A method for intelligent management of construction engineering data information

Abstract

The application discloses a kind of construction engineering data information intelligent management method, it is related to information management technical field, and multi-source heterogeneous data of construction site is collected and aligned, forms engineering alignment dataset, extracts construction state characteristics based on alignment dataset and calculates comprehensive deviation index, constructs dynamic twin scene and generates the health state mapping of key component, generates construction optimization strategy based on health state mapping and deviation index and executes control, the present application is fused multi-source heterogeneous data and dynamic twin model, realize the real-time monitoring, prediction and optimization control of construction site state, quantitatively analyze structure deviation and automatically generate optimal construction strategy, realize model self-evolution and parameter self-optimization by feedback learning, significantly improve the safety and intelligent level of construction engineering, reduce energy consumption and time limit for a project delay rate, realize the change from static monitoring to active decision.

Description

Technical Field

[0001] This invention relates to the field of information management technology, and in particular to an intelligent management method for construction engineering data information. Background Technology

[0002] With the rapid development of complex engineering projects such as large public buildings, rail transit hubs, and hospital complexes, construction sites are characterized by diverse data sources, complex environmental conditions, and numerous participants. Traditional construction management systems mostly rely on static BIM models and schedules, which can only visualize design information and planned tasks, but cannot reflect real-time changes in construction status, let alone dynamically correlate multi-source heterogeneous data such as structural stress, temperature and humidity, and equipment operating conditions.

[0003] Currently, Chinese patent application number CN202410014286.2 discloses a method for water conservancy engineering information management based on big data. The method includes: dividing a line graph of water level data into regions based on the water level change corresponding to the initial water level data, identifying high-frequency and low-frequency water level regions; calculating the wavelet decomposition level for each region in the high-frequency and low-frequency water level regions based on the wavelet layer score factors corresponding to the high-frequency and low-frequency water level regions; and finally, denoising the initial water level data for each region in the high-frequency and low-frequency water level regions based on the wavelet decomposition level and a preset denoising algorithm to confirm the final denoised water level data. By dividing the line graph of water level data into regions and denoising the initial water level data for each region according to the wavelet decomposition level, the accuracy of denoising is improved compared to traditional denoising methods, thereby reducing denoising costs.

[0004] The relevant technologies are insufficient to achieve real-time fusion and dynamic twin modeling of multi-source heterogeneous data during the construction process, and cannot realize intelligent prediction of construction status, self-correction of deviations, and self-evolution management of cross-project knowledge in complex environments. Summary of the Invention

[0005] The technical problem solved by this invention is that existing technologies are unable to achieve real-time fusion and dynamic twin modeling of multi-source heterogeneous data during the construction process, and cannot achieve intelligent prediction of construction status, self-correction of deviations, and self-evolution management of cross-project knowledge in complex environments.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] A method for intelligent management of construction project data includes the following steps:

[0008] Step S1: Collect and align multi-source heterogeneous data from the construction site to form an engineering alignment dataset with spatiotemporal labels;

[0009] Step S2: Extract construction status features based on the aligned dataset and calculate the comprehensive deviation index to determine the deviation status between the structure and the environment;

[0010] Step S3: Construct a construction dynamic twin scenario based on the engineering alignment dataset and structural deviation result set. Use 3D BIM data as the geometric baseline to establish a construction dynamic twin model consisting of a state perception layer, a parameter mapping layer, and a prediction inference layer.

[0011] The state perception layer realizes multi-sensor state estimation based on the Kalman filter fusion model, the parameter mapping layer maps the estimated state to BIM geometric nodes to form a dynamic field distribution, and the prediction inference layer combines LSTM temporal network and Bayesian inference model to generate health status mapping of key components.

[0012] Step S4: Based on the health status mapping and comprehensive deviation index, a construction optimization strategy is generated. The analytic hierarchy process and multi-objective weighted scoring model are used to evaluate the potential for safety improvement, energy consumption impact and schedule disturbance. The strategy with the highest comprehensive score is selected as the control instruction and executed.

[0013] Step S5: Collect environmental parameters, work quality and energy consumption data after control execution; use Z-score detection, gradient descent and genetic algorithm to adaptively correct deviation threshold and monitoring weight; and update construction dynamic twin model parameters through reinforcement learning mechanism.

[0014] Step S6: Version the correction parameters and the engineering alignment dataset, and generate an empirical parameter mapping table by comparing with knowledge snapshots;

[0015] Step S2 includes the following sub-steps:

[0016] Step S201: Read the engineering alignment dataset, use the data analysis engine to perform structured analysis on the original multi-source data, and extract the structural stress change curve, temperature and humidity fluctuation trend and equipment operation cycle characteristics.

[0017] When calculating the stability of key parameters, a time series smoothing algorithm based on a sliding window and a standard deviation evaluation method are used to quantify the fluctuation range and steady-state range of each parameter in the time series.

[0018] Step S202: Analyze the strain difference, temperature difference and displacement gradient between adjacent monitoring points, and construct the spatial distribution feature matrix using the Kriging interpolation algorithm and a three-dimensional finite element analysis model;

[0019] By comparing the consistency of displacement direction and strain coupling at different monitoring points, local stress concentration areas and abnormal temperature difference areas can be identified.

[0020] Step S203: Based on the structural stress change curve, temperature and humidity fluctuation trend, equipment operation cycle characteristics and spatial distribution characteristic matrix, and according to the sensitivity factors of component type and construction stage, the multidimensional features are reduced and fused using fuzzy comprehensive evaluation algorithm and principal component analysis method, the comprehensive deviation index is calculated, and the deviation feature set is output.

[0021] Step S204: When the comprehensive deviation index is detected to exceed the preset range, the abnormality persistence is analyzed based on threshold judgment and ARIMA residual detection, abnormal components are marked and deviation alarm labels are generated to form a structural deviation result set.

[0022] Step S3 includes the following sub-steps:

[0023] Step S301: Based on the engineering alignment dataset and the structural deviation result set, construct a construction dynamic twin scenario in the digital model;

[0024] During the construction process, 3D BIM data is used as the geometric baseline, and a construction dynamic twin model is established driven by real-time collected multi-source alignment data.

