Construction progress dynamic regulation and control system and method for building machine

Abstract

The invention provides a construction progress dynamic regulation and control system and method for a building machine, relates to the technical field of special equipment for building construction, and solves the problem of limitation of an existing building machine regulation and control system in the aspects of dynamics, collaboration and the like. In the system, a sensing execution layer is used for collecting construction data in real time and correspondingly executing a regulation and control instruction, and transmitting the data to an edge calculation layer for processing; the edge computing layer is used for carrying out localized real-time preprocessing and decision generation on the data, and synchronizing key information to the cloud analysis layer to obtain a regulation and control strategy; the cloud analysis layer is used for constructing virtual mapping of a construction process, optimizing a regulation and control strategy and driving a dynamic regulation and control module to perform multi-dimensional decision making; and the dynamic regulation and control module is used for comprehensively predicting the construction progress, coordinating and optimizing the target, handling sudden anomalies, generating regulation and control instructions and feeding back the regulation and control instructions to the sensing execution layer to form a closed loop. According to the method, the construction resource utilization rate and the progress prediction precision of the building machine can be effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of special equipment for building construction, and particularly relates to a construction progress dynamic regulation system and method for a building machine. BACKGROUND

[0002] With the deepening of the urbanization process and the continuous improvement of the level of building industrialization in China, the construction demand for super high-rise buildings and prefabricated buildings is increasingly booming. As a key construction equipment in the modern building industrialization system, the operation efficiency of the building machine directly relates to the overall cost control, construction period guarantee and construction safety level of the project. Under this background, how to optimize the construction progress management of the building machine has become the core issue of the industry. However, the progress regulation mechanism in the current construction practice still faces many challenges, and technical innovation is urgently needed to adapt to the complex and changeable engineering environment.

[0003] Traditional building machine construction progress management mainly relies on manual experience for decision-making, and lacks scientific analysis mechanism based on real-time data. During the construction process, it is difficult for the management personnel to accurately capture the dynamic changes on site, leading to a slow response to the interaction of multi-dimensional factors such as personnel, machinery, materials and environment. The existing system generally uses a static progress planning model, which cannot timely perceive the subtle fluctuations of the construction state, resulting in obvious lag in progress deviation prediction. When encountering sudden weather changes, equipment abnormalities or supply chain fluctuations and other disturbance factors, such management methods lack self-adaptive adjustment ability, which easily leads to significant loss of construction efficiency, and further affects the overall progress rhythm of the project.

[0004] Although new technologies such as building information model and Internet of Things sensing have been gradually introduced into the construction industry, there are still obvious shortcomings in the application of technology in the field of building machine construction. The integration and analysis capabilities of multi-source heterogeneous data are insufficient, such as the real-time fusion of tower crane operation parameters, concrete pumping state and component hoisting sequence, which hinders the support of high-precision construction progress prediction. At the same time, the dynamic regulation algorithm shows weak environmental adaptability in actual working conditions, and cannot realize the efficient dynamic matching of human resources, machinery and materials, leading to the disconnection between process logic and site demand. In addition, the existing system has not formed a complete closed-loop regulation link covering sensing, decision-making, execution and feedback, and the progress correction measures often lag behind the construction situation, which aggravates the risk of multi-linkage coordination failure.

[0005] The cumulative effect of the above problems causes frequent construction delays, not only increasing project costs, but also posing a potential threat to engineering quality and safety. Industry practice shows that the rigidity and fragmentation of progress management have become the main bottleneck restricting the efficient operation of building machines. Therefore, exploring a technical path that can deeply integrate real-time data, intelligently respond to environmental changes, and achieve full-process closed-loop regulation and control has become an urgent need to promote the high-quality development of building industrialization. The shortcomings of the existing technical framework in terms of dynamics and collaboration highlight the necessity and urgency of further innovation. SUMMARY

[0006] The purpose of the present application is to solve the existing limitations of building machine control systems in terms of dynamics and collaboration, such as insufficient data integration and analysis capabilities, weak environmental adaptability, etc. Therefore, a construction progress dynamic regulation system and method for building machines are proposed. The present application fully applies multi-source heterogeneous data from the construction site, combined with digital twin technology, to improve the accuracy of real-time dynamic prediction of construction progress; the present application also provides a reusable intelligent construction progress regulation and control modular solution for building machines, and the utilization rate of construction resources is effectively improved.

[0007] The present application achieves the purpose by adopting the following technical solutions: A construction progress dynamic regulation system for building machines, the system comprises a perception and execution layer, an edge computing layer, a cloud analysis layer and a dynamic regulation module; wherein: The perception and execution layer is used to collect construction environment and structure state data in real time and execute corresponding regulation instructions, and the output data is transmitted to the edge computing layer for processing; The edge computing layer is used to perform real-time preprocessing and decision-making generation on the data provided by the perception and execution layer, and synchronizes key information to the cloud analysis layer to obtain corresponding regulation strategies; The cloud analysis layer is used to construct a virtual mapping of the construction process and optimize the regulation strategy, and the analysis results drive the dynamic regulation module to make multi-dimensional decisions; The dynamic regulation module is used to comprehensively predict construction progress, coordinate and optimize objectives, and handle unexpected abnormalities, and generates regulation instructions that are fed back to the perception and execution layer to form a closed loop.

