Building construction quality information intelligent supervision system based on BIM

Abstract

The invention relates to the technical field of data processing, and discloses a BIM-based building construction quality information intelligent supervision system, which is used for improving the efficiency and accuracy of BIM-based building construction quality information intelligent supervision. The method comprises the following steps: a modeling module performs BIM modeling processing on a building engineering design drawing and building specification data to obtain a building twinborn body; the deployment module carries out field data acquisition system deployment on the building twinborn body to obtain a real-time data acquisition system; the detection module performs anomaly detection processing on the data acquired by the system to obtain a quality anomaly alarm and a priority processing level; the processing module carries out multi-dimensional visualization processing on the quality abnormity alarm and the priority processing level to obtain a visual interface; the decision-making module performs intelligent decision-making analysis on the quality abnormal information in the visual interface to obtain a target processing scheme; and the analysis module performs time sequence analysis and multivariable correlation analysis on the target processing scheme to obtain a dynamic quality evaluation report.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a building construction quality information intelligent supervision system based on BIM. Background Art

[0002] With the rapid development of the construction industry, construction quality management has received more and more attention. Traditional construction quality management methods mainly rely on manual inspections and paper records, which have problems such as low efficiency, scattered data, and poor real-time performance. In recent years, with the advancement of information technology, technologies such as building information modeling (BIM), the Internet of Things (IoT), and big data analysis have gradually been introduced into the field of construction quality management. BIM technology can create a virtual model of a building, providing a visual basis for quality management; IoT technology can realize real-time collection of on-site data; and big data analysis technology can conduct in-depth analysis of the collected data to provide support for decision-making.

[0003] However, the existing construction quality information management system still has some shortcomings. First, most systems lack deep integration of BIM models, IoT data collection and intelligent analysis, resulting in the continued existence of information islands. Second, existing systems usually focus on data collection and storage, but have weak capabilities for intelligent analysis and visualization of data, and are unable to fully tap the value of big data. Third, many systems lack adaptive learning capabilities, making it difficult to continuously optimize the decision-making process based on the specific circumstances of the project and historical experience. Finally, existing systems are inefficient in processing multi-source heterogeneous data, making it difficult to meet the real-time supervision needs of large-scale construction projects. Summary of the invention

[0004] In view of this, an embodiment of the present invention provides a BIM-based intelligent supervision system for building construction quality information, which is used to improve the efficiency and accuracy of BIM-based intelligent supervision of building construction quality information.

[0005] The present invention provides a BIM-based intelligent supervision system for building construction quality information, comprising:

[0006] The modeling module is used to perform BIM modeling on the architectural engineering design drawings and pre-collected building specification data to obtain an intelligent BIM building twin that includes quality parameters and quality control points;

[0007] A deployment module, used to deploy a field data acquisition system for the intelligent BIM building twin to obtain a real-time data acquisition system including a fixed sensor network and mobile acquisition devices;

[0008] A detection module, used to perform abnormality detection processing on the data collected by the real-time data acquisition system to obtain quality abnormality alarms and priority processing levels;

[0009] A processing module, used for performing multi-dimensional visualization processing on the quality abnormality alarm and the priority processing level to obtain a dynamic three-dimensional visualization interface including a heat map and a force-directed map;

[0010] A decision-making module, used to perform intelligent decision-making analysis on the quality abnormality information in the dynamic three-dimensional visualization interface to obtain a target processing solution including corrective measures and resource allocation strategies;

[0011] The analysis module is used to perform time series analysis and multivariate correlation analysis on the target treatment scheme to obtain a dynamic quality assessment report containing trend prediction data and parameter impact weights.

[0012] In the present invention, the modeling module is specifically used to: perform digital conversion processing on the construction engineering design drawings to obtain building structure digital data, and perform parameterization processing on the building structure digital data to obtain initial BIM data containing geometric information and attribute information; perform quality parameter implantation processing on the initial BIM data to obtain parameterized BIM data containing material strength, dimensional error and construction process requirements, and perform quality standard association processing on the parameterized BIM data according to the pre-collected building specification data to obtain standardized BIM data with a quality assessment benchmark; perform quality control point setting processing on the standardized BIM data to obtain monitoring point layout data including key structural nodes and construction quality sensitive areas, and perform data association processing on the monitoring point layout data to obtain data collection rules corresponding to each quality parameter; perform analysis and processing on historical project data through a deep learning algorithm to obtain quality risk prediction data, and perform risk area labeling processing on the standardized BIM data according to the quality risk prediction data to obtain an intelligent BIM building twin containing high-risk area identification.

[0013] In the present invention, the deployment module is specifically used to: analyze and process the quality control points in the intelligent BIM building twin to obtain a sensor layout strategy, and determine the installation position of the fixed sensor according to the sensor layout strategy to obtain a fixed sensor layout plan; perform network topology design processing on the fixed sensor layout plan to obtain a low-power wide area network structure, and perform communication protocol configuration processing on the low-power wide area network structure to obtain a fixed sensor data transmission plan; perform data acquisition task allocation processing on the mobile acquisition device to obtain a mobile acquisition path planning, and perform data fusion processing on the fixed sensor data transmission plan and the mobile acquisition path planning to obtain a comprehensive data acquisition strategy; perform data storage structure design processing on the comprehensive data acquisition strategy to obtain a real-time database architecture, perform data interface definition processing on the real-time database architecture to obtain a real-time data acquisition system including a fixed sensor network and a mobile acquisition device.

[0014] In the present invention, the detection module is specifically used for: preprocessing the data collected by the real-time data acquisition system to obtain cleaned standardized data, and performing time series decomposition processing on the standardized data to obtain trend, seasonal and random component data; performing statistical feature extraction processing on the trend, seasonal and random component data to obtain data feature vectors; performing cluster analysis processing on the data feature vectors through a clustering algorithm to obtain data anomaly degree classification results; performing threshold judgment processing on the data anomaly degree classification results to obtain preliminary quality anomaly alarms; performing context association analysis processing on the preliminary quality anomaly alarms to obtain associated influencing factor data, and performing weight calculation processing on the associated influencing factor data to obtain anomaly severity scores; and prioritizing the quality anomaly alarms according to the anomaly severity scores to obtain quality anomaly alarms and priority processing levels.