[0025] The construction dynamic twin model consists of a state perception layer, a parameter mapping layer, and a prediction and inference layer. The state perception layer uses a Kalman filter fusion model to achieve real-time fusion and state estimation of multi-sensor data on site. The parameter mapping layer maps the estimated state to BIM geometric nodes to form a dynamic field distribution. The prediction and inference layer uses an LSTM time series network and a Bayesian inference model to predict the future state and perform uncertainty analysis.

[0026] The construction dynamic twin model continuously overlays historical data and real-time data streams during operation, forming a spatiotemporal digital mapping of the construction process;

[0027] Step S302: Map the real-time monitoring points to the nodes of the construction dynamic twin model. Use the ICP point cloud registration algorithm and the deep neural network attitude recognition model to match and correct the sensor position information and the model nodes. The rendering engine dynamically generates the state distribution and historical change trajectory of each component and displays the multi-dimensional state parameters in color coding. The multi-dimensional state parameters include component stress, temperature and vibration.

[0028] Step S303: Calculate the component health index using the data evolution sequence in the twin scenario, and use a multidimensional regression model and fuzzy inference system to comprehensively score historical deviation, stress stability, and environmental interference.

[0029] During the health index generation process, an LSTM-based time series prediction model is used to identify future trends, and the output is a health status map to represent the safety level of each component in the current and prediction stages.

[0030] Step S304, the health status mapping includes the real-time safety level, cumulative deviation trend and construction stage confidence score of each component;

[0031] The confidence score was calculated using a dynamic confidence calculation method based on a Bayesian update model.

[0032] Preferably, step S1 includes the following sub-steps:

[0033] Step S101: Collect temperature, humidity, stress, vibration, video images, mechanical operating parameters and personnel positioning information at the construction site, and output raw multi-source data;

[0034] Step S102: Time synchronization and coordinate alignment are performed on the original multi-source data. A unified timestamp mapping is performed based on the construction master control clock, and the output is the preliminary aligned data.

[0035] Step S103: Combining the construction task plan and component number information, the preliminary alignment data is associated with the corresponding component nodes in the digital model to form an engineering alignment dataset with spatiotemporal labels and component identifiers.

[0036] Preferably, step S4 includes the following sub-steps:

[0037] Step S401: Read the health status mapping and structural deviation result set, and prioritize the components with risks.

[0038] The ranking process comprehensively considers the health index in the health status mapping, real-time safety level, cumulative deviation trend and construction stage confidence score, and combines the deviation alarm labels in the structural deviation result set to output a risk priority list.

[0039] Step S402: Based on the priority list and construction schedule, generate a set of candidate optimization strategies;

[0040] The set of candidate optimization strategies includes job sequence adjustment, equipment scheduling, and environmental control parameter adjustment.

[0041] Step S403: Use a multi-objective weighted analysis method to quantitatively evaluate each candidate optimization strategy;

[0042] The multi-objective weighted analysis method combines the analytic hierarchy process and the weighted scoring model to comprehensively score the execution cost, energy consumption impact, safety improvement potential, and the degree of disturbance to the overall construction progress. Based on the comprehensive score, the strategy with the highest score is selected as the optimal control instruction.

[0043] Step S404: Automatically issue optimal control commands and record execution logs within the preset execution window;

[0044] The execution log includes the instruction issuance time, target component identifier, list of participating equipment, environmental control parameter adjustment values, and information on delayed work procedures. The execution log and control instructions together form control execution data.

[0045] Preferably, step S5 includes the following sub-steps:

[0046] Step S501: Collect environmental parameters, operation quality data and energy consumption statistics after control execution to form a feedback dataset;

[0047] The feedback dataset is collected based on an IoT data acquisition system and on-site quality inspection terminals to obtain real-time information on construction environment temperature and humidity, noise intensity, equipment power curves, and work completion.

[0048] Outliers are eliminated using Z-score outlier detection and moving mean smoothing algorithms, and indexed by timestamp and component number.

[0049] Step S502: Correct the threshold of the comprehensive deviation index based on the feedback dataset and readjust the weight coefficients of each monitoring parameter;

[0050] The correction process employs an adaptive threshold correction algorithm and a gradient descent-based parameter optimization model to analyze the difference between the actual improvement magnitude and the predicted improvement value in the feedback dataset, and automatically fine-tunes the comprehensive deviation index threshold based on the difference ratio.

[0051] A multi-parameter optimization method based on genetic algorithm is used to redistribute the weight coefficients of each monitoring parameter;

[0052] Step S503: By analyzing the time difference between control actions and deviation improvements in the execution log, perform lag compensation correction;

[0053] The lag compensation analysis uses a time series correlation analysis model and a dynamic time warping algorithm to extract the time delay characteristics of control actions and deviation changes, and establish lag response curves.

[0054] By combining the error estimation model based on Kalman filtering, the hysteresis compensation coefficient is adjusted, and the execution step size is corrected based on the feedback data;

[0055] Step S504: Output the set of correction parameters. The set of correction parameters is used for model self-learning and threshold update. The set of correction parameters includes the updated comprehensive deviation threshold, monitoring weight vector and hysteresis compensation coefficient. Input the set of correction parameters into the self-learning module of the construction dynamic twin model.

[0056] The self-learning module uses reinforcement learning algorithms to review historical execution data and optimize the model's prediction accuracy and control stability.

[0057] Preferably, step S6 includes the following sub-steps:

[0058] Step S601: Version archive the set of corrected parameters and the engineering alignment dataset to form a knowledge snapshot;

[0059] Step S602: Based on the knowledge snapshot, perform multi-project comparative analysis to generate an experience parameter mapping table and establish an engineering knowledge base;

[0060] Step S603: When a new project is started, extract templates with scene parameter similarity greater than a preset threshold from the engineering knowledge base;

[0061] Step S604: Through a continuous learning mechanism, the construction dynamic twin model is updated in features and calibrated in deviation after new data input.

