor measurements with true values, while the environmental precursor monitoring unit collects data on environmental abrupt changes using a wind speed pulsation monitor and a track vibration spectrum analyzer. All sensors and acquisition units are connected to the data acquisition gateway via an industrial bus, where the gateway performs data filtering, format conversion, and preliminary aggregation, providing standardized data input for subsequent algorithms.

[0041] The local computing system, serving as the core support for the local execution of the algorithm, is deployed near the bridge rotation construction site and mainly consists of edge computing servers and supporting storage units. The edge computing servers are equipped with high-performance processors and large amounts of memory to run local model fine-tuning algorithms, self-organizing evolutionary optimization algorithms, and collaborative computing algorithms in the monitoring methods. Specifically, this includes fine-tuning initial monitoring data based on a global collaborative knowledge model, identifying new working conditions and optimizing dynamic weights and Bayesian early warning parameters, adjusting sampling frequency based on weight parameters, and calculating risk probabilities. The supporting storage units use high-speed solid-state drives to temporarily store locally collected monitoring data, optimized model parameters, and intermediate algorithm results, ensuring that data read / write speeds match the real-time processing requirements of the algorithm and avoiding data storage delays that could affect monitoring response efficiency.

[0042] The cloud-based collaborative system, supporting cross-project knowledge sharing, consists of a remotely deployed federated server cluster and a global knowledge storage array. The federated server cluster comprises multiple high-performance rack-mount servers used to execute the global collaborative knowledge model construction and update algorithms in the monitoring methodology. Specifically, this includes receiving monitoring model parameters uploaded by each project, calculating weights based on project data volume and operational representativeness, aggregating parameters, and updating the global model in conjunction with knowledge timeliness decay factors. The global knowledge storage array employs an all-flash storage architecture to persistently store historical versions of the global collaborative knowledge model, parameter data uploaded by each project, and operational feature libraries. It supports data snapshots and rapid backtracking, ensuring secure storage and efficient retrieval of cross-project knowledge.

[0043] The communication transmission system, supporting data interaction between various hardware systems, employs multi-mode communication technology to construct redundant transmission links. Specifically, the data acquisition system within the construction site is connected to the local computing system via industrial Ethernet to achieve low-latency transmission of monitoring data. The local computing system is connected to the cloud-based collaborative system via 4G / 5G industrial routers or dedicated fiber optic lines, while a LoRa gateway is deployed as a backup communication link. When the primary link is interrupted due to signal interference in the construction environment, it automatically switches to the backup link, ensuring stable transmission of locally optimized model parameters to the cloud and the global collaborative knowledge model from the cloud to the local system, preventing algorithm execution interruptions due to communication disruptions.

[0044] The four hardware support systems mentioned above work together: the data acquisition system provides comprehensive monitoring data input; the local computing system enables real-time processing of algorithms and model optimization; the cloud collaboration system supports cross-project knowledge sharing and global model updates; and the communication transmission system ensures stable data interaction at each stage. Together, they constitute the macroscopic equipment environment required for the execution of the "stability monitoring method for bridge rotation construction," ensuring that the monitoring method can be accurately and efficiently applied to the stability monitoring scenario of bridge rotation construction.

[0045] Secondly, this application discloses a stability monitoring method for bridge rotation construction.

[0046] Figure 2 A flowchart of a stability monitoring method for bridge rotation construction is shown in an embodiment of this application.

[0047] Reference Figure 2 The method specifically includes the following steps:

[0048] S1: Obtain initial monitoring data for the target bridge rotation construction project. The initial monitoring data includes explicit monitoring data reflecting the posture and stress of the rotating structure, as well as implicit correlation data that affect the monitoring accuracy.

[0049] In this step, the explicit monitoring data includes the rotation angle data, stress data, wind speed data, and vibration frequency data of the rotating structure; the implicit correlation data includes the health data of the monitoring equipment and the precursor data of environmental changes; and the synchronous fusion processing of the two types of data is achieved through a multi-source data spatiotemporal alignment algorithm.

[0050] The rotation angle data of the rotating structure was collected using a high-precision GNSS receiver, and the rotation angle data was defined based on the initial attitude of the rotation construction. The angle (in degrees) between the actual posture and the reference posture of the rotating structure must meet the following requirements. , The maximum permissible rotation angle deviation preset for the project (determined based on bridge structural design parameters, such as for continuous beam rotation projects, it is usually taken as...) Stress data is acquired through a MEMS stress sensor, and the stress data is defined. The actual stress values ​​(unit: MPa) at key sections of the rotating structure (such as the mid-span of the main beam and the base of the tower column) should be recorded simultaneously with the structural design stress limits at the sensor installation location during data acquisition. This ensures that it can be subsequently determined whether the stress exceeds the safe range. Wind speed data is collected using an ultrasonic anemometer, and the wind speed data is defined. Real-time instantaneous wind speed in the construction environment (unit: (and record the time series of wind speed) (t is the data acquisition timestamp, in seconds), used for subsequent analysis of wind speed change trends. Vibration frequency data is acquired using a piezoelectric vibration sensor, and the vibration frequency data is defined. The natural vibration frequency (unit: Hz) of the rotating structure is used. The frequency collected must be compatible with the rotation speed during construction, typically taken as 5-10 times the rotation speed. For example, if the rotation speed is... At that time, the vibration frequency sampling frequency was set to 5Hz.