[0015] In the present invention, the processing module is specifically used to: perform three-dimensional scene rendering processing on the intelligent BIM building twin to obtain a basic three-dimensional visualization interface, and perform spatial positioning processing on the quality abnormality alarm to obtain abnormal point position coordinate data; mark the basic three-dimensional visualization interface according to the abnormal point position coordinate data to obtain an abnormal point identification three-dimensional interface, and perform color mapping processing on the priority processing level to obtain an abnormality severity color scheme; perform heat map drawing processing on the abnormal point identification three-dimensional interface according to the abnormality severity color scheme to obtain a heat map superimposed three-dimensional interface, and perform correlation analysis processing on the quality abnormality alarm to obtain relationship data between abnormal points; perform force-directed graph drawing processing on the heat map superimposed three-dimensional interface according to the relationship data between the abnormal points to obtain a force-directed graph superimposed three-dimensional interface; perform interactive function design processing on the force-directed graph superimposed three-dimensional interface to obtain a dynamic three-dimensional visualization interface including a heat map and a force-directed graph.

[0016] In the present invention, the decision module is specifically used to: perform feature extraction processing on the quality abnormality information in the dynamic three-dimensional visualization interface to obtain an abnormal feature data set, and perform data normalization processing on the abnormal feature data set to obtain standardized abnormal feature data; perform classification processing on the standardized abnormal feature data through a decision tree algorithm to obtain an abnormal type identification result; perform similarity matching processing on a historical rectification case library based on the abnormal type identification result to obtain a candidate rectification measure set, perform feasibility assessment processing on the candidate rectification measure set to obtain a preliminary rectification plan; perform resource demand analysis processing on the preliminary rectification plan to obtain a resource demand list, and perform scheduling optimization processing on existing resources based on the resource demand list to obtain a resource allocation strategy; perform collaborative optimization processing on the preliminary rectification plan and the resource allocation strategy to obtain a target processing plan including rectification measures and resource allocation strategies.

[0017] In the present invention, the analysis module is specifically used for: performing time dimension decomposition processing on the corrective measures and resource allocation strategies in the target processing scheme to obtain a time series data set, performing seasonal adjustment processing on the time series data set to obtain deseasonalized time series data; performing trend prediction processing on the deseasonalized time series data by ARIMA algorithm to obtain quality parameter trend prediction data, performing confidence interval calculation processing on the quality parameter trend prediction data to obtain trend prediction results with uncertainty ranges; and performing correlation analysis processing on various measures and strategies in the target processing scheme to obtain a parameter correlation coefficient matrix; performing dimensionality reduction processing on the parameter correlation coefficient matrix by principal component analysis to obtain a set of key influencing factors; performing importance sorting processing on the key influencing factor set to obtain parameter influence weight data; performing comprehensive analysis processing on the trend prediction results and parameter influence weight data with uncertainty ranges to obtain a dynamic quality assessment report containing trend prediction data and parameter influence weights.

[0018] In the technical solution provided by the present invention, by performing BIM modeling on construction engineering design drawings and building specification data, an intelligent BIM building twin is obtained, which provides a unified digital foundation for the entire quality supervision process and effectively solves the problem of information dispersion and difficulty in integration in traditional methods; based on the intelligent BIM building twin, an on-site data acquisition system is deployed to achieve the coordinated work of fixed sensor networks and mobile acquisition equipment, greatly improving the comprehensiveness and real-time nature of data acquisition, so that quality supervision can promptly discover and respond to changes on site; through the anomaly detection module, the collected data is intelligently analyzed, and the system can automatically identify quality anomalies and give priority processing levels, greatly reducing the work of manual inspection The multi-dimensional visualization processing module converts quality abnormality alarms and priority processing levels into intuitive heat maps and force-directed maps, and displays them in a dynamic three-dimensional visualization interface, so that managers can quickly grasp the project quality status and intuitively understand the correlation between quality problems, so as to make more reasonable decisions; the intelligent decision analysis module can automatically generate target processing solutions including corrective measures and resource allocation strategies based on historical experience and current conditions, greatly improving the scientificity and efficiency of decision-making; the time series analysis and multivariate correlation analysis modules conduct in-depth analysis of the target processing solutions and generate dynamic quality assessment reports, which can not only predict the development trend of quality parameters, but also It can identify key influencing factors and provide data support for continuous improvement; the entire system realizes full-process automation from data collection, anomaly detection, visualization to intelligent decision-making, greatly reducing manual intervention and improving the efficiency and consistency of quality management; the modular design of the system enables close collaboration between various functional modules to form a closed-loop quality management system that can continuously learn and optimize to adapt to the characteristics and needs of different projects; by analyzing historical project data through deep learning algorithms, the system can continuously accumulate experience, improve the accuracy of quality risk prediction, and provide a powerful tool for proactively preventing quality problems; the system's real-time data collection and processing capabilities enable managers to grasp project quality at any time The system can identify and solve problems in a timely manner, effectively reducing quality risks. The dynamic 3D visualization interface not only provides an intuitive display of quality status, but also supports interactive operations, making it convenient for managers to conduct in-depth analysis and decision-making. The system's adaptive optimization capability enables it to continuously adjust and optimize supervision strategies according to project progress and changes in the external environment, ensuring the continued effectiveness of quality management. Through comprehensive analysis of multi-source heterogeneous data, the system can reveal potential quality problems and influencing factors, providing comprehensive data support for quality improvement. The system's intelligent report generation function can automatically generate evaluation reports containing trend forecasts and parameter influence weights, providing a scientific basis for management decisions and improving the accuracy and reliability of decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 The present invention relates to a flow chart of a BIM-based intelligent supervision system for building construction quality information in an embodiment of the present invention.

