Construction site risk operation information integrated collaborative management method

Abstract

The application discloses a construction site risk operation data integrated collaborative management method, and belongs to the technical field of data management. The steps comprise the following: constructing and calibrating a space-time risk benchmark database; deducing a set of potential risk transmission links and generating an associated intervention knowledge base; during operation, abnormal signals are confirmed as risk events through multi-sensor collaborative verification; the risk events are matched with the transmission links to calculate the risk escalation level and dynamically adjust the early warning threshold; when the early warning is triggered, a dynamic collaborative response instruction is generated and issued, and the disposal result is fed back to correct the benchmark database, forming a management closed loop. The application adopts technical means of constructing a space-time risk benchmark, deducing a transmission link, collaboratively verifying a risk and dynamically regulating a threshold, and improves the predictability of project risk control, the efficiency of resource collaborative scheduling and the scientificity of overall management decision.

Description

Technical Field

[0001] This invention belongs to the field of data management technology, and in particular relates to an integrated collaborative management method for risk operation data at construction sites. Background Technology

[0002] In construction project management, information technology is commonly used to monitor complex working environments. By deploying various sensors on-site, multi-source heterogeneous data involving personnel, machinery, and environmental conditions can be collected and processed in real time. This data is used to build a risk management system, providing risk warnings to project managers through continuous monitoring of key indicators, which is an important management tool to ensure the smooth progress of project processes.

[0003] However, existing risk management systems generally suffer from the following technical problems: data from different monitoring subsystems are isolated from each other, lacking information fusion and correlation analysis; risk warnings rely heavily on fixed parameter thresholds and cannot dynamically adapt to changes in on-site working conditions; warning information is simplistic and unstructured, lacking effective guidance for subsequent collaborative handling; emergency response processes are inefficient, and the handling results are difficult to use for feedback optimization of management models, failing to form a management closed loop. Summary of the Invention

[0004] To address the aforementioned issues, the present invention aims to provide an integrated collaborative management method for construction site risk operation data. This method employs techniques such as constructing spatiotemporal risk benchmarks, extrapolating transmission links, collaboratively verifying risks, and dynamically adjusting thresholds to improve the predictability of project risk management, the efficiency of resource collaborative scheduling, and the scientific nature of overall management decisions.

[0005] The above objectives can be achieved through the following approach: an integrated collaborative management method for construction site risk operation data, comprising the following steps:

[0006] The static raw sensing data of the construction site is acquired and fused from on-site sensors, the construction site is gridded, a spatiotemporal risk basic database is constructed, and non-core operation areas are selected for pre-monitoring. The normal response of the construction environment is sensed to establish benchmark characteristic parameters. The spatiotemporal risk basic database is initialized and calibrated using the benchmark characteristic parameters to obtain the spatiotemporal risk benchmark database.

[0007] Based on a spatiotemporal risk benchmark database, the causal and temporal correlations among various risk factors are analyzed, a set of potential risk transmission links is deduced, and stress tests and impact assessments are conducted under multiple operating conditions to verify and generate a knowledge base of related and coordinated interventions.

[0008] During construction, the operation data stream is collected in real time through sensors. When any sensor detects an abnormality, other sensors around the abnormal point are scheduled to collect supplementary data in a targeted manner, and the spatiotemporal consistency and logical correlation are compared. The initial abnormal signal is cross-validated, and after confirmation, the abnormality is transformed into a risk event signal.

[0009] The risk event signal is matched with the risk transmission link set in real time to determine whether the current risk status is evolving along the risk transmission link. The real-time risk escalation level of the danger is calculated based on the evolution stage, and the monitoring parameters and early warning thresholds of the link nodes are adjusted accordingly.

[0010] When the real-time risk escalation level triggers the early warning threshold of the monitoring parameters, dynamic collaborative response instructions are matched and generated from the intervention knowledge base according to the risk transmission link and issued. The execution action data of each collaborating party and the on-site status change data are fed back to the spatiotemporal risk benchmark database, forming a closed loop of operation management.

[0011] Preferably, the process of obtaining the spatiotemporal risk benchmark database is as follows:

[0012] Acquire static engineering data of the site's inherent attributes and environmental parameter data of the initial stage, fuse them to obtain static raw perception data, and spatially grid the construction site to build a spatiotemporal risk database;

[0013] Select non-core work areas and conduct periodic continuous monitoring. By analyzing the continuous data stream of sensors in the area, quantify the normal response characteristics of the construction environment under no work interference and establish benchmark characteristic parameters.

[0014] The initial attributes of each grid node in the spatiotemporal risk baseline database are iteratively corrected and assigned values ​​using baseline feature parameters to obtain the spatiotemporal risk baseline database.

[0015] Preferably, the process of verifying and generating the associated collaborative intervention knowledge base is as follows:

[0016] Risk factors are extracted from the spatiotemporal risk benchmark database, and the correlation strength and transmission probability between each risk factor are calculated to obtain a set of potential risk transmission links.

[0017] For each link in the risk transmission link set, a virtual risk event is constructed for initial excitation. Dynamic evolution simulation under multiple working conditions is carried out in the spatiotemporal risk benchmark database. The response status of each link node during the simulation is quantitatively calculated to obtain a multi-dimensional impact assessment vector of the potential impact after the link is excited.

[0018] Based on multi-dimensional impact assessment vectors, structured intervention items are generated for each risk transmission link, and these intervention items are compiled and indexed to generate a collaborative intervention knowledge base.

[0019] Preferably, the risk transmission link set includes a multidimensional risk node set and a weighted directed transmission arc set, wherein:

[0020] The multidimensional risk node set is used to define the individual risk events that constitute the risk transmission chain and to quantify the triggering conditions and state attributes of each event.

[0021] Weighted directed arc sets are used to describe the causal relationships, temporal delays, and impact strengths among risk nodes, and to characterize the evolution logic of risks among different nodes.

[0022] Preferably, the process of converting the anomaly into a risk event signal after confirmation is as follows:

[0023] The system performs instantaneous amplitude determination and dynamic time-series baseline deviation determination on the operation data stream, and after identifying the initial abnormal signal, it schedules nearby sensors to initiate directional collaborative verification and collect supplementary data.