[0051] The health status data of monitoring equipment in implicitly correlated data is quantified through the equipment health status calculation formula, and the equipment health status is defined. The range of values ​​for an indicator characterizing the measurement accuracy of monitoring equipment. The closer the value is to 1, the better the device's health status. The calculation formula is:

[0052] ;

[0053] In the formula, To monitor the actual measured values ​​of the equipment (such as the rotation angle measured by the GNSS receiver, the stress measured by the MEMS sensor). The true value is the result of equipment calibration (obtained through periodic calibration using standard calibration equipment, such as annual baseline calibration of GNSS receivers to obtain the true value of rotation angles; and static loading calibration of MEMS stress sensors to obtain the true value of stress). When If the equipment is found to be in an abnormal health condition, it needs to be replaced or recalibrated to avoid affecting the accuracy of monitoring data due to equipment errors.

[0054] Precursor data for environmental abrupt changes include wind speed pulsation frequency variation data. and track vibration anomaly spectrum data Among them, wind speed fluctuation frequency variation data Defined as the difference (in Hz) between the current wind speed fluctuation frequency and the historical average fluctuation frequency, its calculation formula is:

[0055] ;

[0056] In the formula, The wind speed fluctuation frequency at the current moment (from the wind speed time series collected by the anemometer). (obtained by performing a Fourier transform) The number of historical data samples (usually samples collected within the last hour, e.g., when the collection frequency is 1Hz). ), For the first The frequency of wind speed fluctuations in a historical sample. ( The preset wind speed fluctuation frequency variation limit is determined based on meteorological data statistics of the project location. For example, in coastal areas, it is usually taken as... When this occurs, it is determined that there are precursors to sudden wind speed changes. Track vibration anomaly spectrum data. Data collected using a track vibration spectrum analyzer is defined as vibration frequencies exceeding the normal range (e.g., the normal vibration frequency range of a track during bridge rotation construction is...). The proportion of spectral energy when When the energy proportion of the abnormal frequency band exceeds 30%, it is determined that there are precursors to abnormal track vibration.

[0057] S2: The initial monitoring data is locally fine-tuned based on the global collaborative knowledge model formed by the weighted aggregation of parameters of the multi-project monitoring model to obtain an initial monitoring model that adapts to the basic working conditions of the target project. The aggregation adopts the dynamic allocation of weight values ​​based on the proportion of multi-project data and the representativeness of working conditions.

[0058] In this step, the algorithm for constructing the global collaborative knowledge model includes: collecting historical monitoring parameters from multiple bridge rotation projects and unifying parameter dimensions through feature normalization; calculating the working condition matching coefficients of each project using the analytic hierarchy process (AHP) and generating comprehensive weights based on the data volume ratio; fusing parameters from multiple projects using a weighted average algorithm to form a dynamically callable global knowledge model; and updating the global knowledge model by: receiving the optimized model parameters from the target project and marking the parameters to generate timestamps; calculating the timeliness coefficients of the parameters based on the time decay function; and using an incremental learning algorithm to integrate the weighted new parameters into the global model to achieve incremental evolution of the model. Simultaneously, the local fine-tuning algorithm for the initial monitoring model includes: calculating the feature difference between the basic working conditions of the target project and the typical working conditions globally; adaptively adjusting the initial parameter values ​​of the global model based on the difference; and correcting the error of the adjusted model using the least squares method to keep the model output error within a preset range.

[0059] Multi-source data spatiotemporal alignment algorithms are used to address the asynchrony between explicit monitoring data and implicit correlation data in the temporal and spatial dimensions, ensuring that the fused data has a consistent spatiotemporal reference. Time alignment employs a timestamp-based interpolation method, using the timestamps of a high-precision GNSS receiver as a reference (its time synchronization accuracy can reach the millisecond level) to calibrate the time of data collected by other sensors: Let the timestamp sequence of the GNSS receiver be... The timestamp sequence of a certain sensor is For each If no corresponding Then, the equivalent measurement value of the sensor at that moment is calculated by linear interpolation. The calculation formula is:

[0060] ;

[0061] In the formula, , , The sensors are respectively in , The measured values ​​at any given time. Spatial alignment employs a coordinate transformation-based method, converting all sensor measurement data to a unified construction coordinate system (e.g., a local coordinate system with the center of the rotating ball joint as the origin, the horizontal direction as the X-axis, and the vertical direction as the Y-axis), defining the coordinates of the sensor installation positions as follows. Through coordinate transformation matrix (Based on the sensor installation drawings, including translation and rotation parameters) the raw measurement data of the sensors are converted into unified data in the construction coordinate system to ensure that the monitoring data of sensors at different locations can be directly used for subsequent model calculations.

