A high formwork system whole-process critical point early warning monitoring system and construction method

Abstract

The application relates to a high-formwork-system whole-process critical-point early warning monitoring system and a construction method, wherein the early warning monitoring system comprises a sensor network, an Internet of Things gateway and an early warning monitoring platform; the early warning monitoring platform comprises a dynamic signal intelligent analysis module, a time sequence cumulative damage tracing module, a space coupling deduction module and a risk hub positioning early warning module; the dynamic signal intelligent analysis module is used for identifying abnormal stress points and taking the corresponding time of the abnormal stress points as a time anchor point for load history backtracking; the time sequence cumulative damage tracing module is used for determining cumulative fatigue damage degree, residual service life and damage nodes according to the abnormal stress points and load history data thereof; the space coupling deduction module is used for reducing connection rigidity weight according to the damage nodes and the cumulative fatigue damage degree, and constructing a damage topology network; and the risk hub positioning early warning module is used for determining overall instability probability, instability sensitivity index and a risk hub according to the damage topology network.

Description

Technical Field

[0001] This application relates to the field of building engineering safety monitoring technology, specifically to a critical point early warning monitoring system and construction method for a high formwork system throughout the entire process. Background Technology

[0002] High-formwork systems are widely used in building construction, especially in the construction of large high-rise buildings and complex structures. Due to their high load-bearing capacity and structural stability, they have become an important part of modern building construction. However, the stability and safety of the support system are crucial during the construction of high-formwork systems. If the support system becomes unstable or deformed, it may lead to serious safety accidents. To improve the safety of the construction process, it is particularly important to promptly identify potential risk factors and conduct full-process monitoring and early warning. During the construction and load-bearing process of high-formwork systems, there is a problem of sudden and hidden increases in local stress caused by uneven dynamic load distribution during concrete pouring. Traditional monitoring methods are difficult to effectively capture such micro-change signals. On this basis, the identified stress anomalies can further trigger cumulative fatigue damage at the joint connections. The lack of quantitative ability for the cumulative effect of damage over time in related technologies leads to a significant lag in early warning of sudden joint failures. More seriously, when damage at multiple joints spreads spatially and generates coupling effects, the overall structure faces the risk of cascading overturning due to unclear instability paths. Existing systems cannot simulate this systemic collapse process under the interaction of multiple critical points. Summary of the Invention

[0003] The first aspect of this application provides a whole-process critical point early warning and monitoring system for high formwork systems. This whole-process critical point early warning and monitoring system for high formwork systems can form a whole-process risk identification chain, thereby improving the early warning of high formwork systems under complex construction loads.

[0004] The first aspect of this application provides a high-formwork system full-process critical point early warning monitoring system, which includes a sensor network, an Internet of Things gateway, and an early warning monitoring platform. The sensor network is used to collect nodal stress signals during the construction process of the high-formwork system and transmit them to the early warning monitoring platform via the Internet of Things gateway. The early warning and monitoring platform includes a dynamic signal analysis module, a time-series cumulative damage tracing module, a spatial coupling deduction module, and a risk hub location and early warning module. The dynamic signal analysis module identifies abnormal stress points and uses the corresponding time points as time anchors for load history tracing. The time-series cumulative damage tracing module determines the cumulative fatigue damage degree, remaining service life, and damage nodes based on the abnormal stress points and their historical load data. The spatial coupling deduction module reduces the connection stiffness weight based on the damage nodes and cumulative fatigue damage degree, and constructs a damage topology network. The risk hub location and early warning module determines the overall instability probability, instability sensitivity index, and risk hubs based on the damage topology network. Finally, the early warning and monitoring platform outputs early warning and handling results based on the risk hubs.

[0005] In one alternative embodiment, the dynamic signal analysis module includes a signal cleaning unit, a window feature construction unit, and an abnormal stress identification unit. The signal cleaning unit can perform adaptive Kalman filtering on the node stress signal to obtain clean stress time series data; the window feature construction unit can perform sliding window analysis on the clean stress time series data to extract the stress mean, standard deviation, rate of change, and deviation from the design benchmark model, forming a multi-dimensional feature vector; the abnormal stress identification unit can calculate the local outlier score based on the multi-dimensional feature vector. When the local outlier scores of multiple consecutive windows of the same node all exceed a preset threshold, the node is identified as an abnormal stress point. The abnormal stress point includes the node number, abnormal start time, duration, maximum local outlier score, current stress value, and stress change trend.

[0006] In one optional scheme, the dynamic signal analysis module performs the following steps during operation: By deploying a sensor network at key nodes of high formwork, the original stress signals of the nodes during the concrete pouring process are collected in real time. An adaptive Kalman filter algorithm is used to dynamically denoise and smooth the original stress signals, and clean stress time series data that characterizes the fluctuation of the real load are extracted. A sliding window analysis was performed on the clean stress time series data to extract the statistical features of the data in each window, including the mean stress, standard deviation, rate of change, and deviation from the benchmark model, and to construct a multidimensional feature vector sequence to characterize the local load state of the nodes. Based on the multidimensional feature vector sequence, the outlier score of each node in different time windows is calculated using the local outlier factor algorithm. According to the preset outlier threshold, abnormal data points in which local stress increases in a hidden manner in the spatiotemporal domain are dynamically identified and marked as abnormal stress points.

[0007] In one optional scheme, the time-series cumulative damage tracing module includes a load history tracing unit, a memory network damage modeling unit, and a damage threshold early warning unit. The load history backtracking unit uses the time t0 corresponding to the abnormal stress point as the time anchor point to extract stress time series data within a preset historical period and obtain the stress amplitude sequence, number of cycles, and loading sequence during concrete pouring. The memory network damage modeling unit includes a long short-term memory network encoder, an external memory matrix, and a regression prediction subnetwork. The long short-term memory network encoder can extract deep temporal features, the external memory matrix can retrieve similar load-damage evolution segments, and the regression prediction subnetwork can predict the damage increment based on the deep temporal features and similar load-damage evolution segments. The damage threshold early warning unit can calculate the cumulative fatigue damage degree D based on the damage increment and determine the damage state level and damage node.

[0008] In one alternative approach, the cumulative fatigue damage D is calculated according to the following formula: ; in, To accumulate fatigue damage, The initial cumulative damage level in history. As the current time anchor point, For the predicted length of the future time period, For the stress amplitude level corresponding to the same calculation step size Below, the theoretical number of cycles required for fatigue failure of the node material or connection structure. The variable is the integral variable, representing the time period during which the variable is predicted. The time of continuous change within; Predict the remaining time when the cumulative damage D reaches a preset critical value, and define it as the remaining service life RUL of the node. When D reaches 1, the corresponding node reaches the theoretical fatigue failure state. The remaining useful life (RUL) is calculated according to the following formula: ; ; in, This is the critical time point. To be at the critical time point The cumulative fatigue damage degree, whose value is equal to the critical damage threshold. , The critical damage threshold is when When this occurs, it means that the node has entered a high-risk state, triggering a high-level warning and intervention measures; When D is at the warning level or alarm level, the corresponding node will be marked as a damaged node, and a structural damage assessment report will be generated.

[0009] In one alternative embodiment, the spatial coupling inference module includes a topology network construction unit, a stiffness degradation update unit, and an instability path inference unit. The topology network construction unit can construct an initial topology network based on structural connection relationships and configure node state vectors for each vertex; the stiffness degradation update unit can reduce the connection stiffness weights of adjacent edges based on the cumulative fatigue damage of damaged nodes to form a damaged topology network; the instability path inference unit can transmit messages containing equivalent load, cumulative fatigue damage, and connection stiffness weights between adjacent vertices based on a graph neural network to update the equivalent load of each vertex and identify overloaded and failed nodes. The instability path inference unit starts from the initial damaged node and performs graph search on adjacent overloaded or failed nodes to obtain potential instability propagation paths.

[0010] In one alternative embodiment, the risk hub location and early warning module includes an instability probability simulation unit and a risk hub screening unit. The instability probability simulation unit can randomly sample construction live load, cumulative fatigue damage, and connection stiffness weights based on the damage topology network and potential instability transmission paths to generate multiple load-damage coupling scenarios. When the critical force transmission path fails and causes the high-support formwork system to become a geometrically variable system, the corresponding scenario is recorded as the overall instability sample. The risk hub screening unit calculates the instability sensitivity index based on the number of overloads or failures of each vertex or edge in the overall instability sample.

