Assembly type structure installation process monitoring and management method

Abstract

The invention relates to the technical field of building construction intellectualization, and discloses a fabricated structure installation process monitoring and management method, which comprises the following steps: S1, multi-source data fusion and sensor deployment: firstly arranging a sensor network at a construction site, S2, dynamic sensor scheduling and optimization: optimizing a sensor deployment strategy by means of reinforcement learning, and carrying out dynamic sensor scheduling and optimization; the method comprises the steps of S1, data acquisition, S2, data real-time transmission and cloud processing: uploading all the acquired sensor data to a cloud processing platform in real time, S4, AI intelligent prediction and anomaly detection: cooperatively training an AI model by adopting federal learning, S5, intelligent decision support and risk early warning, S6, automatic construction adjustment and response, and S5, automatic construction management. And S7, cross-project collaboration and central scheduling: integrating different project data through a central scheduling platform. Through deployment of various sensors and combination of unmanned aerial vehicle inspection and computer vision, all-directional acquisition and intelligent analysis of construction site data are realized, and the effects of accurately monitoring the construction state in real time and improving the construction safety are obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent building construction, and in particular to a method for monitoring and managing the installation process of an assembled structure. Background Art

[0002] In modern prefabricated building construction, real-time monitoring of the construction site is crucial for ensuring structural safety and optimizing the construction process. Traditional construction monitoring methods rely primarily on manual inspections, regular measurements, and empirical judgment. The acquisition of monitoring data often lags, making it difficult to meet the demands of high-precision, real-time construction management. Furthermore, manual inspections suffer from limited monitoring range, low data collection efficiency, and difficulty covering high-risk areas. This can lead to delayed detection of construction quality risks and even prevent effective response measures in the event of a safety incident.

[0003] In recent years, intelligent monitoring technologies based on sensor networks have been gradually applied to the construction industry, using sensors such as pressure, vibration, displacement, temperature, and humidity to digitally monitor construction status. However, traditional sensor deployments are mostly fixed, making them difficult to adapt to the dynamic changes in complex construction environments. Furthermore, the data collection range of a single sensor is limited, resulting in incomplete data and limited accuracy. Furthermore, relying solely on sensors makes it difficult to provide a comprehensive overview of the construction site and effectively identify changes in the spatial position and posture of components during construction. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides a monitoring and management method for the installation process of prefabricated structures, which solves the problem that the existing technology mainly relies on fixed sensors and manual inspections. The range and accuracy of these monitoring methods are limited. Fixed sensors can only monitor the status of specific areas and cannot dynamically adapt to the changing needs during the construction process, resulting in insufficient monitoring coverage in certain key areas, low frequency of manual inspections, and inability to achieve 24-hour full-time monitoring, which easily leads to monitoring blind spots. During construction peak periods or high-risk areas, the effectiveness of monitoring is poor.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: A method for monitoring and managing the installation process of an assembled structure, comprising the following steps: S1. Multi-source data fusion and sensor deployment: First, a sensor network is deployed at the construction site. Combined with drone inspections, computer vision recognition, and BIM technology, a data collection network is constructed. Data from different sources is pre-processed using a spatiotemporal data fusion model. S2. Dynamic Sensor Scheduling and Optimization: Utilizing reinforcement learning to optimize sensor deployment strategies, and utilizing drones and robots to dynamically schedule sensors at construction sites, we increase monitoring density in high-risk areas and critical stages, and reduce resource waste in low-risk areas. S3, real-time data transmission and cloud processing: All collected sensor data is uploaded to the cloud processing platform in real time, and the data is directly transmitted on the platform through the data transmission protocol; S4. AI Intelligent Prediction and Anomaly Detection: Based on cloud data, federated learning is used to collaboratively train AI models, improving the predictive capabilities of individual project models while ensuring collaboration and data privacy across multiple projects. Incremental learning methods are used to continuously optimize models, detect anomalies in real time, and make intelligent predictions, providing early warnings. S5. Intelligent Decision Support and Risk Warning: Combining AI model prediction results with expert systems, we automatically generate risk response plans. Using digital twin technology, we conduct virtual simulations of construction sites, modeling structural responses under different risk scenarios and adjusting construction strategies in real time. S6. Automated construction adjustment and response: When a risk warning is triggered, the system automatically generates a detailed adjustment plan and transmits it to the site through intelligent control instructions. Construction adjustments are then executed using automated equipment and robots. S7. Cross-project collaboration and central dispatch: The central dispatch platform integrates data from different projects, implements real-time monitoring and dispatch, and ensures collaborative work between projects. The platform uses blockchain technology for data sharing and evidence storage.