[0062] Through the above data collection and processing, an initial monitoring data set is finally formed. ,in For data sample numbers, The initial monitoring data sample size (usually data collected in the first 12-24 hours of construction to ensure sufficient sample size to support subsequent model fine-tuning) provides a complete and synchronous input data foundation for the local fine-tuning of the global collaborative knowledge model in subsequent steps.

[0063] S3: During the construction of the target bridge rotation, real-time monitoring data is continuously acquired, and the initial monitoring model is dynamically optimized through a self-organizing evolutionary optimization algorithm. The optimization includes the identification of new working conditions and the correction of dynamic weight parameters and Bayesian early warning parameters. The new working conditions are construction conditions of the target project that have not been included in the basic working conditions, thus obtaining a dynamic monitoring model adapted to the real-time working conditions.

[0064] In this step, the self-organizing evolutionary optimization algorithm includes: constructing a feature matrix containing real-time monitoring data and implicitly related data; using density clustering to identify new working conditions not included in the basic working conditions; triggering a parameter optimization process for new working conditions, and iteratively updating dynamic weight parameters and Bayesian early warning parameters using particle swarm optimization; wherein, the dynamic weight parameter optimization algorithm includes: establishing a multi-dimensional evaluation function with the goals of corner control accuracy and risk early warning timeliness; using reinforcement learning algorithm to dynamically adjust the weight ratio of each monitoring parameter; setting operational robustness constraints to adapt the weight adjustment results to the dynamic characteristics of the rotation construction; the Bayesian early warning parameter correction algorithm includes: introducing environmental mutation precursor data as new nodes into the Bayesian network; updating the conditional probabilities between nodes based on the maximum likelihood estimation method; and realizing probabilistic reasoning under complex working conditions through Markov chain Monte Carlo algorithm to improve the accuracy of risk prediction.

[0065] First, the continuously acquired real-time monitoring data is of the same type as the initial monitoring data in S1, including visible monitoring data (corner). ,stress Wind speed Vibration frequency ) and implicitly related data (device health) Wind speed pulsation frequency variation Track vibration anomaly spectrum ),in To collect timestamps in real time (unit: seconds), the collection frequency is dynamically adjusted according to the dynamic weight parameters in S4, typically [value missing]. When constructing the feature matrix, the real-time data is first standardized to eliminate the influence of differences in the magnitude of different parameters. The standardization formula is as follows: ;

[0066] In the formula, This represents the raw, real-time value of a certain monitoring parameter. , These are the minimum and maximum values ​​of this parameter in the initial monitoring data and historical project data of S1, respectively (e.g., turning angle). of , , Same as S1 definition). The standardized parameter values ​​have a range of values. Based on the standardized parameters, a feature matrix is ​​constructed. ,in The number of samples for real-time data (usually samples collected within the last 5 minutes, such as when the sampling frequency is 2Hz). ), To monitor the dimensions of the parameters (here) (corresponding to 7 types of monitoring parameters), the first in the matrix Line 1 Column elements Indicates the first The first sample Standardized values ​​of various monitoring parameters.

[0067] When using the density clustering algorithm (DBSCAN algorithm) to identify new working conditions, two core parameters are first defined: neighborhood radius. The minimum number of samples in the neighborhood, MinPts. Determined using the "k-distance method", the distance from each sample to its k-th ... The distance to the nearest sample (here) (Referring to historical clustering experience values), a k-distance curve is plotted, and the distance value corresponding to the inflection point of the curve is taken as... MinPts is based on the number of samples. Determined, usually taken (MinPts = 14 here) to ensure the stability of the clustering results. During the clustering process, for the feature matrix... Each sample in Calculate its Number of samples in the neighborhood :like ,but As the core sample; if But falling into the core samples Within the neighborhood, Boundary samples are considered boundary samples; otherwise, they are noise samples. Clusters are formed by connecting the density reachability of core samples, with each cluster corresponding to a specific working condition. The working conditions obtained from these clusters are compared with the basic working conditions adapted to the initial monitoring model (clusters obtained from the initial monitoring data of S1), and the center distance between the two types of working condition clusters is calculated. :

[0068] ;

[0069] In the formula, For the new cluster number The center value of the dimension parameter (i.e., the dimensional parameter of all samples within the cluster) (average of dimensional parameters) Based on the working condition clustering cluster number The center value of the dimension parameter, if ( The preset threshold for operating condition differences is determined based on historical project operating condition difference statistics, and is typically set to... If the cluster corresponds to a new operating condition, the subsequent parameter optimization process will be triggered.