[0011] In one alternative approach, the overall instability probability... Calculate using the following formula: ; ; in, The total number of unstable samples. This represents the total number of Monte Carlo simulations. The 95% confidence interval indicates that there is a 95% probability that the true probability of instability will fall within this interval, quantifying the uncertainty in probability estimation caused by a finite number of random samplings; The two-tailed quantiles of a standard normal distribution at a 95% confidence level; The standard error of the estimated value reflects the sampling variability; The formula for calculating the instability sensitivity index is as follows: ; in, As an index of instability sensitivity, For components The number of instability scenarios resulting in failure, and the failure / overload determination: in a single instability simulation, such as the component... If the stress exceeds its reduced bearing capacity, or the displacement exceeds the allowable limit, then it is determined to be in this state.

[0012] This high-formwork system's full-process critical point early warning monitoring system collects nodal stress signals during the construction process of the high-formwork system through a sensor network. The dynamic signal intelligent analysis module identifies abnormal stress points, enabling timely detection of sudden increases in local stress caused by uneven dynamic load distribution during concrete pouring. Simultaneously, the time corresponding to the abnormal stress point is used as the time anchor point for load history backtracking. This allows the time-series cumulative damage tracing module to determine the cumulative fatigue damage degree, remaining service life, and damage nodes based on the abnormal stress point and its load history data. This transforms the latent fatigue damage caused by repeated loading, unloading, and local stress fluctuations at the node connection from a difficult-to-judge state into a quantifiable, gradeable, and predictable risk state, improving the problem of traditional monitoring methods' inability to timely identify the risk of sudden node failure.

[0013] Building upon this foundation, the spatial coupling simulation module reduces the connection stiffness weights based on damaged nodes and cumulative fatigue damage, constructing a damage topology network. This allows the impact of local damage on adjacent connections, force transmission paths, and overall stiffness distribution to be expressed at the spatial topology level. The risk hub location and early warning module then determines the overall instability probability, instability sensitivity index, and risk hubs based on the damage topology network. This elevates the system from single-point anomaly alarms to a comprehensive assessment of multi-node damage coupling, load redistribution, and systemic instability risks. This creates a complete risk identification chain, improving the timeliness, accuracy, and targeted response of early warnings for hidden stress surges, sudden node failures, and systemic collapse risks in high-formwork systems under complex construction loads.

[0014] The second aspect of this application provides a construction method for a high-formwork system throughout the entire process. This construction method employs the high-formwork system critical point early warning and monitoring system provided in the first aspect. The construction method includes the following steps: Prepare a special construction plan for the high formwork system and determine the arrangement of members and stress monitoring points; Complete foundation treatment, positioning and layout, and installation of the support system; Stress sensors are installed at stress monitoring points, and a correspondence between the sensors and node information is established. During the concrete pouring process, zonal and layered loading is carried out. An early warning monitoring platform is used to identify abnormal stress points, cumulative fatigue damage, damage nodes, damage topology network, potential instability transmission paths and risk hubs, and the pouring operation is adjusted according to the early warning results.

[0015] In one alternative approach, when the early warning monitoring platform outputs an early warning level signal, the pouring operation in the area affected by the corresponding damaged node is suspended or slowed down, the connection status of the support system is checked, and the pouring sequence or monitoring frequency is adjusted according to the remaining service life and the trend of damage changes. When an alarm level signal is output and the risk hub is located, pouring operations within the area affected by the risk hub shall be stopped, and temporary support reinforcement, load unloading, or area isolation measures shall be taken. After the handling is completed, data on the risk hub and adjacent nodes are collected again, and the cumulative fatigue damage, instability sensitivity index and overall instability probability are calculated. Construction is resumed when the above indicators return to the allowable range; otherwise, reinforcement or unloading measures are continued.

[0016] This construction method involves developing a specialized construction plan for the high-formwork system before construction, determining the arrangement of members and stress monitoring points, and completing foundation treatment, positioning and layout, support system installation, and establishing the correspondence between stress sensors and node information. This ensures that the actual stress nodes, spatial locations, and monitoring data of the high-formwork system form a traceable data foundation and spatial positioning foundation before construction begins. During concrete pouring, zoned and layered loading is used, and an early warning monitoring platform is employed to identify abnormal stress points, cumulative fatigue damage, damage nodes, damage topology networks, potential instability transmission paths, and risk hubs. This allows the application of pouring loads to be synchronized with structural condition monitoring, damage evolution assessment, and overall instability risk identification, thereby avoiding reliance solely on pre-construction static calculations or manual experience to determine the safety status of the high-formwork system.

[0017] Furthermore, the pouring operation can be adjusted based on the early warning results, enabling the monitoring and identification results to be transformed into a dynamic adjustment basis for the pouring sequence, pouring area, or loading rhythm. This allows for closed-loop control throughout the entire process, improving the proactive control capabilities against sudden increases in local loads, accumulation of joint damage, and multi-node coupled instability risks during the concrete pouring stage, and reducing the risk of instability accidents in high-formwork systems during construction.

[0018] It should be understood that the above general description and the following detailed description are merely exemplary and do not limit this application. Attached Figure Description

[0019] Figure 1 A schematic diagram of the workflow of the critical point early warning and monitoring system for the entire process of a high-support formwork system; Figure 2 This is a schematic diagram of the construction process for the entire construction method of a high-support formwork system.

[0020] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. Detailed Implementation

[0021] To better understand the technical solution of this application, the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0022] It should be understood that the described embodiments are merely some embodiments of this application, and not all embodiments. All other technical solutions obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0023] The terminology used in the embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. The singular forms “a,” “the,” and “the” used in the embodiments of this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0024] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0025] It should be noted that the directional terms such as "upper," "lower," "left," and "right" described in the embodiments of this application are used to describe the angles shown in the accompanying drawings and should not be construed as limiting the embodiments of this application. Furthermore, in the context, it should be understood that when it is mentioned that an element is connected "upper" or "lower" to another element, it can be directly connected to the other element "upper" or "lower," or indirectly connected to the other element "upper" or "lower" through an intermediate element.

[0026] like Figure 1-2 As shown, the first aspect of this application provides a critical point early warning monitoring system for the entire process of a high formwork system. This critical point early warning monitoring system for the entire process of a high formwork system mainly includes a sensor network, an Internet of Things (IoT) gateway, and an early warning monitoring platform. The sensor network is used to collect nodal stress signals during the construction process of the high formwork system and transmits them to the early warning monitoring platform via the IoT gateway. The early warning monitoring platform includes a dynamic signal intelligent analysis module, a time-series cumulative damage tracing module, a spatial coupling inference module, and a risk hub location early warning module.

[0027] The system comprises several modules: a dynamic signal analysis module for identifying abnormal stress points and using the corresponding time points as time anchors for load history tracing; a time-series cumulative damage tracing module for determining cumulative fatigue damage, remaining service life, and damage nodes based on abnormal stress points and their historical load data; a spatial coupling simulation module for reducing connection stiffness weights based on damage nodes and cumulative fatigue damage and constructing a damage topology network; a risk hub location and early warning module for determining the overall instability probability, instability sensitivity index, and risk hubs based on the damage topology network; and finally, an early warning monitoring platform outputs early warning and handling results based on the risk hubs.

[0028] like Figure 1-2 As shown, in one specific embodiment, the dynamic signal intelligent analysis module includes a signal cleaning unit, a window feature construction unit, and an abnormal stress identification unit. The signal cleaning unit can perform adaptive Kalman filtering on the node stress signal to obtain clean stress time series data. The window feature construction unit can perform sliding window analysis on the clean stress time series data to extract the stress mean, standard deviation, rate of change, and deviation from the design benchmark model, forming a multi-dimensional feature vector. The abnormal stress identification unit can calculate the local outlier score based on the multi-dimensional feature vector. When the local outlier scores of multiple consecutive windows of the same node all exceed a preset threshold, the node is identified as an abnormal stress point. The abnormal stress point includes the node number, abnormal start time, duration, maximum local outlier score, current stress value, and stress change trend.

[0029] Specifically, the dynamic signal analysis module uses adaptive Kalman filtering to dynamically denoise and smooth the nodal stress signals collected during the construction of high formwork, effectively separating environmental interference from actual load fluctuations, improving data reliability, obtaining cleaned nodal stress time series data, and performing multi-dimensional outlier analysis on the cleaned nodal stress time series data based on the local outlier factor algorithm to identify abnormal stress points with sudden increases in local stress concealment.