[0062] Preferably, step S203 further includes a dynamic weight adjustment mechanism;

[0063] The dynamic weight adjustment mechanism adjusts the weight ratios of stress, temperature, and humidity in the comprehensive deviation index calculation in real time according to the changing trends of construction activities in different time periods. When a high-intensity operation phase is detected, the weight of stress data is increased to a preset percentage.

[0064] Preferably, step S301 further includes an adaptive model update mechanism;

[0065] The adaptive model update mechanism automatically adjusts the model refresh frequency based on the difference between the actual feedback data and the prediction of the construction dynamic twin model. When the prediction deviation continues to exceed the limit, the refresh cycle is shortened; when the operation is stable, the refresh cycle is extended.

[0066] Preferably, step S503 further includes a delayed learning mechanism when performing hysteresis compensation correction;

[0067] The delayed learning mechanism calculates the optimal lag correction step size by backtracking the time series differences between control execution data and deviation improvement data, and automatically applies it in subsequent control execution.

[0068] The beneficial effects of this invention are as follows: By integrating multi-source heterogeneous data with a dynamic twin model, this invention enables real-time monitoring, prediction, and optimized control of the construction site status, quantitative analysis of structural deviations and automatic generation of optimal construction strategies, and self-evolution and parameter self-optimization of the model through feedback learning. This significantly improves the safety and intelligence level of building engineering, reduces energy consumption and construction delay rates, and realizes the transformation from static monitoring to proactive decision-making. Attached Figure Description

[0069] Figure 1 The flowchart illustrates the steps of an intelligent management method for building engineering data information according to an embodiment of the present invention. Detailed Implementation

[0070] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0071] Example, refer to Figure 1 This paper provides an intelligent management method for construction engineering data information, which includes the following steps:

[0072] Step S1: Collect and align multi-source heterogeneous data from the construction site to form an engineering alignment dataset with spatiotemporal labels;

[0073] Step S2: Extract construction status features based on the aligned dataset and calculate the comprehensive deviation index to determine the deviation status between the structure and the environment;

[0074] Step S3: Construct a construction dynamic twin scenario based on the engineering alignment dataset and structural deviation result set. Use 3D BIM data as the geometric baseline to establish a construction dynamic twin model consisting of a state perception layer, a parameter mapping layer, and a prediction inference layer.

[0075] The state perception layer uses a Kalman filter fusion model to achieve multi-sensor state estimation. The parameter mapping layer maps the estimated state to BIM geometric nodes to form a dynamic field distribution. The prediction inference layer combines an LSTM temporal network and a Bayesian inference model to generate a health status mapping of key components.

[0076] Step S4: Based on the health status mapping and comprehensive deviation index, a construction optimization strategy is generated. The analytic hierarchy process and multi-objective weighted scoring model are used to evaluate the potential for safety improvement, energy consumption impact and schedule disturbance. The strategy with the highest comprehensive score is selected as the control instruction and executed.

[0077] Step S5: Collect environmental parameters, work quality and energy consumption data after control execution; use Z-score detection, gradient descent and genetic algorithm to adaptively correct deviation threshold and monitoring weight; and update construction dynamic twin model parameters through reinforcement learning mechanism.

[0078] Step S6: Version the correction parameters and the engineering alignment dataset, and generate an empirical parameter mapping table by comparing with knowledge snapshots.

[0079] This invention integrates multi-source heterogeneous data with a dynamic twin model to achieve real-time monitoring, prediction, and optimized control of construction site conditions. It quantifies structural deviations and automatically generates optimal construction strategies. Through feedback learning, it enables model self-evolution and parameter self-optimization, significantly improving the safety and intelligence level of building engineering, reducing energy consumption and construction delay rates, and realizing the transformation from static monitoring to proactive decision-making.

[0080] Step S1 includes the following sub-steps:

[0081] Step S101: Collect temperature, humidity, stress, vibration, video images, mechanical operating parameters and personnel positioning information at the construction site, and output raw multi-source data;

[0082] By synchronously collecting temperature, humidity, stress, vibration, video images, mechanical operating parameters and personnel positioning information, we can achieve full coverage perception of multiple physical fields and multiple roles at the construction site, ensuring that the data sources are comprehensive, real-time and traceable.

[0083] Step S102: Time synchronization and coordinate alignment are performed on the original multi-source data. A unified timestamp mapping is performed based on the construction master control clock, and the output is the preliminary aligned data.

[0084] By unifying timestamps and coordinate system mapping, data synchronization and spatial registration between different sampling frequencies and acquisition devices are achieved, eliminating asynchronous errors and enabling data to be comparable across devices and scenarios, laying a consistent foundation for component-level state analysis.

[0085] Step S103: Combining the construction task plan and component number information, the preliminary alignment data is associated with the corresponding component nodes in the digital model to form an engineering alignment dataset with spatiotemporal labels and component identifiers;

[0086] By unifying timestamps and coordinate system mapping, data synchronization and spatial registration between different sampling frequencies and acquisition devices are achieved, eliminating asynchronous errors and enabling data to be comparable across devices and scenarios, laying a consistent foundation for component-level state analysis.

[0087] Step S1 achieves unified collection, time synchronization, and spatial mapping of multi-source heterogeneous data from the construction site, constructing an engineering aligned dataset that can support subsequent analysis. By accurately mapping different types of data in time and space dimensions, a global data baseline for the construction status is formed, providing consistent data support for dynamic twin modeling and deviation analysis.

[0088] Step S2 includes the following sub-steps:

[0089] Step S201: Read the engineering alignment dataset, use the data analysis engine to perform structured analysis on the original multi-source data, and extract the structural stress change curve, temperature and humidity fluctuation trend and equipment operation cycle characteristics.

[0090] When calculating the stability of key parameters, a time series smoothing algorithm based on a sliding window and a standard deviation evaluation method are used to quantify the fluctuation range and steady-state range of each parameter in the time series.

[0091] By performing structured analysis on the engineering alignment dataset, the system extracts key temporal features reflecting changes in construction status, including stress curves, temperature and humidity trends, and equipment operating cycle characteristics. The system uses a sliding window smoothing algorithm and standard deviation evaluation method to quantify the fluctuation range, thereby achieving accurate characterization of the stability and fluctuation trends of each monitoring parameter.