[0070] In the parameter optimization process triggered by new operating conditions, the Particle Swarm Optimization (PSO) algorithm is used to iteratively update the dynamic weight parameters. ( For the first The weights of the monitoring parameters satisfy the following conditions: and ) and Bayesian warning parameters (conditional probability table of Bayesian network). Taking dynamic weight parameter update as an example, the particle dimension of the PSO algorithm is set to Each particle represents a set of candidate solutions with weight parameters. ( (where is the particle index). The particle's fitness function takes the value of the multi-dimensional evaluation function in S3; a higher fitness value indicates better weight parameters. The iterative formula for the PSO algorithm is: ; ;

[0071] In the formula, For the number of iterations, For the first The first particle Dimensional speed, The inertial weight is set at 0.9 initially and decreases linearly to 0.4 during iteration to balance global and local search capabilities. , The learning factor is 2.0 (referencing classic parameters of the PSO algorithm). , for Random numbers within the interval For the first The optimal position of each particle Dimension value (i.e., the dimensional value at which the particle's history fitness is highest) Dimensional weights). The global optimal position of the entire particle swarm. Dimension value (i.e., the dimensional value when the fitness of all particles in history is the highest) (Weights); the second formula normalizes the weights using the softmax function to ensure that it satisfies... The constraints are as follows. The iteration termination condition is set to "the number of iterations reaches a preset maximum value (e.g., 100 times)" or "the global optimal fitness value remains unchanged for 10 consecutive iterations". The output global optimal position is then determined. This refers to the updated dynamic weight parameters.

[0072] Multidimensional evaluation function with dynamic weight parameters optimization Taking into account the accuracy of corner control, the timeliness of risk warning, and operational robustness, the formula is:

[0073] ;

[0074] In the formula, , , Weighting coefficients (satisfying) Set according to project requirements, such as , , ), This refers to the accuracy index for corner control. As an indicator of the timeliness of risk warning, This is used as a robust loss metric. Among them, , The average rotation angle calculated based on the current weight parameters ( , (Sample collection time) The target rotation angle for the rotation construction (determined according to the construction plan). Range of values The closer the value is to 1, the higher the accuracy of the angle control. , The time delay between the generation of a risk signal and the output of an early warning (obtained through simulated risk scenario testing). The preset maximum allowable delay time (e.g., 5 seconds). Range of values The closer the value is to 1, the better the timeliness of the warning. , The standard deviation of the dynamic weight parameters over 10 consecutive iterations. The average value of the weight parameters. The smaller the value, the more stable the weight parameter and the stronger the operational robustness.

[0075] In the Bayesian early warning parameter correction, a Bayesian network structure containing precursory data of environmental mutations is first constructed. Network nodes include "rotation risk status" (target node, with values ​​of "safe," "low risk," "medium risk," and "high risk") and "rotation angle deviation" (parent node, based on...). and The difference in values ​​is used to classify the value levels), and the stress state is defined by the parent node (based on the value level). and The ratio is used to classify the value level), and the "wind speed level" (parent node, based on) (Classification of value levels), "Device Health" (parent node, based on) (Classify value levels), "Abnormal wind speed fluctuations" (add a parent node, based on) Is it excessive? Values ​​"Yes" and "No"), "Track vibration anomaly" (add a parent node, based on...) (Whether the value exceeds 0.3 is marked as "Yes" or "No"). When updating the conditional probabilities between nodes based on the maximum likelihood estimation method, real-time monitoring data under the new operating conditions is used as a sample to statistically analyze the frequency of occurrence of the target node under each combination of parent node values, which is then used as an estimate of the conditional probability. For example, for conditional probability... Its estimated value is "abnormal wind speed fluctuation". Yes and the angle deviation "Risk status" in a large sample The ratio of the number of "high-risk" samples to the total number of samples in that group.

[0076] When implementing probabilistic inference under complex conditions using the Markov Chain Monte Carlo (MCMC) algorithm, Gibbs sampling is employed to generate samples for complex scenarios with multiple coupled parent nodes in the Bayesian network. First, the values ​​of all nodes are initialized. Then, in each iteration, the values ​​of other nodes are fixed, and the value of the current node is sampled and updated based on its conditional probability distribution (obtained by maximum likelihood estimation). The number of iterations is set to 2000. The first 1000 iterations are the burning period (samples are not retained, used to ensure the Markov chain reaches a stationary distribution), and the last 1000 iterations are the sampling period (samples are retained for probability estimation). Finally, the frequency of each value of the target node's "rotation risk state" during the sampling period is statistically analyzed and used as the posterior probability of that state, i.e., the risk probability output after Bayesian warning parameter correction, thus completing the correction of the Bayesian warning parameters.