[0024] Spatiotemporal correlation verification and physical causality verification are performed on the initial anomalous signal and supplementary data to determine whether the initial anomalous signal is correlated with the supplementary data in terms of time series and spatial proximity, and whether the initial anomalous signal is consistent with the data change trend from different types of sensors, so as to obtain multi-dimensional cross-validation results.

[0025] Based on the results of multi-dimensional cross-validation, risk event signals were identified.

[0026] Preferably, after confirming the risk event signal, the risk event signal is correlated with the risk transmission link set to identify and lock the currently activated risk evolution path;

[0027] Based on the risk evolution path and utilizing the state transition probability in the intervention knowledge base, a forward-looking situational simulation is conducted to quantify the future risk intensity, scope of impact, and development speed, generating dynamic risk evolution trajectory data.

[0028] Preferably, the process of adjusting the monitoring parameter early warning threshold of the link node is as follows:

[0029] The risk event signal is matched with the initial node attributes of each link in the risk transmission link set to calculate the probability, the maximum likelihood risk transmission link is calculated, and the current evolution node is determined.

[0030] The inherent risk level of the integrated evolution node and the rate of change of real-time perceived data of the risk event signal are calculated through weighted fusion to obtain the real-time risk escalation level that quantifies the degree of danger.

[0031] Based on the maximum likelihood risk transmission link, the early warning threshold of the monitoring parameters of subsequent risk nodes is dynamically adjusted forward.

[0032] Preferably, the process of forming a closed loop for operation management is as follows:

[0033] When the real-time risk escalation level triggers the monitoring parameter warning threshold, the system combines the risk transmission link and dynamic risk evolution trajectory data to match a collaborative intervention plan covering the current state and high-probability future evolution trends from the intervention knowledge base, and then optimizes the collaborative intervention plan in real time to generate dynamic collaborative response instructions.

[0034] The dynamic collaborative response command is parsed into a set of role-based actions for different on-site roles and actions, and the set of role-based actions is sent to the corresponding execution units to output a dataset of action effect evaluation.

[0035] By analyzing the discrepancy between the treatment effect evaluation dataset and the dynamic risk evolution trajectory data, intervention measures are evaluated in real time, and the parameters of the spatiotemporal risk benchmark database are continuously corrected to complete the operation management closed loop.

[0036] Preferably, the process of continuously correcting the parameters of the spatiotemporal risk benchmark database is as follows:

[0037] By comparing the dataset of the disposal effect assessment with the dynamic risk evolution trajectory data, the risk evolution time, the scope of impact and the probability of escalation are quantified, and a deviation attribution vector is generated.

[0038] Backpropagation is performed based on the bias attribution vector to correct the correlation strength and propagation probability parameters in the spatiotemporal risk benchmark database.

[0039] An integrated collaborative management system for construction site risk operation data is used to implement the above methods, including:

[0040] The spatiotemporal risk benchmark database generation module is used to acquire and fuse static raw sensing data of the construction site from on-site sensors, grid the construction site, construct a spatiotemporal risk basic database, select non-core operation areas for pre-monitoring, establish benchmark characteristic parameters by sensing the normal response of the construction environment, and use the benchmark characteristic parameters to initialize and calibrate the spatiotemporal risk basic database to obtain the spatiotemporal risk benchmark database.

[0041] The risk path deduction and contingency plan generation module is used to analyze the causal and temporal correlations between various risk factors based on a spatiotemporal risk benchmark database, deduce the potential risk transmission link set, conduct stress tests and impact assessments under multiple working conditions, verify and generate a knowledge base of related and coordinated interventions.

[0042] The collaborative verification and event confirmation module is used to collect operation data streams in real time through sensors during construction operations. When any sensor detects an anomaly, it schedules other sensors around the anomaly point to collect supplementary data in a targeted manner, compares spatiotemporal consistency and logical correlation, and cross-verifies the initial anomaly signal. After confirmation, the anomaly is converted into a risk event signal.

[0043] The risk situation assessment and threshold control module is used to match risk event signals with the risk transmission link set in real time, determine whether the current risk status is evolving along the risk transmission link, calculate the real-time risk escalation level of the danger level according to the evolution stage, and feed back to adjust the monitoring parameters and early warning thresholds of the link nodes.

[0044] The collaborative response and closed-loop feedback module is used to match and generate dynamic collaborative response instructions from the intervention knowledge base and issue them when the real-time risk escalation level triggers the warning threshold of the monitoring parameters, based on the risk transmission link. It also feeds back the execution action data of each collaborating party and the on-site status change data to the spatiotemporal risk benchmark database, forming a closed loop for operation management.

[0045] The present invention has the following advantages:

[0046] This invention achieves a shift from passive response to proactive prediction by constructing a spatiotemporal risk benchmark database based on on-site environmental calibration and deducing risk transmission chains. This method can identify potential evolution paths before or in the early stages of risk events, providing a scientific basis for early intervention and prevention, and enhancing the predictability and proactivity of risk management.

[0047] This invention proposes a multi-sensor collaborative verification and dynamic threshold adjustment mechanism, which solves the problems of high false alarm rate and poor adaptability in traditional single-point alarm systems. By performing multi-dimensional cross-verification on initial abnormal signals, the accuracy of risk event confirmation is ensured; at the same time, the early warning threshold is adjusted in a feedforward manner according to the risk evolution stage, enabling the monitoring system to intelligently adapt to changes in on-site conditions, thereby improving the timeliness and reliability of early warnings.

[0048] This invention establishes a complete management process from risk identification to collaborative response, and then to closed-loop feedback learning. Once a risk is confirmed, it can automatically generate and issue role-based, dynamic collaborative response instructions, achieving efficient and precise emergency response. Furthermore, through evaluation and feedback on the response effectiveness, it can continuously optimize its internal risk model and knowledge base, possessing self-evolution capabilities, thereby ensuring its long-term effectiveness. Attached Figure Description

[0049] Figure 1 This is a flowchart illustrating the method of the present invention;

[0050] Figure 2This is a schematic diagram of the feedforward dynamic adjustment of the early warning threshold of the monitoring parameter in an embodiment of the present invention;

[0051] Figure 3 This is a schematic diagram comparing the predicted risk evolution trajectory with the actual treatment effect in an embodiment of the present invention;

[0052] Figure 4 This is a schematic diagram of the system of the present invention. Detailed Implementation

[0053] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0054] Example 1: As Figure 1 As shown, the integrated collaborative management method for construction site risk operation data adopts technical means such as constructing spatiotemporal risk benchmarks, deducing transmission links, collaboratively verifying risks, and dynamically adjusting thresholds, which improves the predictability of project risk control, the efficiency of resource collaborative scheduling, and the scientific nature of overall management decisions.