[0021] Reference numerals:

[0022] 101. Modeling module; 102. Deployment module; 103. Detection module; 104. Processing module; 105. Decision module; 106. Analysis module. DETAILED DESCRIPTION

[0023] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0025] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 , Figure 1 : is a flow chart of a BIM-based intelligent supervision system for building construction quality information according to an embodiment of the present invention. Figure 1 As shown, the following steps are included:

[0026] A modeling module 101 is used to perform BIM modeling processing on the construction engineering design drawings and the pre-collected building specification data to obtain an intelligent BIM building twin including quality parameters and quality control points;

[0027] Specifically, the modeling module 101 digitally converts the design drawings of the construction project. This involves using professional CAD software or BIM software to convert two-dimensional design drawings into three-dimensional models in order to better represent the structure and function of the building. The specific implementation method is to develop a custom plug-in based on the Autodesk Revit API, which first uses the ImportCAD method to import the pre-processed CAD file, then uses ElementClassFilter and ElementCategoryFilter to filter out key components, then calls the CreateFamilyInstance method to create the corresponding 3D family instance, and finally sets the specific parameters of each component through the ParameterSet method. These three-dimensional models not only include the geometric shape of the building, but also cover information such as building materials, component dimensions, and structural layout, forming a preliminary BIM (Building Information Model).

[0028] The pre-collected building specification data are imported into the modeling module 101. These data usually include national or industry-specified building standards, construction specifications, material performance parameters, and safety and environmental protection requirements. The modeling module integrates these specification data with the preliminary BIM model through data processing algorithms, so that the model is not only a collection of geometric information, but also includes various parameters related to quality. The specific implementation method is to develop a building specification database based on SQL Server and a data integration program written in C#. The program interacts with the database through ADO.NET, extracts relevant specification data, and then uses the ExternalCommand interface of the Revit API to write these data into the corresponding parameters of the BIM model.

[0029] These parameters may involve the strength, durability, fire resistance of building materials, as well as the tolerance range and process requirements during the construction process. For example, when processing the BIM data of the wall, the module will implant the corresponding quality parameters according to the building specifications, such as the compressive strength and thermal insulation performance of the wall material, and set the allowable construction error range. Specifically, for concrete walls, the system will set specific values ​​such as the compressive strength of C30 concrete is 30MPa, the thermal conductivity is not more than 0.5W / (m·K), and the verticality error of the wall does not exceed 3mm.

[0030] The setting of quality control points is a key step to ensure construction quality. Modeling module 101 will mark key quality control points in the BIM model. These points are usually set at key locations of the building structure, such as load-bearing walls, beam-column connections, and the connection between the foundation and the top floor. The specific implementation method is to develop a quality control point identification algorithm based on Revit API, which identifies key structural nodes through topological analysis, determines stress concentration areas using finite element analysis, and marks complex construction locations in combination with a construction difficulty assessment algorithm. Through these control points, the construction party can focus on monitoring the construction quality of these locations during the actual construction process to ensure that they meet the design requirements and building specifications.

[0031] In addition, the modeling module can also predict potential quality risks through the analysis of historical project data and mark high-risk areas in the model. These high-risk areas are usually areas where problems or difficulties have occurred in historical data, such as settlement problems caused by improper foundation treatment, or areas where material performance is unstable. The specific implementation method is to use Python scientific computing libraries (such as NumPy and SciPy) to develop machine learning models, such as random forest algorithms, analyze historical project data, establish a risk prediction model, and then apply the model to the current project to identify potential high-risk areas.

[0032] After all the information is integrated, the modeling module 101 generates an intelligent BIM building twin that includes quality parameters and quality control points. This twin is not only a virtual model of a building, but also a digital platform that can be dynamically updated and monitored in real time. The specific implementation includes using Revit API to develop a dynamic link mechanism to achieve real-time data updates; using WPF (Windows Presentation Foundation) to develop a visual interface to support multi-dimensional data display; integrating a rule-based early warning system to issue real-time alerts for parameters that exceed quality thresholds. Construction personnel can obtain detailed information about the building through this platform, including the specifications of each component, quality requirements, and construction schedule.

[0033] For example, if a project requires the use of high-strength concrete to build a load-bearing wall, the modeling module will indicate the specific quality requirements of the wall in the twin, and monitor the concrete mix and pouring process in real time during the construction process to ensure the construction quality. The specific implementation method is to collect real-time data through IoT sensors (such as embedded strain gauges and temperature sensors), which are transmitted to the central server via wireless networks and then analyzed and compared by the data processing program. Once an abnormality is found (such as substandard strength or excessive temperature), the system will immediately send an alarm message to the construction personnel's mobile device through a preset communication protocol (such as MQTT).

[0034] A deployment module 102 is used to deploy a field data acquisition system for the intelligent BIM building twin to obtain a real-time data acquisition system including a fixed sensor network and mobile acquisition devices;

[0035] Specifically, the deployment module 102 will conduct a comprehensive analysis of the needs of the construction site based on the information in the intelligent BIM building twin. This includes determining which key quality control points need to be monitored in real time and selecting appropriate sensor types and data acquisition equipment. The specific implementation method is to develop a Python-based data analysis script that uses the pandas library to process the data exported by the BIM model, identify key monitoring points through clustering algorithms (such as K-means), and select suitable sensor types based on a preset decision tree algorithm. These sensors may include temperature sensors, humidity sensors, vibration sensors, displacement sensors, etc. The specific selection depends on the characteristics of the building and the quality parameters that need to be monitored during the construction process. For example, for the concrete pouring process that requires precise control of temperature and humidity, temperature and humidity sensors (such as DHT22 digital temperature and humidity sensors) will be widely deployed to ensure that the concrete is cured under suitable environmental conditions.

[0036] After determining the sensor type, the deployment module 102 will formulate a layout strategy for the fixed sensor network based on the data in the intelligent BIM building twin. This strategy is implemented by developing an optimization program based on a genetic algorithm, which considers factors such as sensor coverage, signal strength, and installation difficulty to calculate the best sensor layout plan. This strategy includes the optimal installation location, installation method, and wiring plan of the sensor to ensure that the sensor can efficiently collect the required data and can work normally in the complex environment of the construction site. For example, in the construction of a high-rise building, the deployment module may install strain sensors (such as BX120-50AA resistance strain gauges) at key structural nodes such as columns, beams, and floor slabs to monitor the stress state and deformation of these components in real time to ensure construction quality and structural safety.