[0077] After updating the dynamic weight parameters and Bayesian early warning parameters, the updated parameters are substituted into the initial monitoring model to replace the original parameters, resulting in a dynamic monitoring model adapted to the new working conditions. During subsequent construction, if the density clustering algorithm identifies new working conditions again, the above parameter optimization process is repeated to achieve continuous evolution of the dynamic monitoring model and ensure that the model always adapts to the real-time construction conditions of the target project.

[0078] S4: Based on the collaborative calculation of dynamic weight parameters and Bayesian early warning parameters, output the stability monitoring results of the target bridge rotation construction. The collaborative calculation is a linkage process in which dynamic weight parameters guide the priority of monitoring data and Bayesian early warning parameters generate risk judgment results.

[0079] In this step, the collaborative computing algorithm includes: constructing a priority ranking rule for monitoring data based on dynamic weight parameters; dynamically allocating computing resources according to priority; inputting high-priority data into a modified Bayesian model for risk calculation; integrating multi-dimensional calculation results through a weighted fusion algorithm to generate a comprehensive stability assessment value; and the stability monitoring results include real-time dynamic weights, rotation risk probability, early warning level, risk contribution analysis results, and rotation construction control suggestions; the rotation construction control suggestions are generated based on the risk contribution analysis results, including hydraulic traction force adjustment values ​​and cable force optimization amounts; the early warning level is based on the rotation risk probability to classify risk warning levels.

[0080] First, when constructing the priority ranking rule for monitoring data based on dynamic weight parameters, Updated dynamic weight parameters Based on the core basis ( For the first Weights of class monitoring parameters, Corresponding to 7 types of monitoring parameters, meeting the requirements The system compares the weight values ​​of each monitoring parameter with preset priority thresholds to classify priority levels. Priority thresholds are defined. , (Determined based on the statistical impact of monitoring parameters on rotational stability; for example, rotation angle and stress parameters have the greatest impact on stability, and therefore higher thresholds are set): If If the monitoring data corresponding to this parameter is classified as "high priority," it should be collected and calculated first. Classified as "medium priority"; if These are categorized as "low priority". For example, if the corner parameter weight... Wind speed parameter weighting Equipment health parameter weights Therefore, corner data is given high priority, wind speed data medium priority, and equipment health data low priority. Simultaneously, a mapping relationship is established between priority and data collection frequency, with high-priority data collected at a higher frequency. Set to 10Hz, medium priority Set to 5Hz, low priority Set to 1Hz to ensure that the monitoring data of high-weight parameters can more densely reflect changes in the rotation state. The formula is:

[0081] ;

[0082] In the formula, For the first The acquisition frequency of the monitoring parameters is determined according to the above rules. This mapping relationship is implemented through the configuration module of the data acquisition gateway. The gateway adjusts the acquisition frequency of each sensor in real time according to the update of the dynamic weight parameters.

[0083] When dynamically allocating computing resources based on priority, the number of CPU cores on the edge computing server is used as the basis. Memory capacity Allocate resources according to priority proportions based on the total resource volume (e.g., configuring an edge server with 8 CPU cores and 16GB of memory). Define the resource allocation coefficients for high, medium, and low priorities as follows: , , (Based on the computational complexity of priority data, higher priority data requires complex operations such as risk probability calculation, thus requiring more resources.) Therefore, the number of CPU cores allocated to each priority level is... Memory allocation capacity The calculation formula is:

[0084] ;

[0085] In the formula, Representing three priority levels: "high", "mid", and "low". This refers to the resource allocation coefficient corresponding to the priority level. For example, in an 8-core CPU, higher priority allocation... Cores (actually rounded to 5 cores, with core exclusivity achieved through CPU scheduling algorithms), medium priority allocation Cores (rounded to 2 cores), allocated with low priority. Core (rounded to 1 core); high priority allocation within 16GB of memory. 4.8GB is allocated for medium priority tasks and 1.6GB for low priority tasks. Resource allocation is achieved through the operating system resource scheduling module of the edge server to ensure that the calculation process of high priority data is not interfered with by low priority tasks and reduce calculation latency.

[0086] When inputting high-priority data into the modified Bayesian model for risk calculation, the high-priority data must first undergo data preprocessing, including outlier removal and feature extraction. Outlier removal employs... Criteria for real-time data sequences of high-priority parameters ( (For the time of data collection), calculate its mean. with standard deviation If data points satisfy If the value is an outlier, it will be removed. The formula is: ;

[0087] In the formula, This refers to high-priority data after outlier removal. Feature extraction transforms continuous high-priority data into discrete states that can be recognized by a Bayesian model, such as turning corner data. Converted to "small deviation" , "deviation" ( ), "large deviation" Three types of states. The target rotation angle for the rotation construction is defined as in S3. Preprocessed high-priority data is used as input to the Bayesian model, and combined with the corrected conditional probability table from S3, the rotation risk probability is calculated through Bayesian inference. ,in The rotation risk status has values ​​including "safe". "Low risk" "Medium risk" "High risk" The inference formula is based on Bayes' theorem: ;

[0088] In the formula, This is high-priority data after preprocessing (such as "large rotation angle deviation, stress exceeding limits"). Risk status Data The likelihood probability (from the revised conditional probability table). Risk status Prior probabilities (based on historical construction data statistics, such as the prior probability of a "safe" state) ), For data The total probability (through) (Calculation), final output This refers to the posterior probability of various risk states, serving as the core result of the probability of rotational risk.