[0055] The method specifically includes:

[0056] The static raw sensing data of the construction site is acquired and fused from on-site sensors, the construction site is gridded, a spatiotemporal risk basic database is constructed, and non-core operation areas are selected for pre-monitoring. The normal response of the construction environment is sensed to establish benchmark characteristic parameters. The spatiotemporal risk basic database is initialized and calibrated using the benchmark characteristic parameters to obtain the spatiotemporal risk benchmark database.

[0057] Based on a spatiotemporal risk benchmark database, the causal and temporal correlations among various risk factors are analyzed, a set of potential risk transmission links is deduced, and stress tests and impact assessments are conducted under multiple operating conditions to verify and generate a knowledge base of related and coordinated interventions.

[0058] During construction, the operation data stream is collected in real time through sensors. When any sensor detects an abnormality, other sensors around the abnormal point are scheduled to collect supplementary data in a targeted manner, and the spatiotemporal consistency and logical correlation are compared. The initial abnormal signal is cross-validated, and after confirmation, the abnormality is transformed into a risk event signal.

[0059] The risk event signal is matched with the risk transmission link set in real time to determine whether the current risk status is evolving along the risk transmission link. The real-time risk escalation level of the danger is calculated based on the evolution stage, and the monitoring parameters and early warning thresholds of the link nodes are adjusted accordingly.

[0060] When the real-time risk escalation level triggers the early warning threshold of the monitoring parameters, dynamic collaborative response instructions are matched and generated from the intervention knowledge base according to the risk transmission link, and issued to personnel and equipment roles for handling actions. The execution action data of each collaborating party and the on-site status change data are fed back to the spatiotemporal risk benchmark database to form a closed loop of operation management.

[0061] By employing technical means such as constructing spatiotemporal risk benchmarks, simulating transmission links, collaboratively verifying risks, and dynamically adjusting thresholds, the predictability of project risk management, the efficiency of resource collaborative scheduling, and the scientific nature of overall management decisions have been improved.

[0062] The process of obtaining the spatiotemporal risk benchmark database is as follows:

[0063] Acquire static engineering data of the site's inherent attributes and environmental parameter data of the initial stage, fuse them to obtain static raw perception data, and spatially grid the construction site to build a spatiotemporal risk database;

[0064] By retrieving the project's BIM building information model, geological exploration report, and equipment ledger, static engineering data such as the site's three-dimensional geometric information, geotechnical parameters, and structural component attributes were obtained. Simultaneously, before the full commencement of construction activities, weather stations and environmental monitoring sensors were deployed on-site to collect initial environmental parameter data such as temperature, humidity, wind speed, background noise, and atmospheric pressure over a period of time. Subsequently, these two types of data underwent spatiotemporal alignment processing, assigning geographic coordinates to non-spatial attribute data and fusing them to generate a static raw sensing data set containing precise location and multi-dimensional attribute information.

[0065] Based on this, using the digital elevation model and BIM model of the construction site, the entire construction area is divided into a three-dimensional spatial grid, discretized into a series of standard-sized voxel units, each with a unique spatial coordinate index. Finally, the static raw sensing data is mapped to its corresponding grid unit, thereby constructing an initial spatiotemporal risk database.

[0066] Select non-core work areas and conduct periodic continuous monitoring. By analyzing the continuous data stream of sensors in the area, quantify the normal response characteristics of the construction environment under no work interference and establish benchmark characteristic parameters.

[0067] To calibrate potential discrepancies between the initial database and the actual site environment, sensor arrays, including those for vibration, displacement, temperature, and humidity, were deployed in non-core work areas such as material storage yards and office areas, which are less directly affected by construction operations. These sensor arrays perform continuous, uninterrupted periodic monitoring to collect long-term environmental data streams. By performing signal processing and statistical analysis on the collected data streams (e.g., analyzing the dominant frequency of vibration signals through Fourier transform, or calculating the mean and variance of various environmental parameters), numerical values ​​that can quantify the background response characteristics of the specific site environment are extracted, establishing a set of baseline characteristic parameters.

[0068] The initial attributes of each grid node in the spatiotemporal risk baseline database are iteratively corrected and assigned values ​​using baseline feature parameters to obtain the spatiotemporal risk baseline database.

[0069] An iterative correction algorithm is used to calibrate the spatiotemporal risk baseline database. This algorithm uses the baseline characteristic parameters monitored in the previous step as a reference benchmark for the real environment. In each iteration, based on the global baseline characteristic parameters and the spatial location and physical properties of a specific grid cell, the theoretically required calibration attribute value for that cell is calculated. This theoretical value is then compared with the attribute values ​​in the current database. Based on the deviation between the two, and combined with a correction weight coefficient to control the correction rate, the attribute value of that grid node in the database is adjusted. This iterative process is repeated until the error between the response characteristics output by the database model and the real environment benchmark converges and stabilizes. Through this process, a calibrated spatiotemporal risk benchmark database that highly matches the actual field environment is finally output.

[0070] The process of verifying and generating a collaborative intervention knowledge base is as follows:

[0071] Risk factors are extracted from the spatiotemporal risk benchmark database, and the correlation strength and transmission probability between each risk factor are calculated to obtain a set of potential risk transmission links.

[0072] This process is built upon a calibrated spatiotemporal risk benchmark database. Multidimensional risk factors characterizing the safety status of the construction site are extracted from this database. These risk factors are not raw sensor readings, but rather processed indicators with clear physical meaning. For example, risk factors could be the displacement change rate of a specific monitoring point in a foundation pit, the axial force value of a specific upright in a tall formwork support system, or the load-moment ratio during tower crane operation. Subsequently, cross-correlation function statistical analysis techniques from time series analysis are used to mine historical data of these risk factors to calculate the correlation strength and transmission probability between each risk factor. Correlation strength characterizes the significance of a change in one risk factor on another, while transmission probability quantifies the likelihood of this impact occurring. For example, the transmission strength between two risk factors can be comprehensively assessed by combining the time series correlation coefficient representing spatial and physical correlation with the conditional probability representing temporal causality, and then weighted and summed using weighted coefficients. Based on the calculated correlation strength and transmission probability between all risk factors, a directed graph model consisting of a set of risk factor nodes and a set of weighted transmission arcs is constructed, thus obtaining the potential risk transmission link set.

