State early warning method of safety protection device for building construction

Abstract

The invention discloses a state early warning method of a safety protection device for building construction, and belongs to the technical field of building construction. Comprising the following steps: step 1, monitoring the deformation, inclination angle and vibration acceleration of a protection device in real time, and synchronously collecting environmental data of a construction site; 2, preprocessing the collected data, segmenting the data by adopting a sliding time window method, extracting feature parameters in each time window by combining a wavelet transform technology, and generating a standardized feature vector set; 3, calculating the dynamic correlation strength between the features, and analyzing the time sequence evolution features between the indexes by using a recurrent neural network; step 4, self-adaptive adjustment of the weight is realized through the objective function of minimizing the early warning error; step 5, based on the obtained dynamic weight result, establishing a mapping relation between the weight and the threshold value, and realizing collaborative adjustment of the multi-level early warning threshold value; and step 6, evaluating the early warning performance in real time and visually displaying the state and early warning information of the protection device.

Description

Technical Field

[0001] The present application relates to the technical field of building construction, and more specifically, to a state early warning method for a safety protection device used in building construction. Background Art

[0002] Safety protection devices in the construction process are crucial to protecting the lives of construction workers. At the construction site, the protective devices are often in a complex and dynamic environment. Affected by external factors such as temperature and humidity, wind speed, and vibration, they may be deformed, tilted, vibrated, and other problems. If they are not discovered and warned in time, serious safety accidents may occur. Therefore, how to monitor and evaluate the health status of protective devices in real time and warn of their potential failures or instability risks in advance is an important part of ensuring construction safety.

[0003] At present, traditional monitoring methods for protective devices usually rely on manual inspections and regular checks, which cannot achieve real-time and dynamic status monitoring. Even some intelligent monitoring systems only rely on the monitoring results of a single sensor and cannot comprehensively and accurately evaluate the health status of the device. In the actual construction process, the status of the protective device is affected by many factors, especially the dynamic changes in the structure's stress, deformation, tilt, vibration, etc. These factors need to be comprehensively evaluated through complex data processing and correlation analysis.

[0004] Existing early warning systems are mostly based on simple threshold settings, lacking in-depth analysis of dynamic changes in multiple parameters and adaptive adjustment of multi-level early warning thresholds. Traditional methods usually use fixed early warning standards, which are difficult to cope with changes in different construction environments and different construction stages, resulting in early warning results that may deviate from reality, and false alarms or missed alarms occur from time to time, affecting the accuracy and reliability of the early warning system.

[0005] To sum up, how to achieve real-time and accurate monitoring of the status of safety protection devices used in construction, and coordinate the adjustment of multi-level warning thresholds based on dynamic data, has become a technical problem that needs to be solved urgently. Summary of the invention

[0006] In order to overcome a series of defects in the prior art, the purpose of this application is to provide a status warning method for a safety protection device for construction in view of the above problems, including the following steps.

[0007] Step 1: monitor the deformation, tilt angle and vibration acceleration of the protective device in real time, and simultaneously collect environmental data of the construction site.

[0008] Step 2: pre-process the collected data, segment the data using the sliding time window method, extract the characteristic parameters in each time window in combination with the wavelet transform technology, and generate a standardized feature vector set.

[0009] Step 3: Calculate the dynamic correlation strength between features and use recursive neural network to analyze the temporal evolution characteristics between indicators.

[0010] Step 4: Adaptive adjustment of weights is achieved by minimizing the objective function of warning error.

[0011] Step 5: Based on the obtained dynamic weight results, a mapping relationship between weights and thresholds is established to achieve coordinated adjustment of multi-level warning thresholds.

[0012] Step 6: Evaluate the warning performance in real time and visually display the status of the protection device and warning information.

[0013] Furthermore, step 1 includes the following steps.

[0014] Strain sensor arrays are arranged at major stress points, deformation-sensitive areas and key connections to monitor the stress distribution and deformation state of the protective device in real time.

[0015] High-precision angle sensors are installed at the corner nodes and tilt-sensitive areas of the protective device. A two-axis or three-axis measurement solution is used to continuously monitor the spatial posture changes of the protective device. At the same time, a tilt warning threshold system is established to achieve real-time assessment of the stability of the protective device.

[0016] A triaxial acceleration sensor network is arranged on the key components of the protective device, and the vibration response characteristics of the protective device during the construction process are captured in real time through high sampling frequency data acquisition.

[0017] An environmental monitoring network, including temperature sensors, humidity sensors, wind speed sensors, and rainfall sensors, is deployed at the construction site to establish a correlation analysis mechanism between environmental parameters and structural responses and to evaluate the impact of environmental factors on the performance of protective devices.

[0018] Build a data collection and transmission network based on the Industrial Internet of Things, adopt a combination of wired and wireless communication methods to ensure the real-time and reliability of data transmission, and at the same time establish a data backup mechanism and network redundancy design to prevent data loss and communication interruption.

[0019] Furthermore, step 2 includes the following steps.

[0020] The collected raw data are tested for outliers and removed, noise filtering is performed, and data normalization is performed to ensure data consistency and comparability.

[0021] The processed data is segmented based on the sliding window method, and the window size and sliding step parameters are optimized according to the signal characteristics to ensure that the signal characteristics are accurately reflected.

[0022] The data sequence of each time window is subjected to wavelet transform to obtain the multi-scale time-frequency decomposition result of the signal, forming a complete time-frequency domain feature representation.

[0023] Energy features, statistical features and wavelet coefficients are extracted from the time-frequency domain feature representation to ensure that the features can fully reflect the essential characteristics of the data.

[0024] The extracted features are standardized to eliminate the influence of dimension, and the standardized features are combined into feature vectors to finally form a feature vector set.

[0025] Furthermore, the processed data is segmented based on the sliding window method, and the window size and sliding step parameters are optimized according to the signal characteristics to ensure that the signal characteristics are accurately reflected, including the following steps.

