A dam two-way creeping form construction intelligent safety monitoring and early warning system and method

Abstract

The application provides a dam two-way climbing formwork construction intelligent safety monitoring and early warning system and method, relates to the field of dam engineering construction safety monitoring, and the method comprises the following steps: a multi-source data acquisition module, a data preprocessing module, a multi-source data fusion module, a machine learning safety research and judgment module, a hierarchical early warning output module and a background management and control terminal. After collecting six-dimensional data of displacement, deformation, inclination, vibration, load and wind speed, the data are preprocessed by adaptive Kalman filtering and improved 3σ criterion, a dynamic entropy weight fusion algorithm based on mutual information and mechanical prior coefficient is used to construct a nonlinear coupling comprehensive safety characteristic value, and a hierarchical multi-granularity weighted random forest model is used to output a four-level safety risk grade and trigger hierarchical early warning. The technical scheme of the application realizes intelligent monitoring and accurate early warning of all parameters and all time periods of climbing formwork construction.

Description

Technical Field

[0001] This application relates to the field of safety monitoring during dam construction, and in particular to an intelligent safety monitoring and early warning system and method for bidirectional climbing formwork construction of dams. Background Technology

[0002] The construction of dam concrete pouring often employs a two-way climbing formwork construction technique. This technique boasts advantages such as high construction efficiency, strong adaptability, and the ability to perform continuous high-altitude operations, and is widely used in the construction of large and medium-sized water conservancy dam projects. The two-way climbing formwork system is a dynamic temporary load-bearing structure that continuously bears multiple loads during construction, including concrete loads, its own weight, wind loads, and construction vibrations. This makes it highly susceptible to safety hazards such as formwork displacement, structural deformation, excessive tilt angles, and uneven stress. If these hazards are not detected and addressed promptly, they can easily lead to major safety accidents such as formwork slippage and partial collapse, seriously threatening the lives of dam construction personnel and the progress and quality of the project.

[0003] Currently, safety monitoring methods for bidirectional climbing formwork dams remain relatively traditional, relying heavily on periodic manual inspections and single-sensor point monitoring. Manual inspections are affected by factors such as personnel experience, field of vision, and the high-altitude working environment, resulting in problems like large monitoring intervals, poor real-time performance, strong subjectivity, and high rates of missed and false detections, making all-weather dynamic monitoring impossible. Traditional single-parameter monitoring methods can only collect and analyze indicators for a single climbing formwork condition, failing to comprehensively consider the coupled effects of multiple factors such as displacement, deformation, vibration, load, and environmental wind force. This leads to low data utilization and an inability to comprehensively and accurately reflect the overall safety status of the climbing formwork, with extremely poor ability to identify hidden and minor safety hazards. Furthermore, existing monitoring systems often use fixed threshold judgment and early warning methods, which cannot adapt to the complex changes in dynamic climbing formwork construction conditions. Risk classification and assessment accuracy is low, and early warnings are delayed, failing to meet the high-precision and intelligent safety control requirements of modern dam climbing formwork construction. Summary of the Invention

[0004] The purpose of this invention is to address the technical problems in existing climbing formwork construction safety monitoring technologies, such as the single monitoring dimension, fixed thresholds, and fixed weights that cannot adapt to dynamic changes in multiple working conditions, leading to inaccurate risk assessment and delayed early warning. This invention provides an intelligent safety monitoring and early warning system and method for bidirectional climbing formwork construction of dams.

[0005] The above-mentioned objective of this application is achieved through the following technical solution: The system includes: a multi-source data acquisition module, a data preprocessing module, a multi-source data fusion module, a machine learning security assessment module, a hierarchical early warning output module, and a back-end management and control terminal; The multi-source data acquisition module is used to collect raw monitoring data from various sensors; The data preprocessing module is connected to the multi-source data acquisition module and is used to receive raw monitoring data, preprocess it, and output a standardized monitoring dataset. The multi-source data fusion module is connected to the data preprocessing module and is used to receive the standardized monitoring dataset, perform deep fusion of the standardized monitoring data using an improved weighted fusion algorithm, dynamically allocate fusion coefficients according to the impact weight of the monitoring data on the safety risk of climbing formwork, and construct a fusion feature dataset that characterizes the overall safety condition of climbing formwork. The machine learning security assessment module has a built-in pre-trained security level judgment model, which outputs the real-time security risk level of the crawling model based on the fused feature dataset. The graded early warning output module and the background management terminal are both connected to the machine learning security assessment module. The graded early warning output module outputs different levels of early warning signals and triggers the corresponding sound and light early warning devices according to the security risk level. The back-end management terminal is used to display climbing mold working condition data and risk warning information in real time, and to store monitoring data and warning records synchronously.

[0006] Optionally, the sensor group includes a displacement sensor, a deformation sensor, a tilt sensor, a vibration sensor, a load sensor, and an ambient wind speed sensor, used to collect raw monitoring data in real time during the climbing formwork construction process; the raw monitoring data includes: climbing formwork displacement data, structural deformation data, tilt angle data, vibration frequency data, real-time load data, and ambient wind speed data at the construction site.