[0073] For each link in the risk transmission link set, a virtual risk event is constructed for initial excitation. Dynamic evolution simulation under multiple working conditions is carried out in the spatiotemporal risk benchmark database. The response status of each link node during the simulation is quantitatively calculated to obtain a multi-dimensional impact assessment vector of the potential impact after the link is excited.

[0074] Within the digital twin environment of a construction site represented by a spatiotemporal risk benchmark database, stress testing and impact assessment are conducted on the constructed risk transmission link set. This process involves artificially constructing and injecting an initial virtual risk event, such as simulating a momentary exceedance of stress in the support structure of a certain area of ​​the construction site. Then, dynamic simulations are performed according to the transmission rules defined in the risk transmission link set to observe how this initial stimulus propagates and evolves within the link network. During the simulation, the response state of each activated risk factor node is quantitatively calculated, with specific indicators including response delay time, peak impact magnitude, and spatial extent of impact. Integrating these quantitative indicators forms a multi-dimensional impact assessment vector describing the potential impact of the virtual risk event.

[0075] Based on multi-dimensional impact assessment vectors, structured intervention items are generated for each risk transmission link, and these intervention items are compiled and indexed to generate a collaborative intervention knowledge base.

[0076] Based on the multi-dimensional impact assessment vectors obtained from the previous simulation, structured intervention items are generated for each simulated and validated risk transmission link. Each intervention item details the prevention, control, or emergency response measures to be taken when the risk evolves to a certain node in the link. These measures can be further refined, for example, by specifying the responsible personnel, the resources to be allocated, and standardized operating procedures. Finally, all these intervention items are compiled and indexed to form a systematic and rapidly searchable, interconnected intervention knowledge base.

[0077] The risk transmission link set includes a multidimensional risk node set and a weighted directed transmission arc set, wherein:

[0078] The multidimensional risk node set is used to define the individual risk events that constitute the risk transmission chain and to quantify the triggering conditions and state attributes of each event.

[0079] Each node in this multidimensional risk node set is a structured abstraction of a specific risk factor from the aforementioned embodiments. The data structure of each node contains a complete definition of that risk factor. The triggering condition refers to the quantitative standard used to determine whether the risk factor has entered an abnormal state. For example, a risk node representing "the horizontal displacement rate at the top of area A of the foundation pit" can have its triggering condition quantified as "the 24-hour sliding average of the rate exceeds 5 mm / d". The state attributes describe the dynamic characteristics of the node at any given time and may include the node's unique identifier, risk factor name, current monitoring value, risk level, and data update timestamp. This node set defines various potential risks at the construction site using standardized data units.

[0080] Weighted directed arc sets are used to describe the causal relationships, temporal delays, and impact strengths among risk nodes, and to characterize the evolution logic of risks among different nodes.

[0081] This weighted directed transmission arc set defines the connection relationships and quantitative parameters of mutual influence between nodes in a multidimensional risk node set. Each weighted directed transmission arc connects two risk nodes, and the direction indicates the transmission path of the risk. Among the arc's attributes, causal association clarifies the logical relationship between nodes; for example, a "continuous heavy rainfall" risk node will point to a "rising foundation pit water level" risk node. Temporal delay is used to quantify the time lag of this causal relationship; for example, there may be an average delay of 2 hours from the start of "continuous heavy rainfall" to the triggering of the "foundation pit water level" warning. Influence strength, which is the arc's weight, is a quantitative value that characterizes the contribution or influence of the source node's state change on the target node's state change. This value is determined based on the correlation strength and transmission probability between risk factors calculated in the aforementioned embodiment. Through this arc set, isolated risk nodes are connected into a logical network that reflects the evolutionary laws of real-world risks.

[0082] The process of converting an anomaly into a risk event signal after confirmation is as follows:

[0083] The system performs instantaneous amplitude determination and dynamic time-series baseline deviation determination on the operation data stream, and after identifying the initial abnormal signal, it schedules nearby sensors to initiate directional collaborative verification and collect supplementary data.

[0084] The real-time operational data streams from various sensors undergo initial screening. This screening process includes two judgment methods: first, instantaneous amplitude judgment, which checks whether the real-time data point momentarily exceeds the preset static safety threshold for that measurement point; second, dynamic time-series baseline deviation judgment, which uses time-series analysis algorithms such as moving average and exponential smoothing to establish a dynamic baseline reflecting the recent fluctuation characteristics of the measurement point data, and determines whether the current data point continuously and significantly deviates from this dynamic baseline. Once either judgment method identifies an anomaly, the signal is marked as an initial anomaly signal. At this point, no alarm is immediately issued, but a collaborative verification phase is initiated. Based on the location of the initial anomaly signal, other types of sensors within its vicinity are automatically scheduled to perform directional, supplementary data acquisition, thereby obtaining supplementary data for cross-validation. For example, an initial anomaly signal from a displacement sensor may trigger synchronous data acquisition from nearby stress gauges and high-definition cameras.

[0085] Spatiotemporal correlation verification and physical causality verification are performed on the initial anomalous signal and supplementary data to determine whether the initial anomalous signal is correlated with the supplementary data in terms of time series and spatial proximity, and whether the initial anomalous signal is consistent with the data change trend from different types of sensors, so as to obtain multi-dimensional cross-validation results.

[0086] After acquiring supplementary data, the authenticity of the initial anomalous signal is verified in multiple dimensions. First, a spatiotemporal correlation verification is performed. This verification determines whether there are anomalous responses in the supplementary data occurring within the same time window and adjacent spatial range as the initial anomalous signal, thus eliminating isolated and unrelated interference signals. Second, a physical causality verification is performed. This verification utilizes stored engineering knowledge to determine whether the trends in different types of data conform to physical or engineering laws. For example, a valid anomaly in the displacement of the foundation pit retaining structure should be accompanied by changes in soil pressure or stress in the supporting structure at the corresponding location that conform to mechanical logic. Through these verifications, a set of multi-dimensional cross-validation results can be obtained.