[0026] The window size W and sliding step size S are set according to the time domain characteristics and frequency characteristics of the signal.

[0027] Extract continuous windows from the data sequence according to the set window size W and step size S.

[0028] Signal feature extraction is performed on the data in each window to obtain effective signal features.

[0029] Establish a validity verification mechanism for window data and perform quality checks on the data in each window, including data integrity, outlier ratio, and signal-to-noise ratio indicators. When the window data quality does not meet the requirements, automatically adjust the window parameters.

[0030] A dynamic optimization mechanism for window parameters is designed to regularly adjust the window size and sliding step size by evaluating feature extraction effect, computing resource usage, and response time indicators.

[0031] Furthermore, step 3 includes the following steps.

[0032] Construct a multi-dimensional time series association network, take each characteristic parameter as a network node, determine the initial connection relationship between nodes, and form a basic network structure.

[0033] The dynamic correlation strength between features is calculated in different time periods, the degree of correlation between nodes is quantified, and the correlation strength is used as the weight of the network edge to construct a weighted directed graph.

[0034] Use graph theory methods to analyze network structure, calculate the degree centrality, betweenness centrality and eigenvector centrality indicators of nodes, identify key nodes and important paths in the network, and reveal the transmission mechanism between the main influencing factors.

[0035] A long short-term memory network model is constructed, and the key features identified in the time series association network are used as input. The nonlinear time series evolution characteristics between various indicators are captured through the deep learning ability of the recursive neural network.

[0036] Furthermore, a multi-dimensional temporal association network is constructed, each characteristic parameter is used as a network node, the initial connection relationship between nodes is determined, and a basic network structure is formed, including the following steps.

[0037] The preprocessed standardized feature parameters are mapped to network nodes, and a unique identifier is assigned to each node to establish a bijective mapping relationship between the feature parameters and the network nodes.

[0038] Pearson correlation coefficients and mutual information values ​​were calculated between all pairs of nodes.

[0039] The connection threshold λ is set as the criterion for determining the association strength. When the association strength between two nodes exceeds λ, the corresponding connection relationship is recorded in the adjacency matrix.

[0040] The initial network topology is constructed based on the adjacency matrix, and the edge weights are set according to the correlation coefficient or mutual information value.

[0041] Perturbation analysis is performed in multiple time windows to evaluate the evolution of network structure over time and verify the temporal stability of the relationship between nodes.

[0042] Furthermore, graph theory methods are used to analyze the network structure, calculate the degree centrality, betweenness centrality and eigenvector centrality indicators of nodes, identify key nodes and important paths in the network, and reveal the transmission mechanism between the main influencing factors, including the following steps.

[0043] For each node in the network, the number of directly connected edges is counted to obtain the degree centrality of the node.

[0044] The betweenness centrality index of a node is obtained by calculating the proportion of all shortest paths in the network that pass through the node.

[0045] Based on the eigenvalue decomposition of the adjacency matrix, the eigenvector centrality of the node is calculated.

[0046] Comprehensively compare degree centrality, betweenness centrality, and eigenvector centrality to identify key nodes in the network and analyze their characteristic attributes.

[0047] By analyzing the shortest paths and key edges in the network, important propagation paths in the network can be identified.

[0048] Based on the analysis results of key nodes and important paths, a map of the influencing transmission mechanism is drawn to clearly show the transmission relationship between the main influencing factors, including the transmission direction, intensity and timing characteristics.

[0049] Furthermore, step 4 includes the following steps.

[0050] The warning error objective function is constructed, and a comprehensive evaluation standard for measuring warning performance is formed by weighted summing of missed alarm rate and false alarm rate.

[0051] A forgetting factor mechanism is introduced to perform time-decay processing on historical data, so that new data samples have high reference value, while maintaining a reasonable inheritance of historical patterns and ensuring a dynamic balance of data influence.

[0052] Establish a constraint system for feature weights, including non-negative weight constraints, weight normalization constraints, and maximum step size limits for weight changes, to ensure the rationality and stability of weight adjustments.

[0053] An adaptive weight optimization mechanism is designed to optimize the objective function based on the gradient descent method, determine the direction and step size of weight adjustment by calculating partial derivatives, and automatically update the weight parameters when new data samples arrive.

[0054] Furthermore, step 5 includes the following steps.

[0055] Based on the dynamic weight results, the warning threshold optimization objective function is constructed, and the false alarm rate, missed alarm rate and warning timeliness indicators of different warning levels are comprehensively considered to provide optimization direction and evaluation criteria for the particle swarm algorithm.

[0056] Initialize the parameter configuration of the particle swarm algorithm, including the number of particles, number of iterations, learning factor and inertia weight, and assign an initial position and velocity to each particle, where the position represents the threshold combination of different warning levels to ensure that the initial particle swarm has good distribution and diversity.

[0057] A mapping relationship between weights and thresholds is established. By analyzing the impact of changes in characteristic indicator weights on warning thresholds, a mathematical model that can reflect the correlation between the two is constructed, providing a theoretical basis for the coordinated adjustment of thresholds.

[0058] Execute the particle swarm optimization process. In each iteration, the individual optimal position and the global optimal position are updated according to the fitness value of the particle. Combined with the mapping relationship between weight and threshold, the search direction and step size of the particle are dynamically adjusted to ensure that it can quickly converge to the optimal solution.

[0059] A coordinated adjustment mechanism for multi-level warning thresholds is implemented. When the weight of characteristic indicators changes, the thresholds of each warning level are automatically updated based on the established mapping relationship, while ensuring reasonable intervals between different levels.

[0060] Furthermore, step 6 includes the following steps.

[0061] Build a multi-dimensional early warning performance evaluation indicator system, covering early warning accuracy, early warning timeliness and resource consumption, and establish standardized calculation methods and evaluation standards.