[0092] Step S202: Analyze the strain difference, temperature difference and displacement gradient between adjacent monitoring points, and construct the spatial distribution feature matrix using the Kriging interpolation algorithm and a three-dimensional finite element analysis model.

[0093] By comparing the consistency of displacement direction and strain coupling at different monitoring points, local stress concentration areas and abnormal temperature difference areas can be identified.

[0094] By using the Kriging interpolation algorithm and a three-dimensional finite element analysis model, a spatial distribution feature matrix between monitoring points is constructed to achieve a continuous spatial description of structural deformation and thermal stress. By comparing the strain consistency and temperature coupling relationship between monitoring points, local stress concentration areas and abnormal heat distribution areas can be accurately located, providing a spatial basis for early warning of potential structural risks.

[0095] Step S203: Based on the structural stress change curve, temperature and humidity fluctuation trend, equipment operation cycle characteristics, and spatial distribution characteristic matrix, and according to the sensitivity factors of component type and construction stage, the multidimensional features are reduced and fused using fuzzy comprehensive evaluation algorithm and principal component analysis method, the comprehensive deviation index is calculated, and the deviation feature set is output.

[0096] Step S203 further includes a dynamic weight adjustment mechanism.

[0097] The dynamic weight adjustment mechanism adjusts the weight ratios of stress, temperature, and humidity in the comprehensive deviation index calculation in real time according to the changing trends of construction activities in different time periods. When a high-intensity operation phase is detected, the weight of stress data is increased to the preset percentage.

[0098] By using the Kriging interpolation algorithm and a three-dimensional finite element analysis model, a spatial distribution feature matrix between monitoring points is constructed to achieve a continuous spatial description of structural deformation and thermal stress. By comparing the strain consistency and temperature coupling relationship between monitoring points, local stress concentration areas and abnormal heat distribution areas can be accurately located, providing a spatial basis for early warning of potential structural risks.

[0099] Step S204: When the comprehensive deviation index is detected to exceed the preset range, the abnormality persistence is analyzed based on threshold judgment and ARIMA residual detection, abnormal components are marked and deviation alarm labels are generated to form a structural deviation result set.

[0100] By using the Kriging interpolation algorithm and a three-dimensional finite element analysis model, a spatial distribution feature matrix between monitoring points is constructed to achieve a continuous spatial description of structural deformation and thermal stress. By comparing the strain consistency and temperature coupling relationship between monitoring points, local stress concentration areas and abnormal heat distribution areas can be accurately located, providing a spatial basis for early warning of potential structural risks.

[0101] Step S2 enables in-depth analysis and deviation quantification of multi-source aligned data at the construction site. Through time series analysis, spatial modeling, and multi-dimensional feature fusion, a comprehensive deviation index is established and abnormal components are identified. This realizes an intelligent analysis process from data perception to risk quantification, providing a quantitative basis for subsequent dynamic twin modeling and optimization control.

[0102] Step S3 includes the following sub-steps:

[0103] Step S301: Based on the engineering alignment dataset and the structural deviation result set, construct a construction dynamic twin scenario in the digital model.

[0104] During the construction process, 3D BIM data is used as the geometric baseline, and a construction dynamic twin model is established driven by real-time collected multi-source alignment data.

[0105] The construction dynamic twin model consists of a state perception layer, a parameter mapping layer, and a prediction and inference layer. The state perception layer uses a Kalman filter fusion model to achieve real-time fusion and state estimation of multi-sensor data on site. The parameter mapping layer maps the estimated state to BIM geometric nodes to form a dynamic field distribution. The prediction and inference layer uses an LSTM temporal network and a Bayesian inference model to predict the future state and perform uncertainty analysis.

[0106] The construction dynamic twin model continuously overlays historical data and real-time data streams during operation, forming a spatiotemporal digital mapping of the construction process.

[0107] Step S301 further includes an adaptive model update mechanism.

[0108] The adaptive model update mechanism automatically adjusts the model refresh frequency based on the difference between the actual feedback data and the prediction of the construction dynamic twin model. When the prediction deviation continues to exceed the limit, the refresh cycle is shortened, and when the operation is stable, the refresh cycle is extended.

[0109] By establishing a construction dynamic twin model based on 3D BIM data and multi-source aligned data, the system realizes a panoramic digital mapping of the on-site physical state in virtual space.

[0110] The state perception layer uses a Kalman filter fusion model to integrate multi-sensor inputs in real time, improving the accuracy of state estimation; the parameter mapping layer maps the state results to BIM geometric nodes, forming a real-time dynamic field distribution; the prediction and inference layer combines an LSTM temporal network with a Bayesian inference model to achieve predictive analysis of future structural states and uncertainties.

[0111] The adaptive model update mechanism automatically adjusts the model refresh frequency based on prediction deviations, enabling the model to maintain high consistency and stability with the actual construction conditions over a long period of time.

[0112] Step S302: Map the real-time monitoring points to the nodes of the construction dynamic twin model. Use the ICP point cloud registration algorithm and the deep neural network attitude recognition model to match and correct the sensor position information with the model nodes. The rendering engine dynamically generates the state distribution and historical change trajectory of each component and displays the multi-dimensional state parameters in color coding. The multi-dimensional state parameters include component stress, temperature and vibration.

[0113] By spatially mapping real-time monitoring points with twin model nodes, precise alignment between the physical site and the digital model is achieved.

[0114] The ICP point cloud registration algorithm is used to correct sensor pose errors, and a deep neural network pose recognition model is combined to improve matching accuracy, ensuring geometric consistency of model mapping. The rendering engine generates component state distribution and historical trajectory in real time, and dynamically displays key indicators such as stress, temperature, and vibration using color coding, realizing multi-dimensional visualization monitoring and structural state evolution tracking.

[0115] Step S303: Calculate the component health index using the data evolution sequence in the twin scenario, and use a multidimensional regression model and fuzzy inference system to comprehensively score historical deviation, stress stability, and environmental interference.