[0087] Based on the results of multi-dimensional cross-validation, risk event signals were identified.

[0088] To comprehensively assess the authenticity of the initial anomalous signal, a confidence evaluation model is introduced. This model quantifies and fuses the multi-dimensional cross-validation results obtained in the previous step to calculate a final confidence score. For example, this confidence score can be calculated using the following formula:

[0089] ;

[0090] Where V is the calculated final confidence score, which typically ranges from 0 to 1 and is used to characterize the degree of confidence that the initial anomalous signal is a real risk event; It is the significance score of the initial anomalous signal, and its value is determined by the instantaneous amplitude of the signal and the degree of deviation from the dynamic baseline. The greater the deviation, the higher the score. The spatiotemporal correlation verification score is used to evaluate the temporal and spatial correlation between supplementary data and the initial anomalous signal; the higher the correlation, the higher the score. It is a physical causality verification score, used to evaluate whether the changing trends between different data types conform to engineering logic and physical laws. The higher the degree of conformity, the higher the score. , , They are , , The weighting coefficients are set in advance according to the characteristics of different risk scenarios, and the sum of the three is 1. These coefficients are used to balance the importance of each verification result. After the calculation is completed, the confidence score is compared with the confirmation threshold. Only when it exceeds the threshold will the initial abnormal signal be finally confirmed as a valid risk event signal and output to the subsequent risk situation assessment stage.

[0091] After obtaining the risk event signal, the risk event signal is correlated with the risk transmission link set to identify and lock the currently activated risk evolution path;

[0092] Upon receiving a confirmed risk event signal, it is immediately correlated with the set of risk transmission links. Using maximum likelihood estimation or a similar matching algorithm, the posterior probability of the signal generated by each risk transmission link is calculated. The link with the highest probability is selected as the path followed by the current risk evolution, thereby identifying and locking down the currently activated risk evolution path, clarifying the starting point and the already evolved process of the risk.

[0093] Based on the risk evolution path and utilizing the state transition probability in the intervention knowledge base, a forward-looking situational simulation is conducted to quantify the future risk intensity, scope of impact, and development speed, generating dynamic risk evolution trajectory data.

[0094] After identifying the risk evolution path, a forward-looking risk situation simulation is conducted using the state transition probabilities for each risk node in the intervention knowledge base. Here, the state transition probability refers to the likelihood that a risk will evolve from one node in the chain to the next without intervention. Starting with the currently activated risk node, stochastic process models such as Markov chains are used to simulate the risk's continued development along this path over a future period. This simulation quantifies several key parameters of the future risk, such as the potential damage caused when the risk evolves to downstream nodes (i.e., future risk intensity); the spatial area that the risk may spread to (i.e., the impact range); and the time required for the risk to evolve from the current node to the next (i.e., the development speed). Integrating these quantitative parameters generates dynamic risk evolution trajectory data, providing a dynamic and quantitative view of the risk's future development for subsequent intervention decisions.

[0095] The process of adjusting the monitoring parameters and early warning thresholds of link nodes is as follows:

[0096] The risk event signal is matched with the initial node attributes of each link in the risk transmission link set to calculate the probability, the maximum likelihood risk transmission link is calculated, and the current evolution node is determined.

[0097] After a risk event signal is identified, it is probabilistically matched with a set of generated risk transmission links. This process can utilize Bayesian inference or maximum likelihood estimation algorithms to calculate the posterior probability of the current risk event signal being generated by each risk transmission link. Based on this, the link with the highest posterior probability is selected as the maximum likelihood risk transmission link followed by the current risk evolution, and the specific evolution node on this link where the risk event signal is located is simultaneously identified.

[0098] The inherent risk level of the integrated evolution node and the rate of change of real-time perceived data of the risk event signal are calculated through weighted fusion to obtain the real-time risk escalation level that quantifies the degree of danger.

[0099] After identifying the evolutionary node, a quantitative indicator that dynamically reflects the current level of danger is calculated: the real-time risk escalation level. This level calculation integrates information from two dimensions: first, the inherent risk level of the current evolutionary node itself, which is statically assigned a value based on its potential hazard when constructing the risk transmission chain; second, the dynamic level of danger reflected by the rate of change of real-time sensed data that triggered the risk event signal. The final real-time risk escalation level is obtained by weighted and fused calculation of these two dimensions.

[0100] Based on the maximum likelihood risk transmission link, the early warning threshold of the monitoring parameters of subsequent risk nodes is dynamically adjusted forward.

[0101] After calculating the real-time risk escalation level, based on the established maximum likelihood risk transmission path, the monitoring parameter warning thresholds for all unactivated subsequent risk nodes along that path will be dynamically adjusted forward. The magnitude of the adjustment is positively correlated with the calculated real-time risk escalation level; that is, the higher the current risk level, the lower the warning threshold for subsequent nodes, and the higher the monitoring sensitivity. This measure aims to consciously increase the monitoring sensitivity of subsequent risk signs when it is anticipated that the risk may continue to evolve along a specific path, so as to capture signals of further risk deterioration earlier. Figure 2 As shown, when upstream risks occur, how can we proactively and dynamically lower the early warning threshold for downstream risk nodes?

[0102] The process of forming a closed loop in work management is as follows:

[0103] When the real-time risk escalation level triggers the monitoring parameter warning threshold, the system combines the risk transmission link and dynamic risk evolution trajectory data to match a collaborative intervention plan covering the current state and high-probability future evolution trends from the intervention knowledge base, and then optimizes the collaborative intervention plan in real time to generate dynamic collaborative response instructions.

[0104] When the calculated real-time risk escalation level triggers the dynamically adjusted monitoring parameter warning threshold, the collaborative response procedure is immediately initiated. Combining the established risk transmission chain with the dynamic risk evolution trajectory data generated in the previous step, intelligent matching is performed in the intervention knowledge base to retrieve a collaborative intervention plan that simultaneously covers the current risk status and high-probability future evolution trends. To address changes in on-site conditions, the matched collaborative intervention plan is also optimized in real-time based on dynamic information such as the real-time location of on-site personnel and the availability of equipment, ultimately generating a highly customized dynamic collaborative response instruction.