[0007] Optionally, the preprocessing includes: filtering and denoising, outlier removal, missing data interpolation and completion, multi-dimensional unit unification and standardization. The data preprocessing module uses an adaptive Kalman filter algorithm to filter and denoise the raw monitoring data. Its state prediction equation is as follows:

[0008] The covariance adaptive update equation is:

[0009] in, The state prediction value at time k. This is the state estimate at time k-1. Here is the state transition matrix. For the control matrix, This is the control input vector at time k-1. Let be the prediction error covariance matrix at time k. Let be the estimation error covariance matrix at time k-1. This is the transpose of the state transition matrix. The adaptive process noise covariance matrix is ​​dynamically updated based on real-time wind speed and vibration conditions. The data preprocessing module uses a modified 3σ criterion to remove abnormal and mutated data. The anomaly detection formula is as follows:

[0010] in, For the first One monitoring data value, To monitor the mean of the data series, To monitor the standard deviation of the data series, For data exceeding Duration of the interval The preset effective duration threshold for the operating condition; Data that meets the anomaly detection criteria will be identified as invalid outliers and removed. Data with a duration exceeding [a certain duration] will be removed. The effective peak value data of the instantaneous impact load are retained; The data preprocessing module uses piecewise linear interpolation to complete the original monitoring data and applies an extreme value standardization formula. Standardize the units of measurement for all parameters, among which For data to be standardized, The minimum value among similar monitoring data. The maximum value of similar monitoring data. This is the standardized monitoring data.

[0011] Optionally, the multi-source data receives a standardized monitoring dataset and uses an improved weighted fusion algorithm to perform deep fusion on the standardized monitoring data; Based on the impact weight of monitoring data on the safety risks of climbing formwork, a fusion coefficient is dynamically allocated to construct a fusion feature dataset that characterizes the overall safety status of climbing formwork. The multi-source data fusion module constructs a fused feature dataset using the following steps: Calculate the mutual information between monitoring data and climbing model safety risk labels, quantify parameter risk sensitivity, and use the following formula to calculate the univariate information entropy of the parameter:

[0012]

[0013] in, For the first Information entropy of monitoring data For the sample size, For the first Class of monitoring data The proportion of each sample value in the total number of parameters of this type For the first Class of monitoring data Standardized values ​​for each sample; The formula for calculating the mutual information between monitoring data and safety risk labels is as follows:

[0014] in, For the first Class of monitoring data Safety risk label mutual information, For conditional entropy; Prior coefficients of risk contribution for climbing formwork structures obtained by introducing finite element simulation Calculate the basic sensitivity:

[0015] in, For the first The basic sensitivity of the monitoring data. For the first Prior coefficient of structural risk contribution to monitoring data; Set the operating condition identification function Matching piecewise correction coefficient Calculate the sensitivity after dynamic correction:

[0016] in, For dynamically corrected sensitivity, According to the working condition category, For the first Under the first type of working condition Segmentation correction coefficients for monitoring data; Normalized dynamic fusion weights:

[0017] in, For the first Dynamic fusion weights of monitoring data To monitor the total number of data categories, For hierarchical dynamic correction coefficients; Constructing a comprehensive security eigenvalue with nonlinear coupling:

[0018] in, To comprehensively assess safety characteristics, For the first Class 1 monitoring data and the first The mutual information coupling coefficient between monitoring data of different types , The first Class, No. Standardized values ​​of monitoring data For single-parameter basic weighted terms, This is a two-parameter cross-coupling term.

[0019] Optionally, the security level judgment model adopts a hierarchical multi-granularity weighted random forest model, including: A three-layer parallel feature input layer, wherein the three layers of parallel features in the three-layer parallel feature input layer have independently configured hierarchical weights. : The basic feature channel input consists of a single sensor standardized parameter set composed of a standardized monitoring dataset. ; Coupled feature channel input two-parameter cross-coupling term ; Global fusion channel input comprehensive security feature value ; Dual-granularity decision tree parallel layer: Constructs fine-grained decision tree clusters and coarse-grained decision tree clusters in parallel. The fine-grained decision trees are used to identify hidden risks, and the coarse-grained decision trees are used to identify explicit major risks. Dynamic integrated weight calculation unit: weights based on real-time operating conditions and dynamic parameters. Calculate the independent ensemble weights for each decision tree. :

[0020] in, Let be the ensemble weights of the t-th decision tree. For the training samples of the t-th decision tree The mean of the monitoring data. This represents the total number of fine-grained decision trees. This represents the total number of coarse-grained decision trees; The overall risk comprehensive judgment layer has the following judgment formula:

[0021] in, As a comprehensive risk score, Let be the discriminant output of the t-th decision tree. This is a coarse-grained risk amplification factor. ; The comprehensive risk score is mapped to a four-level safety level, which includes: normal, minor risk, general risk, and major risk.

[0022] Optionally, the hierarchical multi-granularity weighted random forest model is trained using a hierarchical weighted loss function:

[0023] in, For training loss values, The total number of training samples, The sample number. For the first The true risk level of each sample Risk level The corresponding hierarchical loss weights, For the first The overall risk score of each sample. For the first The true risk level label for each sample; the graded loss weight for samples with significant risk is higher than that for samples with minor risk.

[0024] Optionally, the graded early warning output module outputs the following early warning signals according to the corresponding safety risk level: Level 0 Normal State: No on-site audible or visual signals; the background control terminal displays operating data in real time and automatically stores monitoring time sequence data. Level 1 warning: A pop-up notification appears on the backend management terminal, and the warning is recorded in the warning log. Level 2 alarm: On-site yellow light continuous audible and visual alarm, and simultaneously push investigation information to the mobile terminal of management personnel; Level 3 Emergency Response: A high-frequency red light alarm will be activated on-site, automatically sending warning information to the project safety manager and construction team leader.