[0062] Through data stream processing technology, various performance indicator data are continuously collected, dynamically evaluated in combination with sliding time windows, and performance warning thresholds are set. When indicators are abnormal, optimization suggestions are issued in a timely manner.

[0063] Based on 3D modeling technology, an accurate digital model of the protective device is constructed to accurately restore the geometric features and spatial relationships of the equipment. The operating status of the equipment is intuitively displayed through visual elements, and dynamic monitoring points are added at key locations to achieve real-time updating of status information.

[0064] The warning information is displayed in the form of floating labels or warning icons superimposed on the 3D model of the equipment, supporting multi-angle viewing and information filtering.

[0065] Compared with the prior art, the present application has the following beneficial effects.

[0066] This application monitors the deformation, tilt and vibration data of the protective device in real time, conducts comprehensive analysis in combination with environmental parameters, and uses dynamic weight adjustment and recursive neural networks to optimize the adjustment of multi-level warning thresholds, thereby improving the accuracy of status monitoring of safety protection devices in construction and the effectiveness of early warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 A schematic flow chart of a method for early warning of a state of a safety protection device for construction disclosed in an embodiment of the present application. DETAILED DESCRIPTION

[0068] In order to make the purpose, technical scheme and advantages of the implementation of the present invention clearer, the technical scheme in the embodiment of the present invention will be described in more detail below in conjunction with the drawings in the embodiment of the present invention. In the drawings, the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The described embodiments are part of the embodiments of the present invention, not all of the embodiments.

[0069] Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in the field without making any creative work shall fall within the scope of protection of the present invention.

[0070] The embodiments and directional terms described below with reference to the accompanying drawings are exemplary and intended to be used to explain the present invention, but should not be construed as limiting the present invention.

[0071] like Figure 1 As shown, a status warning method for a safety protection device for construction includes the following steps.

[0072] Step 1: monitor the deformation, tilt angle and vibration acceleration of the protective device in real time, and simultaneously collect environmental data of the construction site.

[0073] Step 2: pre-process the collected data, segment the data using the sliding time window method, extract the characteristic parameters in each time window in combination with the wavelet transform technology, and generate a standardized feature vector set.

[0074] Step 3: Calculate the dynamic correlation strength between features and use recursive neural network to analyze the temporal evolution characteristics between indicators.

[0075] Step 4: Adaptive adjustment of weights is achieved by minimizing the objective function of warning error.

[0076] Step 5: Based on the obtained dynamic weight results, a mapping relationship between weights and thresholds is established to achieve coordinated adjustment of multi-level warning thresholds.

[0077] Step 6: Evaluate the warning performance in real time and visually display the status of the protection device and warning information.

[0078] The core of step 1 is to monitor the key physical parameters of the protective device and the construction environment in real time. The deformation, tilt angle and vibration acceleration of the protective device are important indicators for evaluating its structural stability and stress conditions, and are directly related to whether the protective device can effectively perform its protective function. Deformation monitoring can reflect the degree of deformation of the safety protective device when it is subjected to external loads, the tilt angle can indicate whether its force is balanced, and the vibration acceleration provides a basis for revealing whether the safety protective device produces unsafe vibrations due to dynamic loads during construction. At the same time, environmental data (such as temperature, humidity, wind speed, etc.) can provide the impact of external conditions on the performance of the protective device. For example, extreme weather may cause excessive deformation or vibration of the structure. Therefore, the synchronous collection of environmental data provides additional information for a comprehensive assessment of the status of the protective device.

[0079] The data preprocessing in step 2 is the basis for ensuring the accuracy and effectiveness of subsequent analysis. The sliding time window method is used to segment the collected data, which can divide the data into several segments in chronological order, thereby realizing dynamic analysis of the data. Wavelet transform technology can effectively extract different frequency components in the signal, help identify small changes in protective devices in different time periods, and then more accurately analyze possible safety hazards. Through wavelet transform, not only can noise be removed, but also the sensitivity to abnormal situations can be enhanced. In addition, the generation of standardized feature vector sets unifies the dimensions of different feature data, avoids analysis errors caused by different data dimensions, and lays the foundation for subsequent machine learning analysis.

[0080] The technical highlight of step 3 is to capture the temporal evolution relationship between features through recursive neural networks (RNNs). RNNs are particularly suitable for processing time series data and can automatically identify temporal patterns in data, so they can effectively analyze the state evolution of protective devices. By calculating the dynamic correlation strength between features, the interaction between different physical quantities and their impact on the safety of the device can be revealed. For example, the tilt angle of the protective device and the vibration acceleration may show a certain temporal dependency, and RNNs can capture these nonlinear temporal characteristics, thereby improving the accuracy and reliability of the early warning system.

[0081] Step 4 optimizes the prediction performance of the system by minimizing the warning error. The design of the objective function can automatically adjust the weight of each feature according to the actual warning effect of historical data. Through iterative optimization, the influence of feature parameters can be flexibly adjusted according to actual conditions to reduce false alarm rate and missed alarm rate. For example, when an erroneous warning is issued, the objective function will adjust the corresponding weight to make the next round of prediction more accurate. The adaptive adjustment mechanism can maintain a high warning performance under different construction sites and different environmental conditions.

[0082] Step 5 establishes a mapping relationship between weights and warning thresholds based on the dynamic weight adjustment in the previous step. This means that the warning threshold can be adjusted dynamically at different construction stages or different safety levels to ensure timely response to potential dangers without overreacting. Through the coordinated adjustment of multi-level thresholds, adaptable and detailed safety protection can be provided at multiple levels. For example, at some key construction nodes (such as high-altitude operations, large-scale machinery operations, etc.), a lower warning threshold can be set, while in other regular operating environments, the threshold can be appropriately increased to reduce unnecessary warnings.