[0089] When integrating multi-dimensional calculation results using a weighted fusion algorithm to generate a comprehensive stability assessment value, the multi-dimensional calculation results include the probability of rotation risk and the standardized deviation values ​​of each monitoring parameter. First, the probability of rotation risk is converted into a risk score. Define "security" as the corresponding "Divided into" and "low-risk" categories. "Medium-risk" "High-risk" The risk score is calculated by weighting risk probabilities.

[0090] ;

[0091] In the formula, The posterior probabilities of various risk states are given (same as the Bayesian inference results mentioned above). Next, the standardized deviation scores of each monitoring parameter are calculated. For the first Class monitoring parameters, their standardized deviation , For real-time monitoring values, Standard values ​​for parameters (e.g., standard angle value) ), , The maximum and minimum values ​​of the parameters (same as the standardized definition of the S3 feature matrix), and the deviation score. value range The smaller the deviation, the higher the score. Finally, based on dynamic weight parameters... The risk score and the deviation scores of each parameter are weighted and fused to generate a comprehensive stability assessment value. :

[0092] ;

[0093] In the formula, The weight of the risk score is set to 0.5 to ensure that the risk status is the core influencing factor. (Dynamic weight parameters) First normalize to a sum of 0.5). Range of values A higher score indicates better stability during the rotation construction, typically based on... For "stability", For "basic stability", It is "unstable".

[0094] The risk contribution analysis results were calculated using the Shapley value algorithm to quantify the impact of each monitoring parameter on the rotation risk. The Shapley value is defined as follows: For the first The risk contribution of the monitoring parameters is calculated based on the cooperative Boyle theory:

[0095] ;

[0096] In the formula, For the set of all monitored parameters, Not including the first A subset of parameters For subset The number of elements, Total number of parameters , For subset The corresponding risk function value (calculated using a subset via a Bayesian model) Risk probability under input The higher the probability of risk, the better. The larger ( For the first Add parameters to subset The change in the post-risk function, Shapley value A larger value indicates a higher contribution of the parameter to the risk of rotation. For example, if the Shapley value of the rotation angle parameter... Stress parameters This indicates that the rotation angle and stress are the core parameters affecting the risk of rotation and require close attention.

[0097] The recommendations for construction rotation control are generated based on the results of risk contribution analysis, targeting the top risks with the highest contribution. Adjustment plans are formulated based on individual parameters. Taking the cornering angle parameter (which has the highest contribution) as an example, if the real-time value of the cornering angle... greater than the target turning angle If so, the hydraulic traction force needs to be reduced. The hydraulic traction force adjustment value is... The calculation formula is:

[0098] ;

[0099] In the formula, The traction force adjustment coefficient (determined based on the stiffness of the bridge's rotation structure, such as...) ), This represents the required reduction in traction force (positive values ​​indicate reduction, negative values ​​indicate increase). For the stress parameter (second highest contribution), if the real-time stress value... Greater than the design limit Then the cable tension needs to be optimized, and the amount of cable tension optimization is... The calculation formula is:

[0100] ;

[0101] In the formula, The cable tension adjustment factor (determined based on the cable material properties, such as...) ), The cable force needs to be increased (positive values ​​indicate an increase, negative values ​​indicate a decrease) to ensure that the structural stress is reduced to a safe range through cable force adjustment.

[0102] The warning level is based on the probability of rotational risk, and the warning level threshold is defined as follows: when the probability of "high risk" is reached... When the risk level is "medium", a "red alert" is triggered; when the risk level is "medium", a "red alert" is triggered. and When the risk level is low, an "orange alert" is triggered; when the risk level is low, an "orange alert" is triggered. and When a "yellow alert" is triggered; when and When the warning level is triggered, a "blue alert" (safety reminder) is activated. The alert level is output through the alert control interaction module of the edge computing server, which simultaneously triggers the audible and visual alarm devices (such as a high-pitched alarm and a red warning light corresponding to a red alert), and pushes the alert information to the mobile terminals of construction management personnel to ensure timely control measures are taken.