[0030] The execution steps of the dynamic signal analysis module during operation are as follows: By deploying a sensor network at key nodes of high formwork, the original stress signals of the nodes are collected in real time during the concrete pouring process. An adaptive Kalman filter algorithm is used to dynamically denoise and smooth the original stress signals, effectively separating environmental vibration and construction machinery interference, extracting clean stress time series data that characterizes the actual load fluctuation, significantly improving the signal-to-noise ratio of the original data, effectively filtering out construction and environmental noise, and ensuring the reliability of subsequent analysis data.

[0031] Sliding window analysis is performed on clean stress time series data to extract statistical features of the data within each window, including stress mean, standard deviation, rate of change, and deviation from the benchmark model. A multidimensional feature vector sequence is constructed to characterize the local load state of nodes, converting the time series data into structured features and quantifying the load state of nodes from multiple dimensions, providing a foundation for accurate anomaly detection.

[0032] Based on multidimensional feature vector sequences, the local outlier factor algorithm is used to calculate the outlier score of each node in different time windows. According to the preset outlier threshold, abnormal data points in which local stress increases in a hidden way in the spatiotemporal domain are dynamically identified and marked as abnormal stress points. The data points are then output to subsequent modules for in-depth source analysis, accurately capturing local stress increase points that are difficult to detect by traditional methods, and achieving precise positioning of hidden early risks.

[0033] The specific work involves: deploying vibrating wire stress sensors at key nodes of the high-support formwork system (mid-span of the main beam, top of the support columns, and intersections of longitudinal and transverse horizontal bars). The sensors have a range of 0-50 MPa, an accuracy of ±0.1%FS, and a sampling frequency of 10Hz. Raw stress signals are collected in real-time via a 4G / 5G IoT gateway and transmitted to the early warning monitoring platform. An adaptive Kalman filter algorithm is used for signal processing, with the process noise covariance Q set to 0.01 and the initial measurement noise covariance R set to 0.1, dynamically adjusted according to signal fluctuations. The filter window length is set to 50 sampling points to effectively suppress tower crane noise. Construction interferences such as operation and pump truck vibration within the 1-30Hz frequency range, as well as baseline drift within ±0.5MPa caused by temperature changes, were used to extract clean stress time-series data with a signal-to-noise ratio of over 20dB. Sliding analysis was performed on the clean stress data with a fixed time window of 120 seconds and a window overlap rate of 50%. Four core statistical features were calculated within each window: mean stress (reflecting the static load level), standard deviation (characterizing the intensity of fluctuation), rate of change (root mean square of the difference between adjacent sampling points), and deviation from the design benchmark model (the difference between the actual stress and the theoretical value was calculated using Euclidean distance).

[0034] Feature vector standardization employs the Z-score method, establishing a 32-dimensional feature vector sequence containing timestamps, node numbers, and spatial coordinates. A continuously updated feature database is maintained, storing the feature sequence from the most recent 72 hours as a dynamic benchmark for comparative analysis in subsequent anomaly detection. Anomaly detection is performed using the local outlier factor algorithm, with a neighborhood number of 20 (k=20) and standardized Mahalanobis distance as the distance metric. Outlier scores for each feature vector in the spatiotemporal dimension are calculated in real-time, with a score threshold of 2.5 (a score > 2.5 indicates an anomaly). Nodes with scores exceeding the threshold for three consecutive time windows are marked as anomalous stress points, and the following metadata is recorded: anomaly start time, duration, maximum outlier score, and spatial distribution pattern. Anomaly reports are automatically generated, including the anomaly point's location coordinates, current stress value, trend curve, and comparative data with neighboring nodes. This structured anomaly information is transmitted to subsequent modules for analysis via a standard data interface. In one specific embodiment, the time-series cumulative damage tracing module includes a load history backtracking unit, a memory network damage modeling unit, and a damage threshold early warning unit. The load history backtracking unit uses the time t0 corresponding to the abnormal stress point as the time anchor point, extracts stress time-series data within a preset historical period, and obtains the stress amplitude sequence, number of cycles, and loading sequence during concrete pouring. The memory network damage modeling unit includes a long short-term memory network encoder, an external memory matrix, and a regression prediction subnetwork. The long short-term memory network encoder can extract deep time-series features, the external memory matrix can retrieve similar load-damage evolution segments, and the regression prediction subnetwork can predict the damage degree increment based on the deep time-series features and similar load-damage evolution segments. The damage threshold early warning unit can calculate the cumulative fatigue damage degree D based on the damage degree increment and determine the damage state level and damage node.

[0035] The temporal cumulative damage tracing module combines memory enhancement networks with fatigue damage evolution mechanisms to quantify the temporal cumulative effect of identified abnormal stress points and assess the degree of cumulative fatigue damage and remaining service life at node connections in order to locate high-risk connection sites and mark them as damaged nodes.

[0036] Among them, the memory network damage modeling unit uses the memory reinforcement network to learn the damage accumulation law in the stress time series data of the nodes, dynamically models the nonlinear relationship between fatigue damage and load history, quantifies the cumulative fatigue damage state of abnormal stress points, predicts the remaining service life, and realizes the temporal visualization of damage evolution.

[0037] The execution steps of the load history backtracking unit and the memory network damage modeling unit during runtime are as follows: Using the abnormal stress points and their associated historical load data as input, the encoder part of the long short-term memory network is used to perform temporal encoding on the stress amplitude, number of cycles and loading sequence experienced by the nodes, extracting deep temporal features of the hidden damage accumulation law, effectively extracting long-term dependence and damage driving mode in the load sequence, laying the foundation for accurate modeling.

[0038] Deep temporal features are input into the memory network component of a memory-enhanced network for storage and association retrieval. Through an attention mechanism, damage evolution segments similar to the current load pattern in the historical data are dynamically weighted to establish a nonlinear mapping model between load history, current state, and future damage growth. This enables personalized damage evolution modeling, improving prediction accuracy and adaptability to different operating conditions. Then, based on the nonlinear mapping model, the cumulative fatigue damage degree of the nodes corresponding to abnormal stress points is calculated, and their remaining service life is extrapolated according to the principles of damage mechanics. The temporal cumulative effect of damage is quantitatively output and visualized in the form of damage degree curves and remaining service life indicators, intuitively presenting the node damage status and development trend, supporting graded early warning and decision-making.

[0039] The specific job duties are as follows: In the early warning and monitoring platform, after the dynamic signal analysis module identifies and outputs abnormal stress points, the input data undergoes standardized preprocessing. Using the occurrence time of the marked abnormal stress point as the time anchor, the stress time series data of that node over the past 24 hours is extracted backward as the load history input sequence. This sequence is constructed at a frequency of one data point per second to ensure a detailed characterization of the loading process. Subsequently, a pre-trained long short-term memory network encoder is used to encode this historical sequence. This encoder contains two layers of LSTM units, each with 64 hidden neurons. The activation functions of its input gate, forget gate, and output gate are Sigmoid and Tanh functions, respectively, to effectively capture the long-term dependencies in the load sequence. The encoding process encodes the stress amplitude sequence experienced by the node, the number of stress cycles (extracted from the sequence in real time using the rainflow counting method), and the temporal information of the loading sequence (loading, unloading, and load maintenance) into a fixed-length 128-dimensional deep temporal feature vector, condensing the damage driving patterns hidden in the load history.

[0040] After obtaining the deep temporal feature vector, it is input into a memory-enhanced network component for further analysis. This component contains a dynamically updated external memory matrix with a size of 500 rows × 128 columns, used to store typical load-damage evolution patterns learned in the past. When a new deep temporal feature vector is input, its similarity with all stored vectors in the memory matrix is ​​calculated using cosine similarity. An attention mechanism based on scaling dot product is used to dynamically retrieve the top K (K=10) memory segments with the highest similarity. The weights of the retrieved memory segments are normalized using Softmax, and their corresponding historical damage evolution trajectories are weighted and aggregated. Then, a regression sub-network with two fully connected layers (64 and 32 neurons respectively) is used to fuse the weighted aggregated historical information with the current deep temporal feature vector. Finally, a nonlinear mapping relationship model representing the relationship from the current state to future damage growth is output. Its parameters are fine-tuned with each memory retrieval and update, realizing accurate and personalized modeling of the damage accumulation law under specific load histories of nodes.