[0006] Preferably, in S1, the sensors include pressure, vibration, displacement, temperature and humidity sensors, the spatiotemporal data fusion model includes LSTM-Transformer, the preprocessing includes cleaning, alignment and modeling, the drone inspection is performed every 1-2 hours according to the route, the flight altitude is 50-60 meters, the shooting resolution is greater than 4000×3000 pixels, and it is used to obtain images and video data of the construction site. The computer vision recognition uses 10 high-definition cameras distributed at different locations of the construction site to collect images in real time, and uses deep learning algorithms to identify component position and posture information. The data cleaning process of the spatiotemporal data fusion model uses a 3σ criterion algorithm based on statistical principles to remove outliers. When modeling, the LSTM layer is set to 3 layers, the number of hidden units is 128, the Transformer layer is set to 2 layers, and the number of heads is 8, which are used to generate complete fusion data.

[0007] Preferably, in S2, the reinforcement learning includes DeepQ-Network, and the reinforcement learning is used to automatically adjust the sensor layout according to the construction progress, the site risk level and the structural health status, the site risk level is divided into low, lower, medium, higher and high, and the structural health status includes stress, strain and displacement data.

[0008] Preferably, in S3, the data transmission protocol includes the MQTT protocol, the data transmission rate is greater than 10Mbps, the data transmission delay is less than 50 milliseconds, and the cloud processing platform is built using elastic computing services and object storage services based on Alibaba Cloud.

[0009] Preferably, in S4, the federated learning includes Federated Learning, the abnormal conditions include structural stress and deformation, the federated learning model is built using the deep learning framework TensorFlow, and the structure is an architecture combining a multi-layer perceptron and a convolutional neural network. The number of training rounds is set to 50 during local training, and the model parameters are uploaded to the central server after every 5 rounds of training. The central server uses a federated averaging algorithm to aggregate the model parameters uploaded by each project, and the aggregation frequency is once after receiving the model parameters of 10 projects.

[0010] Preferably, in S5, the response plan includes adjusting the construction sequence or adding a supporting structure. The expert system is constructed by combining knowledge extraction, representation and storage, and knowledge base construction. The knowledge base includes a rule base and a case base. The prediction results of the AI model are used to predict risks and generate risk response plans. The risk response plans include detailed construction adjustment measures, resource requirements and time arrangements. The digital twin technology is used to construct a time-site mapping model in a virtual environment using three-dimensional modeling software using the collected multi-dimensional data of the construction site. The multi-dimensional data includes the site's geometry, physical and mechanical properties, material properties, and environmental parameters.

[0011] Preferably, in S6, the risk warning includes a structural safety risk warning and a construction process risk warning, the adjustment plan includes a structural safety risk adjustment plan, a construction process risk adjustment plan and an environmental factor risk adjustment plan, and the construction adjustment includes structural adjustment construction, construction process adjustment construction and environmental response construction.

[0012] Preferably, in S7, the evidence is used to make data secure and transparent, prevent data tampering and improve the efficiency of cross-project cooperation. The different project data include project basic information data, construction resource data and construction quality data. The scheduling includes resource scheduling, progress scheduling and technology and experience scheduling. The evidence includes data on-chain evidence storage and evidence data management.