[0025] A method for intelligent safety monitoring and early warning during bidirectional climbing formwork construction of dams, the method comprising: S1. Sensor groups are deployed at key monitoring points of the two-way climbing formwork system of the dam to collect raw monitoring data in real time around the clock and perform preprocessing to generate standardized monitoring datasets. The raw monitoring data includes: climbing formwork displacement data, structural deformation data, tilt angle data, vibration frequency data, real-time load data, and construction site environmental wind speed data. S2. Based on mutual information calculation, the risk sensitivity of each monitoring data in the standardized monitoring dataset is calculated. Mechanical prior coefficient correction is introduced. Combined with the segmented dynamic correction of the weight of each parameter under the construction conditions, a comprehensive safety feature value containing single-parameter basic weighting terms and two-parameter cross-coupling terms is constructed. S3. Input the preprocessed raw monitoring data, the two-parameter cross-coupling term and the comprehensive safety feature value into the hierarchical multi-granularity weighted random forest model. The model outputs a four-level safety risk level. S4. Based on the safety risk level, output different levels of early warning signals and trigger the audible and visual early warning device, and simultaneously display the operating condition data and early warning information on the background management terminal.

[0026] The beneficial effects of the technical solution provided in this application are: 1. This invention uses a network of multiple types of sensors to collect multi-dimensional working condition data of climbing formwork, replacing the traditional single-parameter monitoring mode. Combined with multi-source data fusion technology, it couples the influence of multiple factors, which can comprehensively and truly reflect the overall construction conditions of bidirectional climbing formwork, completely solving the problems of one-sidedness and high missed detection rate of single monitoring data, and greatly improving the monitoring coverage and accuracy.

[0027] 2. This invention breaks through the technical bottlenecks of traditional fixed threshold judgment and fixed weight data fusion. It independently designs a dynamic entropy weight fusion algorithm based on the correlation of working conditions and an improved weighted random forest discrimination model, with original core theoretical formulas. It can adaptively adjust the parameter weights and model discrimination coefficients according to the real-time construction conditions, and can accurately capture small, hidden, and multi-parameter coupled safety hazards in climbing formwork. It can achieve four-level refined risk classification and early warning, and theoretically solve the technical problems of poor working condition adaptability and high false judgment and missed judgment rate of traditional monitoring models. It has significant theoretical innovation and technological advancement.

[0028] 3. This invention realizes the fully automated and intelligent operation of climbing formwork construction data collection, preprocessing, data fusion, risk assessment, hierarchical early warning, and data archiving. It eliminates the need for frequent manual inspections and interventions, significantly reduces labor costs, avoids the safety risks of manual inspections at heights, and achieves uninterrupted safety monitoring around the clock.

[0029] 4. This invention can track construction monitoring data and early warning records in real time, providing data support for quality control, rectification and review of hidden dangers, and optimization of construction technology in dam climbing formwork construction. It effectively ensures the safety and stability of bidirectional climbing formwork construction of dams, and is suitable for climbing formwork construction scenarios of various large and medium-sized water conservancy dams. It has strong versatility and engineering practicality. Attached Figure Description

[0030] The present application will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings: Figure 1 This is a diagram of the overall system architecture for multi-source data acquisition, preprocessing, and fusion in this application embodiment; Figure 2 This is a structural diagram of the security assessment model of hierarchical multi-granularity weighted random forest machine learning in the embodiments of this application; Figure 3 This is a real-time system overview monitoring diagram in the embodiments of this application; Figure 4 This is a system algorithm model configuration diagram in an embodiment of this application. Detailed Implementation

[0031] To provide a clearer understanding of the technical features, objectives, and effects of this application, the specific embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0032] The embodiments of this application provide an intelligent safety monitoring and early warning method for bidirectional climbing formwork construction of dams.

[0033] Please refer to Figure 1 , Figure 1 This application presents a multi-source data acquisition-preprocessing-fusion overall system architecture diagram for an intelligent safety monitoring and early warning method for bidirectional climbing formwork construction of dams, as described in this embodiment. The system includes: Multi-source data acquisition module, data preprocessing module, multi-source data fusion module, machine learning security assessment module, hierarchical early warning output module, and back-end management terminal; The multi-source data acquisition module is used to collect raw monitoring data from various sensors; The data preprocessing module is connected to the multi-source data acquisition module and is used to receive raw monitoring data, preprocess it, and output a standardized monitoring dataset. The multi-source data fusion module is connected to the data preprocessing module and is used to receive the standardized monitoring dataset, perform deep fusion of the standardized monitoring data using an improved weighted fusion algorithm, dynamically allocate fusion coefficients according to the impact weight of the monitoring data on the safety risk of climbing formwork, and construct a fusion feature dataset that characterizes the overall safety condition of climbing formwork. The machine learning security assessment module has a built-in pre-trained security level judgment model, which outputs the real-time security risk level of the crawling model based on the fused feature dataset. The graded early warning output module and the background management terminal are both connected to the machine learning security assessment module. The graded early warning output module outputs different levels of early warning signals and triggers the corresponding sound and light early warning devices according to the security risk level. The back-end management terminal is used to display climbing mold working condition data and risk warning information in real time, and to store monitoring data and warning records synchronously.

