Construction Project Safety Monitoring System

Abstract

This invention relates to the field of building project safety monitoring technology, specifically a building project safety monitoring system. It includes data acquisition, first to fourth processing, and safety analysis modules. The system acquires building structure monitoring parameters and an initial stiffness matrix. Its core is to iteratively correct the stiffness matrix based on the current load and measured displacement, and determine the predicted load by combining historical loads and future construction plans. Then, based on the corrected stiffness matrix and the predicted load, it simulates a progressive failure process. The safety analysis module determines the overall failure probability and the remaining safety window period accordingly, and outputs an early warning level. This invention realizes a shift from passive response to proactive prediction in safety monitoring, solving the problem of traditional systems lacking predictive capabilities.

Description

Technical Field

[0001] This invention relates to the field of building project safety monitoring technology, specifically to a building project safety monitoring system. Background Technology

[0002] In the field of building safety monitoring, traditional monitoring systems mainly rely on sensor measurement data for threshold alarms. Such systems are essentially passive response mechanisms, only issuing alarms after damage or displacement exceeds limits, lacking forward-looking predictive capabilities. At the same time, traditional systems struggle to identify hidden degradation of structural stiffness caused by the accumulation of minor damage, and the linear analysis methods they employ cannot deduce stress redistribution and cascading failure modes triggered by localized damage. Furthermore, their early warning mechanisms are usually singular or binary, failing to quantify future risk probabilities or provide clear safety windows, resulting in a lack of timeliness and accuracy in intervention measures.

[0003] Therefore, how to construct a dynamic model that can synchronize with the actual stiffness degradation, and based on this model to shift from passive response to active prediction, so as to achieve quantitative assessment of the risk of cascading failures and accurate early warning of the safety window, is a technical problem that urgently needs to be solved. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a building project safety monitoring system. Specifically, the technical solution of this invention includes:

[0005] The data acquisition module is used to acquire monitoring parameters of the building structure, including: measured displacement of key nodes, vibration data, ambient temperature, rainfall, current construction log and future construction plan; and to acquire the initial stiffness matrix.

[0006] The first processing module is used to determine the current total equivalent load vector based on the current construction log, vibration data, ambient temperature and rainfall.

[0007] The second processing module is used to determine the predicted load vector based on the historical total equivalent load vector sequence determined by the first processing module at a historical moment and the future construction plan;

[0008] The third processing module is used to iteratively correct the stiffness matrix based on the current total equivalent load vector, measured displacement and current stiffness matrix, where the current stiffness matrix is ​​the initial stiffness matrix or the corrected stiffness matrix obtained by iteration in the previous time step.

[0009] The fourth processing module is used to simulate the progressive failure process based on the corrected stiffness matrix and the predicted load vector, and to determine the total number of simulations, the number of simulations that lead to overall failure, the risk contribution index of the element, and the failure time of each failure simulation.

[0010] The safety analysis module is used to determine the overall failure probability and remaining safety window based on the total number of simulations, the number of simulations that lead to overall failure, the risk contribution index of the unit, and the failure time of each failure simulation, and outputs the warning level.

[0011] Preferably, the first processing module is used to determine the current total equivalent load vector, including:

[0012] Quantify construction activities and vibration data to obtain dynamic loads;

[0013] Quantifying environmental temperature changes and rainfall yields additional loads;

[0014] Based on the principle of load superposition, dynamic loads and additional loads are synthesized to obtain the current total equivalent load vector.

[0015] Preferably, the second processing module is used to determine the predicted load vector, including:

[0016] A time series forecasting model is used to deduce the predicted load vector in response to the historical total equivalent load vector sequence and future construction plans.

[0017] Preferably, the third processing module is used for iterative correction to obtain the corrected stiffness matrix, including:

[0018] Input the current total equivalent load vector into the current stiffness matrix to calculate the simulated displacement;

[0019] Based on simulated displacement and measured displacement, the deviation index of key nodes is calculated.

[0020] Calculate the average deviation index of the critical nodes;

[0021] The current stiffness matrix is ​​corrected based on the average deviation index and the preset correction coefficient to obtain the corrected stiffness matrix.

[0022] Preferably, the fourth processing module simulates a progressive failure process, including:

[0023] The predicted load vector is applied to the modified stiffness matrix, and the predicted stress is obtained by solving.

[0024] Check whether the predicted stress exceeds the preset material failure threshold to identify failure cells;

[0025] When a failure element is identified, the stiffness contribution of the failure element is removed from the current stiffness matrix to obtain a new stiffness matrix. The solution is then repeated based on the new stiffness matrix until the entire structure becomes unstable, and the failure chain is recorded.

[0026] If no failure element is identified, the structure is determined to be stable under the predicted load vector.

[0027] Preferably, the fourth processing module is used to determine the total number of simulations, the number of simulations leading to overall failure, the risk contribution index of the unit, and the failure time of each failure simulation, including:

[0028] By sampling the uncertainty of the predicted load vector through Monte Carlo simulation, the progressive failure process is repeatedly executed to obtain the total number of simulations.

[0029] The number of simulations that resulted in overall structural instability and ultimately overall failure is counted out of the total number of simulations.

[0030] In the number of simulations that lead to overall failure, record the failure time from the start of each simulation to the occurrence of overall failure;

[0031] Count the number of times each unit appears in a failure chain that leads to overall failure in the total number of simulations;

[0032] The risk contribution index of a unit is determined by dividing the number of times each unit appears in a failure chain that leads to overall failure by the total number of simulations.

[0033] Preferably, the security analysis module is used to determine the overall failure probability and the remaining safety window, including:

[0034] The overall failure probability is determined by the ratio of the number of simulations that lead to overall failure to the total number of simulations.

[0035] The minimum failure time from each failure simulation is taken to determine the remaining safety window.