[0041] Based on the established personalized nonlinear mapping relationship model, damage state quantification calculation is performed. The model starts with the current depth time series characteristics and combines the current stress level to iteratively predict the damage increment at each time step in the next period (the next 8 hours). The Miner linear cumulative damage criterion is used as the basic calculation framework to accumulate the predicted damage increment and calculate the cumulative fatigue damage degree D from the historical initial state to the end of the predicted period. This value ranges from 0 to 1 (where D=1 represents theoretical failure). At the same time, based on the principle of damage mechanics, the remaining time corresponding to when the cumulative damage degree D reaches the critical value of 0.8 is predicted by interpolation extrapolation method and defined as the remaining service life of that node. Finally, the results of the entire calculation process are encapsulated and displayed in the form of digital panels and trend charts on the visualization interface of the early warning monitoring platform.

[0042] The formula for calculating cumulative fatigue damage is as follows: ; In the formula: The cumulative fatigue damage refers specifically to the period from the initial historical state to the end of the predicted future period. The cumulative value at that time. This indicates that the node has reached the theoretical fatigue failure state; The initial cumulative damage level in history represents the damage level at the current time anchor point. (i.e., the time when the abnormal stress point is identified) is the fatigue damage degree that the node has accumulated due to past load history; the sum of damage increments during the prediction period is calculated based on the Miner linear cumulative damage criterion. For the stress amplitude level corresponding to the same calculation step size Below, the theoretical number of cycles required for fatigue failure of the node material or connection structure; This is the current time anchor point, i.e., the specific moment when the abnormal stress point is identified and output by the dynamic signal analysis module; This represents the length of the future prediction period, set to 8 hours, meaning the model predicts from... arrive Internal damage evolution; The variable is the integral variable, representing the time period during which the variable is predicted. The time period of continuous change.

[0043] The formula for calculating the remaining service life of the early warning system is as follows: ; ; In the formula: To provide early warning of remaining service life, it refers to the time from the current moment. From the beginning to the node, the cumulative fatigue damage degree Increase to a preset critical damage threshold The estimated remaining time elapsed since the beginning of the month; The critical time point is the cumulative fatigue damage degree obtained from the personalized damage evolution model through interpolation extrapolation. Just reached the critical threshold The future moment; To be at the critical time point The cumulative fatigue damage degree, whose value is equal to the critical damage threshold. ; This is the critical damage threshold, set to 0.8. When this occurs, it means that the node has entered a high-risk state, and a high-level warning must be triggered and intervention measures must be taken.

[0044] In addition, the damage threshold early warning unit constructs a cumulative fatigue damage model based on the fatigue damage evolution mechanism, and combines it with a pre-set damage threshold to provide graded early warning for abnormal stress points that are close to the critical state, and locates high-risk connection parts, marking them as damage nodes, and providing early warning of the risk of sudden failure of nodes.

[0045] The damage threshold early warning unit executes the following steps during operation: Based on the material properties, structural details, and design safety factor of the high formwork joint connection, and combined with the cumulative fatigue damage model constructed based on the fatigue damage evolution mechanism, a three-level damage state level of attention, early warning, and alarm is defined. For each level, a damage threshold range corresponding to the calculated cumulative fatigue damage degree is set, realizing multi-level refined management of damage state and providing a clear quantitative basis for graded early warning.

[0046] The system compares the cumulative fatigue damage of the node under evaluation with the multi-level damage threshold range in real time. When the damage level enters the warning or alarm range, the node is determined to be in a high-risk state and a warning signal of the corresponding level is triggered, realizing automatic and real-time risk assessment and warning, which greatly shortens the response time.

[0047] For nodes that trigger warning signals or higher, their spatial location information is used to locate and mark them as damaged nodes. At the same time, a structural damage assessment report containing the location of the damaged node, the damage status level, and the remaining service life is output, providing a clear input target for subsequent spatial coupling analysis, accurately associating abstract damage with physical location, forming a structured data stream to support subsequent system analysis. The specific job duties are as follows: Based on the mechanical properties and safety standards of the joint connections in the high-support formwork system, a three-level damage threshold range for cumulative fatigue damage is set. Specifically, combining the fatigue limit of Q235 steel, the structural detail coefficient of the joint weld or fastener (taken as 0.7-0.9), and a design safety factor of 1.5, the damage evolution model based on the Miner criterion is used for calibration: the cumulative fatigue damage range [0, 0.3] is defined as the attention level, indicating that the joint is in a healthy state and the damage accumulation is within the elastic recovery range; the cumulative fatigue damage range (0.3, 0.8] is defined as the warning level, indicating that the joint has entered the stable development stage of fatigue damage and microcracks inside the material may begin to propagate; the cumulative fatigue damage range (0.8, 1.0] is defined as the alarm level, representing that the joint is nearing fatigue failure, the damage is accumulating rapidly, and the structural integrity is threatened.

[0048] Real-time damage status assessment and early warning triggering are performed. Each time the cumulative fatigue damage degree D of a node to be assessed is received, it is automatically compared with the preset three-level threshold range. When the D value enters the early warning range for the first time, a yellow early warning signal is immediately generated and issued, indicating that the node has entered a medium-risk state. If the D value further enters the alarm range, a red alarm signal is triggered, indicating that the node is on the verge of high-risk failure. All early warning signals are accompanied by a timestamp, a unique node number, and a precise cumulative fatigue damage degree value, and are simultaneously released through multiple channels such as platform audible and visual alarms and message push.

[0049] Nodes that trigger warning signals or higher are spatially located and their information is encapsulated. Using the spatial coordinates of the nodes (accurate to the centimeter level) pre-recorded when the sensors are deployed, the nodes are automatically highlighted in the 3D structural model and officially marked as damaged nodes. At the same time, a standardized structural damage assessment report is generated. The structural damage assessment report is required to include: the 3D coordinates of the damaged node, the level of damage status (warning or alarm), the calculated remaining service life of the warning, the current cumulative fatigue damage degree and historical change curves, and the structural damage assessment report is output in real time through a standardized JSON data interface.

[0050] In one specific embodiment, the spatial coupling deduction module includes a topology network construction unit, a stiffness degradation update unit, and an instability path deduction unit. The topology network construction unit can construct an initial topology network based on the structural connection relationship and configure node state vectors for each vertex. The stiffness degradation update unit can reduce the connection stiffness weights of adjacent edges based on the cumulative fatigue damage of damaged nodes to form a damaged topology network. The instability path deduction unit can transmit messages containing equivalent load, cumulative fatigue damage, and connection stiffness weights between adjacent vertices based on a graph neural network to update the equivalent load of each vertex and identify overloaded nodes and failed nodes. The instability path deduction unit starts from the initial damaged node and performs graph search on adjacent overloaded nodes or failed nodes to obtain potential instability propagation paths.

[0051] Specifically, the topology network building unit constructs a graph structure topology model based on the actual connection relationship of the high-support formwork structure, taking nodes as vertices and connection relationships as edges, integrating the damage state and load attributes of each node to form a digital spatial mapping of the structural system, providing a dynamic topology basis for load transmission and damage diffusion analysis.

[0052] The execution steps of the topology network building unit during operation are as follows: Based on the construction drawings and on-site measurement data of the high-support formwork system, the connection points between the vertical poles, horizontal poles, and scissor braces are abstracted as vertices in the graph structure, and the physical connection relationships of the components themselves are abstracted as edges. An initial topological network reflecting the true geometry and mechanical force transmission path of the system is established, realizing the accurate mapping of the physical structure to the digital model and providing a unified topological framework for subsequent analysis.

[0053] For each vertex in the initial topology network, its corresponding real-time attribute data is dynamically integrated. The real-time attribute data includes at least the real-time stress from monitoring, the calculated cumulative fatigue damage degree, the damage state marker, and the theoretical mechanical properties of the component, giving the model dynamic perception capabilities and enabling the node state to be updated synchronously with the actual stress and damage.

[0054] Based on integrated node attributes and connection relationships, a weighted directed graph with real-time updated node states is constructed as a digital spatial mapping model of the structural system, i.e., a graph structure topology model. The weights of the edges can be dynamically adjusted according to the connection stiffness, providing a basis for load redistribution simulation and supporting accurate load transfer simulation.