[0105] The dynamic collaborative response command is parsed into a set of role-based actions for different on-site roles and actions, and the set of role-based actions is sent to the corresponding execution units to output a dataset of action effect evaluation.

[0106] The generated dynamic collaborative response instructions are automatically parsed and broken down into specific, actionable sets of actions tailored to different on-site roles and equipment. For example, instructions can be broken down into decision-making instructions for project managers, action instructions for safety officers, or equipment operation instructions for automated equipment such as tower cranes. These action sets are then distributed to the corresponding execution units via mobile terminals, dedicated communication channels, or IoT interfaces. Simultaneously, on-site sensor data and operation records from each execution unit are continuously collected, and this data is integrated and processed to generate a dataset evaluating the effectiveness of the response.

[0107] By analyzing the discrepancy between the treatment effect evaluation dataset and the dynamic risk evolution trajectory data, intervention measures are evaluated in real time, and the parameters of the spatiotemporal risk benchmark database are continuously corrected to complete the operation management closed loop.

[0108] After a risk event is handled, the effectiveness of the intervention measures is evaluated in real time. This evaluation is achieved by analyzing the deviation between the assessment dataset of the handling effect and the pre-predicted dynamic risk evolution trajectory data. Based on the analysis results of this deviation, relevant parameters in the spatiotemporal risk benchmark database are corrected, thereby completing a closed loop of operation management. The specific implementation method of this correction process will be further described in the next embodiment.

[0109] The process of continuously refining the parameters of the spatiotemporal risk benchmark database is as follows:

[0110] By comparing the dataset of the disposal effect assessment with the dynamic risk evolution trajectory data, the risk evolution time, the scope of impact and the probability of escalation are quantified, and a deviation attribution vector is generated.

[0111] The effectiveness evaluation dataset generated during this intervention process was precisely quantitatively compared with the pre-predicted dynamic risk evolution trajectory data. The effectiveness evaluation dataset objectively records the actual situation of the risk after the intervention measures were implemented, including its evolution time, scope of impact, and probability of escalation; while the dynamic risk evolution trajectory data is a prediction of the risk under ideal intervention conditions. By calculating the deviation between these two datasets, the gap between prediction and reality was identified, and these deviations were integrated to generate a multi-dimensional deviation attribution vector. Each component of this vector points to a possible cause of prediction inaccuracies, such as... Figure 3 As shown, by comparing the "predicted risk evolution" with the "actual situation after intervention", biases can be identified and the model can be corrected.

[0112] Backpropagation is performed based on the bias attribution vector to correct the correlation strength and propagation probability parameters in the spatiotemporal risk benchmark database.

[0113] Using this bias attribution vector, a backpropagation algorithm similar to that in neural network training is employed to provide feedback corrections to the knowledge system. This algorithm traces the source of the bias and locates the correlation strength and propagation probability parameters related to that bias in the spatiotemporal risk benchmark database. For example, if the actual response of a risk factor node is found to be significantly weaker than predicted, the algorithm will correspondingly lower the propagation probability parameters of its upstream nodes. In this way, experience in risk management is transformed into a deeper understanding of risk patterns and solidified into the parameters of the spatiotemporal risk benchmark database, thereby enabling informed judgments in future risk prediction and management decisions.

[0114] To verify the feasibility and effectiveness of the method of this invention, it was applied to the curtain wall installation project of a super high-rise building. This project involved high altitude operations, was greatly affected by high-altitude wind speeds, and involved multiple trades working simultaneously, making coordination difficult. Traditional risk management relies on manual observation and walkie-talkie communication, which is insufficient for timely and coordinated responses to sudden risks such as gusts of wind.

[0115] In the initial stage of the project, the first step was to generate a spatiotemporal risk baseline database. By importing the project's building information model, precise three-dimensional spatial information of the building structure and curtain wall units was obtained. Simultaneously, anemometers, thermometers, hygrometers, and barometers were deployed at the highest points on the site, and torque, tilt angle, and hook position sensors were installed on the tower crane to collect initial environmental and equipment static parameters, thus constructing the initial spatiotemporal risk baseline database. Subsequently, through continuous environmental monitoring during non-operational windows, baseline characteristic parameters reflecting the normal response of the site under no-operational interference were established, and the baseline database was iteratively corrected and calibrated based on these parameters.

[0116] Based on the calibrated database, key risk factors such as instantaneous increase in high-altitude wind speed, tower crane load torque ratio, and hoisting unit attitude angle deviation were extracted. By analyzing historical data, the correlation strength and transmission probability between these factors were calculated, constructing a risk transmission link set. For example, a high-probability link was identified: instantaneous increase in high-altitude wind speed → surge in tower crane load torque ratio → excessive deviation of hoisting unit attitude angle. Finally, an intervention knowledge base containing various intervention measures was generated through dynamic simulation.

[0117] At 10:32 AM on July 17, 2025, while monitoring the real-time operation data stream, an initial anomaly signal was detected in the reading of the wind speed sensor W-01, installed at a height of 250 meters, which surged from 5 m / s to 15 m / s within 3 seconds. Instead of immediately triggering an alarm, a collaborative verification was initiated. Details are shown in Table 1.

[0118] Table 1. Collaborative Verification Data of Risk Event Signals

[0119]

[0120] Upon confirming the risk event signal, it was immediately matched with the risk transmission link set, identifying the aforementioned instantaneous increase in high-altitude wind speed → surge in tower crane load torque ratio → excessive deviation of lifting unit attitude angle as the maximum likelihood risk transmission link. Considering the urgency of the event, the real-time risk escalation level was calculated to be high. Based on this, the warning thresholds for the monitoring parameters of the tower crane load torque ratio and lifting unit attitude angle deviation that had not yet been activated on this link were dynamically lowered by 20%, improving monitoring sensitivity. Simultaneously, dynamic risk evolution trajectory data was generated, predicting that without intervention, the tower crane load torque would reach a dangerous value within 2 minutes.

[0121] At 10:33 AM, the tower crane's load torque ratio triggered the adjusted warning threshold. A coordinated response was immediately initiated, establishing a closed-loop operation management system. Details are shown in Table 2.