[0036] Preferably, the security analysis module is used to output early warning levels, including:

[0037] The overall failure probability and the remaining safety window are compared with preset first probability threshold, second probability threshold, first time threshold and second time threshold.

[0038] When the overall failure probability is less than or equal to the first probability threshold and the remaining safety window is greater than or equal to the first time threshold, output the safety level.

[0039] When the overall failure probability is greater than the first probability threshold and less than or equal to the second probability threshold, or when the remaining safe window period is less than the first time threshold and greater than or equal to the second time threshold, the attention level is output.

[0040] When the overall failure probability is greater than the second probability threshold or the remaining safety window is less than the second time threshold, an early warning level is output.

[0041] Preferably, the security analysis module is further used for:

[0042] When a level of concern or warning is output, it indicates that for units whose risk contribution index is greater than the preset risk threshold, key inspections or contingency reinforcement plans should be implemented.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. This system predicts future loads by constructing a dynamically corrected digital twin model and combining it with the analysis of future construction plans and historical loads. It performs forward-looking failure simulations of the structure, realizing the transformation of safety monitoring from passive response to active prediction, and solving the problem that traditional systems lack predictive capabilities and can only issue alarms after the fact.

[0045] 2. This system calculates the deviation between the measured displacement of key structural nodes and the simulated displacement of the model in real time, and iteratively corrects the current stiffness matrix accordingly. This ensures that the digital twin model can keep pace with the stiffness degradation of the real structure, and solves the problem that traditional systems cannot identify the hidden stiffness degradation of the structure caused by the accumulation of minor damage.

[0046] 3. This system simulates the progressive failure process. When an element fails in the simulation, its stiffness contribution is removed and the solution is resolved. This can effectively simulate the load redistribution caused by local failure, thereby revealing the cascading failure modes that traditional linear analysis methods cannot detect, and improving the realism of risk assessment.

[0047] 4. This system determines the overall failure probability by simulating the ratio of failure times to the total number of failures, and determines the remaining safety window by taking the minimum failure time. This transforms the safety issue from a binary judgment into a quantitative assessment of probability and time. At the same time, by identifying key weak units and combining them with tiered early warning, it achieves precise and targeted intervention for high-risk units, solving the problems of unclear early warning and low intervention efficiency in traditional systems. Attached Figure Description

[0048] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0049] Figure 1 This is a structural diagram of the system of the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0051] Example 1:

[0052] Please see Figure 1 Construction project safety monitoring system, including:

[0053] The data acquisition module is used to acquire monitoring parameters of the building structure, including: measured displacement of key nodes, vibration data, ambient temperature, rainfall, current construction log and future construction plan; and to acquire the initial stiffness matrix.

[0054] The first processing module is used to determine the current total equivalent load vector based on the current construction log, vibration data, ambient temperature and rainfall.

[0055] The second processing module is used to determine the predicted load vector based on the historical total equivalent load vector sequence determined by the first processing module at a historical moment and the future construction plan;

[0056] The third processing module is used to iteratively correct the stiffness matrix based on the current total equivalent load vector, measured displacement, and current stiffness matrix. The current stiffness matrix is ​​either the initial stiffness matrix or the corrected stiffness matrix obtained from the previous iteration. That is, in... At time t, the current stiffness matrix equal to the initial stiffness matrix ;exist At time t, the current stiffness matrix equal The corrected stiffness matrix output by the third processing module at time step ;

[0057] The fourth processing module is used to simulate the progressive failure process based on the corrected stiffness matrix and the predicted load vector, and to determine the total number of simulations, the number of simulations that lead to overall failure, the risk contribution index of the element, and the failure time of each failure simulation.

[0058] The safety analysis module is used to determine the overall failure probability and remaining safety window based on the total number of simulations, the number of simulations that lead to overall failure, the risk contribution index of the unit, and the failure time of each failure simulation, and outputs the warning level.

[0059] This invention provides a building project safety monitoring system, the purpose of which is to construct a dynamically corrected digital twin model and combine it with the prediction of future loads to perform progressive failure simulation of the structure, thereby realizing safety monitoring from passive response to active prediction.

[0060] The system includes a data acquisition module; this module is the perception foundation of the system, and its purpose is to comprehensively acquire raw data reflecting the current state of the structure, external disturbances and future trends; in this embodiment, the module is implemented through sensors deployed at key nodes of the building structure, such as beam-column connections or the center of large-span structures, such as high-precision GPS, laser displacement gauges, accelerometers, strain gauges, as well as data interfaces of on-site weather stations and construction management systems.

[0061] This module is used to acquire monitoring parameters, including the measured displacement of key nodes. The data, collected via GPS or laser displacement gauges, is used for subsequent comparison with the simulated displacement in the model and is the core basis for model correction; vibration data It collects data via accelerometers to quantify dynamic disturbances caused by construction activities such as pile driving and blasting; ambient temperature. and rainfall It collects data through on-site weather stations to quantify the additional loads caused by thermal expansion and contraction and rainwater accumulation; current construction log. It obtains information from the construction management system interface, recording the type and location of currently ongoing construction activities, as well as future construction plans. This information, also obtained from the construction management system, is an important boundary condition for predicting future loads.

[0062] This module is also used to obtain the initial stiffness matrix; initial stiffness matrix It refers to the theoretical stiffness of a structure when it is undamaged; its source is a high-fidelity benchmark physical model established using the finite element method (FEM) based on Building Information Modeling (BIM), design drawings, and geological survey reports.

[0063] The system also includes a first processing module; the purpose of this module is to quantify the multi-source heterogeneous data acquired by the data acquisition module, namely construction, vibration and environmental data, into a unified finite element model, namely a digital twin model, and an identifiable mechanical input, namely the current total equivalent load vector.