[0083] In step 6, the early warning performance is evaluated in real time and the status and early warning information of the protective device are displayed intuitively, so that construction personnel and safety management personnel can understand the current safety situation in a timely manner. This link displays the early warning results through a graphical interface or alarm system to help staff respond quickly. The real-time evaluation function can further optimize the early warning mechanism and continuously improve the system performance through the feedback mechanism. This measure not only improves the efficiency of safety management, but also enhances the safety awareness of on-site staff.

[0084] In general, this state warning method realizes dynamic monitoring and early warning of safety protection devices in construction through the integration of multiple advanced technologies, such as sliding time window method, wavelet transform, recursive neural network, etc. The technical measures of each step cooperate with each other, from data collection, preprocessing to feature analysis, to early warning model optimization and real-time feedback, forming a complete closed-loop mechanism. The application of this method can not only effectively improve the safety of the construction process and reduce the possibility of accidents, but also ensure that the device can maintain an efficient and reliable working state in a complex and changeable construction environment through adaptive adjustment and multi-level early warning mechanism. It has high practical value and technical prospects.

[0085] Furthermore, step 1 includes the following steps.

[0086] Strain sensor arrays are arranged at major stress points, deformation-sensitive areas and key connections to monitor the stress distribution and deformation state of the protective device in real time.

[0087] High-precision angle sensors are installed at the corner nodes and tilt-sensitive areas of the protective device. A two-axis or three-axis measurement solution is used to continuously monitor the spatial posture changes of the protective device. At the same time, a tilt warning threshold system is established to achieve real-time assessment of the stability of the protective device.

[0088] A triaxial acceleration sensor network is arranged on the key components of the protective device, and the vibration response characteristics of the protective device during the construction process are captured in real time through high sampling frequency data acquisition.

[0089] An environmental monitoring network, including temperature sensors, humidity sensors, wind speed sensors, and rainfall sensors, is deployed at the construction site to establish a correlation analysis mechanism between environmental parameters and structural responses and to evaluate the impact of environmental factors on the performance of protective devices.

[0090] Build a data collection and transmission network based on the Industrial Internet of Things, adopt a combination of wired and wireless communication methods to ensure the real-time and reliability of data transmission, and at the same time establish a data backup mechanism and network redundancy design to prevent data loss and communication interruption.

[0091] In summary, by arranging strain sensor arrays at the main stress points, deformation sensitive areas and key connections, the stress distribution and deformation state of the protective device can be monitored in real time; high-precision angle sensors are installed at corner nodes and tilt sensitive areas to continuously monitor spatial posture changes and establish a tilt warning threshold system; a three-axis acceleration sensor network is arranged on key components to capture vibration response characteristics in real time; an environmental monitoring network is deployed to evaluate the impact of environmental factors on the performance of protective devices; and a data acquisition and transmission network based on the industrial Internet of Things is built to ensure the real-time and reliability of data transmission. This series of technical measures jointly builds a comprehensive, efficient and reliable state monitoring and early warning system for construction safety protective devices. Each part conducts detailed monitoring and analysis on the safety, stability and reliability of the protective device. Through real-time data acquisition, dynamic evaluation and early warning mechanisms, it can effectively prevent the occurrence of safety accidents and ensure the life safety of construction workers and project quality.

[0092] Furthermore, the vibration response characteristics of the protective device during construction are expressed as follows: , where a total (t) represents the comprehensive acceleration signal of the protective device in all directions, which is the composite value of the three-axis acceleration signal and is used to measure the overall vibration response; a x (t) represents the acceleration signal of the protective device in the x-axis direction; a y (t) represents the acceleration signal of the protective device in the y-axis direction; a z (t) represents the acceleration signal of the protective device in the z-axis direction.

[0093] In summary, the integrated acceleration signal can fully reflect the vibration response of the protective device during the construction process. This signal can not only effectively capture the vibration changes in all directions, but also provide data support for vibration source positioning, dynamic load identification and structural safety assessment. With this formula, the status of the protective device can be analyzed in real time, and an early warning can be issued when the vibration exceeds the safety threshold, thereby providing effective protection for construction safety.

[0094] Furthermore, step 2 includes the following steps.

[0095] The collected raw data are tested for outliers and removed, noise filtering is performed, and data normalization is performed to ensure data consistency and comparability.

[0096] The processed data is segmented based on the sliding window method, and the window size and sliding step parameters are optimized according to the signal characteristics to ensure that the signal characteristics are accurately reflected.

[0097] The data sequence of each time window is subjected to wavelet transform to obtain the multi-scale time-frequency decomposition result of the signal, forming a complete time-frequency domain feature representation.

[0098] Energy features, statistical features and wavelet coefficients are extracted from the time-frequency domain feature representation to ensure that the features can fully reflect the essential characteristics of the data.

[0099] The extracted features are standardized to eliminate the influence of dimension, and the standardized features are combined into feature vectors to finally form a feature vector set.

[0100] In summary, step 2 provides high-quality data support for subsequent analysis and prediction by performing detailed preprocessing, segmentation, transformation and feature extraction on the collected data. The quality of the data is ensured by removing outliers and noise; the sliding time window method and wavelet transform effectively extract valuable time-frequency domain features from the signal; and feature standardization ensures the consistency between different features, laying a solid foundation for further pattern recognition and modeling. Finally, after this series of processing, the standardized feature vector set obtained can fully reflect the dynamic characteristics of the protective device and support subsequent early warning and risk assessment work.

[0101] Furthermore, the processed data is segmented based on the sliding window method, and the window size and sliding step parameters are optimized according to the signal characteristics to ensure that the signal characteristics are accurately reflected, including the following steps.

[0102] The window size W and sliding step size S are set according to the time domain characteristics and frequency characteristics of the signal.

[0103] Extract continuous windows from the data sequence according to the set window size W and step size S.

[0104] Signal feature extraction is performed on the data in each window to obtain effective signal features.