[0055] The specific work involves: based on the design and construction drawings of the high-support formwork system and the on-site total station measurement and layout data, performing topological abstraction on the physical structure. Key force transmission points, such as the connection points of the uprights with couplers, the connection points of the horizontal bars and the cross couplers of the uprights, and the intersection points of the scissor braces with the rotating couplers of the members, are abstracted as vertices in the diagram structure. Each vertex is assigned a globally unique ID number. Physical members such as uprights, horizontal bars, and scissor braces are abstracted as edges connecting their two vertices. This ensures that the generated initial topology network completely covers the entire force transmission path from the foundation to the top support and accurately reflects the geometric spatial relationship and node connection logic of the structure.

[0056] During network construction, the node types and connection methods are classified and labeled according to the "Safety Technical Standard for Socket-type Disc-lock Steel Pipe Scaffolding in Building Construction" (JGJ / T231-2021) to form a systematic topological skeleton. Based on the initial topological network, multi-source real-time and quasi-static attribute data are dynamically integrated for each vertex to construct a comprehensive state vector of the node. Real-time data is accessed through an IoT gateway, including: axial stress (range 0-50MPa, sampling frequency 10Hz) collected by a vibrating wire sensor, clean stress value after filtering, and cumulative fatigue damage degree and damage status markers (health, early warning, alarm) calculated and output by the damage tracing module. Quasi-static attributes are entered according to design documents and material testing reports, including: the component section characteristics corresponding to the vertex, the material elastic modulus (2.06×105MPa for Q235 steel), and the initial value of connection stiffness calibrated based on the fastener tightening torque (standard requirement 40-65N·m). All attributes are managed in a time-series database, supporting state backtracking and real-time updates by timestamp.

[0057] After integrating the attributes, a weighted directed graph with real-time updated node states is constructed as a digital twin model of the high-support formwork system. In this model, the direction of the edges is set according to the main force transmission direction, and their weights are defined as connection stiffness. The initial values ​​are calculated and determined based on the fastener type and tightening torque. During system operation, the weights are adaptively adjusted based on the dynamic attributes of the vertices. For example, when the cumulative fatigue damage degree D of a vertex enters the warning range, the connection stiffness of all edges associated with it will be reduced and updated according to the preset damage-stiffness degradation model, thus fully characterizing the stiffness distribution and time-varying characteristics of the system.

[0058] In addition, the stiffness degradation updating unit and the instability path deduction unit are configured to use a graph neural network to simulate the interaction of damaged nodes in the graph structure topological model, deduce the load redistribution process and the potential instability propagation chain, analyze the instability diffusion path under multi-node damage coupling, clarify the critical propagation relationship of systemic collapse, and break through the limitations of traditional single-node analysis.

[0059] The execution steps when the stiffness degradation updating unit and the instability path deduction unit work are as follows: Vertices marked as damaged nodes and their current damage states and damage state levels are taken as key inputs, injected into the graph structure topological model, and a damage topology network containing local weak points is initialized, so as to establish a dynamic network reflecting the real damage distribution in the system and lay a foundation for subsequent load redistribution simulation.

[0060] With the damage topology network as the input, the trained graph structure topological model is used to simulate the dynamic redistribution process of load transferring from damaged nodes to adjacent nodes under the action of continuous external load, and the information of neighbor nodes is aggregated through a message passing mechanism to iteratively update the equivalent load state of all nodes in the whole network, realizing real-time dynamic simulation and visualization of the internal load transfer path and stress state of the structure.

[0061] Analyze the chain reaction sequence of sharp load increase or node state failure occurring in the simulation process, identify the continuous key node chain caused by the interaction of multiple damaged nodes and leading to the failure of the whole structure, define this chain as a potential instability propagation path, evaluate the critical conditions for its formation, reveal the weak chain where the structure is most likely to suffer cascading failure in advance, and provide an accurate spatial positioning basis for targeted reinforcement.

[0062] The specific work content is as follows: After receiving the structural damage assessment report from the damage threshold early warning unit, locate the vertex matching the node number in the report from the graph structure topological model, update the state vector of this vertex, forcibly write its current damage state mark (early warning or alarm) and the quantified cumulative fatigue damage value, and formally change the attribute of this vertex in the digital twin model from the healthy or attention state to the damaged state. At the same time, according to the embedded damage-stiffness degradation model, dynamically reduce the connection stiffness weights of all edges associated with this vertex automatically. For example, for a node entering the early warning state (0.3<D≤0.8), its connection stiffness will be updated according to a linear or nonlinear reduction coefficient (for example, multiplied by a coefficient ranging from 0.6 to 0.9), so as to accurately characterize the bearing capacity attenuation of the connection part in the model. After completing the state and weight update of all identified damaged nodes, a damage topology network containing clear local weak points is initialized and generated.

[0063] Using the initialized damaged topology network as the computational graph, a pre-trained graph neural network model is invoked to simulate load redistribution. The simulation condition is set as the continuous application of a combination of dead load and construction live load as specified in the design code. In each iteration of the simulation, the GNN operates through a message passing mechanism: each vertex (node) sends a message to all its neighboring vertices. This message is a feature vector containing the sending node's current equivalent load state, damage degree, and connection stiffness weights. The receiving node aggregates messages from all its neighbors and calculates and updates its own equivalent load state by integrating its own state and neighbor information through a learnable nonlinear update function. This process is synchronized and iterative across all vertices in the network. Through multiple iterations, the load gradually shifts from weak nodes with high damage and low stiffness to relatively healthy neighboring nodes, thereby dynamically deducing the load redistribution result within the entire high-support formwork system under the condition of damage, and outputting the final equivalent load state of each node after the simulation stabilizes.

[0064] The entire process data of load redistribution simulation is scanned and analyzed. The analysis algorithm locates the vertex in the simulation iteration process where the equivalent load state growth rate exceeds a preset threshold or the final load value exceeds the material yield strength multiplied by a reduction factor. The vertex determined by the scan analysis is marked as an overload node or a failure node. Then, a graph search algorithm (depth-first search) is used to trace and reconstruct the causal sequence triggered by the initial damage node, which leads to the successive overload or failure of a series of neighboring nodes. This forms the longest critical node chain that is spatially continuous and may cause geometric variability of the overall structure. This chain is defined as a potential instability transmission path. At the same time, the critical conditions for the formation of this path are evaluated through parameter sensitivity analysis.

[0065] In one specific embodiment, the risk hub location and early warning module includes an instability probability simulation unit and a risk hub screening unit. The instability probability simulation unit can randomly sample construction live load, cumulative fatigue damage degree, and connection stiffness weights based on the damage topology network and potential instability transmission paths to generate multiple load-damage coupling scenarios. When the key force transmission path fails and causes the high-support formwork system to become a geometrically variable system, the corresponding scenario is recorded as the overall instability sample. The risk hub screening unit calculates the instability sensitivity index based on the number of overloads or failures of each vertex or edge in the overall instability sample.

[0066] Specifically, the execution steps of the entire risk hub location and early warning module are as follows: Based on the damage topology network and the deduced potential instability propagation path, the Monte Carlo method is used to simulate the overall structural instability process under multiple working conditions and multiple critical point interactions. The overall instability probability of the structure under different damage coupling scenarios is statistically calculated, the macroscopic risk level of the system under uncertain load and damage coupling is evaluated, and the overall safety margin is quantified.

[0067] By analyzing simulation results, the contribution of the instability sensitivity index of each node or connection edge in the topology network to the instability propagation in most instability scenarios is calculated. The key nodes and connections with the highest contribution, i.e., whose state changes have the most significant impact on the overall instability probability, are selected and marked as risk hubs. This allows for the precise identification of the transmission bottleneck of systemic risks and provides a clear objective for the efficient allocation of emergency resources.

[0068] For each identified risk hub, based on its type, damage mechanism, and location in the topology network, specific intervention measures are automatically generated and recommended. These measures include temporary reinforcement of the hub, load unloading, adjustment of pouring sequence, or increased monitoring, forming a precise prevention and control plan to curb the spread of systemic collapse. This achieves an intelligent closed loop from risk warning to targeted treatment, proactively cutting off the chain of instability spread. The specific job duties are as follows: Based on the established damage topology network and potential instability propagation paths, the Monte Carlo method is used to conduct multi-scenario random sampling simulation. Specifically, based on the current monitoring data, sampling is performed within the load condition parameter space: the construction live load fluctuates randomly according to a normal distribution within the allowable range of the specification (the mean is taken as the design value, and the coefficient of variation is set to 0.1); at the same time, random disturbances are introduced into the cumulative fatigue damage of all nodes in the network that are in the warning state or above, and the disturbance range is controlled within ±10% of the current measurement value to simulate the uncertainty of damage development. Each sampling constitutes a complete load-damage coupled scenario, and the integrated finite element calculation kernel is automatically called.