[0008] Preferably, the perception and execution layer comprises a multi-modal sensor array, a UWB and laser radar fusion positioning module, and a hydraulic servo closed-loop execution mechanism; wherein: The multimodal sensor array is used to collect real-time construction environment data and structural status data of the building construction machine. The multimodal sensor array includes vibration sensors, tilt sensors, temperature and humidity sensors, and vision sensors. The vibration sensors are used to monitor the hydraulic impact and structural resonance of the building construction machine's lifting system; the tilt sensors are used to monitor the plane tilt angle of the building construction machine's own frame; the temperature and humidity sensors are used to monitor the temperature and humidity of the working environment where the building construction machine is located; and the vision sensors are used to identify the progress of rebar binding, formwork alignment error, and component hoisting trajectory on the corresponding working surface of the building construction machine. The UWB and LiDAR fusion positioning module is used for the three-dimensional spatial positioning of building construction machines and construction components. It establishes a global coordinate system for the construction area based on the UWB base station network and performs real-time calibration using LiDAR point cloud data. The module uses a particle filter algorithm to fuse UWB ranging data and LiDAR point cloud data during positioning. The hydraulic servo closed-loop actuator is used to adjust the working status of the building machine's lifting system, formwork system, and concrete placing machine under corresponding control commands.

[0009] Preferably, the edge computing layer includes an edge computing gateway, a local real-time decision engine, and a data caching module; wherein: The edge computing gateway is deployed in the building machine body to perform real-time preprocessing of the data transmitted from the perception and execution layer. The preprocessing process uses wavelet denoising and Kalman filtering algorithms. The local real-time decision engine is used to generate corresponding preliminary control instructions based on the preprocessed data; The data caching module is used to asynchronously transmit and interact with the cloud analysis layer in a 5G-MEC network environment, thereby synchronizing key information to the cloud analysis layer to obtain corresponding control strategies.

[0010] Preferably, the cloud-based analytics layer includes a digital twin simulation platform, a deep reinforcement learning training module, and a construction knowledge graph database; wherein: The digital twin simulation platform is used to construct a virtual mapping of the building construction process based on the BIM model, which is driven by bidirectional data between the physical and virtual spaces. The deep reinforcement learning training module is used to optimize construction progress prediction and control strategies. Its training environment integrates meteorological API, supply chain data and historical construction logs. The module decomposes the construction actions of the building machine into multiple preset discrete operation instructions and then trains the strategies for each instruction. It also constructs a multi-dimensional reward mechanism by combining progress deviation rate, safety risk index and resource utilization rate. At the same time, it forms a knowledge graph based on historical construction logs to initialize the trained model. The construction knowledge graph database is used to store knowledge graph data, including process logic relationships, resource constraint rules, safety specifications, and historical construction logs.

[0011] Preferably, the dynamic control module includes a progress prediction unit, a multi-objective optimization unit, and an anomaly handling unit; wherein: The progress prediction unit is implemented using an LSTM-GRU hybrid neural network model to output the construction phase time prediction and the corresponding dynamic confidence interval; The multi-objective optimization unit is used to perform real-time Pareto front analysis on multiple preset dimensions of objectives based on the NSGA-III algorithm; The anomaly handling unit includes a fault tree analysis module, a Bayesian network inference engine, and an emergency plan library. The fault tree analysis module is used to define multiple preset fault node contents. The Bayesian network inference engine is used to update the fault probability based on construction environment and structural status data and output the optimal handling path. The emergency plan library is used to store standardized emergency procedures for multiple preset emergency scenarios.

[0012] This invention also provides a method for dynamic control of construction progress in a building construction machine, the hardware basis of which is the aforementioned dynamic control system for construction progress; the method includes the following steps: S1. Real-time collection and preprocessing of construction environment and structural status data; spatiotemporal alignment of preprocessed data with virtual BIM model; construction of digital mapping of construction elements; and identification of risk characteristics. S2. Synchronize the physical construction status to the virtual BIM model to generate the process network topology, simulate different environmental conditions and random variables to predict the progress and quantify the impact of disturbances. S3. Deploy a progress prediction unit based on multi-objective optimization, generate a set of control strategies, decompose them into executable control instructions, issue them for execution, monitor the execution effect in real time, and trigger the parameter update of the prediction model in the progress prediction unit to achieve closed-loop control.