[0037] At the same time, the deployment module 102 also handles the layout of mobile collection equipment, which usually includes drones, mobile robots, handheld scanners, etc. The specific implementation method is to develop a mobile device control program based on ROS (Robot Operating System), which can automatically plan the movement path and data collection tasks of the equipment. The use of these devices can flexibly cover areas that are difficult to monitor with fixed sensors, or be used for temporary and detailed monitoring of specific processes. For example, during the foundation excavation process at the construction site, a drone equipped with a high-precision laser scanner (such as Velodyne VLP-16) may be used to regularly photograph and measure the entire excavation area to monitor surface deformation and ensure safety during the excavation process.

[0038] Path planning for mobile acquisition equipment is one of the important tasks of the deployment module. It will optimize the movement route of the equipment according to the model of the building twin to ensure that all key areas are covered while avoiding interference with the construction process. The specific implementation method is to use the A* algorithm combined with a dynamic obstacle avoidance strategy to plan the optimal path in real time. After the fixed sensors and mobile acquisition equipment are deployed, the deployment module 102 integrates them into a unified real-time data acquisition system. The system adopts a distributed message queue architecture based on Apache Kafka to ensure low-latency and high-reliability data transmission capabilities to ensure that the field data collected from sensors and mobile devices can be quickly transmitted to the central monitoring platform.

[0039] To this end, the deployment module will design an efficient network topology based on the network conditions and data volume requirements of the construction site. Low-power wide area network (LPWAN) technologies such as LoRaWAN or 5G networks are usually used to ensure the stability and timeliness of data transmission. The specific implementation includes the use of LoRa gateways (such as Multitech Conduit) and LoRa node modules (such as RFM95W) to build a wireless sensor network covering the entire construction site. In addition, the deployment module also needs to design a data storage and processing architecture, using a distributed storage system based on Apache Hadoop and a real-time data processing framework based on Apache Spark to ensure that data can be stored and analyzed in real time, providing a reliable basis for subsequent quality monitoring and decision-making.

[0040] For example, in a large bridge construction project, the deployment module 102 may arrange fiber Bragg grating strain sensors (such as os3600 fiber strain sensors) at key stress points of the bridge, such as piers and bridge towers, and arrange drones equipped with laser scanners (such as DJI Matrice 300 RTK) to regularly fly and monitor the construction progress and geometric changes of the bridge deck. These data are transmitted to the central monitoring platform (based on Siemens MindSphere IoT operating system) through a real-time data acquisition system (using OPC UA protocol). Construction managers can view the stress status and construction progress of each part of the bridge in real time through a customized visualization interface, and promptly discover and deal with potential quality problems to ensure the safety and quality of bridge construction.

[0041] The detection module 103 is used to perform abnormal detection processing on the data collected by the real-time data acquisition system to obtain quality abnormality alarms and priority processing levels;

[0042] Specifically, the detection module 103 receives real-time data from various sensors and mobile acquisition devices. These data may include information in multiple dimensions such as temperature, humidity, strain, displacement, vibration, etc. Since these data often contain noise or incomplete parts, the module first preprocesses the data, including data cleaning, removing outliers, and filling missing data. The purpose is to convert the original data into a clean and standardized data set to ensure the accuracy of subsequent analysis. The specific implementation method is to use Python's Pandas library for data cleaning, use the moving median method to remove outliers, and use the KNN (K-Nearest Neighbors) algorithm to interpolate missing data. For example, for temperature data, a threshold of ±3σ (standard deviation) can be set to identify outliers, and data points outside this range will be removed or replaced with the average of neighboring valid values.

[0043] After the data preprocessing is completed, the detection module 103 starts to perform time series decomposition analysis, and decomposes the collected standardized data into three parts: trend, seasonality and random components. The specific implementation uses the seasonal_decompose function in the statsmodels library of Python, and uses an additive model or a multiplicative model for decomposition. The trend component reflects the overall change direction of the data over time, the seasonal component shows periodic fluctuations, and the random component includes unpredictable abnormal fluctuations. This decomposition analysis helps the module better understand the essential changes of the data and provides a basis for anomaly detection. For example, when monitoring the stress of a building structure, the trend component may show that the pressure on the structure gradually increases as the construction progresses, while the seasonal component may reflect the impact of the temperature difference between day and night on the structural stress. By identifying these normal change patterns, the module can more accurately detect abnormal changes that exceed the expected range.

[0044] The detection module 103 then uses a clustering algorithm to perform feature extraction and cluster analysis on the data after time series decomposition, grouping similar data points to determine the degree of abnormality of the data. The specific implementation uses the K-means clustering algorithm in Python's scikit-learn library, combined with the silhouette coefficient to automatically determine the optimal number of clusters. For example, when monitoring the wind load of a high-rise building, cluster analysis may distinguish normal wind pressure change patterns from abnormal wind pressure surges. This distinction can help the module identify potential quality problems, such as structural offset or component deformation due to abnormal wind pressure.

[0045] After the cluster analysis is completed, the detection module further performs threshold judgment and context association analysis on the abnormal data. This involves comparing each abnormal data point with the preset threshold. If the data exceeds the threshold range, it is preliminarily determined to be a quality abnormality. The specific implementation uses Python's NumPy library to perform matrix operations and quickly compare data with thresholds. In order to improve the accuracy of detection, the module also performs context association analysis to link abnormal data with relevant influencing factors. The implementation method is to use the Granger causality test to analyze the causal relationship between different time series. For example, when a crack is found in a section of wall, the module will not only determine the severity of the crack, but also associate and analyze factors such as recent construction activities, material strength, and weather changes to comprehensively evaluate the root cause of the abnormality.