[0105] Establish a validity verification mechanism for window data and perform quality checks on the data in each window, including data integrity, outlier ratio, and signal-to-noise ratio indicators. When the window data quality does not meet the requirements, automatically adjust the window parameters.

[0106] A dynamic optimization mechanism for window parameters is designed to regularly adjust the window size and sliding step size by evaluating feature extraction effect, computing resource usage, and response time indicators.

[0107] In summary, by segmenting the data based on the sliding window method and optimizing the window size and sliding step parameters, accurate extraction of signal features is achieved. First, the window size and sliding step are set according to the time domain and frequency characteristics of the signal, and then continuous windows are extracted from the data sequence, and features are extracted from the data in each window. At the same time, a window data validity verification mechanism is established to check the data quality, and automatically adjust the window parameters if it does not meet the requirements. In addition, a dynamic optimization mechanism is designed to regularly adjust the window parameters according to indicators such as feature extraction effect, computing resource occupancy, and response time, so as to ensure the accuracy and efficiency of signal feature extraction, while optimizing the use of computing resources.

[0108] Furthermore, wavelet transform is performed on the data sequence of each time window to obtain a multi-scale time-frequency decomposition result of the signal, thereby forming a complete time-frequency domain feature representation, including the following steps.

[0109] For the data sequence in each time window, multi-scale decomposition is performed using the selected wavelet basis function to decompose the signal into approximate components and detail components of different frequency components.

[0110] Set the number of decomposition layers according to analysis requirements and obtain the wavelet coefficients corresponding to each decomposition layer.

[0111] The obtained wavelet coefficients are reorganized and arranged according to the two dimensions of time and frequency to construct a complete time-frequency feature representation matrix to achieve joint representation of the signal in the time and frequency domains.

[0112] In summary, by performing wavelet transform on the data sequence of each time window, the multi-scale time-frequency decomposition of the signal can effectively extract different frequency components in the signal, including both the stable trend of low frequency and the instantaneous change of high frequency. Wavelet transform decomposes the signal into approximate components and detail components, so that the time information and frequency information of the signal can be fully expressed at multiple scales. By setting the number of decomposition layers according to the analysis requirements and extracting the corresponding wavelet coefficients, the characteristics of the signal at different scales can be captured more accurately. Finally, these wavelet coefficients are reorganized and arranged according to the time and frequency dimensions to form a complete time-frequency domain feature representation matrix, realizing the time-frequency joint representation of the signal. This extraction and representation of time-frequency domain features can not only effectively improve the recognition accuracy of complex signals, but also provide rich information support for subsequent signal analysis, anomaly detection and early warning.

[0113] Furthermore, the energy characteristics are expressed as: , where E j is the energy of the j-th layer wavelet transform; W j (k) is the value of the kth wavelet coefficient in the jth wavelet transform; N is the number of signal sampling points after wavelet transform.

[0114] The statistical characteristics are expressed as follows: ,in, is the kurtosis of the j-th layer wavelet coefficient, indicating the sharpness of the data distribution; σ j is the variance of the j-th layer wavelet coefficient, indicating the discrete degree or fluctuation range of the data.

[0115] The wavelet coefficients are expressed as: , where x(n) is the original signal; is the wavelet basis function.

[0116] In summary, through the analysis of the above formulas, energy features, statistical features and wavelet coefficients provide rich information for the analysis of vibration signals. Energy features can reflect the intensity changes of signals, statistical features reveal the distribution pattern and fluctuation characteristics of signals, and wavelet coefficients capture the time-frequency characteristics of signals through multi-scale decomposition. The extraction of these features provides necessary support for subsequent pattern recognition, anomaly detection and early warning, helping to identify potential risks and hidden faults in the construction process, thereby ensuring the safety of protective devices.

[0117] Furthermore, step 3 includes the following steps.

[0118] Construct a multi-dimensional time series association network, take each characteristic parameter as a network node, determine the initial connection relationship between nodes, and form a basic network structure.

[0119] The dynamic correlation strength between features is calculated in different time periods, the degree of correlation between nodes is quantified, and the correlation strength is used as the weight of the network edge to construct a weighted directed graph.

[0120] Use graph theory methods to analyze network structure, calculate the degree centrality, betweenness centrality and eigenvector centrality indicators of nodes, identify key nodes and important paths in the network, and reveal the transmission mechanism between the main influencing factors.

[0121] A long short-term memory network model is constructed, and the key features identified in the time series association network are used as input. The nonlinear time series evolution characteristics between various indicators are captured through the deep learning ability of the recursive neural network.

[0122] In summary, step 3 effectively reveals the dynamic correlation between the characteristics of the protective device and the transmission mechanism of its impact on safety by constructing a multidimensional time series association network and combining graph theory analysis. By calculating the dynamic correlation strength between features and constructing a weighted graph, the relationship between features can be quantified, and then key features and important paths can be identified. Combined with the long short-term memory network model, complex nonlinear time series evolution characteristics can be captured, providing strong technical support for the status warning of the protective device. This method has important application value in real-time monitoring, risk prediction and structural safety assessment, and can provide more accurate safety protection for construction.

[0123] Furthermore, a multi-dimensional temporal association network is constructed, each characteristic parameter is used as a network node, the initial connection relationship between nodes is determined, and a basic network structure is formed, including the following steps.

[0124] The preprocessed standardized feature parameters are mapped to network nodes, and a unique identifier is assigned to each node to establish a bijective mapping relationship between the feature parameters and the network nodes.

[0125] Pearson correlation coefficients and mutual information values ​​were calculated between all pairs of nodes.

[0126] The connection threshold λ is set as the criterion for determining the association strength. When the association strength between two nodes exceeds λ, the corresponding connection relationship is recorded in the adjacency matrix.

[0127] The initial network topology is constructed based on the adjacency matrix, and the edge weights are set according to the correlation coefficient or mutual information value.