[0116] The health index generation process incorporates an LSTM-based time-series prediction model to identify future trends, and outputs a health status map to represent the safety level of each component in the current and prediction stages.

[0117] By calculating the component health index through the data evolution sequence in a twin scenario, a quantitative assessment of the component's safety status can be achieved.

[0118] The multidimensional regression model and fuzzy inference system comprehensively consider factors such as historical bias, stress stability and environmental disturbance, and output a comprehensive health score. Combined with the LSTM time series prediction model, it predicts future trends and forms a health status mapping, providing a scientific basis for structural risk assessment and early intervention.

[0119] Step S304, the health status mapping includes the real-time safety level, cumulative deviation trend and construction stage confidence score of each component.

[0120] The confidence score was calculated using a dynamic confidence calculation method based on a Bayesian update model.

[0121] A complete indicator system is generated in the health status mapping, including real-time safety level, cumulative deviation trend and construction stage confidence score.

[0122] The confidence score is dynamically calculated based on the Bayesian update model, which comprehensively considers the reliability of monitoring data and sensor redundancy. It measures the accuracy of the twin model in fitting the actual state, providing a reliable evaluation benchmark for subsequent optimization control and risk decision-making.

[0123] Step S3, by constructing a dynamic construction twin model, achieves real-time mapping and collaborative evolution between the on-site physical state and the digital model, enabling the system to dynamically perceive, predict, and assess the health status of building components. By integrating multi-source aligned data and structural deviation result sets, the system reflects construction progress and structural changes in real time within the virtual model, realizing the transformation from static BIM display to intelligent twin cognition, and providing high-precision and predictable digital basis for construction optimization decisions.

[0124] Step S4 includes the following sub-steps:

[0125] Step S401: Read the health status mapping and structural deviation result set, and prioritize the components with risks.

[0126] The ranking process comprehensively considers the health index in the health status mapping, real-time safety level, cumulative deviation trend and construction stage confidence score, and combines the deviation alarm labels in the structural deviation result set to output a risk priority list.

[0127] By reading the health status mapping and structural deviation result set, the system prioritizes potentially risky components in the construction structure.

[0128] The ranking takes into account health index, real-time safety level, cumulative deviation trend and construction stage confidence score, and combines deviation alarm labels for weighted evaluation to form a risk priority list.

[0129] This process enables the automatic identification and classification of key risk points, allowing subsequent optimization strategies to focus on high-risk areas and reduce errors from subjective human judgment.

[0130] Step S402: Based on the priority list and construction schedule, generate a set of candidate optimization strategies.

[0131] The set of candidate optimization strategies includes job sequence adjustment, equipment scheduling, and environmental control parameter adjustment.

[0132] Based on the risk priority list and construction schedule, a set of candidate optimization strategies is automatically generated.

[0133] Candidate strategies cover three major intervention areas: job sequence adjustment, equipment scheduling, and environmental control.

[0134] Reduce the risk of structural superposition loads by adjusting the sequence of work processes; optimize construction energy consumption and machinery utilization by scheduling equipment; and improve temperature and humidity conditions and vibration levels by controlling environmental parameters.

[0135] This process combines a historical scene retrieval model with a rule-based reasoning engine to achieve intelligent reuse and rapid generation of strategy solutions.

[0136] Step S403: Use a multi-objective weighted analysis method to quantitatively evaluate each candidate optimization strategy.

[0137] The multi-objective weighted analysis method combines the analytic hierarchy process (AHP) and a weighted scoring model to comprehensively score the execution cost, energy consumption impact, safety improvement potential, and degree of disturbance to the overall construction progress. Based on the comprehensive score, the strategy with the highest score is selected as the optimal control instruction.

[0138] The generated candidate optimization strategies are quantitatively evaluated from multiple dimensions, and multi-objective optimization calculations are performed using the analytic hierarchy process and a weighted scoring model.

[0139] A standardized scoring matrix is ​​formed by comprehensively considering four indicators: execution cost, energy consumption impact, safety improvement potential, and schedule disturbance degree.

[0140] Finally, the optimal control command is automatically selected based on the comprehensive scoring results, realizing algorithm-driven strategy selection and resource allocation optimization.

[0141] Step S404: Automatically issue optimal control commands and record execution logs within the preset execution window.

[0142] The execution log includes the instruction issuance time, target component identification, list of participating equipment, environmental control parameter adjustment values, and information on delayed work procedures. The execution log and control instructions together form control execution data.

[0143] Within the preset execution window, the system automatically issues optimal control commands to the corresponding construction equipment and control system, and simultaneously generates execution logs.

[0144] The log records include the instruction issuance time, target components, equipment list, parameter adjustment values, and process change information, which are used for subsequent feedback learning and model correction.

[0145] This step enables automated execution of optimization decisions and full-process traceability management, providing high-quality data support for closed-loop system learning and continuous improvement.

[0146] Step S4 achieves closed-loop control throughout the entire process, from risk identification to strategy generation, and from decision selection to automatic execution, through intelligent analysis and multi-objective optimization of the construction status. By integrating health status mapping and deviation result sets, high-risk components are automatically identified and optimal construction adjustment schemes are generated, shifting the construction process from experience-driven to data-driven and algorithm-based decision-making, significantly improving construction safety, efficiency, and resource utilization.

[0147] Step S5 includes the following sub-steps:

[0148] Step S501: Collect environmental parameters, operation quality data and energy consumption statistics after control execution to form a feedback dataset.

[0149] The feedback dataset is collected based on an IoT data acquisition system and on-site quality inspection terminals to obtain real-time information on construction environment temperature and humidity, noise intensity, equipment power curves, and work completion.

[0150] Outliers are eliminated using Z-score outlier detection and moving mean smoothing algorithms, and indexed by timestamp and component number.

[0151] By collecting data such as temperature, humidity, noise, equipment power, and operation quality after control execution, a feedback dataset with time and component association indexes is formed.

[0152] Z-score outlier detection and moving mean smoothing algorithms are used to clean up outlier data, ensuring that the feedback information is accurate and reliable, and providing high-quality input for subsequent threshold correction and hysteresis analysis.