[0034] The sensor group includes displacement sensors, deformation sensors, tilt sensors, vibration sensors, load sensors, and environmental wind speed sensors, which are used to collect raw monitoring data in real time during the climbing formwork construction process. The raw monitoring data includes: climbing formwork displacement data, structural deformation data, tilt data, vibration frequency data, real-time load data, and environmental wind speed data at the construction site.

[0035] The preprocessing includes: filtering and denoising, outlier removal, missing data interpolation and completion, and multi-dimensional unit unification and standardization. The data preprocessing module uses an adaptive Kalman filter algorithm to filter and denoise the raw monitoring data. Its state prediction equation is as follows:

[0036] The covariance adaptive update equation is:

[0037] in, The state prediction value at time k. This is the state estimate at time k-1. Here is the state transition matrix. For the control matrix, This is the control input vector at time k-1. Let be the prediction error covariance matrix at time k. Let be the estimation error covariance matrix at time k-1. This is the transpose of the state transition matrix. The adaptive process noise covariance matrix is ​​dynamically updated based on real-time wind speed and vibration conditions. As one example, the collected raw monitoring data is preprocessed, abandoning the traditional single-filter fixed parameter processing method, and adopting an adaptive Kalman filter algorithm to achieve dynamic noise suppression, adapting to the dynamic disturbance conditions of climbing formwork at high altitude.

[0038] The data preprocessing module uses a modified 3σ criterion to remove abnormal and mutated data. The anomaly detection formula is as follows:

[0039] in, For the first One monitoring data value, To monitor the mean of the data series, To monitor the standard deviation of the data series, For data exceeding Duration of the interval The preset effective duration threshold for the operating condition; Data that meets the anomaly detection criteria will be identified as invalid outliers and removed. Data with a duration exceeding [a certain duration] will be removed. The effective peak value data of the instantaneous impact load are retained; As one embodiment, an improved The criteria remove abnormal and sudden data. Effective peak data generated by instantaneous impact loads during climbing formwork construction are not mistakenly removed. Only outliers with short-term sudden changes and no corresponding working conditions are removed.

[0040] The data preprocessing module uses piecewise linear interpolation to complete the original monitoring data and applies an extreme value standardization formula. Standardize the units of measurement for all parameters, among which For data to be standardized, The minimum value among similar monitoring data. The maximum value of similar monitoring data. This is the standardized monitoring data.

[0041] The multi-source data receives a standardized monitoring dataset and uses an improved weighted fusion algorithm to deeply fuse the standardized monitoring data. Based on the impact weight of monitoring data on the safety risks of climbing formwork, a fusion coefficient is dynamically allocated to construct a fusion feature dataset that characterizes the overall safety status of climbing formwork. The multi-source data fusion module constructs a fused feature dataset using the following steps: Calculate the mutual information between monitoring data and climbing model safety risk labels, quantify parameter risk sensitivity, and use the following formula to calculate the univariate information entropy of the parameter:

[0042]

[0043] in, For the first Information entropy of monitoring data For the sample size, For the first Class of monitoring data The proportion of each sample value in the total number of parameters of this type For the first Class of monitoring data Standardized values ​​for each sample; The formula for calculating the mutual information between monitoring data and safety risk labels is as follows:

[0044] in, For the first Class of monitoring data Safety risk label mutual information, For conditional entropy; Prior coefficients of risk contribution for climbing formwork structures obtained by introducing finite element simulation Calculate the basic sensitivity:

[0045] in, For the first The basic sensitivity of the monitoring data. For the first Prior coefficient of structural risk contribution to monitoring data; Set the operating condition identification function Matching piecewise correction coefficient Calculate the sensitivity after dynamic correction:

[0046] in, For dynamically corrected sensitivity, According to the working condition category, For the first Under the first type of working condition Segmentation correction coefficients for monitoring data; Normalized dynamic fusion weights:

[0047] in, For the first Dynamic fusion weights of monitoring data To monitor the total number of data categories, For hierarchical dynamic correction coefficients; Constructing a comprehensive security eigenvalue with nonlinear coupling:

[0048] in, To comprehensively assess safety characteristics, For the first Class 1 monitoring data and the first The mutual information coupling coefficient between monitoring data of different types , The first Class, No. Standardized values ​​of monitoring data For single-parameter basic weighted terms, This is a two-parameter cross-coupling term.

[0049] As one example, deep coupling and fusion of multi-source data: abandoning the traditional fixed-weight linear entropy weight fusion, it implements mechanical prior correction, dynamic weighting of working conditions, and nonlinear coupling feature construction step by step. It calculates the mutual information between each monitoring data point and the climbing model risk label, quantifying the univariate information entropy of the parameter risk sensitivity parameter and the mutual information between the parameter and the safety risk label. As conditional entropy, historical climbing formwork collapse and slippage accident sample labels are introduced to characterize the degree of information change of parameters under risky working conditions.