[0064] In this embodiment, the module synthesizes disturbances of different natures into a current total equivalent load vector based on the physical principle of load superposition. To achieve this goal, the following calculation method is introduced: ;

[0065] in, This is the current total equivalent load vector, with the dimension of force. The data is obtained from calculations in this module and is used as a test load input to the third processing module to evaluate the accuracy of the current model. Dynamic loads introduced by construction activities and vibrations, with the dimension of force. The source is based on the current construction log. For example, identifying it as pile driving, and vibration data. The amplitude and frequency are obtained through a specific dynamic model; The additional load caused by environmental temperature changes and rainfall is expressed in force. The source is based on ambient temperature. Calculate thermal stress based on changes in rainfall. The water load is calculated and superimposed; the motivation behind this formula is to transform complex site disturbances, such as construction and weather, into accurate FEM inputs. ;

[0066] The system also includes a second processing module; the purpose of this module is to look to the future, extrapolating or predicting specific future time windows based on historical trends and future plans. Within, for example, the predicted load vector that may be encountered in the next 24 hours;

[0067] In this embodiment, the module utilizes the historical total equivalent load vector sequence determined by the first processing module at a historical moment, such as the past 72 hours. And combined with the future construction plan obtained by the data acquisition module For example, tomorrow's plan is to carry out tower crane hoisting;

[0068] To achieve this prediction, this embodiment employs a time series forecasting model. For example, LSTM or RNN can be used to perform the inference: ;in, To predict the load vector, its dimension is force. The source is calculated by this module, and its function is to serve as the boundary condition for risk simulation, which is then input into the fourth processing module. The predictive model can be a Long Short-Term Memory (LSTM) network, obtained through offline training using historical monitoring data and construction logs. This is the historical total equivalent load vector sequence, sourced from the historical output of the first processing module; The data source is the data acquisition module, which is for future construction plans; the motivation behind this formula design is to obtain the maximum or most unfavorable loads that may occur in the future. This serves as the input for the fourth processing module to perform stress testing;

[0069] The system also includes a third processing module; this module is the core of the system and its purpose is to realize the self-adaptive evolution of the digital twin model; it uses measured data to iteratively correct the stiffness matrix of the model so that it can synchronously reflect the stiffness degradation of the hidden physical layer in the real structure caused by damage accumulation or stress redistribution.

[0070] This module receives the current total equivalent load vector. Its source is the first processing module and the measured displacement. Its source is the data acquisition module; at the same time, it maintains the current stiffness matrix. ;

[0071] Current stiffness matrix It refers to The digital twin model stiffness at time 10:00; at system startup, it is equal to the initial stiffness matrix. During operation, it equals the corrected stiffness matrix obtained from the previous iteration. ;

[0072] The correction process for this module is as follows: enter The simulated displacement of the model was calculated. To quantify the discrepancy between the model and reality, an i-node deviation index is introduced. The calculation formula is as follows:

[0073]

[0074] in, The deviation index of node i is dimensionless, i.e., m / m, and is obtained from the calculation of this module. The measured displacement vector of node i is derived from the data acquisition module. The simulated displacement vector for node i is derived from the module's... and Calculated; where, To prevent division by zero, a preset displacement threshold is set, such as the sensor's minimum resolution or a very small positive number, to ensure computational stability; this formula, namely the standardized L2 norm relative error, is motivated by the quantization of digital twin models. With real structure Deviation between;

[0075] To obtain the global bias, the average bias exponent δ of the Ns key sensing nodes is calculated:

[0076]

[0077] Where δ is the average deviation index, which is dimensionless and is obtained from the calculation in this module; The total number of critical nodes is derived from system presets; the motivation behind this formula, i.e., the arithmetic mean, is to provide a single scalar input for subsequent global stiffness correction.

[0078] According to the average deviation index The corrected stiffness matrix is ​​obtained through iterative correction. :

[0079]

[0080] in, The corrected stiffness matrix, with dimensions of force / displacement, is derived from calculations within this module and will be used as the stiffness matrix for the next time step. ; The correction factor is dimensionless; is a gain parameter between 0 and 1, and its value is determined based on the source of the data, calibrated through the Kalman filter algorithm or convergence tests based on historical data, to ensure stable convergence of the model correction; where, Set a preset lower limit for stiffness scaling to ensure Always maintain a positive attitude and prevent [the occurrence of certain events]. The abrupt change caused the model calculation to crash; correction coefficients Average deviation and stiffness scaling lower limit All are dimensionless scalars. The function returns a dimensionless scaling factor applied to the stiffness matrix. This formula is an iterative feedback control algorithm; its motivation is to create a closed-loop correction logic: it utilizes the previous step's... As a feedback signal, dynamically adjust get ,make The matrix can reflect the stiffness changes caused by damage accumulation in the real world;

[0081] The system also includes a fourth processing module; the purpose of this module is to perform destructive virtual simulations; it is based on the true stiffness corrected by the third processing module, i.e., the corrected stiffness matrix. The future load predicted by the second processing module, i.e., the predicted load vector. By simulating the progressive failure process, potential failure paths and key weak units can be identified;

[0082] The execution process of this module is as follows:

[0083] Perform damage simulation: Apply to Solving the finite element equations The predicted stress is obtained. ;

[0084] Perform failure path identification: check all units of Does it exceed its material failure threshold? Material failure threshold This refers to the yield strength of materials, etc., which is derived from material test data or relevant design specifications.

[0085] Perform progressive failure: when the cell When it fails, simulate its exit from the job, that is, from Remove the stiffness contribution of this element. The new stiffness matrix is ​​obtained. The motivation for this step is to simulate the chain reaction of load transfer to neighboring elements caused by the failure of one element; maintain Unchanged, based on Solve again and check if there is a cascading failure in a neighboring element k; repeat this process until the entire structure becomes unstable and record the failure chain.

[0086] Risk quantification: Monte Carlo simulation is used to quantify the risk. Uncertainty The above damage simulation, failure path identification, and progressive failure steps are repeated for the next sampling.