[0069] Over 5000 independent structural stability analyses were performed in batches using a GPU parallel computing architecture. Each analysis determined whether the structure had experienced overall instability (defined as the failure of a critical force transmission path leading to a geometrically variable system). The frequency of instability occurrences was statistically analyzed, and the overall instability probability of the structure in the current state was calculated. This probability value was presented as a percentage with a 95% confidence interval. Post-processing analysis was performed on all instability samples generated by the Monte Carlo simulation. The contribution of each node (vertices) and connecting edge (link) in the topology network was quantified by calculating its instability sensitivity index. This index is defined as the proportion of scenarios in which the element (node ​​or edge) is in a failed or overloaded state among all simulation scenarios that lead to instability. The elements with the top 5% sensitivity indices were selected and officially marked as risk hubs. Risk hubs are located at the intersection of potential instability transmission paths, critical nodes of load redistribution, or critical links connecting different stable substructures. Their common characteristic is that a small deterioration in their own state leads to a significant jump of more than 20% in the overall instability probability, indicating that risk hubs are the bottlenecks and key valves for the spread of systemic collapse.

[0070] Based on the type of risk hub, the current damage mechanism, and its structural location in the graph topology, specific measures are generated by matching and combining from a pre-set strategy library. For risk hubs at the top of poles that are in an alarm state, temporary reinforcement measures are recommended first. The specific parameters are as follows: additional vertical support is provided by using disc-lock steel pipes with a rated bearing capacity of not less than 60kN. The support points are arranged within 1 meter of the node and connected to at least three adjacent stable nodes. Load unloading is performed simultaneously. It is recommended to immediately stop concrete pouring operations in the area affected by the hub (within a radius of 3 meters) and to perform surface diversion and pressure reduction on the poured but not yet solidified concrete. All recommended measures are presented in the form of an operation instruction list, along with the three-dimensional coordinates of the risk hub, current monitoring data, and expected effect assessment, and are pushed to on-site technical personnel and safety managers through the early warning monitoring platform, forming a complete chain from risk early warning to closed-loop management.

[0071] The formula for calculating the overall instability probability is as follows: ; ; In the formula: The overall instability probability represents the estimated likelihood that the structure will experience overall instability at some point in the future or under specific load conditions, given the current monitored state of structural damage. The total number of instability samples refers to the number of simulation scenarios in which the structure is judged to have experienced overall instability in all Monte Carlo simulations. The judgment criterion (instability event) is: if the structure becomes a geometrically variable system due to the failure of the critical force transmission path in a simulation, it is counted as an instability event. This represents the total number of Monte Carlo simulations, i.e., the total number of independent structural stability analyses performed in batches by the system. The 95% confidence interval indicates that there is a 95% probability that the true probability of instability will fall within this interval, quantifying the uncertainty in probability estimation caused by a finite number of random samplings; The two-tailed quantiles of a standard normal distribution at a 95% confidence level; This represents the standard error of the estimate, reflecting sampling variability.

[0072] The formula for calculating the instability sensitivity index is as follows: ; In the formula: This is an instability sensitivity index used to quantify the instability sensitivity of the first instability layer in a topological network. The contribution or criticality of an individual component (node ​​or connecting edge) to the overall instability of the system; the higher the value, the more critical the role that component plays in the process of causing system instability. For components The number of instability scenarios that fail refers to the total number of failures. In a simulation sample that led to overall system instability, the components The number of samples that are in a state of failure or overload; failure / overload determination: in a single instability simulation, if the component... If the stress exceeds its reduced bearing capacity, or the displacement exceeds the allowable limit, then it is determined to be in this state.

[0073] The following is the specific working principle of the critical point early warning and monitoring system for the entire process of this high-support formwork system: First, a high-precision vibrating wire sensor network is deployed at key force transmission nodes such as the top of the pole and the main beam span to collect raw stress signals in real time at a frequency of 10Hz. Faced with the complex construction site environment, an adaptive Kalman filter algorithm with adjustable parameters is used to deeply clean the raw signals, filtering out mechanical vibration interference in the 1-30Hz frequency band and temperature drift of ±0.5MPa, thereby extracting clean time-series data that characterizes the actual load fluctuation. Then, a sliding analysis is performed on the data with a window of 120 seconds to extract four-dimensional statistical features such as mean and standard deviation, and a 32-dimensional feature vector containing spatiotemporal information is constructed. The outlier score of each feature vector is calculated by the local outlier factor algorithm, and nodes with scores exceeding the threshold of 2.5 for three consecutive windows are marked, accurately identifying abnormal stress points with sudden increases in local stress that are difficult to capture by traditional methods. Based on the accurate identification of anomalies, the system uses these anomalies as anchors to trace back their historical load data over the past 24 hours. A pre-trained two-layer LSTM encoder encodes these data into 128-dimensional deep temporal features, refining their historical load patterns. This data is then input into a memory-enhanced network containing a 500×128-dimensional external memory matrix. An attention mechanism retrieves the 10 most similar damage evolution segments from history, constructing a personalized load-damage nonlinear mapping model. Based on this model, the Miner criterion is used to iteratively predict the damage increment over the next 8 hours, accumulating a quantitative index – cumulative fatigue damage. According to material mechanics and safety standards, the damage state is divided into three levels: attention, warning, and alarm. When the cumulative fatigue damage of a node enters the warning or alarm range, a graded alarm (yellow / red) is immediately triggered, and the node is highlighted as a damaged node in the 3D model. Simultaneously, the remaining service life of the node is calculated through model extrapolation, generating a standardized assessment report containing coordinates, damage level, and remaining service life. This transforms isolated anomaly signals into precise quantification and prediction of node damage and remaining service life, achieving a fundamental shift in risk warning from post-event alerts to pre-event predictions.

[0074] This high-formwork system's full-process critical point early warning monitoring system collects nodal stress signals during the construction process of the high-formwork system through a sensor network. The dynamic signal intelligent analysis module identifies abnormal stress points, enabling timely detection of sudden increases in local stress caused by uneven dynamic load distribution during concrete pouring. Simultaneously, the time corresponding to the abnormal stress point is used as the time anchor point for load history backtracking. This allows the time-series cumulative damage tracing module to determine the cumulative fatigue damage degree, remaining service life, and damage nodes based on the abnormal stress point and its load history data. This transforms the latent fatigue damage caused by repeated loading, unloading, and local stress fluctuations at the node connection from a difficult-to-judge state into a quantifiable, gradeable, and predictable risk state, improving the problem of traditional monitoring methods' inability to timely identify the risk of sudden node failure.

[0075] Building upon this foundation, the spatial coupling simulation module reduces the connection stiffness weights based on damaged nodes and cumulative fatigue damage, constructing a damage topology network. This allows the impact of local damage on adjacent connections, force transmission paths, and overall stiffness distribution to be expressed at the spatial topology level. The risk hub location and early warning module then determines the overall instability probability, instability sensitivity index, and risk hubs based on the damage topology network. This elevates the system from single-point anomaly alarms to a comprehensive assessment of multi-node damage coupling, load redistribution, and systemic instability risks. This creates a complete risk identification chain, improving the timeliness, accuracy, and targeted response of early warnings for hidden stress surges, sudden node failures, and systemic collapse risks in high-formwork systems under complex construction loads.

[0076] The second aspect of this application provides a construction method for a high-formwork system throughout the entire process. This construction method employs the high-formwork system critical point early warning monitoring system provided in the first aspect. The construction method includes the following steps: Prepare a special construction plan for the high formwork system and determine the arrangement of members and stress monitoring points; Complete foundation treatment, positioning and layout, and installation of the support system; Stress sensors are installed at stress monitoring points, and a correspondence between the sensors and node information is established. During the concrete pouring process, zonal and layered loading is carried out. An early warning monitoring platform is used to identify abnormal stress points, cumulative fatigue damage, damage nodes, damage topology network, potential instability transmission paths and risk hubs, and the pouring operation is adjusted according to the early warning results.