[0013] The present invention provides a method for monitoring and managing the installation process of an assembled structure. It has the following beneficial effects: 1. The present invention deploys multiple sensors and combines drone inspections and computer vision to achieve all-round collection and intelligent analysis of construction site data, thereby achieving real-time and accurate monitoring of construction status and improving construction safety.

[0014] 2. The present invention ensures the security and transparency of construction data through data chain storage technology. At the same time, it combines intelligent scheduling to optimize resources, progress and experience sharing, achieving traceability throughout the construction process and efficient cross-project collaboration, thereby preventing data tampering, improving construction efficiency and optimizing resource allocation.

[0015] 3. The present invention uses a federated learning model, combined with a multi-layer perceptron and a convolutional neural network, to train locally and aggregate the global model through a federated averaging algorithm to achieve multi-project data sharing and risk prediction, thereby improving the accuracy of anomaly detection while ensuring data privacy and security.

[0016] 4. Through DeepQ-Network reinforcement learning, the present invention dynamically adjusts sensor deployment according to construction progress, risk level and structural health status, achieving intelligent allocation and optimization of monitoring resources, thereby improving data collection efficiency and enhancing monitoring accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 The present invention is a flowchart of a method for monitoring and managing the installation process of an assembled structure. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solution of the present invention in conjunction with the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0019] Please see the attached Figure 1 The embodiment of the present invention provides a method for monitoring and managing the installation process of an assembled structure, comprising the following steps: S1. Multi-source data fusion and sensor deployment: First, a sensor network is deployed at the construction site. Combined with drone inspections, computer vision recognition, and BIM technology, a data collection network is constructed. Data from different sources is pre-processed using a spatiotemporal data fusion model. S2. Dynamic Sensor Scheduling and Optimization: Utilizing reinforcement learning to optimize sensor deployment strategies, and utilizing drones and robots to dynamically schedule sensors at construction sites, we increase monitoring density in high-risk areas and critical stages, and reduce resource waste in low-risk areas. S3, real-time data transmission and cloud processing: All collected sensor data is uploaded to the cloud processing platform in real time, and the data is directly transmitted on the platform through the data transmission protocol; S4. AI Intelligent Prediction and Anomaly Detection: Based on cloud data, federated learning is used to collaboratively train AI models, improving the predictive capabilities of individual project models while ensuring collaboration and data privacy across multiple projects. Incremental learning methods are used to continuously optimize models, detect anomalies in real time, and make intelligent predictions, providing early warnings. S5. Intelligent Decision Support and Risk Warning: Combining AI model prediction results with expert systems, we automatically generate risk response plans. Using digital twin technology, we conduct virtual simulations of construction sites, modeling structural responses under different risk scenarios and adjusting construction strategies in real time. S6. Automated construction adjustment and response: When a risk warning is triggered, the system automatically generates a detailed adjustment plan and transmits it to the site through intelligent control instructions. Construction adjustments are then executed using automated equipment and robots. S7. Cross-project collaboration and central dispatch: The central dispatch platform integrates data from different projects, implements real-time monitoring and dispatch, and ensures collaborative work between projects. The platform uses blockchain technology for data sharing and evidence storage.

[0020] In S1, the sensors include pressure, vibration, displacement, temperature and humidity sensors, and the spatiotemporal data fusion model includes LSTM-Transformer. Preprocessing includes cleaning, alignment and modeling. UAV inspections are carried out every 1-2 hours according to the route, with a flight altitude of 50-60 meters and a shooting resolution of >4000×3000 pixels, which are used to obtain image and video data of the construction site. Computer vision recognition uses 10 high-definition cameras distributed at different locations on the construction site to collect images in real time, and uses deep learning algorithms to identify component position and posture information. The data cleaning process of the spatiotemporal data fusion model uses the 3σ criterion algorithm based on statistical principles to remove outliers. When modeling, the LSTM layer is set to 3 layers, the number of hidden units is 128, and the Transformer layer is set to 2 layers, the number of heads is 8, which are used to generate complete fusion data.