[0087] This module ultimately determines and outputs the following parameters for use by the security analysis module:

[0088] Total number of simulations This is the total number of executions of the Monte Carlo simulation, for example. ;

[0089] Number of simulations leading to overall failure This is a statistic. The total number of times the overall structural instability occurred in this simulation;

[0090] The risk contribution index πi of a unit is calculated using the following formula:

[0091]

[0092] Wherein, πi is the risk contribution index of unit i, and its dimensionless nature; risk contribution index It is an indicator used to quantify and locate the key weak units most likely to trigger a chain reaction; The number of times unit i appears in the failure chain is derived from statistics. In this simulation, the number of times unit i appears in the failure chain that leads to overall failure;

[0093] and the failure time of each failure simulation. This is in In each simulation that led to failure, record the results of that simulation. The time from the beginning to the occurrence of overall instability ,in ;

[0094] The system also includes a security analysis module; this module is the system's decision output layer; its purpose is to analyze the results of the fourth processing module, i.e. This transforms complex simulation data into clear risk indicators and early warning signals that can guide on-site actions;

[0095] The execution process of this module is as follows:

[0096] The overall failure probability Pf is determined by the following formula:

[0097]

[0098] Wherein, Pf is the overall failure probability, and its dimension is dimensionless; The number of simulations that led to the overall failure was determined by the fourth processing module. The total number of simulations originates from the fourth processing module; the motivation behind this formula, i.e., the definition of frequency probability, is to transform the binary question of safety into a question of probability. The probability of insecurity;

[0099] Determine the remaining safe window period When the number of simulations leading to overall failure When, the calculation formula is:

[0100]

[0101] in, The remaining safety window period is measured in time. The failure time in the k-th failure simulation originates from the fourth processing module; when When, it means at No failures occurred in any of the simulations, therefore the structure was deemed safe. Defined as greater than the prediction time window or greater than the highest time threshold The motivation behind this formula, namely the conservative estimation definition, is to adopt a worst-case strategy and provide an actionable, most urgent early warning window.

[0102] Output warning level This step is based on and The calculation results are used to implement dynamic hierarchical early warning; the system presets probability thresholds. and time threshold Threshold The settings are based on the project's risk tolerance and the needs of on-site personnel to implement intervention or evacuation, for example... It can be set to 12 hours, the shortest operable time;

[0103] The early warning logic includes: Level 1, namely, safety. The judgment condition is: and Level 2, i.e., attention The judgment condition is: or Level 3, i.e., early warning The judgment condition is: or ;

[0104] This embodiment constructs a complete technical closed loop from data acquisition to model correction, then to load prediction and risk simulation, and finally to decision-making and early warning through the collaborative work of the above-mentioned data acquisition module, first processing module, second processing module, third processing module, fourth processing module and security analysis module.

[0105] This system solves the technical problems of traditional monitoring systems that can only respond passively, such as only knowing about accidents after they occur, cannot predict the hidden stiffness degradation of structures, such as the accumulation of micro-cracks, and cannot quantify the risk of cascading failures.

[0106] This system ensures the digital twin model's integrity through the data assimilation function of the third processing module. The synchronization with the stiffness degradation of the actual structure ensures the accuracy of the deduction starting point; the progressive failure simulation through the fourth processing module reveals the cascading failure modes caused by local damage and stress redistribution that traditional linear analysis cannot detect; and finally, the output of the safety analysis module... That is, the probability of failure and The two core indicators, namely the safety window period, have enabled a shift from passive response to proactive prediction, providing a forward-looking and actionable basis for safety decision-making at construction sites.

[0107] Example 2:

[0108] The first processing module is used to determine the current total equivalent load vector, including:

[0109] Quantify construction activities and vibration data to obtain dynamic loads;

[0110] Quantifying environmental temperature changes and rainfall yields additional loads;

[0111] Based on the principle of load superposition, dynamic loads and additional loads are synthesized to obtain the current total equivalent load vector.

[0112] As a further refinement of Embodiment 1, this embodiment clarifies how the first processing module determines the current total equivalent load vector. The detailed process;

[0113] This process includes: quantifying construction activities and vibration data to obtain dynamic loads; the purpose of this step is to measure mechanical activities, such as... Identified heavy trucks passing by, and physical vibrations. Unified as mechanical load In this embodiment, the first processing module has a built-in dynamic response model library, which is used when the current construction log... When the instruction is for pile driving, the module retrieves the corresponding impact load model and combines it with vibration data. The actual amplitude is used to calculate the dynamic load applied to the corresponding node. ;Specifically, The calculation method is as follows: ,in, Based on the current construction log The corresponding influence coefficient matrix or weight vector is retrieved from the dynamic response model library; This is a dynamic transformation function used to convert the acquired vibration data. Converted into equivalent nodal dynamic forces through integration or frequency domain analysis;

[0114] The process also includes quantifying environmental temperature changes and rainfall to obtain additional loads; the purpose of this step is to transform non-construction-related environmental factors into mechanical loads. In this embodiment, the module determines the ambient temperature based on the collected data. Changes are observed, and the thermal stress of the structure is calculated using the coefficient of thermal expansion of the material; simultaneously, rainfall is considered. The potential static load of water deposition is calculated based on the catchment area of ​​the structure; the additional load is obtained by superimposing the two. ;Specifically, The calculation method is as follows Among them, thermal stress load , The coefficient of thermal expansion of the material. For elastic modulus, For cross-sectional area, The current temperature. Reference temperature; rainwater load , For the density of water, It is the acceleration due to gravity. For structural catchment area, Rainfall;