[0046] The detection module 103 prioritizes all identified anomalies by scoring the severity of the anomalies. The implementation uses a weighted scoring system and uses a custom function in Python to calculate the overall score. The module calculates the severity score for each anomaly based on the results of the contextual analysis, taking into account factors such as the scope of the anomaly, potential consequences, and the urgency of resolving the problem. For example, the detection module may mark a structural crack in a load-bearing wall as a high priority issue because such issues, if not addressed in a timely manner, may pose a serious threat to the safety of the entire building.

[0047] For example, in a construction project of a large underground parking lot, the detection module 103 may monitor the change of the groundwater level in real time. If the groundwater level sensor (such as the HOBO U20L-04 water level recorder) detects a sudden rise in the water level, the detection module first pre-processes and decomposes the anomaly, and uses Python's Prophet library to perform time series prediction to determine whether the change is a normal seasonal fluctuation or a sudden abnormal phenomenon. If it is identified as an anomaly, the detection module 103 will analyze relevant factors, such as the recent rainfall (obtained through the meteorological API) and the operating status of the underground drainage system (through the drainage system sensor data). Finally, the detection module 103 prioritizes the water level anomaly and uses a rule-based expert system to make decisions. It may determine that it is a high priority issue and trigger an immediate alarm through a preset communication protocol (such as MQTT), suggesting that immediate measures be taken to prevent flooding at the construction site.

[0048] The processing module 104 is used to perform multi-dimensional visualization processing on the quality abnormality alarm and the priority processing level to obtain a dynamic three-dimensional visualization interface including a heat map and a force-directed map;

[0049] Specifically, the processing module 104 receives quality anomaly alarms and priority level data from the detection module 103, which cover multiple parts of the building and different types of abnormal situations. In order to convert these complex information into easy-to-understand visualization content, the processing module first needs to perform three-dimensional scene rendering processing on the intelligent BIM building twin to generate a basic three-dimensional visualization interface. The specific implementation method is to use Autodesk Forge API to render the three-dimensional model and use WebGL technology to achieve high-performance 3D graphics display in the browser. This interface shows the full picture of the building, including its geometric structure, key components, and basic information such as construction progress. After the basic three-dimensional visualization interface is generated, the processing module 104 will perform spatial positioning processing based on the received abnormal alarms, and accurately mark each abnormal point in the three-dimensional model. This step is implemented by developing a custom plug-in based on Three.js, which can convert 2D coordinates into precise positions in 3D space. This step not only includes determining the specific location of the abnormal point, but also involves superimposing multi-dimensional information on the abnormal point. For example, if a building column detects an abnormal alarm of excessive structural stress, the processing module will map the location information of the abnormal alarm to the specific location of the column and mark it with a specific symbol or color on the 3D interface. This spatial positioning allows construction workers to quickly identify the specific location of the problem and focus on inspection and repair.

[0050] The processing module 104 further visualizes these marked abnormal points, first by visualizing the priority level as different color levels through color mapping. The specific implementation uses the color scale function of the D3.js library to create a continuous color mapping from green to red. Usually, the processing module uses a color gradient to indicate priority, for example, marking low-priority abnormalities as green, medium-priority ones as yellow, and high-priority ones as red. Through this intuitive color mapping, construction managers can see at a glance which problems are the most urgent and need to be handled as a priority.

[0051] After the color mapping is completed, the processing module 104 further visually enhances the three-dimensional interface by drawing a heat map. The implementation of the heat map uses WebGL shader technology, and the gradient effect of color intensity is achieved through a custom fragment shader program. The heat map represents the density and severity of abnormalities in different areas through changes in color intensity. For example, in the construction of a building, if the quality abnormalities of multiple components are concentrated on a certain floor or a certain area, the heat map will show a darker color in the area, which prompts the construction management personnel that there may be systemic problems in the area and a detailed inspection must be carried out immediately. The superposition of the heat map enables the three-dimensional visualization interface to not only display a single abnormal point, but also provide the distribution of quality problems at the macro level, helping the construction team to grasp the quality risks of the project as a whole. The processing module 104 also uses a force-directed graph to further reveal the potential associations between abnormal points. The implementation of the force-directed graph uses the force-directed layout algorithm of the D3.js library, combined with WebGL technology for efficient rendering. The force-directed graph is a graphical network structure, where nodes represent abnormal points and edges represent the relationship or impact path between these abnormal points. For example, if the abnormalities of some columns are detected, which may be caused by problems in the foundation construction, the processing module will connect these abnormal points through a force-directed graph to show the causal relationship or common characteristics between them. This kind of diagram can help engineers understand the complex network of abnormal problems and provide ideas for troubleshooting and solving related problems.

[0052] Finally, the processing module 104 designs dynamic interactive functions for the entire three-dimensional visualization interface, allowing users to freely switch perspectives, zoom in on specific areas, or click on abnormal points to view detailed information in the interface. Such interactive design is achieved by developing an orbit controller (OrbitControls) based on Three.js and a custom click event handler. For example, construction personnel can click on an abnormal point to immediately view the detailed data of the abnormality, such as the nature of the abnormality, detection time, severity score, etc., and can also call up related historical data or rectification suggestions. This dynamic interaction enables construction managers to quickly obtain the required information and make timely decisions when dealing with complex engineering quality issues.

[0053] For example, in a construction project of a large commercial complex, the processing module 104 will display the quality anomalies of each floor and each component in real time. If the stress of multiple support beams on the 10th floor is detected to be abnormal and has a high priority, the processing module will mark the support beams on the 10th floor as red in the 3D visualization interface (using the material system of Three.js to set the component color) and highlight the high-risk areas of the floor through a heat map. In addition, through the force-directed graph, the module may show the relationship between these beams and other components with similar construction time on the floor slab, prompting managers to check potential problems in the floor slab construction. Construction personnel can use the interactive function of the interface to view the specific information of these anomalies in depth, and even call historical data to analyze whether such problems have occurred on other floors, so as to make more targeted decisions. These functions are implemented through real-time connection and data query with the back-end database (such as MongoDB) to ensure the timeliness and accuracy of information.