[0050] Calculation of prior sensitivity of mechanical foundations: Finite element simulation is introduced to obtain the prior coefficient of risk contribution of climbing formwork structure. (load, deformation parameters) Larger values ​​indicate higher mechanical risks): Dynamic correction settings for construction conditions segmentation, including the condition identification function. , Pouring conditions Horizontal movement condition Climbing conditions Static operating condition, matching segmented correction coefficient During the pouring stage, the weights of load and deformation are amplified; during the climbing stage, the weights of tilt angle and vibration are amplified; during the stationary stage, the weights of wind speed and displacement are amplified; normalized dynamic fusion weights (layered minimum value correction to avoid low-sensitivity parameter weights returning to zero); parameters The hierarchical dynamic correction coefficients have differentiated values ​​for the safety and warning intervals. A nonlinear coupled comprehensive safety characteristic value (distinct from the general linear weighted formula, possessing originality) introduces a parameter mutual information coupling coefficient. To characterize the risk amplification caused by the synergistic effect of two types of parameters, a new dual-parameter cross-coupling term has been added. The first term is a single-parameter basic weighted term, and the second term is a multi-indicator synchronous anomaly coupling risk term. F is the comprehensive safety fusion feature of the crawling model, which fully characterizes the two types of safety hazards: single-parameter anomalies and multi-parameter synergistic anomalies.

[0051] The security level determination model adopts a hierarchical multi-granularity weighted random forest model, the structure of which includes: A three-layer parallel feature input layer, wherein the three layers of parallel features in the three-layer parallel feature input layer have independently configured hierarchical weights. : The basic feature channel input consists of a single sensor standardized parameter set composed of a standardized monitoring dataset. ; Coupled feature channel input two-parameter cross-coupling term ; Global fusion channel input comprehensive security feature value ; Dual-granularity decision tree parallel layer: Constructs fine-grained decision tree clusters and coarse-grained decision tree clusters in parallel. The fine-grained decision trees are used to identify hidden risks, and the coarse-grained decision trees are used to identify explicit major risks. Dynamic integrated weight calculation unit: weights based on real-time operating conditions and dynamic parameters. Calculate the independent ensemble weights for each decision tree. :

[0052] in, Let be the ensemble weights of the t-th decision tree. For the training samples of the t-th decision tree The mean of the monitoring data. This represents the total number of fine-grained decision trees. This represents the total number of coarse-grained decision trees; The overall risk comprehensive judgment layer has the following judgment formula:

[0053] in, As a comprehensive risk score, Let be the discriminant output of the t-th decision tree. This is a coarse-grained risk amplification factor. ; The comprehensive risk score is mapped to a four-level safety level, which includes: normal, minor risk, general risk, and major risk.

[0054] As one example, the coarse-grained dual-granularity decision tree hierarchical training constructs two types of decision trees: fine-grained weak trees (deep trees with fine segmentation thresholds, identifying hidden risks such as micro-deformation and weak vibration) and coarse-grained strong trees (deep trees with loose thresholds, identifying explicit major risks such as overload and significant tilt); the hierarchical constraint discrimination output of a single decision tree is as follows:

[0055] in The three layers of features are independently layered with weights, I is the indicator function, and c corresponds to the four security levels.

[0056] The hierarchical multi-granularity weighted random forest model is trained using a hierarchical weighted loss function:

[0057] in, For training loss values, The total number of training samples, The sample number. For the first The true risk level of each sample Risk level The corresponding hierarchical loss weights, For the first The overall risk score of each sample. For the first The true risk level label for each sample; the graded loss weight for samples with significant risk is higher than that for samples with minor risk.

[0058] The graded early warning output module outputs the following early warning signals according to the security risk level: Level 0 Normal State: No on-site audible or visual signals; the background control terminal displays operating data in real time and automatically stores monitoring time sequence data. Level 1 warning: A pop-up notification appears on the backend management terminal, and the warning is recorded in the warning log. Level 2 alarm: On-site yellow light continuous audible and visual alarm, and simultaneously push investigation information to the mobile terminal of management personnel; Level 3 Emergency Response: A high-frequency red light alarm will be activated on-site, automatically sending warning information to the project safety manager and construction team leader.

[0059] Figure 3 It is a real-time overview monitoring chart of the system; Figure 4This is the system algorithm model configuration diagram.

[0060] A method for intelligent safety monitoring and early warning during bidirectional climbing formwork construction of dams, the method comprising: S1. Sensor groups are deployed at key monitoring points of the two-way climbing formwork system of the dam to collect raw monitoring data in real time around the clock and perform preprocessing to generate standardized monitoring datasets. The raw monitoring data includes: climbing formwork displacement data, structural deformation data, tilt angle data, vibration frequency data, real-time load data, and construction site environmental wind speed data. S2. Based on mutual information calculation, the risk sensitivity of each monitoring data in the standardized monitoring dataset is calculated. Mechanical prior coefficient correction is introduced. Combined with the segmented dynamic correction of the weight of each parameter under the construction conditions, a comprehensive safety feature value containing single-parameter basic weighting terms and two-parameter cross-coupling terms is constructed. S3. Input the preprocessed raw monitoring data, the two-parameter cross-coupling term and the comprehensive safety feature value into the hierarchical multi-granularity weighted random forest model. The model outputs a four-level safety risk level. S4. Based on the safety risk level, output different levels of early warning signals and trigger the audible and visual early warning device, and simultaneously display the operating condition data and early warning information on the background management terminal.