[0115] The process also includes: synthesizing dynamic loads and additional loads based on the principle of load superposition to obtain the current total equivalent load vector; this step is a summary of the above loads; the first processing module follows the physical principle of load superposition to... and Vector summation is performed at the nodes of the finite element model, and the final output is... That is, the current total equivalent load vector; this embodiment adopts The calculation method is based on the superposition principle in structural mechanics and elasticity, which states that under the assumptions of small deformation and linear elasticity, the response of a structure to multiple loads acting simultaneously is equal to the algebraic sum of the responses when each load acts alone. In this system, although the physical causes of dynamic loads and additional loads are different, they can be regarded as independent force vectors acting on nodes within the mechanical solution space of the finite element model, and they satisfy linear additivity when the structure is in the elastic working stage. This calculation method is based on the fact that the nonlinear coupling effect of the structural response is small compared to the main effect of the two loads within the range of conventional construction disturbances and environmental changes. Ignoring this coupling term is a reasonable simplification adopted to achieve the calculation efficiency required for real-time monitoring while ensuring engineering accuracy. Its error can be compensated by the closed-loop correction of the third processing module.

[0116] This embodiment uses dynamic loads. and additional loads The refined differentiation and quantification ensured The high accuracy of this differentiation process avoids conflating disturbances of different natures, such as dynamic effects caused by vibration and thermal stress caused by temperature changes. It can more realistically reflect the complex combined loads currently being borne by the structure, thereby providing a basis for the third processing module to calculate deviations. It provides more reliable input, improving the convergence speed and accuracy of subsequent model corrections.

[0117] Example 3:

[0118] The second processing module is used to determine the predicted load vector, including:

[0119] A time series forecasting model is used to deduce the predicted load vector in response to the historical total equivalent load vector sequence and future construction plans.

[0120] As a further refinement of Embodiment 1, this embodiment clarifies how the second processing module determines the predicted load vector. The way;

[0121] The method is as follows: a time series prediction model is used to deduce the predicted load vector in response to the historical total equivalent load vector sequence and the future construction plan.

[0122] In this embodiment, the time series prediction model used It can be a Long Short-Term Memory (LSTM) network;

[0123] The LSTM model is a type of recurrent neural network (RNN). Its construction method, i.e., its source, is based on historical monitoring data of the project. Sequence, and corresponding historical construction logs Obtained through offline training;

[0124] The input to this LSTM model consists of two parts: one is the historical total equivalent load vector sequence. The historical sequence refers to the sequence of events processed by the first processing module in the past. A time step, for example This represents the calculation over the past 72 hours. The sequence reflects the periodicity and fluctuation trend of the load; the second is the future construction plan. ; It refers to the future A time window, such as 24 hours, containing construction schedules like concrete pouring, is input into the model as a condition or disturbance term for the prediction; to achieve this fused prediction, the model ℳ, when processing the inputs, will include the future construction schedule as a conditional input. Convert into numerical feature vectors using word embedding or one-hot encoding. At each time step of the LSTM network The feature vector With the load vector at that time step The vectors are concatenated to form an augmented input vector. The feature vector is input into the LSTM unit; specifically, this feature vector... Represents the future The overall construction plan within the window, it Every step of each historical time step (from arrive All of them are related to that moment. The data is spliced ​​together and used as input conditions to enable the model to learn the evolution pattern of historical trends under specific future plans.

[0125] This enables the model to learn specific future construction plans. ) on the historical load time series trend ({ The dynamic influence pattern of});

[0126] LSTM model In response to these two inputs, I learned The inherent temporal dependencies of the sequence were learned, and specific... right The typical impact patterns of the sequence are analyzed, and the predicted load vector is ultimately derived and output. ;

[0127] This embodiment employs time series forecasting models such as LSTM, which are not merely simple linear extrapolations but can capture complex nonlinear time series relationships; more importantly, they integrate future construction plans... This crucial information significantly improved the accuracy of predictions; it enabled... Instead of blindly assuming the worst-case scenario, it is based on adverse loads that are highly probable to occur in the plan, which makes the risk simulation results of the fourth processing module more realistic and instructive, and avoids false warnings caused by overly conservative load predictions.

[0128] Example 4:

[0129] The third processing module is used for iterative correction to obtain the corrected stiffness matrix, including:

[0130] Input the current total equivalent load vector into the current stiffness matrix to calculate the simulated displacement;

[0131] Based on simulated displacement and measured displacement, the deviation index of key nodes is calculated.

[0132] Calculate the average deviation index of the critical nodes;

[0133] The current stiffness matrix is ​​corrected based on the average deviation index and the preset correction coefficient to obtain the corrected stiffness matrix.

[0134] As a further refinement of Example 1, this example clarifies the detailed steps of obtaining the corrected stiffness matrix through iterative correction by the third processing module;

[0135] The process includes: inputting the current total equivalent load vector into the current stiffness matrix to calculate the simulated displacement; in At that moment, the third processing module obtains Its source is the first processing module, and Its source is the previous cycle or By solving the static equations of the finite element method Calculate the model in Simulated displacement under action ;

[0136] The process also includes: calculating the deviation index of key nodes based on simulated and measured displacements; module acquisition. Its source is the data acquisition module; deviation index It is an indicator used to quantify the deviation between the model and reality; as described in Example 1, it uses a standardized L2 norm relative error formula. The calculation yielded, where A preset displacement threshold to prevent division by zero is used to ensure the measured displacement. Stability of calculations near zero;

[0137] The process also includes: calculating the average deviation index of key nodes; the average deviation index... It is all Key nodes Perform the arithmetic mean, that is ;this As a single scalar feedback signal for global stiffness correction;

[0138] The process also includes: correcting the current stiffness matrix based on the average deviation index and a preset correction coefficient to obtain the corrected stiffness matrix; this step is the core of the correction; correction coefficients... This is a key gain parameter, and its value is determined through Kalman filtering or historical data convergence testing to ensure the stability of the iteration; the module applies an iterative feedback control algorithm. ,use and right Scaling is performed to obtain the corrected stiffness matrix. ,in, A preset lower limit for stiffness scaling is used to ensure that the corrected stiffness matrix... Always maintain positive constant to prevent deviation due to average bias Excessive instantaneous magnitude leads to computational instability;