[0021] Specifically, pressure, vibration, displacement, temperature and humidity sensors are deployed at the construction site. Pressure sensors monitor changes in structural stress, vibration sensors identify the impact of construction vibrations, displacement sensors detect structural deformation, and temperature and humidity sensors ensure suitable environmental conditions. All data is processed through a spatiotemporal data fusion model (LSTM-Transformer). Data is first cleaned using the 3σ criterion to remove outliers. Time series features are then extracted in the LSTM layer (3 layers, 128 hidden units). Finally, feature fusion is performed in the Transformer layer (2 layers, 8 heads) to improve monitoring accuracy. Drone inspections are carried out every 1-2 hours, with a flight altitude of 50-60 meters and a shooting resolution of no less than 4000×3000 pixels, acquiring high-definition images and video data of the construction site. Computer vision recognition relies on 10 distributed high-definition cameras, combined with deep learning algorithms, to identify the position and posture information of components in real time, providing accurate data support for the construction process. By combining sensors, drone inspections and computer vision, efficient monitoring and intelligent early warning of the construction process are achieved, thereby improving the safety and accuracy of prefabricated building construction.

[0022] In S2, reinforcement learning includes DeepQ-Network. Reinforcement learning is used to automatically adjust sensor layout based on construction progress, site risk level and structural health status. Site risk level is divided into low, relatively low, medium, relatively high and high. Structural health status includes stress, strain and displacement data.

[0023] Specifically, site risk levels are categorized into five levels: low, relatively low, medium, relatively high, and high. These levels are dynamically assessed based on factors such as the stress concentration in the construction area, the complexity of the construction process, and environmental impacts. Structural health status, including stress, strain, and displacement data, characterizes the current safety and stability of the structure. The DQN model utilizes this data to make decisions and continuously optimize sensor placement to enhance monitoring accuracy in critical areas. The DQN first extracts current status information from sensor data at the construction site, including the location of deployed sensors, risk level distribution, and structural health status. Then, using a reinforcement learning policy network, the optimal sensor adjustment plan is predicted based on long-term monitoring results and historical data. For example, the system automatically increases the density of stress and displacement sensors in high-risk areas, while reducing monitoring equipment in low-risk areas to optimize resource allocation. This dynamic optimization of sensor placement through reinforcement learning improves the adaptability of construction monitoring, ensuring construction safety and efficient data collection.

[0024] In S3, the data transmission protocol includes the MQTT protocol, the data transmission rate is greater than 10Mbps, and the data transmission delay is less than 50 milliseconds. The cloud processing platform is built using the elastic computing service and object storage service based on Alibaba Cloud.

[0025] Specifically, data transmission uses the MQTT protocol, which is lightweight and low-bandwidth, making it suitable for real-time transmission of sensor data. The data transmission rate is >10Mbps, ensuring stable transmission of high-frequency monitoring data, while the data transmission delay is kept to <50 milliseconds, meeting real-time monitoring requirements. The cloud-based processing platform is built on Alibaba Cloud's Elastic Compute Service (ECS) and Object Storage Service (OSS). ECS is used for efficient computing and real-time data analysis, while OSS is used to store massive amounts of monitoring data during construction. Once data is transmitted to the cloud, the system automatically performs preprocessing, feature extraction, and risk assessment, providing real-time monitoring results through a visual interface. This efficient data transmission and cloud-based computing enable low-latency, high-speed, and stable monitoring data management at the construction site, improving the intelligence and responsiveness of prefabricated structure construction.

[0026] In S4, federated learning includes Federated Learning, and abnormal conditions include structural stress and deformation. The federated learning model is built using the deep learning framework TensorFlow. The structure is a combination of multi-layer perceptron and convolutional neural network. The number of training rounds is set to 50 during local training. After every 5 rounds of training, the model parameters are uploaded to the central server. The central server uses the federated averaging algorithm to aggregate the model parameters uploaded by each project. The aggregation frequency is once after receiving the model parameters of 10 projects.