[0061] In one embodiment, this is applied to a bidirectional climbing formwork construction scenario for a rockfill concrete dam in a reservoir. The maximum pouring height of the dam is 62.5m. A bidirectional climbing formwork system is used, with the climbing formwork main beam, attachment supports, climbing cylinders, and formwork panels as the core stress-prone areas. A complete set of sensor groups is deployed in different areas at key nodes of the climbing formwork to complete multi-source data acquisition across the entire area. Displacement sensors: a total of 8 units, respectively arranged on the top, middle and bottom panels of the left and right templates, to collect data on the horizontal offset and vertical settlement displacement of the templates; Strain sensors: 12 in total, attached to the stress-bearing sections of the climbing formwork main beam and supporting columns, to collect the micro-deformation strain values ​​of the structure in real time; Inclination sensors: 4 units, installed on the back ribs of the template and the main body of the climbing frame, to monitor the tilt of the template and the tilt angle of the frame. Vibration sensors: 6 units, arranged on the climbing cylinder, template panel, and attachment support, to capture the vibration frequency and amplitude generated by pouring vibration and climbing impact; Load sensors: 4 sets, installed on the support bearing of the climbing formwork, to collect concrete pouring load and frame self-weight load in real time. Wind speed sensors: 2 units, respectively installed on the left and right sides of the top of the climbing formwork, to collect real-time wind speed and wind pressure environmental parameters at high altitude.

[0062] All sensors are connected to the field data acquisition box via a wired bus. The acquisition box is electrically connected to the back-end control terminal and the graded audible and visual early warning device. The entire system collects raw monitoring data 24 hours a day without interruption.

[0063] Data preprocessing implementation process: The collected raw monitoring data includes interference data such as high-altitude gusts, equipment electromagnetic interference, instantaneous impact loads during pouring, and short-term signal interruptions. The entire preprocessing process is completed using the adaptive Kalman filter, improved 3σ criterion, piecewise linear interpolation, and extreme value standardization of this invention. The implementation process of dynamic fusion of multi-source data: This embodiment abandons the fixed-weight linear fusion method and completes the deep fusion of multi-source heterogeneous data in four steps: mutual information sensitivity calculation → mechanical prior coefficient correction → dynamic weight correction under working conditions → nonlinear coupling feature construction. The mutual information between each monitoring parameter and the climbing model safety risk label is calculated based on the information entropy formula. Quantify the risk sensitivity of a single parameter; combine the prior coefficients of structural risk output from finite element simulation. To obtain the basic sensitivity The mechanical contribution of load and deformation parameters is greater, and the foundation sensitivity is significantly higher than that of wind speed and vibration parameters.

[0064] System built-in operating condition identification function Automatically distinguishes four types of construction conditions and matches segmented correction coefficients. Dynamically adjust sensitivity: Casting condition (k=1): Increase the weight of load and deformation parameters; Climbing condition (k=2): Increase tilt angle and vibration parameter weights; Horizontal mold moving case (k=3): Balance the adjustment of displacement and tilt angle weights; Stationary high wind condition (k=4): The weight of ambient wind speed is significantly increased. The dynamic weighting results measured on-site are shown in Table 1: Table 1

[0065] A hierarchical minimum correction is introduced to normalize the weights, preventing the weights of low-sensitivity parameters from dropping to zero. A comprehensive safety feature value F containing a two-parameter cross-coupling term is constructed to characterize both single-parameter anomalies and multi-parameter collaborative coupling risks. A control experiment was conducted with 60 sets of coupled hazard samples where "all single parameters are within limits and both parameters show slight synchronous deviations": using only the single-parameter weighting scheme, only 21 hazard samples were identified, with an identification rate of 35%; after introducing the coupling feature term of this invention, 57 hazard samples were identified, increasing the identification rate of coupled hazards to 95%, effectively identifying collaborative instability hazards missed by traditional monitoring systems.

[0066] Implementation of a hierarchical, multi-granular machine learning security assessment model: This invention improves the weighted random forest model, which has a five-layer structure, corresponding to the attached... Figure 3The overall architecture of the hierarchical, multi-granularity weighted random forest safety assessment model. The offline training loss calculation unit uses a hierarchical weighted loss function to train the model, setting a loss weight for Level 3 major collapse and overturning accident samples that is much higher than that for minor risk samples, thus solving the problem of imbalanced engineering hazard samples.

[0067] Model grading output and comparative test: The model output score was mapped to four safety levels: Level 0 (normal), Level 1 (Level 1 minor warning), Level 2 (Level 2 general alarm), and Level 3 (Level 3 emergency response). 72 hours of continuous on-site monitoring data were selected, and 120 sets of hazard samples were manually implanted (40 sets of minor hidden hazards, 40 sets of coupled hazards, and 40 sets of major emergencies). Two existing technical solutions were set up for simultaneous comparison. The overall performance comparison results are shown in Table 2.

[0068] Table 2

[0069] The experimental data shows that traditional fixed threshold monitoring relies on a single indicator, resulting in serious omissions of minute deformations and multi-parameter coupled hazards, and significant delays in early warning. Although the fixed weight fusion scheme has improved upon this, it cannot adapt to the dynamic adjustment of parameter weights under various construction conditions, and there are still shortcomings in the identification of hidden hazards. The hierarchical multi-granularity weighted random forest model of this invention can accurately capture various minute, coupled, and major safety hazards, with zero omissions in major emergencies, and a significant reduction in false alarm rate and early warning delay.