[0139] This embodiment is illustrated by... , arrive The complete closed-loop correction steps achieve an efficient data assimilation; it does not simply replace the model, but utilizes... The model-reality discrepancy represents a negative feedback signal that dynamically brings the digital twin model closer to reality. Compared with the actual structural stiffness; the stiffness used in this embodiment is... The formula implicitly assumes that structural stiffness only degrades and does not increase during the monitoring period; at the same time, the formula uses a global average deviation. For stiffness matrix Uniform scaling is a simplified model used to ensure computational efficiency and convergence, but it sacrifices physical fidelity. In other embodiments, model correction methods based on sensitivity analysis or extended Kalman filtering can also be used, utilizing local displacement deviations. To identify and correct The corresponding local stiffness element is used to obtain higher physical fidelity; this allows It can capture stiffness degradation caused by the accumulation of minor damage or stress redistribution that traditional monitoring methods, such as looking at displacement alone, cannot detect. This greatly improves the fidelity of the digital twin base and lays the foundation for the accuracy of subsequent risk simulation, i.e., the fourth processing module.

[0140] Example 5:

[0141] The fourth processing module simulates the progressive failure process, including:

[0142] The predicted load vector is applied to the modified stiffness matrix, and the predicted stress is obtained by solving.

[0143] Check whether the predicted stress exceeds the preset material failure threshold to identify failure cells;

[0144] When a failure element is identified, the stiffness contribution of the failure element is removed from the current stiffness matrix to obtain a new stiffness matrix. The solution is then repeated based on the new stiffness matrix until the entire structure becomes unstable, and the failure chain is recorded.

[0145] If no failure element is identified, the structure is determined to be stable under the predicted load vector.

[0146] The fourth processing module is used to determine the total number of simulations, the number of simulations leading to overall failure, the risk contribution index of the unit, and the failure time of each failure simulation, including:

[0147] By sampling the uncertainty of the predicted load vector through Monte Carlo simulation, the progressive failure process is repeatedly executed to obtain the total number of simulations.

[0148] The number of simulations that resulted in overall structural instability and ultimately overall failure is counted out of the total number of simulations.

[0149] In the number of simulations that lead to overall failure, record the failure time from the start of each simulation to the occurrence of overall failure;

[0150] Count the number of times each unit appears in a failure chain that leads to overall failure in the total number of simulations;

[0151] The risk contribution index of a unit is determined by dividing the number of times each unit appears in a failure chain that leads to overall failure by the total number of simulations.

[0152] As a further refinement of Example 1, this example combines and explains how the fourth processing module simulates the progressive failure process and how to determine various risk indicators;

[0153] The fourth processing module receives Its source is the third processing module, and Its source is the second processing module;

[0154] The module performs a single progressive failure process simulation:

[0155] The predicted load vector is applied to the modified stiffness matrix, and the predicted stress is obtained by solving the module; this module solves the static finite element equations. This is to solve the structure under future loads. and current true stiffness Underlying mechanical response, such as predicting stress The standard method;

[0156] The module checks whether the predicted stress exceeds the preset material failure threshold to identify failed elements; the module iterates through all elements. Comparing its Is it greater than the preset material failure threshold? ; The source is the yield strength in the material test data or design specifications;

[0157] When a failed element is identified, its stiffness contribution is removed from the current stiffness matrix to obtain a new stiffness matrix. The solution is then repeated based on this new stiffness matrix until the entire structure becomes unstable, and the failure chain is recorded. This is the core of progressive failure. At that time, the unit The element was identified as a failure; the module immediately updated the stiffness matrix. ,in It is a unit The stiffness contribution; the motivation for this is to simulate the failure of an element, causing a stress redistribution of the load to neighboring elements; the module maintains Unchanged, based on Repeat the solution process, checking if any neighboring elements k fail due to increased stress; repeat this process until a large number of elements fail. It becomes a singular matrix, meaning the entire structure becomes unstable; the sequence of unit failures throughout the process is recorded as a failure chain.

[0158] If no failed elements are identified, the structure is deemed stable under the predicted load vector; if, after the initial solution or any round of redistribution, all elements... All less than If the structure is stable, the simulation ends.

[0159] The module quantifies the above process:

[0160] Monte Carlo simulations are used to sample the uncertainty of the predicted load vector and repeatedly execute the progressive failure process to obtain the total number of simulations. The purpose of Monte Carlo simulations is to handle... Uncertainty; module to The probability distribution, for example, assuming it follows a mean of The predicted values ​​and variances are based on the normal distribution of historical prediction errors. Second sampling; That is, the total number of simulations; The method for determining the value is as follows: During the system debugging phase, dynamically increase... The value of and monitor the calculated overall failure probability. The convergence of; for example, when from Increase to At that time, the calculated and The relative change between them is less than the preset convergence threshold. At that time, it is determined It has converged; at this point... That is, the total number of simulations that satisfy the statistical convergence requirement; for each sampling... All of them repeatedly execute the progressive failure process;

[0161] The statistics include the number of simulations in which overall structural instability led to overall failure within the total number of simulations; module statistics are also included. The number of times the overall structural instability was triggered in this simulation is denoted as . That is, the number of simulations that led to overall failure;

[0162] In the number of simulations that lead to overall failure, record the failure time from the start of each simulation to the occurrence of overall failure; In the secondary failure simulation, if It is over time For varying scenarios, such as simulating a load process over 24 hours, the module records each simulation. Failure time from initiation to the occurrence of overall instability ;

[0163] Count the number of times each unit appears in a failure chain leading to overall failure in the total number of simulations; the module in In this simulation, the number of times each unit i appears in the failure chain leading to overall failure is counted. ;

[0164] The risk contribution index of a unit is determined by dividing the number of times each unit appears in the failure chain leading to overall failure by the total number of simulations. Subsequently identified as ; The motivation is to quantitatively locate the key weak unit most likely to trigger a chain reaction;

[0165] This embodiment uses progressive failure simulation, which goes beyond traditional one-time failure assessment. It can reveal the cascading failure modes caused by stress redistribution due to local damage in the structure, which is crucial for assessing the true safety of highly redundant structures.