[0103] The final stability monitoring results are presented in a structured report format, including real-time dynamic weights. (List of parameter weight values), probability of rotation risk (Posterior probabilities of four risk states), warning levels (red / orange / yellow / blue), risk contribution analysis results (ranking of Shapley values ​​for each parameter), and recommendations for rotation construction control (adjustment value of hydraulic traction force). Optimization of cable tension The report is uploaded to the cloud collaborative layer storage through the network interface of the edge computing server and displayed in real time on the monitoring terminal at the construction site, providing complete data support for the stability control of bridge rotation construction.

[0104] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, because according to the embodiments of this application, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this application.

[0105] In the method of this application embodiment, firstly, in the data input stage, by collecting "explicit parameters of the rotating structure (rotation angle, stress, etc.) + implicit related data (equipment health status, environmental abrupt change precursors)" and executing a multi-source data spatiotemporal alignment algorithm, on the one hand, the explicit parameters cover the core dimensions of the rotating structure's attitude and stress, and the implicit data supplements the equipment status and risk precursor information that affect the monitoring accuracy, which can avoid the one-sidedness of monitoring caused by a single data dimension; on the other hand, the time alignment based on timestamp interpolation and the spatial alignment based on coordinate transformation can eliminate the spatiotemporal deviation of data from different sensors, reduce data errors, and ultimately provide high-quality input data with "full dimensions and high synchronization" for subsequent model calculations, thereby avoiding monitoring deviations caused by incomplete or asynchronous data from the source.

[0106] Secondly, in the initial model building stage, a combination of "weighted aggregation of multi-project monitoring parameters to build a global collaborative knowledge model + local fine-tuning optimization" is used. The aggregation of multi-project parameters can reuse historical monitoring experience of different bridge rotation constructions, eliminating the need to train the model from scratch for the target project and directly shortening the initial model adaptation cycle. At the same time, the dynamic weighting based on "data volume + working condition representativeness" during the aggregation process ensures that the global model takes into account both the universality of historical experience and the adaptability to the basic working conditions of the target project. Combined with the "working condition difference calculation + least squares error correction" in the local fine-tuning stage, the working condition deviation between the global model and the target project can be further corrected, enabling the initial monitoring model to quickly achieve the effect of "adapting to basic working conditions and controllable output error", solving the problems of long adaptation cycle and low initial accuracy of single-project monitoring models.

[0107] Furthermore, in the dynamic optimization stage of the model, relying on the self-organizing evolutionary logic of "density clustering for new working conditions identification + particle swarm optimization for parameter updates + Bayesian early warning parameter correction", the density clustering algorithm can automatically identify new working conditions not covered in the construction of the target project through the scientific setting of neighborhood radius and minimum sample number, triggering parameter optimization without manual intervention, breaking through the limitation of traditional models that can only adapt to preset working conditions; the particle swarm optimization algorithm can efficiently iterate and update dynamic weight parameters through the coordinated regulation of inertial weights and learning factors, so that the weight ratio of each monitoring parameter matches the changes in working conditions in real time; the Bayesian network introduces environmental mutation precursor data as new nodes, and updates conditional probabilities through maximum likelihood estimation and realizes complex working condition reasoning through Markov chain Monte Carlo algorithm, which can capture risk causes in advance and improve the accuracy of risk probability calculation. The synergy of the three enables the model to have the ability of "self-identification of new working conditions, dynamic parameter updates, and early prediction of risks", solving the problems of poor response to dynamic working conditions and delayed risk warning of traditional models.

[0108] Finally, in the collaborative management and control phase, the approach of "dynamic weighting guiding data priority and computational resource allocation + Bayesian risk calculation + weighted fusion assessment + Shapley value risk tracing and control suggestion generation" ensures that high-impact monitoring data (such as rotation angle and stress) receives priority in collection and computational resources, reducing data processing latency and improving real-time monitoring. Bayesian inference accurately calculates risk probabilities based on preprocessed high-priority data, while the weighted fusion algorithm integrates risk scores and parameter deviation scores, avoiding the one-sidedness of single-dimensional assessment and making the stability assessment results more comprehensive. The Shapley value algorithm quantifies the risk contribution of each parameter, accurately locating core risk sources and generating targeted control suggestions such as hydraulic traction force adjustment and cable force optimization. Ultimately, this achieves the effect of "high real-time monitoring, accurate risk judgment, and highly targeted control schemes," providing full-process technical support for the stability of bridge rotation construction.