[0077] When the early warning monitoring platform outputs an early warning level signal, the pouring operation in the area affected by the corresponding damaged node is suspended or slowed down. The connection status of the support system is checked, and the pouring sequence or monitoring frequency is adjusted according to the remaining service life and the trend of damage changes. When an alarm level signal is output and the risk hub is located, the pouring operation in the area affected by the risk hub is stopped, and temporary support reinforcement, load unloading, or area isolation measures are taken. After the handling is completed, the data of the risk hub and adjacent nodes are collected again, and the cumulative fatigue damage degree, instability sensitivity index, and overall instability probability are calculated. Construction is resumed when the above indicators return to the allowable range; otherwise, reinforcement or unloading measures are continued.

[0078] Specifically, the construction method for this high-formwork system includes the following steps: 1) Based on the engineering structural drawings and load requirements, complete the specialized construction scheme design for the high formwork system, determine the planar layout and vertical spacing of the uprights, horizontal bars, and scissor braces, and plan the key node locations for sensor placement; according to the building structure construction drawings and design load requirements, conduct a specialized construction scheme design for the high formwork support system, clarifying the longitudinal and transverse spacing of the uprights (covering 0.9m, 1.2m, and 1.5m), the horizontal bar spacing (not exceeding 1.5m), and the arrangement of the scissor braces (continuous vertically, with one set every 4-6 spans horizontally), ensuring that the overall stability calculation meets the requirements of the "Building Construction Standard". The "Technical Specification for Formwork Safety" (JGJ162) standard requires that, based on this, a sensor layout scheme for the monitoring system be planned. According to the mechanical analysis results, the key node locations for stress monitoring should be determined, including: the bottom of the mid-span section of the main beam and at 1 / 4 of the span; the top of the support column (where the axial force is the greatest) and at 1 / 3 of the height from the top; the cross-coupler nodes connecting the longitudinal and transverse horizontal bars to the column; and the swivel coupler nodes where the vertical scissor brace intersects with the horizontal bar. The sensor layout should form a complete coverage of the main force transmission path, and a detailed monitoring point layout diagram should be drawn as the basis for subsequent construction and installation.

[0079] 2) The foundation of the support system shall be compacted, leveled, and its bearing capacity tested. A total station shall be used for precise positioning and layout, marking the plane position and elevation control points of all uprights. Before erecting the support system, the foundation within the support area shall be treated. The backfilled foundation shall be compacted in layers with a compaction coefficient of not less than 0.94, and a hardened cushion layer (using C15 or higher strength grade concrete) with a thickness of not less than 50mm shall be laid. After the foundation treatment, static cone penetration test or plate load test shall be conducted to verify whether its bearing capacity characteristic value meets the requirements of the special plan. The requirements are as follows (not less than 150 kPa). Subsequently, using a total station with an accuracy of not less than 2″, based on the monitoring point layout diagram determined in step 1, the plane center position of each pole is accurately measured and marked on site. The points are fixed with conspicuous marks (intersection of ink lines or steel nails). At the same time, the structural elevation benchmark is established. Using the elevation measurement function of the total station, the elevation control line of the support system (the bottom elevation of the template minus the thickness of the template and keel) is marked on the adjacent stable structure to ensure that the final elevation error of the top of the support is controlled within ±5mm.

[0080] 3) Install the uprights, horizontal bars, and scissor braces layer by layer according to the layout points. Tighten the fasteners to the specified requirements of 40-65 N·m using a torque wrench to ensure that the joint connection stiffness meets the design requirements. Erect the support system according to the layout points. The uprights must be installed vertically, and their verticality deviation should not exceed 1 / 200 of the scaffold height. Ensure that the connecting sleeves of the upper and lower uprights are accurately inserted and firmly connected. The horizontal bars and scissor braces should be installed simultaneously. The horizontal bars should be set in both directions and locked to the uprights with cross couplers to form a stable horizontal structure. For scissor bracing, swivel couplers should be used to fix the bracing to the intersecting horizontal or vertical poles. The distance from the center of the coupler to the main node should not exceed 150mm. All couplers should be checked and inspected using a calibrated torque wrench to ensure that the tightening torque of right-angle couplers and swivel couplers meets the specification requirements of 40N·m to 65N·m. The anti-slip performance of butt couplers must meet the standards. During installation, check layer by layer and span by span to ensure that there are no missing members and no loose nodes, forming a spatial grid structure with design rigidity.

[0081] 4) Install vibrating wire stress sensors at key nodes, including the mid-span of the main beam and the top of the uprights, and connect them to the IoT gateway. Activate the early warning monitoring platform and perform initial calibration and communication debugging of the sensors. Simultaneously install vibrating wire stress sensors when the support system reaches the key node positions. Before installation, the sensors must be initially calibrated in the laboratory to determine their range (0-50MPa), sensitivity coefficient, and temperature compensation coefficient. During installation, for axial stress monitoring points on the members, the sensor should be tightly fitted to the centerline of the member surface using a special clamp to ensure that the force axis is aligned with the sensor's sensing direction. For the monitoring points at the bottom of the beam, surface-adhesive installation is adopted. All sensor cables are laid and fixed along the pole according to specifications and connected to the nearest waterproof intelligent data acquisition box. The data acquisition box uploads data to the early warning monitoring platform through a 4G / 5G industrial IoT gateway. After all sensors are installed, system integration testing is required: the mapping relationship between sensor number and spatial position coordinates is checked one by one on the platform, zero-point acquisition and initial state setting are performed, and the stability and real-time performance of data transmission are tested to ensure that the sampling frequency (set to 1-10Hz) and data packet reception are normal. The system then enters standby monitoring state.

[0082] 5) During concrete pouring, loading is carried out in zones and layers according to a predetermined sequence. Stress changes and damage status are monitored in real time through an early warning monitoring platform, and the construction process is dynamically adjusted based on system warning commands. Concrete pouring is conducted in the sequence determined by the specific plan, using a symmetrical and uniform layered pouring method from the mid-span to both ends. The thickness of each layer is controlled at 300-500mm. Throughout the pouring process, the early warning monitoring platform receives and processes stress data from each node in real time. Construction management personnel closely monitor the platform's visual interface, paying attention to the real-time stress curve and the damage status prompts automatically issued by the system. When the platform issues a yellow warning... When a color-coded warning signal is issued, it indicates that a specific node has entered a warning state. On-site pouring in the affected area of ​​that node should be suspended, the support condition should be checked, and the technical supervisor should assess whether to slow down the pouring speed or adjust the pouring sequence. When the platform issues a red alarm signal and locates the risk hub, all pouring operations in the affected area (within a radius of 3 meters) of the alarm point should be stopped immediately, and emergency measures should be implemented according to the targeted handling instructions pushed by the platform (temporary support reinforcement or unloading). Through real-time linkage between monitoring data and construction actions, active and dynamic control of construction loads can be achieved, ensuring the safety of the high formwork system throughout the entire load-bearing process.

[0083] The construction method in this embodiment involves developing a specialized construction plan for the high-formwork system before construction, determining the arrangement of members and stress monitoring points, and completing foundation treatment, positioning and layout, support system installation, and establishing the correspondence between stress sensors and node information. This ensures that the actual stress nodes, spatial locations, and monitoring data of the high-formwork system can form a traceable data foundation and spatial positioning foundation before construction begins. During concrete pouring, zoned and layered loading is used, and an early warning monitoring platform is employed to identify abnormal stress points, cumulative fatigue damage, damage nodes, damage topology networks, potential instability transmission paths, and risk hubs. This allows the application of pouring loads to be synchronized with structural state monitoring, damage evolution judgment, and overall instability risk identification, thereby avoiding reliance solely on pre-construction static calculations or manual experience to determine the safety status of the high-formwork system.

[0084] Furthermore, the pouring operation can be adjusted based on the early warning results, enabling the monitoring and identification results to be transformed into a dynamic adjustment basis for the pouring sequence, pouring area, or loading rhythm. This allows for closed-loop control throughout the entire process, improving the proactive control capabilities against sudden increases in local loads, accumulation of joint damage, and multi-node coupled instability risks during the concrete pouring stage, and reducing the risk of instability accidents in high-formwork systems during construction.