[0027] Specifically, the present invention uses federated learning technology to achieve collaborative training of distributed data on construction sites, improving the accuracy of structural health status prediction while ensuring data privacy and security. The federated learning model is built on the TensorFlow deep learning framework and adopts an architecture that combines a multi-layer perceptron (MLP) and a convolutional neural network (CNN) to fully learn key features such as structural stress and deformation. During the local training phase, data from each construction project is trained on the local device for 50 rounds. After every five rounds, the model parameters are uploaded to a central server to reduce communication overhead. The central server uses a federated averaging algorithm to aggregate the model parameters uploaded by each project. The server performs aggregation after receiving model parameters for every 10 projects, updates the global model, and distributes the optimized parameters to each construction project, ensuring that the performance of all project models is simultaneously improved without sharing original data. Through federated learning technology, while ensuring the security of construction data, multi-project collaborative training is achieved, improving the intelligent detection capabilities of construction anomalies (such as structural stress and deformation), and enhancing the generalization and real-time performance of the construction monitoring system.

[0028] In S5, the response plan includes adjusting the construction sequence or adding supporting structures. The expert system is constructed by combining knowledge extraction, representation and storage, and building a knowledge base. The knowledge base contains a rule base and a case base. The AI model prediction results are used to predict risks and generate risk response plans. The risk response plan includes detailed construction adjustment measures, resource requirements and time arrangements. Digital twin technology is used to build a time-site mapping model in a virtual environment through the collected multi-dimensional data of the construction site and three-dimensional modeling software. The multi-dimensional data includes the site's geometry, physical and mechanical properties, material properties, and environmental parameters.

[0029] Specifically, through expert systems, AI models, and digital twin technology, intelligent risk prediction and response can be achieved during the construction process to ensure the safety and efficiency of prefabricated structure construction. The expert system is built based on knowledge extraction, representation, and storage, and includes a rule library and a case library for storing construction safety specifications, empirical rules, and historical cases. The AI model combines construction monitoring data to predict risks, identify abnormal conditions such as structural stress and deformation, and automatically generate risk response plans, including adjusting the construction sequence (such as optimizing the installation sequence to reduce structural stress concentration), adding support structures (adding temporary supports in high-risk areas), detailed construction adjustment measures (such as adjusting connection methods or reinforcement), resource requirements (required materials and equipment), and time arrangements (rationally adjusting the construction progress). Digital twin technology is used to collect multidimensional construction site data (including geometry, physical and mechanical properties, material attributes, and environmental parameters) and construct a virtual construction site mapping model within 3D modeling software. This model can be used to simulate the effects of construction adjustments, optimize construction strategies, and rehearse the construction process in a virtual environment to ensure the feasibility and safety of construction adjustment measures. The collaborative application of AI, expert systems, and digital twins enhances the intelligence level of construction monitoring and risk response, enabling efficient and accurate construction management.

[0030] In S6, risk warning includes structural safety risk warning and construction process risk warning, adjustment plan includes structural safety risk adjustment plan, construction process risk adjustment plan and environmental factor risk adjustment plan, construction adjustment includes structural adjustment construction, construction process adjustment construction and environmental response construction.

[0031] Specifically, through the intelligent monitoring system, early warnings are issued for structural safety risks and construction process risks that occur during the construction process, and corresponding adjustment plans are automatically generated to ensure the safety and stability of the construction process. Risk warnings include structural safety risk warnings (such as abnormal component stress, excessive displacement, loose connections, etc.) and construction process risk warnings (such as installation errors, construction step deviations, quality defects, etc.). The system uses sensor data, computer vision and AI analysis models for real-time detection, and triggers early warning mechanisms when abnormalities are found. In response to the early warning situation, the system automatically generates adjustment plans, including: structural safety risk adjustment plan: such as optimizing component connection methods, adding support structures, adjusting force distribution, etc., construction process risk adjustment plan: optimizing construction steps, adjusting construction sequence or process, reducing errors, and improving construction accuracy, environmental factor risk adjustment plan: adjusting construction time or taking environmental protection measures in response to changes in environmental parameters such as temperature, humidity, and wind speed. When the adjustment plan is implemented on the construction site, the system provides construction adjustment measures, including: structural adjustment construction: optimizing component installation methods, reducing uneven stress distribution, and adjusting assembly accuracy, construction process adjustment construction: optimizing construction processes, such as modifying welding, fastening or bonding processes to ensure construction quality, environmental response construction: increasing protective measures, such as adjusting construction time under severe weather conditions, adding temporary shielding or temperature control equipment, and improving construction safety and reliability through precise risk warnings and intelligent adjustment plans, while reducing potential quality problems and safety hazards during construction.