[0013] Specifically, in step S1, the vibration, tilt angle, and temperature and humidity data of the building machine are first acquired in real time through the multimodal sensor array of the perception execution layer, as well as the construction scene images collected by the vision sensor; the position coordinates of the construction components are tracked in real time through the UWB and LiDAR fusion positioning module, and the corresponding spatiotemporal label data is generated; then the edge computing gateway of the edge computing layer performs wavelet denoising and Kalman filtering on the acquired data to remove abnormal noise interference. Subsequently, the preprocessed data is spatiotemporally aligned with the BIM model in the cloud analysis layer via the 5G-MEC network to construct a digital mapping of construction elements that includes equipment status, resource distribution, and environmental parameters. Finally, based on the anomaly handling unit of the dynamic control module, the risk characteristics in the mapped data are analyzed to construct a risk situation awareness matrix and complete the safety level classification and schedule deviation threshold calculation.

[0014] Specifically, in step S2, the construction status data of the physical space is injected into the digital twin simulation platform of the cloud analysis layer in real time to drive the virtual BIM model to run synchronously with the physical equipment; then the process logic relationship in the construction knowledge graph database is called to generate a process network topology including concrete pouring, formwork erection and rebar binding, and complete the synchronization of the virtual BIM model. Real-time weather data from a meteorological API is loaded into the deep reinforcement learning training module to simulate the impact of various preset weather scenarios on construction efficiency. For construction progress simulation, Monte Carlo simulation is adopted, with the standard deviation of concrete setting time and the fluctuation of tower crane hoisting efficiency as input random variables, and the confidence interval of progress prediction is output. Finally, disturbance propagation analysis is performed through a digital twin simulation platform to quantify the duration of the chain effect of material supply delays on subsequent processes.

[0015] Specifically, in step S3, the multi-objective optimization unit of the dynamic control module calls the NSGA-III algorithm to generate a Pareto optimal policy set with multiple preset dimensions as optimization objectives; the local real-time decision engine based on the edge computing layer uses the AHP-entropy weight method to dynamically prioritize the control policies in the optimal policy set, and the ranking is performed according to the preset weight allocation rules. After sorting, the obtained optimal control strategy is decomposed into an executable control instruction set, and the control instruction set is asynchronously transmitted to the corresponding device in the perception execution layer through the data caching module of the edge computing layer. Finally, various sensor data and images from the vision sensor are collected in real time after the control instruction set is executed to evaluate the consistency between the actual construction progress and the virtual simulation. If the deviation between the actual construction progress and the virtual simulation exceeds a preset threshold, the incremental learning mechanism of the deep reinforcement learning training module is triggered to update the parameters of the prediction model in the progress prediction unit, and the updated model parameters are synchronized to the local real-time decision engine of the edge computing layer.

[0016] Preferably, the method further includes: S4. During the construction process, conduct construction compliance verification, compare the current construction status with the preset quality acceptance standards in the virtual BIM model, and when the comparison result does not meet the preset standard conditions, lock the building machine hydraulic system and issue a rectification instruction to the outside.

[0017] In summary, due to the adoption of this technical solution, the beneficial effects of this invention are as follows: This invention integrates perception-based execution, edge computing, and cloud-based intelligent analysis to achieve dynamic and refined management of the entire construction process of high-rise buildings. The system constructs a real-time interactive mechanism between physical construction and virtual mapping, significantly improving the accuracy of progress prediction and the timeliness of decision-making response. It effectively solves the problems of large progress deviations and delayed adjustments caused by information fragmentation in traditional construction, providing reliable technical support for the construction of complex high-rise buildings.

[0018] In practical applications, the system can proactively coordinate resource allocation and process connections, optimize construction paths, and mitigate potential risks, significantly reducing downtime losses caused by sudden environmental changes or equipment malfunctions. Its closed-loop control mechanism can instantly identify structural safety and quality hazards, ensuring that the construction process is both efficient and compliant, thereby improving the overall quality and safety of the project while saving manpower and material costs, achieving a dual improvement in economic benefits and construction efficiency.

[0019] The promotion of this invention will strongly drive the transformation and upgrading of the construction industry towards intelligence and digitalization, setting a new benchmark for smart construction site construction. By continuously accumulating construction data and optimizing control strategies, it can not only shorten the construction period and ensure delivery quality, but also promote the upgrading of industry standards. It has broad application prospects and high social value, providing strong technical support for future green construction and sustainable development. Attached Figure Description

[0020] The present invention will be further described in detail with reference to the following figures, which specifically include two figures as follows: Figure 1 This is a schematic diagram illustrating the composition and logical principle of the construction progress dynamic control system of the present invention; Figure 2 This is a schematic diagram of the overall process steps of the construction progress dynamic control method of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0022] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0023] Example 1 A dynamic control system for construction progress of building construction machinery, such as Figure 1 As shown, the system includes a perception and execution layer, an edge computing layer, a cloud analysis layer, and a dynamic control module; among which: The perception and execution layer is used to collect construction environment and structural status data in real time and execute corresponding control commands. Its output data is transmitted to the edge computing layer for processing. The edge computing layer is used to perform local real-time preprocessing and decision generation on the data provided by the perception and execution layer, and synchronize key information to the cloud analysis layer to obtain corresponding control strategies; The cloud-based analysis layer is used to construct a virtual mapping of the construction process and optimize control strategies. Its analysis results drive the dynamic control module to make multi-dimensional decisions. The dynamic control module is used to comprehensively predict construction progress, coordinate and optimize objectives, and handle sudden anomalies. After generating control instructions, it feeds them back to the perception and execution layer to form a closed loop.