[0109] In summary, the entire technical approach, through a progressive process of "high-quality data support - rapidly adaptable initial model - dynamically evolving optimization mechanism - precise and collaborative control logic," ultimately achieves the core objective of "solving the problems of slow adaptation of single-project monitoring models, poor real-time operational condition response, and inaccurate risk warnings," significantly improving the efficiency and accuracy of stability monitoring during bridge rotation construction. The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this application is not limited to technical solutions formed by specific combinations of the above-mentioned technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-mentioned technical features or their equivalents without departing from the aforementioned disclosed concept. For example, technical solutions formed by substituting the above-mentioned features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A stability monitoring method for bridge rotation construction, characterized in that, Includes the following steps: Acquire initial monitoring data for the target bridge rotation construction project. The initial monitoring data includes explicit monitoring data reflecting the posture and stress of the rotating structure, as well as implicit correlation data that affect the monitoring accuracy. The initial monitoring data is locally fine-tuned based on a global collaborative knowledge model formed by weighted aggregation of parameters from multiple project monitoring models, resulting in an initial monitoring model adapted to the basic working conditions of the target project. The aggregation uses a dynamic weighting system that assigns weights based on the proportion of data from multiple projects and the representativeness of operating conditions. During the construction of the target bridge rotation, real-time monitoring data is continuously acquired, and the initial monitoring model is dynamically optimized through a self-organizing evolutionary optimization algorithm. The self-organizing evolutionary optimization algorithm includes the identification of new working conditions and the correction of dynamic weight parameters and Bayesian early warning parameters, so as to obtain a dynamic monitoring model that adapts to the real-time working conditions. Based on the collaborative calculation of the dynamic weight parameters and Bayesian early warning parameters, the stability monitoring results of the target bridge rotation construction are output.

2. The method according to claim 1, characterized in that, The algorithms for constructing a global collaborative knowledge model include: Historical monitoring parameters from multiple bridge rotation projects were collected, and the parameter dimensions were unified through feature normalization. The working condition matching coefficient of each project is calculated using the analytic hierarchy process, and a comprehensive weight is generated by combining the proportion of data volume. A global knowledge model that can be dynamically invoked is formed by fusing parameters from multiple items using a weighted average algorithm.

3. The method according to claim 1, characterized in that: The visible monitoring data includes the rotation angle data, stress data, wind speed data, and vibration frequency data of the rotating structure; Implicit correlation data includes health data of monitoring equipment and data on precursors of environmental changes; The synchronous fusion processing of two types of data is achieved through a multi-source data spatiotemporal alignment algorithm.

4. The method according to claim 1, characterized in that, The local fine-tuning algorithm for the initial monitoring model includes: Calculate the characteristic difference between the target project's basic operating conditions and the global typical operating conditions; The initial values ​​of the parameters of the global model are adaptively adjusted based on the degree of difference. The least squares method is used to correct the error of the adjusted model, so that the model output error is controlled within a preset range.

5. The method according to claim 1, characterized in that, The self-organizing evolutionary optimization algorithm includes: Construct a feature matrix that includes real-time monitoring data and implicit correlation data; Density clustering algorithm is used to identify new operating conditions not included in the basic operating conditions; For the new operating condition trigger parameter optimization process, the dynamic weight parameters and Bayesian early warning parameters are iteratively updated using the particle swarm optimization algorithm.

6. The method according to claim 5, characterized in that, Dynamic weight parameter optimization algorithms include: Establish a multi-dimensional evaluation function with the objectives of corner control accuracy and risk warning timeliness; The weight ratio of each monitoring parameter is dynamically adjusted using a reinforcement learning algorithm; Set operational robustness constraints to adapt the weight adjustment results to the dynamic characteristics of the rotation construction.

7. The method according to claim 5, characterized in that, The Bayesian early warning parameter correction algorithm includes: Introduce environmental aberration precursor data as new nodes into the Bayesian network; Update the conditional probabilities between nodes based on the maximum likelihood estimation method; By implementing probabilistic reasoning under complex working conditions using Markov chain Monte Carlo algorithms, the accuracy of risk prediction can be improved.

8. The method according to claim 1, characterized in that, The collaborative computing includes: Construct a priority ranking rule for monitoring data based on dynamic weight parameters; Computing resources are dynamically allocated based on priority. High-priority data is input into the modified Bayesian model for risk calculation. By integrating multi-dimensional calculation results through a weighted fusion algorithm, a comprehensive stability assessment value representing the stability monitoring results is generated.

9. The method according to claim 2, characterized in that, Global knowledge model update algorithms include: Receive the optimized model parameters for the target project and mark the parameters to generate timestamps; The timeliness coefficient of the parameters is calculated based on the time decay function; An incremental learning algorithm is used to incorporate the weighted new parameters into the global model, thereby achieving incremental evolution of the model.

10. A stability monitoring system for bridge rotation construction, characterized in that, include: The data acquisition layer contains a monitoring network composed of various sensors, used to collect various monitoring data during the rotation construction. An edge computing layer, including a local server deployed at the construction site, is used to execute the method of any one of claims 1 to 9 to realize real-time processing of monitoring data and model optimization; The cloud-based collaboration layer, including a remote server cluster, is used to store the global collaborative knowledge model and provide cross-project data collaboration services. The communication transmission layer employs multi-mode communication technology to enable data interaction between different layers, ensuring data transmission requirements during algorithm execution.

Images