[0122] Table 2. Risk Collaborative Response and Closed-Loop Feedback Data

[0123]

[0124] Through the application of this invention, a major safety risk caused by a sudden high-altitude gust of wind was successfully warned and handled collaboratively, preventing potential curtain wall unit collisions or detachments. The entire process achieved early risk identification, intelligent resource scheduling, and closed-loop optimization of the handling process, verifying the effectiveness of this invention in improving the safety, collaboration, and intelligence levels of high-risk work sites.

[0125] Table 1 details the confirmation process of a real risk event triggered by high-altitude gusts. Data shows that by integrating multi-source information such as on-site sensors, external data interfaces, and equipment status for collaborative verification, an isolated anomalous signal was confirmed as a real risk event with a confidence level of 0.91 within one minute, supporting the subsequent rapid response.

[0126] Table 2 presents the complete process from early warning triggering to handling completion and self-learning. Data shows that the automated collaborative handling process controls response and handling time to within 2 minutes. More importantly, after the event, by comparing the deviation between prediction and reality, the internal model parameters were successfully corrected in reverse, demonstrating self-evolution capability and proving the technical advantages of the method in this embodiment in achieving an intelligent closed-loop risk management system.

[0127] Example 2: Figure 4 As shown, the integrated collaborative management system for construction site risk operation data is used to implement the method in Example 1, including:

[0128] The spatiotemporal risk benchmark database generation module is used to acquire and fuse static raw sensing data of the construction site from on-site sensors, grid the construction site, construct a spatiotemporal risk basic database, select non-core operation areas for pre-monitoring, establish benchmark characteristic parameters by sensing the normal response of the construction environment, and use the benchmark characteristic parameters to initialize and calibrate the spatiotemporal risk basic database to obtain the spatiotemporal risk benchmark database.

[0129] The risk path deduction and contingency plan generation module is used to analyze the causal and temporal correlations between various risk factors based on a spatiotemporal risk benchmark database, deduce the potential risk transmission link set, conduct stress tests and impact assessments under multiple working conditions, verify and generate a knowledge base of related and coordinated interventions.

[0130] The collaborative verification and event confirmation module is used to collect operation data streams in real time through sensors during construction operations. When any sensor detects an anomaly, it schedules other sensors around the anomaly point to collect supplementary data in a targeted manner, compares spatiotemporal consistency and logical correlation, and cross-verifies the initial anomaly signal. After confirmation, the anomaly is converted into a risk event signal.

[0131] The risk situation assessment and threshold control module is used to match risk event signals with the risk transmission link set in real time, determine whether the current risk status is evolving along the risk transmission link, calculate the real-time risk escalation level of the danger level according to the evolution stage, and feed back to adjust the monitoring parameters and early warning thresholds of the link nodes.

[0132] The collaborative response and closed-loop feedback module is used to match and generate dynamic collaborative response instructions from the intervention knowledge base according to the risk transmission link when the real-time risk escalation level triggers the early warning threshold of the monitoring parameters. It then issues these instructions to personnel and equipment roles to take action and feeds back the execution data of each collaborating party and the data on changes in the on-site status to the spatiotemporal risk benchmark database, thus forming a closed loop for operation management.

[0133] The functional division and information interaction among the various modules described above are logical, but in terms of physical implementation, they can be integrated on the same software platform or deployed in a distributed manner. The connections between them represent data flow and control flow, designed to collaboratively achieve the objectives of this invention.

Claims

1. A method for integrated collaborative management of risk operation data at construction sites, characterized in that: Includes the following steps: The static raw sensing data of the construction site is acquired and fused from on-site sensors, and the construction site is gridded to build a spatiotemporal risk basic database. Non-core operation areas are selected for pre-monitoring and periodic monitoring. Through signal processing and statistical analysis, the numerical values ​​of the background response characteristics in a specific site environment are extracted and quantified. The normal response of the construction environment is sensed to establish benchmark characteristic parameters. The spatiotemporal risk basic database is initialized and calibrated using the benchmark characteristic parameters to obtain the spatiotemporal risk benchmark database. The initial calibration of the spatiotemporal risk database using benchmark feature parameters includes: calculating the theoretical calibration attribute value of the specific grid cell based on the benchmark feature parameters and the spatial location and physical properties of the specific grid cell; comparing the calibration attribute value with the attribute value in the current database; and adjusting the attribute value in the current database based on the comparison deviation. Based on a spatiotemporal risk benchmark database, the causal and temporal correlations among various risk factors are analyzed, a set of potential risk transmission links is deduced, and stress tests and impact assessments are conducted under multiple operating conditions to verify and generate a knowledge base of related and coordinated interventions. During construction, the operation data stream is collected in real time through sensors. When any sensor detects an abnormality, other sensors around the abnormal point are scheduled to collect supplementary data in a targeted manner, and the spatiotemporal consistency and logical correlation are compared. The initial abnormal signal is cross-validated, and after confirmation, the abnormality is transformed into a risk event signal. The risk event signal is matched with the risk transmission link set in real time to determine whether the current risk status is evolving along the risk transmission link. The real-time risk escalation level of the danger is calculated based on the evolution stage, and the monitoring parameters and early warning thresholds of the link nodes are adjusted accordingly. When the real-time risk escalation level triggers the early warning threshold of the monitoring parameters, dynamic collaborative response instructions are matched and generated from the intervention knowledge base according to the risk transmission link and issued. The execution action data of each collaborating party and the on-site status change data are fed back to the spatiotemporal risk benchmark database, forming a closed loop of operation management.

2. The integrated collaborative management method for construction site risk operation data according to claim 1, characterized in that, The process of obtaining the spatiotemporal risk benchmark database is as follows: Acquire static engineering data of the site's inherent attributes and environmental parameter data of the initial stage, fuse them to obtain static raw perception data, and spatially grid the construction site to build a spatiotemporal risk database; Select non-core work areas and conduct periodic continuous monitoring. By analyzing the continuous data stream of sensors in the area, quantify the normal response characteristics of the construction environment under no work interference and establish benchmark characteristic parameters. The initial attributes of each grid node in the spatiotemporal risk baseline database are iteratively corrected and assigned values ​​using baseline feature parameters to obtain the spatiotemporal risk baseline database.