[0024] Overall, the perception and execution layer provides raw input to the edge computing layer, the edge computing layer transmits processing information to the cloud analysis layer, the cloud analysis layer outputs strategy guidance to the dynamic control module, and the dynamic control module directly acts on the perception and execution layer to achieve precise execution; see also Figure 1 This is a schematic diagram of the logical principle. This embodiment will describe in detail the details and optimal configurations of each component level and module in the system.

[0025] The perception and execution layer includes a multimodal sensor array, a UWB and LiDAR fusion positioning module, and a hydraulic servo closed-loop actuator.

[0026] A multimodal sensor array is used to collect real-time construction environment and structural status data of the building construction machine. This array includes vibration sensors, tilt sensors, temperature and humidity sensors, and vision sensors. Vibration sensors, implemented using triaxial accelerometers with a frequency response range of 0.1Hz to 10kHz, monitor hydraulic shock and structural resonance in the building construction machine's lifting system. Tilt sensors monitor the plane tilt angle of the building construction machine's frame, ensuring that the machine and its auxiliary equipment remain horizontally stable during construction, preventing tilting and ensuring construction safety. Temperature and humidity sensors monitor the temperature and humidity of the working environment. Vision sensors identify the rebar binding progress, formwork alignment errors, and component hoisting trajectories on the corresponding working surface. These sensors can be constructed using infrared cameras and RGB-D depth cameras, ensuring sufficient resolution for the application.

[0027] The UWB and LiDAR fusion positioning module is used for the three-dimensional spatial positioning of the building construction machine and construction components. In this embodiment, the overall positioning accuracy is recommended to be ±2cm. This module establishes a global coordinate system for the construction area based on the UWB base station network and performs real-time calibration using LiDAR point cloud data. During positioning, a particle filter algorithm is used to fuse UWB ranging data and LiDAR point cloud data to eliminate multipath interference. The particle filter algorithm is a nonlinear filtering method based on Monte Carlo simulation. It estimates the target state through a large number of randomly sampled "particles," effectively fusing multi-source sensor data in dynamic environments to improve the robustness and accuracy of positioning.

[0028] The hydraulic servo closed-loop actuator is only used to adjust the working status of the building machine lifting system, the formwork system, and the concrete placing machine under the corresponding control commands; its response time is recommended to be less than 50ms in this embodiment.

[0029] The edge computing layer includes an edge computing gateway, a local real-time decision engine, and a data caching module.

[0030] The edge computing gateway uses industrial-grade equipment deployed within the building machine itself to perform real-time preprocessing of data transmitted from the perception and execution layer. The preprocessing process employs wavelet denoising and Kalman filtering algorithms. Wavelet denoising is a signal processing technique that decomposes signals into different frequency levels through wavelet transform, effectively distinguishing noise from useful signals. This removes random interference while preserving key signal features, making it particularly suitable for processing non-stationary or abrupt engineering signals. Kalman filtering is a recursive estimation algorithm that can estimate the state of a dynamic system from noisy measurement data. Through a continuous prediction and update cycle, it effectively smooths the data and improves the stability of time series.

[0031] The local real-time decision engine is used to generate corresponding preliminary control instructions based on the preprocessed data. In this embodiment, the decision delay time is recommended to be less than 200ms.

[0032] The data caching module is used for asynchronous data transmission and interaction with the cloud analysis layer in a 5G-MEC network environment, thereby synchronizing key information to the cloud analysis layer to obtain corresponding control strategies. 5G-MEC is a technical architecture that pushes cloud computing capabilities down to the network edge. 5G provides high-speed, low-latency, and massive connectivity wireless communication support, while MEC (Multi-access Edge Computing) deploys computing, storage, and network resources close to users and devices at the network edge, such as near base stations, to achieve localized data processing and response. In this embodiment, the combined use of 5G-MEC can significantly reduce data transmission latency, improve real-time performance, and alleviate the burden on the core network, supporting efficient and stable data synchronization and strategy interaction between the building machine and the cloud in complex construction environments.

[0033] The cloud-based analytics layer includes a digital twin simulation platform, a deep reinforcement learning training module, and a construction knowledge graph database.

[0034] The digital twin simulation platform is used to construct a virtual mapping of the building construction process based on the BIM model, which is driven by bidirectional data from physical and virtual spaces.