[0128] Perturbation analysis is performed in multiple time windows to evaluate the evolution of network structure over time and verify the temporal stability of the relationship between nodes.

[0129] In summary, by constructing a multidimensional temporal association network and taking each feature parameter as a network node, the association between each feature parameter can be effectively captured to form a dynamic network structure. By mapping the standardized feature parameters to network nodes and assigning a unique identifier to each node, the precise correspondence between nodes and features is ensured, which is convenient for subsequent analysis. After calculating the Pearson correlation coefficient and mutual information value between node pairs, a suitable connection threshold is set to ensure that only those nodes with significant associations are connected, thereby avoiding over-association and maintaining the simplicity of the network. By establishing an adjacency matrix and using the correlation coefficient or mutual information value as the weight of the edge, the topological structure of the network can truly reflect the strong and weak associations between features. Furthermore, evaluating the evolution of the network over time through perturbation analysis helps to understand the temporal stability of the association between nodes and provides deep insights at the temporal level for anomaly detection and prediction.

[0130] Furthermore, the degree of association between nodes is expressed as: , where W(X, Y, μ) represents the weighted dynamic correlation strength of feature X and feature Y in time period μ; T is the total length of the time period; α is the attenuation factor, which controls the influence of data in a more distant time period on the current correlation strength; a T-τ is the dynamic adjustment of the attenuation factor α at time τ, where: when τ is close to T, a T-τ will be close to 1, which means that the data at the recent time point has a greater impact on the current correlation strength; when τ is far away from T, a T-τ will be close to 0, which means that the data at a distant time point has less influence on the current correlation strength; r(X τ , Y τ ) is the Pearson correlation coefficient calculated over the time period τ.

[0131] In summary, the weighted dynamic correlation strength between nodes provides a flexible and efficient method to quantify the temporal relationship between features. By introducing a decay factor to reduce the weight of data in distant time periods, the impact of data in different periods on the current correlation strength can be more accurately controlled, which is especially important in the dynamic analysis of time series data. Combined with the Pearson correlation coefficient, the linear dependency between features can be effectively quantified, thereby identifying key features and influencing paths. This method not only improves the timeliness and accuracy of the protective device status warning, but also provides a theoretical basis for more complex time series data analysis.

[0132] Furthermore, graph theory methods are used to analyze the network structure, calculate the degree centrality, betweenness centrality and eigenvector centrality indicators of nodes, identify key nodes and important paths in the network, and reveal the transmission mechanism between the main influencing factors, including the following steps.

[0133] For each node in the network, the number of directly connected edges is counted to obtain the degree centrality of the node.

[0134] The betweenness centrality index of a node is obtained by calculating the proportion of all shortest paths in the network that pass through the node.

[0135] Based on the eigenvalue decomposition of the adjacency matrix, the eigenvector centrality of the node is calculated.

[0136] Comprehensively compare degree centrality, betweenness centrality, and eigenvector centrality to identify key nodes in the network and analyze their characteristic attributes.

[0137] By analyzing the shortest paths and key edges in the network, important propagation paths in the network can be identified.

[0138] Based on the analysis results of key nodes and important paths, a map of the influencing transmission mechanism is drawn to clearly show the transmission relationship between the main influencing factors, including the transmission direction, intensity and timing characteristics.

[0139] In summary, by using graph theory to analyze network structure, we can deeply identify key nodes and important paths in the network and reveal the transmission mechanism between the main influencing factors. By calculating the degree centrality, betweenness centrality and eigenvector centrality of the node, we can comprehensively evaluate the importance of the node in the network and its impact on the overall structure. Degree centrality reflects the direct influence of the node in the network by counting the number of edges directly connected to the node; betweenness centrality reveals the importance of the node as an information dissemination intermediary by calculating the participation of the node in the shortest path; eigenvector centrality measures the global importance of the node in the network through the eigenvalue decomposition of the adjacency matrix. The combination of these indicators enables the key nodes in the network to be accurately identified, providing data support for further analysis. In addition, analyzing the shortest path and key edges helps to identify the most important propagation path in the network, thereby revealing the transmission mechanism between the influencing factors. Finally, by drawing a map of the influencing transmission mechanism, the direction, intensity and timing characteristics of the transmission can be clearly displayed, providing a strong basis for network optimization, abnormal warning and information dissemination management.

[0140] Furthermore, step 4 includes the following steps.

[0141] The warning error objective function is constructed, and a comprehensive evaluation standard for measuring warning performance is formed by weighted summing of missed alarm rate and false alarm rate.

[0142] A forgetting factor mechanism is introduced to perform time-decay processing on historical data, so that new data samples have high reference value, while maintaining a reasonable inheritance of historical patterns and ensuring a dynamic balance of data influence.

[0143] Establish a constraint system for feature weights, including non-negative weight constraints, weight normalization constraints, and maximum step size limits for weight changes, to ensure the rationality and stability of weight adjustments.

[0144] An adaptive weight optimization mechanism is designed to optimize the objective function based on the gradient descent method, determine the direction and step size of weight adjustment by calculating partial derivatives, and automatically update the weight parameters when new data samples arrive.

[0145] In summary, step 4 effectively optimizes the warning performance by constructing the warning error objective function and introducing the forgetting factor mechanism. The objective function of weighted false alarm rate and false negative rate provides a comprehensive evaluation criterion, which helps to find the best balance between different error types. The introduction of the forgetting factor ensures that the impact of new data on the model is more prominent, while maintaining the reasonable inheritance of historical patterns, ensuring the adaptability and stability of the model. The constraint system of feature weights and the adaptive weight optimization mechanism further enhance the robustness and self-learning ability of the model, so that it can respond to changes in real time and optimize the prediction results.

[0146] Furthermore, step 5 includes the following steps.

[0147] Based on the dynamic weight results, the warning threshold optimization objective function is constructed, and the false alarm rate, missed alarm rate and warning timeliness indicators of different warning levels are comprehensively considered to provide optimization direction and evaluation criteria for the particle swarm algorithm.