[0054] The decision module 105 is used to perform intelligent decision analysis on the quality abnormality information in the dynamic three-dimensional visualization interface to obtain a target processing solution including rectification measures and resource allocation strategies;

[0055] Specifically, the decision module 105 extracts quality anomaly information from the dynamic three-dimensional visualization interface, including the type, severity, spatial location, and correlation with other anomalies. The module converts this data into an anomaly feature data set through a feature extraction algorithm, which contains detailed information and feature vectors of each anomaly. The specific implementation method is to use the PCA (principal component analysis) algorithm in Python's scikit-learn library to extract features, and combine custom feature engineering functions to capture domain-specific information. In order to eliminate the dimensional differences between different anomaly features, the module normalizes these data to ensure the accuracy of subsequent analysis. This step is implemented using the StandardScaler class of scikit-learn, which converts all features to a standard normal distribution with a mean of 0 and a variance of 1. For example, if the stress data (in MPa) and temperature data (in °C) involved in an anomaly point have different dimensions, the normalization process will make these data comparable and analyzed within the same range to ensure the rationality of the analysis results. The decision module 105 uses a decision tree algorithm to classify the normalized anomaly feature data. The specific implementation uses the DecisionTreeClassifier class in the scikit-learn library, and controls the model complexity and prevents overfitting by setting parameters such as max_depth (maximum depth of the tree) and min_samples_split (minimum number of samples required for internal node re-division). The decision tree algorithm is a classification method based on feature selection. It automatically determines which category the current anomaly belongs to by learning the feature patterns in historical anomaly cases. For example, a crack in a beam column may be caused by insufficient material strength, while another similar crack may be caused by improper temperature control during construction. The decision tree algorithm can accurately classify different types of anomalies based on these features, helping the construction team quickly identify the root cause of the problem.

[0056] After the abnormal type is identified, the decision module 105 will call the historical rectification case library for similarity matching analysis. The historical rectification case library is implemented using MongoDB, which supports efficient document storage and retrieval. The module compares the current abnormality with the problem features in the case library through the cosine similarity algorithm (implemented using the cosine function of the scipy library), finds the most similar historical cases, and extracts possible rectification measures from them. This process ensures that the treatment plan generated by the module has a practical basis and high feasibility. For example, if a certain type of wall crack has been successfully repaired in a high-rise building construction, the decision module will use the repair plan as a candidate rectification measure for the current similar problem.

[0057] After generating candidate corrective measures, the module also conducts a feasibility assessment based on the specific conditions of the current construction site, available resources, time constraints and other factors. The specific implementation adopts the multi-criteria decision analysis (MCDA) method and uses Python's AHP (analytic hierarchy process) library to calculate the comprehensive score of each solution. The module analyzes the implementation difficulty, required resources and possible effects of each candidate solution and selects the most appropriate measure as the preliminary corrective solution. For example, if a solution requires a large amount of manpower and special equipment in the current construction phase, and these resources are difficult to allocate in the short term, the module may tend to choose an alternative solution that requires less resources and can be implemented quickly. The decision module 105 conducts a resource demand analysis on the preliminary corrective solution to determine the various resources required to implement the solution. This includes requirements for materials, equipment, manpower and time. The module generates a detailed resource demand list and performs scheduling optimization based on the actual situation of existing resources. The specific implementation adopts a linear programming algorithm and uses Python's PuLP library to solve the resource allocation problem, with the goal of minimizing cost or maximizing efficiency. By reasonably allocating currently available resources, the module can ensure maximum resource utilization, reduce waste and avoid construction delays caused by resource shortages. For example, if resources for a certain type of equipment are tight at a construction site, the module may recommend adjusting equipment usage for other low-priority tasks to ensure that critical issues are addressed first.

[0058] The decision module 105 will coordinately optimize the preliminary rectification plan and resource allocation strategy to generate the final target treatment plan. This process is implemented using a genetic algorithm, using Python's DEAP (Distributed Evolutionary Algorithm Framework) library for optimization calculations. This plan not only includes specific rectification measures, but also specifies in detail the allocation and use order of various resources to ensure that the plan is highly operational and efficient during implementation. For example, in the construction of a large complex, the decision module may suggest mobilizing a batch of equipment for emergency foundation reinforcement work, while arranging another group of workers for concrete pouring, to ensure that the various processes can be closely connected without wasting time and resources.

[0059] For example, suppose that during the construction of a high-rise building, the detection module found multiple wall cracks, and the heat map showed that these cracks were mainly concentrated in certain floors. The decision module 105 will first extract and classify the characteristics of these cracks, and may find that some of the cracks are similar to similar cracks caused by sudden temperature changes in history. The module then calls the historical case library and uses the ElasticSearch full-text search engine to quickly match a solution that has successfully repaired a similar problem, that is, adding insulation to the crack area and re-grouting. Then, the decision module 105 evaluates the feasibility of this repair solution under current conditions and generates a list of resource requirements, such as the required insulation materials, grouting equipment, and number of workers. Finally, the decision module 105 integrates all the information and uses the API of the project management software (such as Microsoft Project) to generate a detailed target treatment plan, including a work breakdown structure (WBS) and a Gantt chart, and presents it to the construction team through a 3D visualization interface (using Autodesk Forge Viewer) to ensure that they can execute it quickly and effectively.

[0060] The analysis module 106 is used to perform time series analysis and multivariate correlation analysis on the target processing solution to obtain a dynamic quality assessment report including trend prediction data and parameter impact weights.