[0032] In S7, evidence storage is used to make data secure and transparent, prevent data tampering, and improve the efficiency of cross-project cooperation. Different project data include basic project information data, construction resource data, and construction quality data. Scheduling includes resource scheduling, progress scheduling, and technology and experience scheduling. Evidence storage includes data on-chain storage and evidence data management.

[0033] Specifically, evidence storage uses data on-chain evidence storage and evidence data management, using blockchain technology to encrypt and store construction data to prevent tampering and ensure the authenticity and traceability of the data. After the data is on-chain, the evidence management system provides query, verification, and permission management functions, supports cross-project data sharing, and improves the transparency of construction management. The evidence storage mechanism can effectively improve the efficiency of cross-project cooperation, ensure the consistency of data information between different projects, and prevent management chaos or waste of resources due to inconsistent data; The data of different projects include: basic project information data: key project information such as project name, location, construction unit, and person in charge; construction resource data: equipment, materials, and personnel allocation to ensure rational use of resources; construction quality data: inspection data and quality assessment records during the construction process to ensure that construction quality meets standards. The scheduling system optimizes resource scheduling, progress scheduling, and technology and experience scheduling through intelligent analysis. Resource scheduling: Rationally allocate construction machinery, materials, and personnel to improve resource utilization and reduce waste. Progress scheduling: Dynamically adjust the process according to the progress of construction, optimize the construction rhythm, and ensure that the construction period is controllable. Technology and experience scheduling: Improve construction quality and management level through knowledge sharing between different projects. By combining blockchain evidence storage and intelligent scheduling, the security and credibility of construction data can be improved, resource allocation can be optimized, cross-project collaboration can be achieved, and the overall management efficiency of prefabricated building construction can be improved.

[0034] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring and managing the installation process of an assembled structure, characterized in that: The following steps are involved: S1. Multi-source data fusion and sensor deployment: First, a sensor network is deployed at the construction site. Combined with drone inspections, computer vision recognition, and BIM technology, a data collection network is constructed. Data from different sources is pre-processed using a spatiotemporal data fusion model. S2. Dynamic Sensor Scheduling and Optimization: Utilizing reinforcement learning to optimize sensor deployment strategies, and utilizing drones and robots to dynamically schedule sensors at construction sites, we increase monitoring density in high-risk areas and critical stages, and reduce resource waste in low-risk areas. S3, real-time data transmission and cloud processing: All collected sensor data is uploaded to the cloud processing platform in real time, and the data is directly transmitted on the platform through the data transmission protocol; S4. AI Intelligent Prediction and Anomaly Detection: Based on cloud data, federated learning is used to collaboratively train AI models, improving the predictive capabilities of individual project models while ensuring collaboration and data privacy across multiple projects. Incremental learning methods are used to continuously optimize models, detect anomalies in real time, and make intelligent predictions, providing early warnings. S5. Intelligent Decision Support and Risk Warning: Combining AI model prediction results with expert systems, we automatically generate risk response plans. Using digital twin technology, we conduct virtual simulations of construction sites, modeling structural responses under different risk scenarios and adjusting construction strategies in real time. S6. Automated construction adjustment and response: When a risk warning is triggered, the system automatically generates a detailed adjustment plan and transmits it to the site through intelligent control instructions. Construction adjustments are then executed using automated equipment and robots. S7. Cross-project collaboration and central dispatch: The central dispatch platform integrates data from different projects, implements real-time monitoring and dispatch, and ensures collaborative work between projects. The platform uses blockchain technology for data sharing and evidence storage.