[0148] Initialize the parameter configuration of the particle swarm algorithm, including the number of particles, number of iterations, learning factor and inertia weight, and assign an initial position and velocity to each particle, where the position represents the threshold combination of different warning levels to ensure that the initial particle swarm has good distribution and diversity.

[0149] A mapping relationship between weights and thresholds is established. By analyzing the impact of changes in characteristic indicator weights on warning thresholds, a mathematical model that can reflect the correlation between the two is constructed, providing a theoretical basis for the coordinated adjustment of thresholds.

[0150] Execute the particle swarm optimization process. In each iteration, the individual optimal position and the global optimal position are updated according to the fitness value of the particle. Combined with the mapping relationship between weight and threshold, the search direction and step size of the particle are dynamically adjusted to ensure that it can quickly converge to the optimal solution.

[0151] A coordinated adjustment mechanism for multi-level warning thresholds is implemented. When the weight of characteristic indicators changes, the thresholds of each warning level are automatically updated based on the established mapping relationship, while ensuring reasonable intervals between different levels.

[0152] In summary, step 5 constructs the objective function of warning threshold optimization, initializes the parameters of the particle swarm optimization algorithm, and successfully implements the coordinated adjustment mechanism of multi-level warning thresholds through the mapping relationship between weights and thresholds and the execution of the particle swarm optimization process. This process can ensure that the warning threshold is automatically updated when the weights of different features change, thereby improving the response speed and accuracy. Through the application of the particle swarm optimization algorithm, the optimal combination of warning thresholds can be efficiently searched, and the state monitoring and warning capabilities of the protective device can be further improved, thereby enhancing the level of safety protection for construction.

[0153] Furthermore, step 6 includes the following steps.

[0154] Build a multi-dimensional early warning performance evaluation indicator system, covering early warning accuracy, early warning timeliness and resource consumption, and establish standardized calculation methods and evaluation standards.

[0155] Through data stream processing technology, various performance indicator data are continuously collected, dynamically evaluated in combination with sliding time windows, and performance warning thresholds are set. When indicators are abnormal, optimization suggestions are issued in a timely manner.

[0156] Based on 3D modeling technology, an accurate digital model of the protective device is constructed to accurately restore the geometric features and spatial relationships of the equipment. The operating status of the equipment is intuitively displayed through visual elements, and dynamic monitoring points are added at key locations to achieve real-time updating of status information.

[0157] The warning information is displayed in the form of floating labels or warning icons superimposed on the 3D model of the equipment, supporting multi-angle viewing and information filtering.

[0158] In summary, step 6 provides comprehensive technical support for the status monitoring and early warning of protective devices by building a multi-dimensional early warning performance evaluation index system, real-time data stream processing and dynamic evaluation, accurate three-dimensional modeling technology, and intuitive early warning information display. The application of these technologies not only improves the accuracy and timeliness of early warnings, but also optimizes resource consumption and enhances the visual management of equipment. By integrating these technologies, construction site personnel can monitor and manage protective devices more efficiently and accurately, ensuring the maximum safety and efficiency of construction.

[0159] Finally, it should be pointed out that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them. Although the present invention has been described in detail with reference to the above embodiments, a person skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A state warning method for a safety protection device for construction, characterized in that: The following steps are involved: Step 1: monitor the deformation, tilt angle and vibration acceleration of the protective device in real time, and simultaneously collect environmental data of the construction site; Step 2: preprocess the collected data, segment the data using a sliding time window method, extract the characteristic parameters in each time window using wavelet transform technology, and generate a standardized feature vector set; Step 3: Calculate the dynamic correlation strength between features and use recursive neural network to analyze the temporal evolution characteristics between indicators; Step 4, adaptively adjust the weights by minimizing the objective function of the warning error; Step 5: Based on the obtained dynamic weight results, a mapping relationship between weights and thresholds is established to achieve coordinated adjustment of multi-level warning thresholds; Step 6: Evaluate the warning performance in real time and visually display the status of the protection device and warning information.

2. A state warning method for a safety protection device for construction according to claim 1, characterized in that: Step 1 includes the following steps: Arrange strain sensor arrays at major stress points, deformation-sensitive areas and key connections to monitor the stress distribution and deformation state of the protective device in real time; Install high-precision angle sensors at the corner nodes and tilt-sensitive areas of the protective device, use a two-axis or three-axis measurement solution to continuously monitor the spatial posture changes of the protective device, and establish a tilt warning threshold system to achieve real-time evaluation of the stability of the protective device; A triaxial acceleration sensor network is arranged on the key components of the protective device to capture the vibration response characteristics of the protective device during the construction process in real time through high sampling frequency data acquisition; Deploy an environmental monitoring network at the construction site, including temperature sensors, humidity sensors, wind speed sensors, and rainfall sensors, establish a correlation analysis mechanism between environmental parameters and structural responses, and evaluate the impact of environmental factors on the performance of protective devices; Build a data collection and transmission network based on the Industrial Internet of Things, adopt a combination of wired and wireless communication methods to ensure the real-time and reliability of data transmission, and at the same time establish a data backup mechanism and network redundancy design to prevent data loss and communication interruption.

3. A state warning method for a safety protection device for construction according to claim 1, characterized in that: Step 2 includes the following steps: The collected raw data is tested for outliers and removed, noise filtering is performed, and data normalization is performed to ensure data consistency and comparability; The processed data is segmented based on the sliding window method, and the window size and sliding step parameters are optimized according to the signal characteristics to ensure that the signal characteristics are accurately reflected; Perform wavelet transform on the data sequence of each time window to obtain the multi-scale time-frequency decomposition result of the signal and form a complete time-frequency domain feature representation; Extract energy features, statistical features and wavelet coefficients from the time-frequency domain feature representation to ensure that the features can fully reflect the essential characteristics of the data; The extracted features are standardized to eliminate the influence of dimension, and the standardized features are combined into feature vectors to finally form a feature vector set.