[0061] Specifically, the analysis module 106 extracts relevant data from the target treatment plan generated by the decision module 105, including the implementation time of the rectification measures, resource allocation, and historical records of various quality parameters. These data usually show the characteristics of time series, that is, data points that change over time. In order to effectively analyze these time series data, the module first performs a decomposition process in the time dimension, decomposing the data into components such as trend, seasonality, periodicity, and random fluctuations, so as to identify long-term trends and periodic changes in the data. The specific implementation method is to use the seasonal_decompose function in the statsmodels library of Python, which can decompose the time series data into trend, seasonality, and residual components. After the time series is decomposed, the analysis module 106 uses the ARIMA (autoregressive integrated moving average model) algorithm to predict the trend of the time series after removing the seasonal effect. The specific implementation uses the ARIMA class in the statsmodels library to automatically select the optimal p, d, and q parameters through the grid search method. ARIMA is a powerful statistical tool that can predict future trends based on past time series data. The module analyzes historical quality parameter data to identify potential patterns and predict future parameter change trends. For example, when monitoring the time series data of concrete strength, the module may find that the strength increases rapidly in the early stage and then gradually stabilizes. The ARIMA model can use this information to predict the trend of concrete strength changes in the next few days or even weeks, providing a reference for project scheduling.

[0062] While generating trend forecast data, the analysis module 106 also calculates the confidence interval of the forecast data, that is, the reliability range of the forecast results. The confidence interval is calculated using the get_forecast method of the ARIMA model, and the upper and lower bounds of the forecast value are obtained by setting the confidence level (such as 95%). The calculation of the confidence interval takes into account the volatility of historical data and the prediction accuracy of the model, thereby providing managers with a reasonable expected range. For example, the module may predict that the bearing capacity of a key structural component will maintain a steady growth in a certain period of time in the future, but due to the uncertainty of environmental factors, the predicted confidence interval may show a certain range of fluctuations, which prompts managers to consider this uncertainty when making decisions. In addition to time series analysis, the analysis module 106 also performs multivariate correlation analysis to reveal the mutual influence relationship between different quality parameters. The specific implementation uses the corr function of the pandas library to calculate the Pearson correlation coefficient. By constructing the correlation coefficient matrix between parameters, the module can identify which parameters have a greater impact on the overall quality assessment and whether these parameters are positively or negatively correlated. For example, the module may find that there is a significant negative correlation between construction temperature and concrete strength, that is, the higher the temperature, the lower the concrete strength. This correlation analysis helps construction managers optimize construction conditions in actual operations, such as taking necessary cooling measures under high temperature conditions to ensure that the strength of the concrete meets the requirements.

[0063] To further simplify the decision-making process for managers, the analysis module 106 performs principal component analysis (PCA) on these correlation coefficient matrices. The specific implementation uses the PCA class in the scikit-learn library, and the number of principal components retained is controlled by setting the n_components parameter. Principal component analysis is a dimensionality reduction technique that can highlight the factors that have the greatest impact on quality assessment by converting multiple related variables into a few comprehensive indicators (i.e., principal components). This process helps managers focus on the most critical quality parameters and improve management efficiency. For example, through principal component analysis, the module may identify that material quality, construction technology, and ambient temperature are the three most important factors affecting the stability of a certain type of structure, while other minor factors can be temporarily ignored. Finally, the analysis module 106 combines the trend prediction data, confidence intervals, parameter correlations, and the results of principal component analysis to generate a dynamic quality assessment report. The report generation process uses Python's reportlab library to automatically create a PDF document containing charts and text descriptions. This report not only provides a forecast of future quality status, but also reveals the key factors and their weights that affect these forecast results. For example, in the construction of a large bridge project, the analysis module may generate a report predicting the fatigue damage that may occur in certain key components of the bridge in the coming months, and point out that ambient temperature, material fatigue limit and construction load are the main factors affecting these damages. The report will also recommend managers to conduct regular inspections under specific temperature conditions or avoid overloading during construction.

[0064] Therefore, the analysis module 106 converts the data of the target treatment scheme into a dynamic quality assessment report with predictive and guiding significance through time series analysis and multivariate correlation analysis. This report can not only provide real-time quality monitoring and prediction for construction management, but also help managers understand the specific impact of different factors on quality, so as to take more scientific and accurate control measures in construction to ensure the dual guarantee of project quality and progress. The visualization part of the report uses matplotlib and seaborn libraries to create various charts, such as time series prediction charts, correlation heat maps, and principal component contribution rate bar charts, to intuitively display the analysis results.

[0065] Through the collaborative work of the above modules, the intelligent BIM building twin is obtained by BIM modeling of the construction engineering design drawings and building specification data, which provides a unified digital foundation for the entire quality supervision process and effectively solves the problem of information dispersion and difficulty in integration in traditional methods; the on-site data acquisition system is deployed based on the intelligent BIM building twin, which realizes the collaborative work of fixed sensor networks and mobile acquisition equipment, greatly improving the comprehensiveness and real-time nature of data acquisition, so that quality supervision can promptly discover and respond to changes on site; the system can automatically identify quality anomalies and give priority processing levels through the intelligent analysis of the collected data by the anomaly detection module, which greatly reduces the workload of manual inspection. The multi-dimensional visualization processing module converts quality abnormality alarms and priority processing levels into intuitive heat maps and force-directed maps, and displays them in a dynamic three-dimensional visualization interface, so that managers can quickly grasp the project quality status and intuitively understand the correlation between quality problems, so as to make more reasonable decisions; the intelligent decision-making analysis module can automatically generate target processing solutions including corrective measures and resource allocation strategies based on historical experience and current conditions, greatly improving the scientificity and efficiency of decision-making; the time series analysis and multivariate correlation analysis modules conduct in-depth analysis of the target processing solutions and generate dynamic quality assessment reports, which can not only predict the development trend of quality parameters, It can also identify key influencing factors and provide data support for continuous improvement; the entire system realizes full-process automation from data collection, anomaly detection, visualization to intelligent decision-making, greatly reducing manual intervention and improving the efficiency and consistency of quality management; the modular design of the system enables close collaboration between various functional modules to form a closed-loop quality management system that can continuously learn and optimize to adapt to the characteristics and needs of different projects; by analyzing historical project data through deep learning algorithms, the system can continuously accumulate experience, improve the accuracy of quality risk prediction, and provide a powerful tool for proactively preventing quality problems; the system's real-time data collection and processing capabilities enable managers to grasp project quality at any time The system can identify and solve problems in a timely manner, effectively reducing quality risks. The dynamic 3D visualization interface not only provides an intuitive display of quality status, but also supports interactive operations, making it convenient for managers to conduct in-depth analysis and decision-making. The system's adaptive optimization capability enables it to continuously adjust and optimize supervision strategies according to project progress and changes in the external environment, ensuring the continued effectiveness of quality management. Through comprehensive analysis of multi-source heterogeneous data, the system can reveal potential quality problems and influencing factors, providing comprehensive data support for quality improvement. The system's intelligent report generation function can automatically generate evaluation reports containing trend forecasts and parameter influence weights, providing a scientific basis for management decisions and improving the accuracy and reliability of decisions.