[0085] The above are merely preferred embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A critical point early warning and monitoring system for a high-support formwork system throughout its entire process, characterized in that, It includes a sensor network, an Internet of Things (IoT) gateway, and an early warning and monitoring platform. The sensor network is used to collect nodal stress signals during the construction process of the high formwork system and transmits them to the early warning and monitoring platform via the IoT gateway. The early warning and monitoring platform includes a dynamic signal analysis module, a time-series cumulative damage tracing module, a spatial coupling deduction module, and a risk hub location and early warning module. The dynamic signal analysis module identifies abnormal stress points and uses the corresponding time points as time anchors for load history tracing. The time-series cumulative damage tracing module determines the cumulative fatigue damage degree, remaining service life, and damage nodes based on the abnormal stress points and their historical load data. The spatial coupling deduction module reduces the connection stiffness weight based on the damage nodes and cumulative fatigue damage degree, and constructs a damage topology network. The risk hub location and early warning module determines the overall instability probability, instability sensitivity index, and risk hubs based on the damage topology network. Finally, the early warning and monitoring platform outputs early warning and handling results based on the risk hubs.

2. The high-support formwork system full-process critical point early warning monitoring system according to claim 1, characterized in that, The dynamic signal intelligent analysis module includes a signal cleaning unit, a window feature construction unit, and an abnormal stress identification unit; The signal cleaning unit can perform adaptive Kalman filtering on the nodal stress signal to obtain clean stress time series data. The window feature construction unit can perform sliding window analysis on clean stress time series data, extract the stress mean, standard deviation, rate of change, and deviation from the design benchmark model to form a multi-dimensional feature vector; the abnormal stress identification unit can calculate the local outlier score based on the multi-dimensional feature vector. When the local outlier scores of multiple consecutive windows of the same node all exceed a preset threshold, the node is identified as an abnormal stress point. The abnormal stress point includes the node number, abnormal start time, duration, maximum local outlier score, current stress value, and stress change trend.

3. The high-support formwork system full-process critical point early warning monitoring system according to claim 1 or 2, characterized in that, The execution steps of the dynamic signal analysis module during operation are as follows: By deploying a sensor network at key nodes of high formwork, the original stress signals of the nodes during the concrete pouring process are collected in real time. An adaptive Kalman filter algorithm is used to dynamically denoise and smooth the original stress signals, and clean stress time series data that characterizes the fluctuation of the real load are extracted. A sliding window analysis was performed on the clean stress time series data to extract the statistical features of the data in each window, including the mean stress, standard deviation, rate of change, and deviation from the benchmark model, and to construct a multidimensional feature vector sequence to characterize the local load state of the nodes. Based on the multidimensional feature vector sequence, the outlier score of each node in different time windows is calculated using the local outlier factor algorithm. According to the preset outlier threshold, abnormal data points in which local stress increases in a hidden manner in the spatiotemporal domain are dynamically identified and marked as abnormal stress points.

4. The high-support formwork system full-process critical point early warning monitoring system according to claim 1, characterized in that, The time-series cumulative damage tracing module includes a load history tracing unit, a memory network damage modeling unit, and a damage threshold early warning unit. The load history backtracking unit uses the time t0 corresponding to the abnormal stress point as the time anchor point, extracts the stress time series data within the preset historical period, and obtains the stress amplitude sequence, the number of cycles, and the loading sequence during the concrete pouring process. The memory network damage modeling unit includes a long short-term memory network encoder, an external memory matrix, and a regression prediction subnetwork. The long short-term memory network encoder can extract deep temporal features, the external memory matrix can retrieve similar load-damage evolution segments, and the regression prediction subnetwork can predict the damage increment based on the deep temporal features and similar load-damage evolution segments. The damage threshold early warning unit can calculate the cumulative fatigue damage degree D based on the damage degree increment, and determine the damage state level and damage node.

5. The high-support formwork system full-process critical point early warning monitoring system according to claim 4, characterized in that, The cumulative fatigue damage degree D is calculated according to the following formula: ； in, To accumulate fatigue damage, The initial cumulative damage level in history. As the current time anchor point, For the predicted length of the future time period, For the stress amplitude level corresponding to the same calculation step size Below, the theoretical number of cycles required for fatigue failure of the node material or connection structure. The variable is the integral variable, representing the time period during which the variable is predicted. The time of continuous change within; Predict the remaining time when the cumulative damage D reaches a preset critical value, and define it as the remaining service life RUL of the node. When D reaches 1, the corresponding node reaches the theoretical fatigue failure state. The remaining useful life (RUL) is calculated according to the following formula: ； ； in, This is the critical time point. To be at the critical time point The cumulative fatigue damage degree, whose value is equal to the critical damage threshold. , The critical damage threshold is when When this occurs, it means that the node has entered a high-risk state, triggering a high-level warning and intervention measures. When D is at the warning level or alarm level, the corresponding node will be marked as a damaged node, and a structural damage assessment report will be generated.

6. The critical point early warning and monitoring system for a high-support formwork system according to any one of claims 1-2 or 4-5, characterized in that, The spatial coupling inference module includes a topology network construction unit, a stiffness degradation update unit, and an instability path inference unit. The topology network construction unit can construct an initial topology network based on structural connection relationships and configure node state vectors for each vertex; the stiffness degradation update unit can reduce the connection stiffness weights of adjacent edges based on the cumulative fatigue damage of damaged nodes to form a damaged topology network; the instability path inference unit can transmit messages containing equivalent load, cumulative fatigue damage, and connection stiffness weights between adjacent vertices based on a graph neural network to update the equivalent load of each vertex and identify overloaded and failed nodes. The instability path inference unit starts from the initial damaged node and performs graph search on adjacent overloaded or failed nodes to obtain potential instability propagation paths.

7. The critical point early warning and monitoring system for the entire process of a high-support formwork system according to any one of claims 1-2 or 4-5, characterized in that, The risk hub location and early warning module includes an instability probability simulation unit and a risk hub screening unit; The instability probability simulation unit can randomly sample construction live load, cumulative fatigue damage, and connection stiffness weights based on the damage topology network and potential instability transmission paths to generate multiple load-damage coupling scenarios. When the critical force transmission path fails and causes the high-support formwork system to become a geometrically variable system, the corresponding scenario is recorded as the overall instability sample. The risk hub screening unit calculates the instability sensitivity index based on the number of overloads or failures of each vertex or edge in the overall instability sample.

8. The high-support formwork system full-process critical point early warning monitoring system according to claim 7, characterized in that, Overall instability probability Calculate using the following formula: ； ； in, The total number of unstable samples. This represents the total number of Monte Carlo simulations. The 95% confidence interval indicates that there is a 95% probability that the true probability of instability will fall within this interval, quantifying the uncertainty in probability estimation caused by a finite number of random samplings; The two-tailed quantiles of a standard normal distribution at a 95% confidence level; The standard error of the estimated value reflects the sampling variability; The formula for calculating the instability sensitivity index is as follows: ； in, As an index of instability sensitivity, For components The number of instability scenarios resulting in failure, and the failure / overload determination: in a single instability simulation, such as the component... If the stress exceeds its reduced bearing capacity, or the displacement exceeds the allowable limit, then it is determined to be in this state.

9. A construction method for a high-formwork system throughout the entire process, characterized in that, The high formwork system full-process critical point early warning monitoring system according to any one of claims 1-8, the full-process construction method of the high formwork system includes the following steps: Prepare a special construction plan for the high formwork system and determine the arrangement of members and stress monitoring points; Complete foundation treatment, positioning and layout, and installation of the support system; Stress sensors are installed at stress monitoring points, and a correspondence between the sensors and node information is established. During the concrete pouring process, zonal and layered loading is carried out. An early warning monitoring platform is used to identify abnormal stress points, cumulative fatigue damage, damage nodes, damage topology network, potential instability transmission paths and risk hubs, and the pouring operation is adjusted according to the early warning results.

10. The construction method for the entire process of a high-formwork system according to claim 9, characterized in that, When the early warning monitoring platform outputs an early warning level signal, the pouring operation in the area affected by the corresponding damaged node is suspended or slowed down, the connection status of the support system is checked, and the pouring sequence or monitoring frequency is adjusted according to the remaining service life and the trend of damage change. When an alarm level signal is output and the risk hub is located, pouring operations within the area affected by the risk hub shall be stopped, and temporary support reinforcement, load unloading, or area isolation measures shall be taken. After the handling is completed, data on the risk hub and adjacent nodes are collected again, and the cumulative fatigue damage, instability sensitivity index and overall instability probability are calculated. Construction is resumed when the above indicators return to the allowable range; otherwise, reinforcement or unloading measures are continued.

Images