3. The integrated collaborative management method for construction site risk operation data according to claim 1, characterized in that, The process of verifying and generating a collaborative intervention knowledge base is as follows: Risk factors are extracted from the spatiotemporal risk benchmark database, and the correlation strength and transmission probability between each risk factor are calculated to obtain a set of potential risk transmission links. For each link in the risk transmission link set, a virtual risk event is constructed for initial excitation. Dynamic evolution simulation under multiple working conditions is carried out in the spatiotemporal risk benchmark database. The response status of each link node during the simulation is quantitatively calculated to obtain a multi-dimensional impact assessment vector of the potential impact after the link is excited. Based on multi-dimensional impact assessment vectors, structured intervention items are generated for each risk transmission link, and these intervention items are compiled and indexed to generate a collaborative intervention knowledge base.

4. The integrated collaborative management method for construction site risk operation data according to claim 3, characterized in that, The risk transmission link set includes a multidimensional risk node set and a weighted directed transmission arc set, wherein: The multidimensional risk node set is used to define the individual risk events that constitute the risk transmission chain and to quantify the triggering conditions and state attributes of each event. Weighted directed arc sets are used to describe the causal relationships, temporal delays, and impact strengths among risk nodes, and to characterize the evolution logic of risks among different nodes.

5. The integrated collaborative management method for construction site risk operation data according to claim 1, characterized in that, The process of converting the anomaly into a risk event signal after confirmation is as follows: The system performs instantaneous amplitude determination and dynamic time-series baseline deviation determination on the operation data stream, and after identifying the initial abnormal signal, it schedules nearby sensors to initiate directional collaborative verification and collect supplementary data. Spatiotemporal correlation verification and physical causality verification are performed on the initial anomalous signal and supplementary data to determine whether the initial anomalous signal is correlated with the supplementary data in terms of time series and spatial proximity, and whether the initial anomalous signal is consistent with the data change trend from different types of sensors, so as to obtain multi-dimensional cross-validation results. Based on the results of multi-dimensional cross-validation, risk event signals were identified.

6. The integrated collaborative management method for construction site risk operation data according to claim 5, characterized in that, After confirming the risk event signal, the risk event signal is correlated with the risk transmission link set to identify and lock the currently activated risk evolution path; Based on the risk evolution path and utilizing the state transition probability in the intervention knowledge base, a forward-looking situational simulation is conducted to quantify the future risk intensity, scope of impact, and development speed, generating dynamic risk evolution trajectory data.

7. The integrated collaborative management method for construction site risk operation data according to claim 1, characterized in that, The process of adjusting the monitoring parameters and early warning thresholds of the link nodes is as follows: The risk event signal is matched with the initial node attributes of each link in the risk transmission link set to calculate the probability, the maximum likelihood risk transmission link is calculated, and the current evolution node is determined. The inherent risk level of the integrated evolution node and the rate of change of real-time perceived data of the risk event signal are calculated through weighted fusion to obtain the real-time risk escalation level that quantifies the degree of danger. Based on the maximum likelihood risk transmission link, the early warning threshold of the monitoring parameters of subsequent risk nodes is dynamically adjusted forward.

8. The integrated collaborative management method for construction site risk operation data according to claim 6, characterized in that, The process of forming a closed loop in operation management is as follows: When the real-time risk escalation level triggers the monitoring parameter warning threshold, the system combines the risk transmission link and dynamic risk evolution trajectory data to match a collaborative intervention plan covering the current state and high-probability future evolution trends from the intervention knowledge base, and then optimizes the collaborative intervention plan in real time to generate dynamic collaborative response instructions. The dynamic collaborative response command is parsed into a set of role-based actions for different on-site roles and actions, and the set of role-based actions is sent to the corresponding execution units to output a dataset of action effect evaluation. By analyzing the discrepancy between the treatment effect evaluation dataset and the dynamic risk evolution trajectory data, intervention measures are evaluated in real time, and the parameters of the spatiotemporal risk benchmark database are continuously corrected to complete the operation management closed loop.

9. The integrated collaborative management method for construction site risk operation data according to claim 8, characterized in that, The process of continuously refining the parameters of the spatiotemporal risk benchmark database is as follows: By comparing the dataset of the disposal effect assessment with the dynamic risk evolution trajectory data, the risk evolution time, the scope of impact and the probability of escalation are quantified, and a deviation attribution vector is generated. Backpropagation is performed based on the bias attribution vector to correct the correlation strength and propagation probability parameters in the spatiotemporal risk benchmark database.

10. An integrated collaborative management system for construction site risk operation data, used to implement the method described in any one of claims 1-9, characterized in that, include: The spatiotemporal risk benchmark database generation module is used to acquire and fuse static raw sensing data of the construction site from on-site sensors, grid the construction site, construct a spatiotemporal risk basic database, select non-core operation areas for pre-monitoring, establish benchmark characteristic parameters by sensing the normal response of the construction environment, and use the benchmark characteristic parameters to initialize and calibrate the spatiotemporal risk basic database to obtain the spatiotemporal risk benchmark database. The risk path deduction and contingency plan generation module is used to analyze the causal and temporal correlations between various risk factors based on a spatiotemporal risk benchmark database, deduce the potential risk transmission link set, conduct stress tests and impact assessments under multiple working conditions, verify and generate a knowledge base of related and coordinated interventions. The collaborative verification and event confirmation module is used to collect operation data streams in real time through sensors during construction operations. When any sensor detects an anomaly, it schedules other sensors around the anomaly point to collect supplementary data in a targeted manner, compares spatiotemporal consistency and logical correlation, and cross-verifies the initial anomaly signal. After confirmation, the anomaly is converted into a risk event signal. The risk situation assessment and threshold control module is used to match risk event signals with the risk transmission link set in real time, determine whether the current risk status is evolving along the risk transmission link, calculate the real-time risk escalation level of the danger level according to the evolution stage, and feed back to adjust the monitoring parameters and early warning thresholds of the link nodes. The collaborative response and closed-loop feedback module is used to match and generate dynamic collaborative response instructions from the intervention knowledge base and issue them when the real-time risk escalation level triggers the warning threshold of the monitoring parameters, based on the risk transmission link. It also feeds back the execution action data of each collaborating party and the on-site status change data to the spatiotemporal risk benchmark database, forming a closed loop for operation management.

Images