[0035] The deep reinforcement learning training module is used to optimize construction progress prediction and control strategies. Its training environment integrates meteorological APIs, supply chain data, and historical construction logs. This module decomposes the construction actions of the building machine into multiple preset discrete operation commands, such as jacking height, material placement rate, and formwork opening and closing degree, and then trains strategies for each command separately. In its reward function design, a multi-dimensional reward mechanism is constructed by integrating the progress deviation rate, safety risk index, and resource utilization rate. To form a transfer learning framework, a knowledge graph is generated based on historical construction logs to initialize the trained model, which can shorten the training cycle for deploying the model to new projects. The trained model here is the LSTM-GRU hybrid neural network model of the progress prediction unit, which will be introduced in the dynamic control module.

[0036] The construction knowledge graph database is used to store knowledge graph data, including process logic relationships, resource constraint rules, safety specifications, and historical construction logs.

[0037] The dynamic control module includes a progress prediction unit, a multi-objective optimization unit, and an anomaly handling unit.

[0038] The schedule prediction unit is implemented using an LSTM-GRU hybrid neural network model to output the predicted construction phase time and the corresponding dynamic confidence interval. LSTM (Long Short-Term Memory) and GRU (Gated Recurrent Unit) are both variants of Recurrent Neural Networks (RNNs) in deep learning, adept at handling data with time-series characteristics. LSTM, by introducing "memory units" and three gating structures—input gate, forget gate, and output gate—effectively captures long-term dependencies, avoiding the gradient vanishing problem of traditional RNNs. GRU is a simplified version of LSTM, controlling information flow through update and reset gates, reducing computational complexity while maintaining high prediction accuracy. This embodiment uses an LSTM-GRU hybrid model, combining the advantages of both, to more accurately learn the time evolution patterns of each process during construction. While predicting the construction phase time, it outputs a dynamic confidence interval, reflecting the uncertainty of the prediction and providing more robust and interpretable decision support for project management.

[0039] The multi-objective optimization unit is used to perform real-time Pareto front analysis on seven dimensions of objectives—schedule, cost, safety, quality, energy consumption, resources, and environmental protection—based on the NSGA-III algorithm. NSGA-III (Non-dominated sorting genetic algorithm III) is an evolutionary algorithm for high-dimensional multi-objective optimization problems, particularly suitable for simultaneously optimizing three or more conflicting objectives. In the system of this embodiment, the seven objectives—schedule, cost, safety, quality, energy consumption, resource utilization, and environmental protection—have trade-offs. For example, shortening the schedule may increase costs or affect quality, making it difficult to obtain a single optimal solution using traditional methods. NSGA-III, by simulating the natural selection mechanism, evolves a set of "non-dominated solutions," or Pareto fronts, in the solution space. These solutions represent feasible solutions that achieve the best trade-offs among different objectives. Through real-time analysis of the Pareto fronts, construction managers can dynamically select the optimal strategy based on current priorities (such as urgent work or energy conservation and emission reduction), achieving collaborative optimization and intelligent decision support for multi-dimensional objectives in complex construction environments.

[0040] The anomaly handling unit itself is used to diagnose faults using an FTA-Bayesian network, generate adaptive emergency plans, and feed them back to the perception and execution layer through a human-machine interface. FTA provides a top-down logical reasoning framework, treating potential major system failures, such as equipment malfunctions or structural anomalies, as the top event, and decomposing them layer by layer into basic fault events such as sensor malfunctions and control signal interruptions, clarifying the logical relationships between various faults. The Bayesian network, building on this, introduces a probabilistic reasoning mechanism, using real-time environmental and operational data collected on-site to dynamically update the probability of occurrence of each fault node, thereby identifying the most likely fault path and root cause under uncertain conditions. This unit combines the advantages of both, not only accurately locating the fault source but also generating adaptive emergency plans based on current operating conditions and promptly feeding them back to the perception and execution layer through the human-machine interface, achieving rapid closed-loop control of fault response and improving the safety and operational reliability of the building machine system.

[0041] The anomaly handling unit specifically includes a fault tree analysis (FTA) module, a Bayesian network inference engine, and an emergency plan library. The FTA module is used to define multiple preset fault node contents, such as hydraulic system leakage, loss of positioning signals, and structural overload. The Bayesian network inference engine is used to update the fault probability based on construction environment and structural status data and output the optimal handling path. The emergency plan library is used to store standardized emergency procedures for multiple preset emergency scenarios, including typhoons, power outages, and supply chain disruptions.