4. A state warning method for a safety protection device for construction according to claim 1, characterized in that: Step 3 includes the following steps: Construct a multi-dimensional temporal association network, take each characteristic parameter as a network node, determine the initial connection relationship between nodes, and form a basic network structure; Calculate the dynamic correlation strength between features in different time periods, quantify the degree of correlation between nodes, and use the correlation strength as the weight of the network edge to construct a weighted directed graph; Use graph theory to analyze network structure, calculate node degree centrality, betweenness centrality and eigenvector centrality, identify key nodes and important paths in the network, and reveal the transmission mechanism between major influencing factors; A long short-term memory network model is constructed, and the key features identified in the time series association network are used as input. The nonlinear time series evolution characteristics between various indicators are captured through the deep learning ability of the recursive neural network.

5. A state warning method for a safety protection device for construction according to claim 4, characterized in that: Construct a multi-dimensional temporal association network, take each characteristic parameter as a network node, determine the initial connection relationship between nodes, and form a basic network structure, including the following steps: Map the preprocessed standardized feature parameters to network nodes, assign a unique identifier to each node, and establish a bijective mapping relationship between the feature parameters and the network nodes; Calculate the Pearson correlation coefficient and mutual information value between all node pairs; Set the connection threshold λ as the criterion for determining the association strength. When the association strength between two nodes exceeds λ, the corresponding connection relationship is recorded in the adjacency matrix. The initial network topology is constructed based on the adjacency matrix, and the edge weights are set according to the correlation coefficient or mutual information value; Perturbation analysis is performed in multiple time windows to evaluate the evolution of network structure over time and verify the temporal stability of the relationship between nodes.

6. A state warning method for a safety protection device for construction according to claim 4, characterized in that: Use graph theory to analyze the network structure, calculate the degree centrality, betweenness centrality and eigenvector centrality of nodes, identify key nodes and important paths in the network, and reveal the transmission mechanism between the main influencing factors, including the following steps: For each node in the network, count the number of directly connected edges to obtain the degree centrality of the node; The betweenness centrality index of a node is obtained by calculating the proportion of all shortest paths in the network that pass through the node; Based on the eigenvalue decomposition of the adjacency matrix, the eigenvector centrality of the node is calculated; Comprehensively compare degree centrality, betweenness centrality and eigenvector centrality to identify key nodes in the network and analyze their characteristic attributes; By analyzing the shortest paths and key edges in the network, important propagation paths in the network can be identified; Based on the analysis results of key nodes and important paths, a map of the influencing transmission mechanism is drawn to clearly show the transmission relationship between the main influencing factors, including the transmission direction, intensity and timing characteristics.

7. The state warning method of a safety protection device for construction according to claim 1 is characterized in that: Step 4 includes the following steps: Construct an early warning error objective function, and form an evaluation standard for comprehensively measuring early warning performance by weighted summing of missed alarm rate and false alarm rate; Introducing the forgetting factor mechanism to perform time decay processing on historical data, so that new data samples have high reference value, while maintaining reasonable inheritance of historical patterns and ensuring a dynamic balance of data influence; Establish a constraint system for feature weights, including non-negative weight constraints, weight normalization constraints, and maximum step size limits for weight changes to ensure the rationality and stability of weight adjustments; An adaptive weight optimization mechanism is designed to optimize the objective function based on the gradient descent method, determine the direction and step size of weight adjustment by calculating partial derivatives, and automatically update the weight parameters when new data samples arrive.

8. The state warning method of a safety protection device for construction according to claim 1 is characterized in that: Step 5 includes the following steps: Based on the dynamic weight results, the warning threshold optimization objective function is constructed, and the false alarm rate, missed alarm rate and warning timeliness index of different warning levels are comprehensively considered to provide optimization direction and evaluation criteria for the particle swarm algorithm; Initialize the parameter configuration of the particle swarm algorithm, including the number of particles, number of iterations, learning factor and inertia weight, and assign an initial position and velocity to each particle, where the position represents the combination of thresholds of different warning levels to ensure that the initial particle swarm has good distribution and diversity; Establish a mapping relationship between weights and thresholds, and by analyzing the impact of changes in characteristic indicator weights on warning thresholds, build a mathematical model that can reflect the correlation between the two, providing a theoretical basis for the coordinated adjustment of thresholds; Execute the particle swarm optimization process. In each iteration, the individual optimal position and the global optimal position are updated according to the fitness value of the particle. Combined with the mapping relationship between weight and threshold, the search direction and step size of the particle are dynamically adjusted to ensure that it can quickly converge to the optimal solution. A coordinated adjustment mechanism for multi-level warning thresholds is implemented. When the weight of characteristic indicators changes, the thresholds of each warning level are automatically updated based on the established mapping relationship, while ensuring reasonable intervals between different levels.

9. The state warning method of a safety protection device for construction according to claim 1 is characterized in that: Step 6 includes the following steps: Construct a multi-dimensional early warning performance evaluation indicator system, covering early warning accuracy, early warning timeliness and resource consumption, and establish standardized calculation methods and evaluation standards; Through data stream processing technology, various performance indicator data are continuously collected, and dynamic evaluation is performed in combination with sliding time windows. Performance warning thresholds are set, and optimization suggestions are issued in a timely manner when indicators are abnormal. Based on 3D modeling technology, an accurate digital model of the protective device is constructed to accurately restore the geometric features and spatial relationships of the equipment. The operating status of the equipment is intuitively displayed through visual elements, and dynamic monitoring points are added to key locations to achieve real-time updates of status information. The warning information is displayed in the form of floating labels or warning icons superimposed on the 3D model of the equipment, supporting multi-angle viewing and information filtering.

Images