2. A method for monitoring and managing the installation process of an assembled structure according to claim 1, characterized in that: In S1, the sensors include pressure, vibration, displacement, temperature and humidity sensors, the spatiotemporal data fusion model includes LSTM-Transformer, the preprocessing includes cleaning, alignment and modeling, the drone inspection is performed every 1-2 hours according to the route, the flight altitude is 50-60 meters, the shooting resolution is greater than 4000×3000 pixels, and it is used to obtain images and video data of the construction site. The computer vision recognition uses 10 high-definition cameras distributed at different locations on the construction site to collect images in real time, and uses deep learning algorithms to identify component position and posture information. The data cleaning process of the spatiotemporal data fusion model uses a 3σ criterion algorithm based on statistical principles to remove outliers. When modeling, the LSTM layer is set to 3 layers, the number of hidden units is 128, and the Transformer layer is set to 2 layers, the number of heads is 8, which are used to generate complete fusion data.

3. A method for monitoring and managing the installation process of an assembled structure according to claim 1, characterized in that: In S2, the reinforcement learning includes DeepQ-Network, and the reinforcement learning is used to automatically adjust the sensor layout based on the construction progress, the site risk level and the structural health status. The site risk level is divided into low, relatively low, medium, relatively high and high. The structural health status includes stress, strain and displacement data.

4. A method for monitoring and managing the installation process of an assembled structure according to claim 1, characterized in that: In S3, the data transmission protocol includes the MQTT protocol, the data transmission rate is greater than 10Mbps, and the data transmission delay is less than 50 milliseconds. The cloud processing platform is built using elastic computing services and object storage services based on Alibaba Cloud.

5. The method for monitoring and managing the installation process of an assembled structure according to claim 1, wherein: In S4, the federated learning includes Federated Learning, the abnormal conditions include structural stress and deformation, and the federated learning model is built using the deep learning framework TensorFlow. The structure is an architecture combining a multi-layer perceptron and a convolutional neural network. The number of training rounds is set to 50 during local training. After every 5 rounds of training, the model parameters are uploaded to the central server. The central server uses a federated averaging algorithm to aggregate the model parameters uploaded by each project. The aggregation frequency is once after receiving the model parameters of 10 projects.

6. The method for monitoring and managing the installation process of an assembled structure according to claim 1, wherein: In S5, the response plan includes adjusting the construction sequence or adding a supporting structure. The expert system is constructed by combining knowledge extraction, representation and storage, and knowledge base construction. The knowledge base includes a rule base and a case base. The prediction results of the AI model are used to predict risks and generate risk response plans. The risk response plans include detailed construction adjustment measures, resource requirements and time arrangements. The digital twin technology is used to construct a time-site mapping model in a virtual environment using three-dimensional modeling software using the collected multi-dimensional data of the construction site. The multi-dimensional data includes the site's geometry, physical and mechanical properties, material properties, and environmental parameters.

7. The method for monitoring and managing the installation process of an assembled structure according to claim 1, wherein: In S6, the risk warning includes structural safety risk warning and construction process risk warning, the adjustment plan includes structural safety risk adjustment plan, construction process risk adjustment plan and environmental factor risk adjustment plan, and the construction adjustment includes structural adjustment construction, construction process adjustment construction and environmental response construction.

8. The method for monitoring and managing the installation process of an assembled structure according to claim 1, wherein: In S7, the evidence is used to make data safe and transparent, prevent data tampering and improve the efficiency of cross-project cooperation. The different project data include basic project information data, construction resource data and construction quality data. The scheduling includes resource scheduling, progress scheduling and technology and experience scheduling. The evidence includes data on-chain evidence storage and evidence data management.

Images