[0042] Example 2 Based on Example 1, this example provides a method for dynamic control of construction progress in a building construction machine. The hardware foundation of this method is the dynamic control system for construction progress in Example 1. (See also...) Figure 2 As illustrated, the method includes the following steps: S1. Real-time collection and preprocessing of construction environment and structural status data; spatiotemporal alignment of preprocessed data with virtual BIM model; construction of digital mapping of construction elements; and identification of risk characteristics. S2. Synchronize the physical construction status to the virtual BIM model to generate the process network topology, simulate different environmental conditions and random variables to predict the progress and quantify the impact of disturbances. S3. Deploy a progress prediction unit based on multi-objective optimization, generate a set of control strategies, decompose them into executable control instructions, issue them for execution, monitor the execution effect in real time, and trigger the parameter update of the prediction model in the progress prediction unit to achieve closed-loop control.

[0043] For step S1, this embodiment first uses a multimodal sensor array in the perception execution layer to acquire in real time the vibration, tilt angle, and temperature and humidity data of the building machine, as well as construction scene images collected by a vision sensor; then, through a UWB and LiDAR fusion positioning module, the position coordinates of the construction components are tracked in real time to generate corresponding spatiotemporal label data; then, the edge computing gateway in the edge computing layer performs wavelet denoising and Kalman filtering on the acquired data to remove abnormal noise interference.

[0044] Subsequently, the preprocessed data is spatiotemporally aligned with the BIM model in the cloud analysis layer via the 5G-MEC network to construct a digital mapping of construction elements that includes equipment status, resource distribution, and environmental parameters. Finally, based on the anomaly handling unit of the dynamic control module, the risk characteristics in the mapped data are analyzed to construct a risk situation awareness matrix and complete the safety level classification and schedule deviation threshold calculation.

[0045] In this embodiment, the safety level classification is divided into 4 levels based on the pressure threshold of the hydraulic system of the building machine and the amount of structural deformation; the calculation of the progress deviation threshold is based on the prediction results of the LSTM-GRU hybrid neural network model, and the allowable deviation range is controlled within ±5%.

[0046] In this embodiment, when constructing the risk situation awareness matrix, four levels of risk indicators are defined, including: environmental risk, equipment risk, human risk, and supply chain risk. Environmental risks include wind speed > level 8 and temperature < -10℃; equipment risks include hydraulic pressure exceeding limits and positioning error > 5cm; human risks include violations of regulations; and supply chain risks include material delays > 24 hours. A comprehensive risk index is then calculated based on the fuzzy comprehensive evaluation method to trigger different levels of control responses.

[0047] For step S2, this embodiment injects the construction status data of the physical space into the digital twin simulation platform of the cloud analysis layer in real time, driving the virtual BIM model to run synchronously with the physical equipment; then, it calls the process logic relationship in the construction knowledge graph database to generate a process network topology including concrete pouring, formwork erection, and rebar tying, thus completing the synchronization of the virtual BIM model.

[0048] When performing multi-scenario simulation and prediction in step S2, real-time weather data from the meteorological API is loaded into the deep reinforcement learning training module to simulate the impact of various preset weather scenarios on construction efficiency, such as extreme weather conditions like heavy rain and high temperatures. For construction progress simulation, the Monte Carlo simulation method is adopted, using the standard deviation of concrete setting time (controlled within ±15%) and the fluctuation of tower crane hoisting efficiency (controlled within ±20%) as input random variables, and outputting the 90% confidence interval of the progress prediction. At the same time, a critical path sensitivity analysis report can be output for management personnel reference. Finally, disturbance propagation analysis is performed through a digital twin simulation platform to quantify the duration of the chain reaction effect of material supply delays on subsequent processes.

[0049] In this embodiment, in addition to the standard deviation of concrete setting time and the fluctuation of tower crane hoisting efficiency, the input variables for the Monte Carlo simulation method may also include worker attendance rate (range 85%-95%); the output result is the generation of a 90% confidence interval for the progress prediction, and the marking of sensitive nodes on the critical path.

[0050] For step S3, in this embodiment, the multi-objective optimization unit of the dynamic control module calls the NSGA-III algorithm to generate the Pareto optimal strategy set with the construction period, cost, safety, quality, energy consumption, resources and environmental protection as optimization objectives; based on the local real-time decision engine of the edge computing layer, the AHP-entropy weight method is used to dynamically prioritize the control strategies in the optimal strategy set, and the ranking is performed according to the preset weight allocation rules.

[0051] In this embodiment, the weight allocation rules include subjective weight and objective weight. The subjective weight is the safety weight set by experts, accounting for ≥30%. The objective weight automatically increases the corresponding resource utilization weight according to the resource inventory warning level. In the subjective weight, the core weight ratio of construction period, cost and safety is determined to be 4:3:3 through expert scoring. The objective weight is specifically calculated based on the dynamic weight adjustment coefficient of resources and environmental protection based on historical data entropy values. Finally, the subjective and objective weights are integrated to generate the strategy priority ranking.