[0166] Based on Monte Carlo simulations, this scheme incorporates uncertainty and outputs [results]. , and Key quantitative indicators; especially It no longer relies on experience to make judgments, but instead uses a large number of simulations to calculate weaknesses, which makes the early warning actions of the security analysis module, such as key inspections, have precise targets, greatly improving the efficiency and pertinence of security intervention.

[0167] Example 6:

[0168] The security analysis module is used to determine the overall probability of failure and the remaining safety window, including:

[0169] The overall failure probability is determined by the ratio of the number of simulations that lead to overall failure to the total number of simulations.

[0170] The minimum failure time from each failure simulation is taken to determine the remaining safety window.

[0171] As a further refinement of Example 1, this example clarifies how the security analysis module determines the overall failure probability and the remaining security window.

[0172] This module determines the overall failure probability based on the ratio of the number of simulations leading to overall failure to the total number of simulations; the security analysis module receives data from the fourth processing module. That is, the number of simulations that lead to overall failure, and That is, the total number of simulations; the overall failure probability. It is about the future A quantitative assessment of security over a given time period; its determination method is based on the standard frequency probability definition, calculated... The motivation behind this design is to transform the binary problem of safety / insecurity into a question of probability. The probability of insecurity provides a basis for tiered early warning systems.

[0173] This module is also used to determine the remaining safety window; the module receives data from the fourth processing module. and Failure time series corresponding to the second failure simulation Remaining safety window It is a time indicator used to guide emergency response; when When, its determination method is to take That is, in all In each failure scenario, the shortest time required for overall instability to occur; this design, i.e., taking the minimum value, is motivated by adopting a conservative estimate or worst-case strategy to provide on-site personnel with a warning time window before the most urgent intervention or evacuation must be completed; when When, it means at No failures occurred in any of the simulations, indicating that the structure is safe under the predicted loads. Defined as greater than the prediction time window or greater than the highest time threshold ;

[0174] This embodiment calculates... and It provides two complementary and highly condensed decision indicators; This answers the question of how great the risk is in terms of probability, and This dual-indicator output answers the question of how close the risk is in terms of time. This allows subsequent early warning decisions to no longer be triggered by a single threshold, but to formulate a more refined probability-time two-dimensional risk matrix, which greatly improves the scientific nature and operability of the early warning strategy.

[0175] Example 7:

[0176] The security analysis module is used to output warning levels, including:

[0177] The overall failure probability and the remaining safety window are compared with preset first probability threshold, second probability threshold, first time threshold and second time threshold.

[0178] When the overall failure probability is less than or equal to the first probability threshold and the remaining safety window is greater than or equal to the first time threshold, output the safety level.

[0179] When the overall failure probability is greater than the first probability threshold and less than or equal to the second probability threshold, or when the remaining safe window period is less than the first time threshold and greater than or equal to the second time threshold, the attention level is output.

[0180] When the overall failure probability is greater than the second probability threshold or the remaining safety window is less than the second time threshold, an early warning level is output.

[0181] The security analysis module is also used for:

[0182] When a level of concern or warning is output, it indicates that for units whose risk contribution index is greater than the preset risk threshold, key inspections or contingency reinforcement plans should be implemented.

[0183] This embodiment combines and explains how the security analysis module outputs warning levels and performs corresponding actions;

[0184] The module executes tiered early warning logic:

[0185] The overall failure probability and remaining safety window are compared with preset first probability thresholds, second probability thresholds, first time thresholds, and second time thresholds; the module receives... and And compare it with four preset thresholds: ;

[0186] Preset threshold The basis for determining this value, i.e., its source, is the project's risk tolerance and the minimum operable time required for on-site personnel to implement intervention or evacuation; for example... If a single on-site reinforcement operation takes 12 hours, then It can be set to 12 hours, and It can be set to 48 hours;

[0187] The comparison logic is as follows:

[0188] When the overall failure probability is less than or equal to the first probability threshold and the remaining safe window is greater than or equal to the first time threshold, i.e. and The system outputs the security level; if the system determines the risk is low and there is sufficient time, it outputs the security level. ;

[0189] When the overall failure probability is greater than the first probability threshold and less than or equal to the second probability threshold, or when the remaining safety window is less than the first time threshold and greater than or equal to the second time threshold, i.e. or The system outputs the level of concern; if the system determines the risk to be moderate or the time window to be shortening, it outputs the level of concern. ;

[0190] When the overall failure probability is greater than the second probability threshold or the remaining safety window is less than the second time threshold, i.e. or The system outputs an early warning level; if the system determines the risk to be high (i.e., high probability) or the situation to be urgent (i.e., short time), it outputs an early warning level. ;

[0191] The security analysis module is also used to execute early warning actions:

[0192] When a level of concern or a level of warning is output, it indicates that units whose risk contribution index to the unit is greater than the preset risk threshold should be subject to key inspections or contingency reinforcement.

[0193] This step completes the early warning loop; a preset risk threshold is set. Used to filter key units; The calibration method involves statistical analysis based on failure data from similar historical projects or a high-fidelity simulation database; for example, receiver operating characteristic (ROC) curves or precision-recall curves can be used for analysis by traversing all possible... The selected value is used as the candidate threshold. Precision and recall are calculated for each candidate threshold, and the threshold that maximizes precision while maintaining high recall is chosen. Value as For example, choosing the option that maximizes the F1 score. Values;

[0194] When the module outputs the attention level At that time, the system will instruct on-site personnel to process the risk contribution index identified by the fourth processing module. The units were subject to key inspections;

[0195] When the module outputs a warning level At that time, the system will instruct to immediately target these high-risk individuals. The unit was reinforced according to the contingency plan, and construction in the relevant area was suspended.