[0153] This process enables comprehensive quantification of the post-construction state and establishes a true mapping relationship between control execution and environmental response.

[0154] Step S502: Based on the feedback dataset, correct the threshold of the comprehensive deviation index and readjust the weight coefficients of each monitoring parameter.

[0155] The correction process employs an adaptive threshold correction algorithm and a gradient descent-based parameter optimization model. It performs a difference analysis between the actual improvement magnitude and the predicted improvement value in the feedback dataset and automatically fine-tunes the comprehensive deviation index threshold based on the difference ratio.

[0156] A multi-parameter optimization method based on genetic algorithms was used to redistribute the weight coefficients of each monitoring parameter.

[0157] By performing a difference analysis on the feedback dataset, the comprehensive deviation index threshold was corrected and the weights of the monitoring parameters were reallocated.

[0158] The adaptive threshold correction algorithm automatically fine-tunes the deviation limit based on the difference between the actual improvement and the predicted improvement value; the gradient descent optimization model achieves dynamic and smooth updates of the threshold; and the genetic algorithm optimizes the weight vector of the monitoring parameters globally.

[0159] This process ensures that the deviation judgment criteria are adjusted in real time as the construction status changes, making the deviation calculation results more consistent with the actual site conditions.

[0160] Step S503: By analyzing the time difference between control actions and deviation improvements in the execution log, lag compensation correction is performed.

[0161] The lag compensation analysis uses a time series correlation analysis model and a dynamic time warping algorithm to extract the time delay characteristics of control actions and deviation changes, and establish lag response curves.

[0162] By combining the error estimation model based on Kalman filtering, the hysteresis compensation coefficient is adjusted, and the execution step size is corrected based on the feedback data.

[0163] Step S503 further includes a delayed learning mechanism when performing hysteresis compensation correction.

[0164] The delayed learning mechanism improves the time series difference between control execution data and deviation by backtracking control execution data and deviation data, calculates the optimal lag correction step size, and automatically applies it in subsequent control execution.

[0165] By analyzing the time difference between control actions and deviation improvements in the execution log, a lag relationship between control and response can be established.

[0166] Time series correlation analysis and dynamic time warping (DTW) algorithm are used to extract time delay features. Combined with Kalman filtering model, error is estimated and compensation coefficient is corrected, thereby optimizing the execution step size of subsequent control commands.

[0167] The delayed learning mechanism further utilizes historical data to backtrack and calculate the optimal hysteresis correction step size, and automatically applies it in subsequent execution, making the control response more accurate, real-time and smooth.

[0168] Step S504: Output the set of corrected parameters. The set of corrected parameters is used for model self-learning and threshold update. The set of corrected parameters includes the updated comprehensive deviation threshold, monitoring weight vector and hysteresis compensation coefficient. Input the set of corrected parameters into the self-learning module of the construction dynamic twin model.

[0169] The self-learning module uses reinforcement learning algorithms to review historical execution data and optimize the model's prediction accuracy and control stability.

[0170] The output set of corrected parameters includes the updated deviation threshold, monitoring weight vector, and hysteresis compensation coefficient, which are input into the self-learning module of the construction dynamic twin model.

[0171] The self-learning module is based on the reinforcement learning (Q-learning) method. It reviews the strategy and updates the parameters based on historical execution data, thereby continuously optimizing the model's prediction accuracy and control stability.

[0172] Ultimately, the twin model gradually acquires self-correction and self-optimization capabilities after multiple rounds of feedback, forming an intelligent evolution mechanism oriented towards the actual construction environment.

[0173] Step S5, by collecting and analyzing feedback data after the execution of construction control, realizes dynamic correction of deviation threshold, optimization of monitoring parameter weights, and self-learning of control strategies, enabling the construction dynamic twin model to have continuous evolution capabilities.

[0174] The entire process forms a closed loop of execution, feedback, learning, and correction, enabling the deviation analysis and control model to adaptively adjust in the actual construction environment, maintain high-precision prediction and stable control, and significantly improve the reliability and energy efficiency of the construction process.

[0175] Step S6 includes the following sub-steps:

[0176] Step S601: Version the set of corrected parameters and the engineering alignment dataset to form a knowledge snapshot.

[0177] By versioning and archiving the modified parameter set with the engineering alignment dataset, a structured knowledge snapshot is formed, enabling time traceability and version comparability of data, models, and parameters during the construction phase. This snapshot records the basis for model modifications and deviation calibrations at each stage, providing a data foundation for subsequent knowledge extraction.

[0178] Step S602: Based on the knowledge snapshot, perform multi-project comparative analysis to generate an experience parameter mapping table and establish an engineering knowledge base.

[0179] By comparing and analyzing knowledge snapshots from different projects, the correlation patterns between deviation thresholds, monitoring weights, and lag compensation parameters in various construction scenarios are extracted, generating an empirical parameter mapping table. This mapping table forms the core of the engineering knowledge base, enabling the systematic organization and long-term accumulation of parameter patterns.

[0180] Step S603: When a new project is started, extract templates with scene parameter similarity greater than a preset threshold from the engineering knowledge base.

[0181] When a new project is launched, scene parameter templates are extracted from the engineering knowledge base using a similarity matching algorithm. When the similarity exceeds a preset threshold, the relevant parameter set is automatically loaded. This process significantly shortens the model initialization and configuration cycle, enabling new projects to have a stable dynamic twin model foundation from an early stage.

[0182] Step S604: Through a continuous learning mechanism, the construction dynamic twin model is updated in features and calibrated in deviation after new data input.

[0183] When a new project is launched, scene parameter templates are extracted from the engineering knowledge base using a similarity matching algorithm. When the similarity exceeds a preset threshold, the relevant parameter set is automatically loaded. This process significantly shortens the model initialization and configuration cycle, enabling new projects to have a stable dynamic twin model foundation from an early stage.

[0184] Step S6 enables the knowledge accumulation and cross-project reuse of construction data and models, giving the intelligent management system of building engineering the ability to continuously accumulate and self-evolve. By archiving, analyzing, and extracting modified parameters and historical aligned data in a templated manner, a transferable engineering knowledge base is formed, allowing subsequent projects to quickly call upon experience parameters and realize the transformation from project data-driven to knowledge-driven.