[0052] After sorting, the obtained optimal control strategy is decomposed into an executable control instruction set, and the control instruction set is asynchronously transmitted to the corresponding device in the perception and execution layer through the data caching module of the edge computing layer. The executable control instruction set may include: lifting speed adjustment instruction of hydraulic servo closed-loop actuator, component hoisting path replanning instruction based on UWB positioning, concrete delivery priority adjustment instruction of supply chain system, etc.

[0053] Finally, real-time data from various sensors and images from the vision sensor are collected after the control command set is executed to evaluate the consistency between the actual construction progress and the virtual simulation. If the deviation between the actual construction progress and the virtual simulation exceeds a preset threshold, such as a deviation exceeding 3%, the incremental learning mechanism of the deep reinforcement learning training module is triggered to update the parameters of the LSTM-GRU model in the progress prediction unit. The updated model parameters are then synchronized to the local real-time decision engine of the edge computing layer via OTA.

[0054] As a preferred embodiment, the method further includes: S4. During construction, conducting construction compliance verification, comparing the current construction status with the preset quality acceptance standards in the virtual BIM model, and locking the building machine hydraulic system and issuing rectification instructions when the comparison result does not meet the preset standard conditions. The preset standard conditions, such as detecting a template verticality deviation > 3‰ or exceeding the concrete curing temperature limit, can be preset according to actual construction quality requirements.

Claims

1. A dynamic control system for construction progress of a building construction machine, characterized in that, The system comprises a perception and execution layer, an edge computing layer, a cloud analysis layer, and a dynamic control module; wherein: The perception and execution layer is used to collect construction environment and structural status data in real time and execute corresponding control commands. Its output data is transmitted to the edge computing layer for processing. The edge computing layer is used to perform local real-time preprocessing and decision generation on the data provided by the perception and execution layer, and synchronize key information to the cloud analysis layer to obtain corresponding control strategies; The cloud-based analysis layer is used to construct a virtual mapping of the construction process and optimize control strategies. Its analysis results drive the dynamic control module to make multi-dimensional decisions. The dynamic control module is used to comprehensively predict construction progress, coordinate and optimize objectives, and handle sudden anomalies. After generating control instructions, it feeds them back to the perception and execution layer to form a closed loop.

2. The construction progress dynamic control system according to claim 1, characterized in that, The perception and execution layer includes a multimodal sensor array, a UWB and LiDAR fusion positioning module, and a hydraulic servo closed-loop actuator; among which: The multimodal sensor array is used to collect real-time construction environment data and structural status data of the building construction machine. The multimodal sensor array includes vibration sensors, tilt sensors, temperature and humidity sensors, and vision sensors. The vibration sensors are used to monitor the hydraulic impact and structural resonance of the building construction machine's lifting system; the tilt sensors are used to monitor the plane tilt angle of the building construction machine's own frame; the temperature and humidity sensors are used to monitor the temperature and humidity of the working environment where the building construction machine is located; and the vision sensors are used to identify the progress of rebar binding, formwork alignment error, and component hoisting trajectory on the corresponding working surface of the building construction machine. The UWB and LiDAR fusion positioning module is used for the three-dimensional spatial positioning of building construction machines and construction components. It establishes a global coordinate system for the construction area based on the UWB base station network and performs real-time calibration using LiDAR point cloud data. The module uses a particle filter algorithm to fuse UWB ranging data and LiDAR point cloud data during positioning. The hydraulic servo closed-loop actuator is used to adjust the working status of the building machine's lifting system, formwork system, and concrete placing machine under corresponding control commands.

3. The construction progress dynamic control system according to claim 1, characterized in that, The edge computing layer includes an edge computing gateway, a local real-time decision engine, and a data caching module; among which: The edge computing gateway is deployed in the building machine body to perform real-time preprocessing of the data transmitted from the perception and execution layer. The preprocessing process uses wavelet denoising and Kalman filtering algorithms. The local real-time decision engine is used to generate corresponding preliminary control instructions based on the preprocessed data; The data caching module is used to asynchronously transmit and interact with the cloud analysis layer in a 5G-MEC network environment, thereby synchronizing key information to the cloud analysis layer to obtain corresponding control strategies.

4. The construction progress dynamic control system according to claim 1, characterized in that, The cloud-based analytics layer includes a digital twin simulation platform, a deep reinforcement learning training module, and a construction knowledge graph database; among which: The digital twin simulation platform is used to construct a virtual mapping of the building construction process based on the BIM model, which is driven by bidirectional data between the physical and virtual spaces. The deep reinforcement learning training module is used to optimize construction progress prediction and control strategies. Its training environment integrates meteorological API, supply chain data and historical construction logs. The module decomposes the construction actions of the building machine into multiple preset discrete operation instructions and then trains the strategies for each instruction. It also constructs a multi-dimensional reward mechanism by combining progress deviation rate, safety risk index and resource utilization rate. At the same time, it forms a knowledge graph based on historical construction logs to initialize the trained model. The construction knowledge graph database is used to store knowledge graph data, including process logic relationships, resource constraint rules, safety specifications, and historical construction logs.