[0196] This embodiment constructs a hierarchical response and precise intervention intelligent decision-making mechanism; the system utilizes and Two dimensions enable a refined classification of risks: safety, attention, and early warning, avoiding the inefficiency of traditional black-and-white alarms and the "boy who cried wolf" effect.

[0197] More importantly, it will be an early warning action. Risk contribution index calculated by the fourth processing module Strong correlation enables precise, guided safety intervention; it no longer vaguely indicates a structural risk, but explicitly instructs that units A and B should be reinforced immediately because they are high-strength units. Units are the root cause of cascading failures; this greatly improves the targeting and efficiency of on-site handling and is the core of achieving proactive predictive maintenance.

[0198] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A building project safety monitoring system, characterised in that, The method comprises the following steps: a data acquisition module is configured to acquire monitoring parameters of the building structure, wherein the monitoring parameters comprise measured displacements of key nodes, vibration data, environmental temperature, rainfall, current construction logs and future construction plans, and an initial stiffness matrix is acquired; a first processing module is configured to determine a current total equivalent load vector based on the current construction logs, the vibration data, the environmental temperature and the rainfall; a second processing module is configured to determine a predicted load vector based on a historical total equivalent load vector sequence determined by the first processing module at a historical time and the future construction plans; a third processing module is configured to iteratively correct a current stiffness matrix to obtain a corrected stiffness matrix, wherein the current stiffness matrix is the initial stiffness matrix or a corrected stiffness matrix obtained at a previous time through iteration based on the current total equivalent load vector, the measured displacements and the current stiffness matrix; a fourth processing module is configured to simulate a progressive failure process based on the corrected stiffness matrix and the predicted load vector, to determine a total simulation number, a simulation number leading to overall failure, a risk contribution index of a unit and a failure time of each failure simulation; a safety analysis module is configured to determine an overall failure probability and a remaining safety window period based on the total simulation number, the simulation number leading to overall failure, the risk contribution index of the unit and the failure time of each failure simulation, and to output a warning level; the fourth processing module simulates the progressive failure process, which comprises: applying the predicted load vector to the corrected stiffness matrix to obtain a predicted stress; checking whether the predicted stress exceeds a preset material failure threshold to identify a failed unit; when a failed unit is identified, removing the stiffness contribution of the failed unit from the current stiffness matrix to obtain a new stiffness matrix, and repeatedly solving based on the new stiffness matrix until the overall structure is unstable, and recording a failure chain; when no failed unit is identified, determining that the structure is stable under the predicted load vector; the third processing module is configured to iteratively correct the current stiffness matrix to obtain the corrected stiffness matrix, which comprises: inputting the current total equivalent load vector into the current stiffness matrix to calculate a simulated displacement; calculating a deviation index of the key nodes based on the simulated displacement and the measured displacement; calculating an average deviation index of the key nodes; correcting the current stiffness matrix based on the average deviation index and a preset correction coefficient to obtain the corrected stiffness matrix.

2. The construction project safety monitoring system of claim 1, wherein, The first processing module is configured to determine the current total equivalent load vector, which comprises: quantifying construction activities and vibration data to obtain a dynamic load; quantifying environmental temperature changes and rainfall to obtain an additional load; combining the dynamic load and the additional load based on the load superposition principle to obtain the current total equivalent load vector.

3. The construction project safety monitoring system of claim 1, wherein, The second processing module is configured to determine the predicted load vector, which comprises: using a time series prediction model to deduce the predicted load vector in response to the historical total equivalent load vector sequence and the future construction plans.

4. The construction project safety monitoring system of claim 1, wherein, The fourth processing module is configured to determine the total simulation number, the simulation number leading to overall failure, the risk contribution index of the unit and the failure time of each failure simulation, which comprises: sampling the uncertainty of the predicted load vector through Monte Carlo simulation, repeatedly executing the progressive failure process to obtain the total simulation number; Counting the number of simulations leading to overall failure in the total number of simulations; Recording the failure time of each simulation from the beginning to the overall failure in the number of simulations leading to overall failure; Counting the number of times each unit appears in the failure chain leading to overall failure in the total number of simulations; Dividing the number of times each unit appears in the failure chain leading to overall failure by the total number of simulations to determine the risk contribution index of the unit.

5. The construction project safety monitoring system of claim 1, wherein, The safety analysis module is configured to determine the overall failure probability and the remaining safety window period, and includes: Determining the overall failure probability according to the ratio of the number of simulations leading to overall failure to the total number of simulations; Determining the remaining safety window period by taking the minimum value of the failure time of each failure simulation.

6. The construction project safety monitoring system of claim 5, wherein, The safety analysis module is configured to output the warning level, and includes: Comparing the overall failure probability and the remaining safety window period with preset first probability threshold, second probability threshold, first time threshold and second time threshold; Outputting a safety level when the overall failure probability is less than or equal to the first probability threshold and the remaining safety window period is greater than or equal to the first time threshold; Outputting an attention level when the overall failure probability is greater than the first probability threshold and less than or equal to the second probability threshold, or the remaining safety window period is less than the first time threshold and greater than or equal to the second time threshold; Outputting a warning level when the overall failure probability is greater than the second probability threshold or the remaining safety window period is less than the second time threshold.

7. The construction project safety monitoring system of claim 6, wherein, The safety analysis module is further configured to: When the attention level or the warning level is output, instructing to perform key inspection or preplan reinforcement on the unit whose risk contribution index is greater than the preset risk threshold.

Images