[0185] This invention establishes a spatiotemporally labeled engineering alignment dataset, combining it with the construction master clock for time synchronization and coordinate registration. This enables the fusion of data such as temperature and humidity, stress, vibration, mechanical status, and personnel positioning at the same time scale, providing a unified data baseline for subsequent twin modeling and deviation calculation. A construction dynamic twin model is constructed, using 3D BIM data as the geometric baseline. A Kalman filter fusion model is employed for multi-sensor state estimation, and an LSTM temporal network and Bayesian inference model are combined to predict future trends and uncertainties, achieving a 3D dynamic mapping between real-time construction status, predicted status, and structural safety status. Fuzzy comprehensive evaluation and principal component analysis (PCA) algorithms are introduced to integrate structural stress changes, environmental fluctuations, and equipment operating characteristics to establish a comprehensive deviation index. A dynamic weight adjustment mechanism automatically corrects feature weights based on changes in construction stages, achieving adaptive risk enhancement identification during high-intensity operation stages. An AHP layer is utilized. This method employs sub-analysis, a multi-objective weighted model, and a genetic algorithm to quantitatively evaluate candidate optimization strategies and generate optimal control commands. This enables automatic optimization of work sequence, equipment scheduling, and environmental control parameters. Combined with Z-score detection, DTW time warping, and Q-learning reinforcement learning algorithms, a closed-loop learning mechanism is formed to continuously improve decision-making quality and response speed. Through versioned archiving and multi-project comparative analysis, the corrected parameter set and construction process experience are mapped into reusable templates. Similar scenario parameters can be directly loaded in subsequent projects, achieving intelligent accumulation from single-project learning to cross-project migration. In practical applications, this method can achieve real-time assessment and predictive warning of the health status of key components on construction sites, avoiding structural overload and environmental abrupt changes. Simultaneously, through algorithm-driven energy consumption scheduling and process optimization, energy consumption is reduced, and project delays are decreased, realizing a shift from traditional passive monitoring to proactive decision-making management.

[0186] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0187] 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, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for intelligent management of construction project data information, characterized in that, Includes the following steps: Step S1: Collect and align multi-source heterogeneous data from the construction site to form an engineering alignment dataset with spatiotemporal labels; Step S2: Extract construction status features based on the aligned dataset and calculate the comprehensive deviation index to determine the deviation status between the structure and the environment; Step S3: Construct a construction dynamic twin scenario based on the engineering alignment dataset and structural deviation result set. Use 3D BIM data as the geometric baseline to establish a construction dynamic twin model consisting of a state perception layer, a parameter mapping layer, and a prediction inference layer. The state perception layer realizes multi-sensor state estimation based on the Kalman filter fusion model, the parameter mapping layer maps the estimated state to BIM geometric nodes to form a dynamic field distribution, and the prediction inference layer combines LSTM temporal network and Bayesian inference model to generate health status mapping of key components. Step S4: Based on the health status mapping and comprehensive deviation index, a construction optimization strategy is generated. The analytic hierarchy process and multi-objective weighted scoring model are used to evaluate the potential for safety improvement, energy consumption impact and schedule disturbance. The strategy with the highest comprehensive score is selected as the control instruction and executed. Step S5: Collect environmental parameters, work quality and energy consumption data after control execution; use Z-score detection, gradient descent and genetic algorithm to adaptively correct deviation threshold and monitoring weight; and update construction dynamic twin model parameters through reinforcement learning mechanism. Step S6: Version the correction parameters and the engineering alignment dataset, and generate an empirical parameter mapping table by comparing with knowledge snapshots; Step S2 includes the following sub-steps: Step S201: Read the engineering alignment dataset, use the data analysis engine to perform structured analysis on the original multi-source data, and extract the structural stress change curve, temperature and humidity fluctuation trend and equipment operation cycle characteristics. When calculating the stability of key parameters, a time series smoothing algorithm based on a sliding window and a standard deviation evaluation method are used to quantify the fluctuation range and steady-state range of each parameter in the time series. Step S202: Analyze the strain difference, temperature difference and displacement gradient between adjacent monitoring points, and construct the spatial distribution feature matrix using the Kriging interpolation algorithm and a three-dimensional finite element analysis model; By comparing the consistency of displacement direction and strain coupling at different monitoring points, local stress concentration areas and abnormal temperature difference areas can be identified. Step S203: Based on the structural stress change curve, temperature and humidity fluctuation trend, equipment operation cycle characteristics and spatial distribution characteristic matrix, and according to the sensitivity factors of component type and construction stage, the multidimensional features are reduced and fused using fuzzy comprehensive evaluation algorithm and principal component analysis method, the comprehensive deviation index is calculated, and the deviation feature set is output. Step S204: When the comprehensive deviation index is detected to exceed the preset range, the abnormality persistence is analyzed based on threshold judgment and ARIMA residual detection, abnormal components are marked and deviation alarm labels are generated to form a structural deviation result set. Step S3 includes the following sub-steps: Step S301: Based on the engineering alignment dataset and the structural deviation result set, construct a construction dynamic twin scenario in the digital model; During the construction process, 3D BIM data is used as the geometric baseline, and a construction dynamic twin model is established driven by real-time collected multi-source alignment data. The construction dynamic twin model consists of a state perception layer, a parameter mapping layer, and a prediction and inference layer. The state perception layer uses a Kalman filter fusion model to achieve real-time fusion and state estimation of multi-sensor data on site. The parameter mapping layer maps the estimated state to BIM geometric nodes to form a dynamic field distribution. The prediction and inference layer uses an LSTM time series network and a Bayesian inference model to predict the future state and perform uncertainty analysis. The construction dynamic twin model continuously overlays historical data and real-time data streams during operation, forming a spatiotemporal digital mapping of the construction process; Step S302: Map the real-time monitoring points to the nodes of the construction dynamic twin model. Use the ICP point cloud registration algorithm and the deep neural network attitude recognition model to match and correct the sensor position information and the model nodes. The rendering engine dynamically generates the state distribution and historical change trajectory of each component and displays the multi-dimensional state parameters in color coding. The multi-dimensional state parameters include component stress, temperature and vibration. Step S303: Calculate the component health index using the data evolution sequence in the twin scenario, and use a multidimensional regression model and fuzzy inference system to comprehensively score historical deviation, stress stability, and environmental interference. During the health index generation process, an LSTM-based time series prediction model is used to identify future trends, and the output is a health status map to represent the safety level of each component in the current and prediction stages. Step S304, the health status mapping includes the real-time safety level, cumulative deviation trend and construction stage confidence score of each component; The confidence score was calculated using a dynamic confidence calculation method based on a Bayesian update model.