5. The construction progress dynamic control system according to claim 1, characterized in that, The dynamic control module includes a progress prediction unit, a multi-objective optimization unit, and an anomaly handling unit; among which: The progress prediction unit is implemented using an LSTM-GRU hybrid neural network model to output the construction phase time prediction and the corresponding dynamic confidence interval; The multi-objective optimization unit is used to perform real-time Pareto front analysis on multiple preset dimensions of objectives based on the NSGA-III algorithm; The anomaly handling unit includes a fault tree analysis module, a Bayesian network inference engine, and an emergency plan library. The fault tree analysis module is used to define multiple preset fault node contents. The Bayesian network inference engine is used to update the fault probability based on construction environment and structural status data and output the optimal handling path. The emergency plan library is used to store standardized emergency procedures for multiple preset emergency scenarios.

6. A method for dynamic control of construction progress for a building construction machine, characterized in that, The hardware basis of the method is the construction progress dynamic control system as described in any one of claims 1-5; the method includes the following steps: S1. Real-time collection and preprocessing of construction environment and structural status data; spatiotemporal alignment of preprocessed data with virtual BIM model; construction of digital mapping of construction elements; and identification of risk characteristics. S2. Synchronize the physical construction status to the virtual BIM model to generate the process network topology, simulate different environmental conditions and random variables to predict the progress and quantify the impact of disturbances. S3. Deploy a progress prediction unit based on multi-objective optimization, generate a set of control strategies, decompose them into executable control instructions, issue them for execution, monitor the execution effect in real time, and trigger the parameter update of the prediction model in the progress prediction unit to achieve closed-loop control.

7. The method for dynamic control of construction progress according to claim 6, characterized in that: In step S1, the vibration, tilt angle, and temperature and humidity data of the building machine are first acquired in real time through the multimodal sensor array of the perception execution layer, as well as the construction scene images collected by the vision sensor; the position coordinates of the construction components are tracked in real time through the UWB and LiDAR fusion positioning module, and the corresponding spatiotemporal label data is generated; then the edge computing gateway of the edge computing layer performs wavelet denoising and Kalman filtering on the acquired data to remove abnormal noise interference. Subsequently, the preprocessed data is spatiotemporally aligned with the BIM model in the cloud analysis layer via the 5G-MEC network to construct a digital mapping of construction elements that includes equipment status, resource distribution, and environmental parameters. Finally, based on the anomaly handling unit of the dynamic control module, the risk characteristics in the mapped data are analyzed to construct a risk situation awareness matrix and complete the safety level classification and schedule deviation threshold calculation.

8. The method for dynamic control of construction progress according to claim 6, characterized in that: In step S2, the construction status data of the physical space is injected into the digital twin simulation platform of the cloud analysis layer in real time to drive the virtual BIM model to run synchronously with the physical equipment; then the process logic relationship in the construction knowledge graph database is called to generate a process network topology including concrete pouring, formwork erection and rebar binding, and complete the synchronization of the virtual BIM model. Real-time weather data from a meteorological API is loaded into the deep reinforcement learning training module to simulate the impact of various preset weather scenarios on construction efficiency. For construction progress simulation, Monte Carlo simulation is adopted, with the standard deviation of concrete setting time and the fluctuation of tower crane hoisting efficiency as input random variables, and the confidence interval of progress prediction is output. Finally, disturbance propagation analysis is performed through a digital twin simulation platform to quantify the duration of the chain effect of material supply delays on subsequent processes.

9. The method for dynamic control of construction progress according to claim 6, characterized in that: In step S3, the multi-objective optimization unit of the dynamic control module calls the NSGA-III algorithm to generate a Pareto optimal policy set with multiple preset dimensions as optimization objectives; the local real-time decision engine based on the edge computing layer uses the AHP-entropy weight method to dynamically prioritize the control policies in the optimal policy set, and the ranking is performed according to the preset weight allocation rules. After sorting, the obtained optimal control strategy is decomposed into an executable control instruction set, and the control instruction set is asynchronously transmitted to the corresponding device in the perception execution layer through the data caching module of the edge computing layer. Finally, various sensor data and images from the vision sensor are collected in real time after the control instruction set is executed to evaluate the consistency between the actual construction progress and the virtual simulation. If the deviation between the actual construction progress and the virtual simulation exceeds a preset threshold, the incremental learning mechanism of the deep reinforcement learning training module is triggered to update the parameters of the prediction model in the progress prediction unit, and the updated model parameters are synchronized to the local real-time decision engine of the edge computing layer.

10. The method for dynamic control of construction progress according to claim 6, characterized in that: The method also includes: S4. During the construction process, conduct construction compliance verification, compare the current construction status with the preset quality acceptance standards in the virtual BIM model, and when the comparison result does not meet the preset standard conditions, lock the building machine hydraulic system and issue a rectification instruction to the outside.

Images