[0070] Tiered early warning and back-end control implementation process: After the model outputs the risk level, the tiered early warning output module executes a tiered handling mechanism: Level 0 Normal Operating Condition: No on-site audible or visual signals; the back-end terminal displays operating condition data in real time and automatically stores monitoring time sequence data. Level 1 Minor Risk (Level 1 Warning): Only a pop-up notification is displayed on the backend management terminal, the warning log is recorded, and a self-inspection reminder for potential hazards is pushed to the system. Level 2 General Risk (Level 2 Alarm): On-site yellow light continuous audible and visual alarm, and simultaneously push text investigation reminders to the project management personnel's mobile devices; Level 3 Major Risk (Level 3 Emergency Response): On-site high-frequency red light alarm automatically sends SMS warnings to the project safety manager and construction team leader, along with real-time key abnormal data such as climbing formwork displacement, load, and tilt angle, urging on-site work to be stopped immediately for handling.

[0071] The back-end management terminal has functions such as full-cycle data storage, historical working condition backtracking, early warning record export, and hidden danger statistical analysis. Construction management personnel can retrieve monitoring curves and early warning records for any time period for review of hidden danger rectification and optimization of climbing formwork construction process, providing complete data support for dam construction safety and quality control.

[0072] The above are merely exemplary embodiments of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure.

[0073] This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure. The specification and embodiments are to be considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

Claims

1. An intelligent safety monitoring and early warning system for bidirectional climbing formwork construction of dams, characterized in that, The system includes: a multi-source data acquisition module, a data preprocessing module, a multi-source data fusion module, a machine learning security assessment module, a hierarchical early warning output module, and a back-end management and control terminal; The multi-source data acquisition module is used to collect raw monitoring data from various sensors; The data preprocessing module is connected to the multi-source data acquisition module and is used to receive raw monitoring data, preprocess it, and output a standardized monitoring dataset. The multi-source data fusion module is connected to the data preprocessing module and is used to receive the standardized monitoring dataset, perform deep fusion of the standardized monitoring data using an improved weighted fusion algorithm, dynamically allocate fusion coefficients according to the impact weight of the monitoring data on the safety risk of climbing formwork, and construct a fusion feature dataset that characterizes the overall safety condition of climbing formwork. The machine learning security assessment module has a built-in pre-trained security level judgment model, which outputs the real-time security risk level of the crawling model based on the fused feature dataset. The graded early warning output module and the background management terminal are both connected to the machine learning security assessment module. The graded early warning output module outputs different levels of early warning signals and triggers the corresponding sound and light early warning devices according to the security risk level. The back-end management terminal is used to display climbing mold working condition data and risk warning information in real time, and to store monitoring data and warning records synchronously.

2. The intelligent safety monitoring and early warning system for bidirectional climbing formwork construction of a dam as described in claim 1, characterized in that, The sensor group includes displacement sensors, deformation sensors, tilt sensors, vibration sensors, load sensors, and environmental wind speed sensors, which are used to collect raw monitoring data in real time during the climbing formwork construction process. The raw monitoring data includes: climbing formwork displacement data, structural deformation data, tilt data, vibration frequency data, real-time load data, and environmental wind speed data at the construction site.

3. The intelligent safety monitoring and early warning system for bidirectional climbing formwork construction of a dam as described in claim 1, characterized in that, The preprocessing includes: filtering and denoising, outlier removal, missing data interpolation and completion, and multi-dimensional unit unification and standardization. The data preprocessing module uses an adaptive Kalman filter algorithm to filter and denoise the raw monitoring data. Its state prediction equation is as follows: The covariance adaptive update equation is: in, The state prediction value at time k. This is the state estimate at time k-1. Here is the state transition matrix. For the control matrix, This is the control input vector at time k-1. Let be the prediction error covariance matrix at time k. Let be the estimation error covariance matrix at time k-1. This is the transpose of the state transition matrix. The adaptive process noise covariance matrix is ​​dynamically updated based on real-time wind speed and vibration conditions. The data preprocessing module uses a modified 3σ criterion to remove abnormal and mutated data. The anomaly detection formula is as follows: in, For the first One monitoring data value, To monitor the mean of the data series, To monitor the standard deviation of the data series, For data exceeding Duration of the interval The preset effective duration threshold for the operating condition; Data that meets the anomaly detection criteria will be identified as invalid outliers and removed. Data with a duration exceeding [a certain duration] will be removed. The effective peak value data of the instantaneous impact load are retained; The data preprocessing module uses piecewise linear interpolation to complete the original monitoring data and applies an extreme value standardization formula. Standardize the units of measurement for all parameters, among which For data to be standardized, The minimum value among similar monitoring data. The maximum value of similar monitoring data. This is the standardized monitoring data.