2. The intelligent management method for construction engineering data information as described in claim 1, characterized in that, Step S1 includes the following sub-steps: Step S101: Collect temperature, humidity, stress, vibration, video images, mechanical operating parameters and personnel positioning information at the construction site, and output raw multi-source data; Step S102: Time synchronization and coordinate alignment are performed on the original multi-source data. A unified timestamp mapping is performed based on the construction master control clock, and the output is the preliminary aligned data. Step S103: Combining the construction task plan and component number information, the preliminary alignment data is associated with the corresponding component nodes in the digital model to form an engineering alignment dataset with spatiotemporal labels and component identifiers.

3. The intelligent management method for construction engineering data information as described in claim 2, characterized in that, Step S4 includes the following sub-steps: Step S401: Read the health status mapping and structural deviation result set, and prioritize the components with risks. The ranking process comprehensively considers the health index in the health status mapping, real-time safety level, cumulative deviation trend and construction stage confidence score, and combines the deviation alarm labels in the structural deviation result set to output a risk priority list. Step S402: Based on the priority list and construction schedule, generate a set of candidate optimization strategies; The set of candidate optimization strategies includes job sequence adjustment, equipment scheduling, and environmental control parameter adjustment. Step S403: Use a multi-objective weighted analysis method to quantitatively evaluate each candidate optimization strategy; The multi-objective weighted analysis method combines the analytic hierarchy process and the weighted scoring model to comprehensively score the execution cost, energy consumption impact, safety improvement potential, and the degree of disturbance to the overall construction progress. Based on the comprehensive score, the strategy with the highest score is selected as the optimal control instruction. Step S404: Automatically issue optimal control commands and record execution logs within the preset execution window; The execution log includes the instruction issuance time, target component identifier, list of participating equipment, environmental control parameter adjustment values, and information on delayed work procedures. The execution log and control instructions together form control execution data.

4. The intelligent management method for construction engineering data information as described in claim 3, characterized in that, Step S5 includes the following sub-steps: Step S501: Collect environmental parameters, operation quality data and energy consumption statistics after control execution to form a feedback dataset; The feedback dataset is collected based on an IoT data acquisition system and on-site quality inspection terminals to obtain real-time information on construction environment temperature and humidity, noise intensity, equipment power curves, and work completion. Outliers are eliminated using Z-score outlier detection and moving mean smoothing algorithms, and indexed by timestamp and component number. Step S502: Correct the threshold of the comprehensive deviation index based on the feedback dataset and readjust the weight coefficients of each monitoring parameter; The correction process employs an adaptive threshold correction algorithm and a gradient descent-based parameter optimization model to analyze the difference between the actual improvement magnitude and the predicted improvement value in the feedback dataset, and automatically fine-tunes the comprehensive deviation index threshold based on the difference ratio. A multi-parameter optimization method based on genetic algorithm is used to redistribute the weight coefficients of each monitoring parameter; Step S503: By analyzing the time difference between control actions and deviation improvements in the execution log, perform lag compensation correction; The lag compensation analysis uses a time series correlation analysis model and a dynamic time warping algorithm to extract the time delay characteristics of control actions and deviation changes, and establish lag response curves. By combining the error estimation model based on Kalman filtering, the hysteresis compensation coefficient is adjusted, and the execution step size is corrected based on the feedback data; Step S504: Output the set of correction parameters. The set of correction parameters is used for model self-learning and threshold update. The set of correction parameters includes the updated comprehensive deviation threshold, monitoring weight vector and hysteresis compensation coefficient. Input the set of correction parameters into the self-learning module of the construction dynamic twin model. The self-learning module uses reinforcement learning algorithms to review historical execution data and optimize the model's prediction accuracy and control stability.

5. The intelligent management method for construction engineering data information as described in claim 4, characterized in that, Step S6 includes the following sub-steps: Step S601: Version archive the set of corrected parameters and the engineering alignment dataset to form a knowledge snapshot; Step S602: Based on the knowledge snapshot, perform multi-project comparative analysis to generate an experience parameter mapping table and establish an engineering knowledge base; Step S603: When a new project is started, extract templates with scene parameter similarity greater than a preset threshold from the engineering knowledge base; Step S604: Through a continuous learning mechanism, the construction dynamic twin model is updated in features and calibrated in deviation after new data input.

6. The intelligent management method for construction engineering data information as described in claim 5, characterized in that, Step S203 further includes a dynamic weight adjustment mechanism; The dynamic weight adjustment mechanism adjusts the weight ratios of stress, temperature, and humidity in the comprehensive deviation index calculation in real time according to the changing trends of construction activities in different time periods. When a high-intensity operation phase is detected, the weight of stress data is increased to a preset percentage.

7. The intelligent management method for construction engineering data information as described in claim 6, characterized in that, Step S301 further includes an adaptive model update mechanism; The adaptive model update mechanism automatically adjusts the model refresh frequency based on the difference between the actual feedback data and the prediction of the construction dynamic twin model. When the prediction deviation continues to exceed the limit, the refresh cycle is shortened; when the operation is stable, the refresh cycle is extended.

8. The intelligent management method for construction engineering data information as described in claim 7, characterized in that, Step S503 further includes a delayed learning mechanism when performing hysteresis compensation correction. The delayed learning mechanism calculates the optimal lag correction step size by backtracking the time series differences between control execution data and deviation improvement data, and automatically applies it in subsequent control execution.

Images