4. The intelligent safety monitoring and early warning system for bidirectional climbing formwork construction of a dam as described in claim 3, characterized in that, The multi-source data receives a standardized monitoring dataset and uses an improved weighted fusion algorithm to deeply fuse the standardized monitoring data; Based on the impact weight of monitoring data on the safety risks of climbing formwork, a fusion coefficient is dynamically allocated to construct a fusion feature dataset that characterizes the overall safety status of climbing formwork. The multi-source data fusion module constructs a fused feature dataset using the following steps: Calculate the mutual information between monitoring data and climbing model safety risk labels, quantify parameter risk sensitivity, and use the following formula to calculate the univariate information entropy of the parameter: in, For the first Information entropy of monitoring data For the sample size, For the first Class of monitoring data The proportion of each sample value in the total number of parameters of this type For the first Class of monitoring data Standardized values ​​for each sample; The formula for calculating the mutual information between monitoring data and safety risk labels is as follows: in, For the first Class of monitoring data Safety risk label mutual information, For conditional entropy; Prior coefficients of risk contribution for climbing formwork structures obtained through finite element simulation Calculate the basic sensitivity: in, For the first The basic sensitivity of the monitoring data. For the first Prior coefficient of structural risk contribution to monitoring data; Set the operating condition identification function Matching piecewise correction coefficient Calculate the sensitivity after dynamic correction: in, For dynamically corrected sensitivity, According to the working condition category, For the first Under the first type of working condition Segmentation correction coefficients for monitoring data; Normalized dynamic fusion weights: in, For the first Dynamic fusion weights of monitoring data To monitor the total number of data categories, For hierarchical dynamic correction coefficients; Constructing a comprehensive security eigenvalue with nonlinear coupling: in, To comprehensively assess safety characteristics, For the first Class 1 monitoring data and the first The mutual information coupling coefficient between monitoring data of different types , The first Class, No. Standardized values ​​of monitoring data For single-parameter basic weighted terms, This is a two-parameter cross-coupling term.

5. The intelligent safety monitoring and early warning system for bidirectional climbing formwork construction of a dam as described in claim 4, characterized in that, The security level determination model adopts a hierarchical multi-granularity weighted random forest model, including: A three-layer parallel feature input layer, wherein the three layers of parallel features in the three-layer parallel feature input layer have independently configured hierarchical weights. : The basic feature channel input consists of a single sensor standardized parameter set composed of a standardized monitoring dataset. ; Coupled feature channel input two-parameter cross-coupling term ; Global fusion channel input comprehensive security feature value ; Dual-granularity decision tree parallel layer: Constructs fine-grained decision tree clusters and coarse-grained decision tree clusters in parallel. The fine-grained decision trees are used to identify hidden risks, and the coarse-grained decision trees are used to identify explicit major risks. Dynamic integrated weight calculation unit: weights based on real-time operating conditions and dynamic parameters. Calculate the independent ensemble weights for each decision tree. : in, Let be the ensemble weights of the t-th decision tree. For the training samples of the t-th decision tree The mean of the monitoring data. This represents the total number of fine-grained decision trees. This represents the total number of coarse-grained decision trees; The overall risk comprehensive judgment layer has the following judgment formula: in, As a comprehensive risk score, Let be the discriminant output of the t-th decision tree. This is a coarse-grained risk amplification factor. ; The comprehensive risk score is mapped to a four-level safety level, which includes: normal, minor risk, general risk, and major risk.

6. The intelligent safety monitoring and early warning system for bidirectional climbing formwork construction of a dam as described in claim 5, characterized in that, The hierarchical multi-granularity weighted random forest model is trained using a hierarchical weighted loss function: in, For training loss values, The total number of training samples, The sample number. For the first The true risk level of each sample Risk level The corresponding hierarchical loss weights, For the first The overall risk score of each sample. For the first The true risk level label for each sample; the graded loss weight for samples with significant risk is higher than that for samples with minor risk.

7. The intelligent safety monitoring and early warning system for bidirectional climbing formwork construction of a dam as described in claim 1, characterized in that, The graded early warning output module outputs the following early warning signals according to the security risk level: Level 0 Normal State: No on-site audible or visual signals; the background control terminal displays operating data in real time and automatically stores monitoring time sequence data. Level 1 warning: A pop-up notification appears on the backend management terminal, and the warning is recorded in the warning log. Level 2 alarm: A continuous yellow light alarm is activated on-site, and investigation information is simultaneously pushed to the mobile terminal of the management personnel; Level 3 Emergency Response: A high-frequency red light alarm will be activated on-site, automatically sending warning information to the project safety manager and construction team leader.

8. A method for intelligent safety monitoring and early warning of bidirectional climbing formwork construction for dams, implemented based on an intelligent safety monitoring and early warning system for bidirectional climbing formwork construction for dams as described in any one of claims 1-7, characterized in that, The method includes: S1. Sensor groups are deployed at key monitoring points of the two-way climbing formwork system of the dam to collect raw monitoring data in real time around the clock and perform preprocessing to generate standardized monitoring datasets. The raw monitoring data includes: climbing formwork displacement data, structural deformation data, tilt angle data, vibration frequency data, real-time load data, and construction site environmental wind speed data. S2. Based on mutual information calculation, the risk sensitivity of each monitoring data in the standardized monitoring dataset is calculated. Mechanical prior coefficient correction is introduced. Combined with the segmented dynamic correction of the weight of each parameter under the construction conditions, a comprehensive safety feature value containing single-parameter basic weighting terms and two-parameter cross-coupling terms is constructed. S3. Input the preprocessed raw monitoring data, the two-parameter cross-coupling term and the comprehensive safety feature value into the hierarchical multi-granularity weighted random forest model. The model outputs a four-level safety risk level. S4. Based on the safety risk level, output different levels of early warning signals and trigger the audible and visual early warning device, and simultaneously display the operating condition data and early warning information on the background management terminal.

Images