[0066] In a specific embodiment, the modeling module 101 is specifically used to

[0067] (1) Digitally convert the architectural engineering design drawings to obtain digital data of the building structure, and then parametrically process the digital data of the building structure to obtain initial BIM data containing geometric information and attribute information;

[0068] (2) Perform quality parameter implantation processing on the initial BIM data to obtain parametric BIM data including material strength, dimensional error and construction process requirements. Perform quality standard association processing on the parametric BIM data based on the pre-collected building specification data to obtain standardized BIM data with quality assessment benchmarks.

[0069] (3) Perform quality control point setting processing on the standardized BIM data to obtain monitoring point layout data including key structural nodes and construction quality sensitive areas, and perform data association processing on the monitoring point layout data to obtain data collection rules corresponding to each quality parameter;

[0070] (4) Through deep learning algorithms, historical project data is analyzed and processed to obtain quality risk prediction data. Based on the quality risk prediction data, risk area annotation is performed on standardized BIM data to obtain an intelligent BIM building twin that contains high-risk area identification.

[0071] Specifically, the architectural engineering design drawings are digitally converted to obtain digital data of the building structure, and the digital data of the building structure is parametrically processed to obtain initial BIM data containing geometric information and attribute information: Modeling module 101 first uses the automation script developed in Python, combined with the COM interface of AutoCAD, to pre-process the design drawings in CAD format (.dwg). This includes deleting useless layers, unifying layer naming, and correcting non-standard elements. Subsequently, the Autodesk Revit API is used for secondary development to create a custom plug-in to achieve automatic conversion from 2D to 3D. The specific steps include: using the ImportCAD method to import the pre-processed CAD file; applying ElementClassFilter and ElementCategoryFilter to filter key components; calling the CreateFamilyInstance method to create a 3D family instance based on 2D geometric information; and setting component parameters (such as height, width, material, etc.) through the ParameterSet method.

[0072] To achieve parametric processing, Modeling Module 101 developed a C#-based Revit plug-in, which includes a custom parameter definer, an associated constraint generator, and a parametric family template generator. This enables the various components and material properties of the building to be converted into adjustable parameters, such as the height of the column, the span of the beam, the thickness of the wall, etc., thereby obtaining initial BIM data containing geometric information and attribute information. The initial BIM data is processed by implanting quality parameters to obtain parametric BIM data containing material strength, dimensional error, and construction process requirements. The parametric BIM data is then processed by associating quality standards based on the pre-collected building specification data to obtain standardized BIM data with a quality assessment benchmark:

[0073] Modeling module 101 developed a SQL Server-based building specification database to store various quality standards and parameters. The data integration program written in C# interacts with the database through ADO.NET to extract relevant specification data. Then, the ExternalCommand interface of the Revit API is used to write these data into the corresponding parameters of the BIM model. For example, in concrete components, the material strength is expressed as the compressive strength value of the concrete (such as the compressive strength of C30 concrete is 30MPa), and the construction process requirements include the temperature control range during concrete pouring. These parameters are associated with the initial BIM data through standardization to form standardized BIM data with a quality assessment benchmark. The standardized BIM data is processed by setting quality control points to obtain the monitoring point layout data including key structural nodes and construction quality sensitive areas, and the monitoring point layout data is processed by data association to obtain the data collection rules corresponding to each quality parameter:

[0074] The modeling module 101 has developed a quality control point identification algorithm based on the Revit API. The algorithm identifies key structural nodes (such as load-bearing walls, beam-column connections, and foundation parts) through topological analysis, uses finite element analysis to determine stress concentration areas, and combines the construction difficulty assessment algorithm to mark complex construction sites. These identified quality control points are set in the standardized BIM data to form monitoring point layout data. Subsequently, the modeling module 101 associates the monitoring point layout data with various quality parameters and defines corresponding data collection rules. For example, for quality control points that require real-time monitoring of concrete strength, the system will set the frequency of data collection, the location of data storage, and the early warning mechanism for abnormal situations. In this way, the monitoring point not only has a physical identification function, but can also actively participate in the entire process of quality monitoring.

[0075] The historical project data is analyzed and processed through deep learning algorithms to obtain quality risk prediction data, and the standardized BIM data is labeled with risk areas based on the quality risk prediction data to obtain an intelligent BIM building twin containing high-risk area identification: Modeling module 101 uses Python's TensorFlow or PyTorch library to develop a deep learning model (such as an LSTM network). By learning from a large amount of historical project data, the model can identify potential quality risks and predict quality problems that may arise in future construction processes. For example, by analyzing a large amount of data from bridge construction projects, the system may find that the foundation of a bridge pier is more likely to have settlement problems under certain soil conditions.

[0076] The trained model is applied to the current project to predict high-risk areas. Then, the prediction results are visualized and annotated in the BIM model through the Revit API. In the standardized BIM data, these high-risk areas will be specially marked, so that construction workers can pay special attention to these parts during construction and take corresponding preventive measures. Finally, the modeling module 101 integrates all information, including geometric information, material properties, quality parameters, control point layout and risk prediction, to generate an intelligent BIM building twin. This twin achieves 3D visualization and real-time data updates through a web application based on the Autodesk Forge platform, and integrates an early warning system based on a rule engine (such as Drools) for quality threshold monitoring and alarms.

[0077] In a specific embodiment, the deployment module 102 is specifically used to:

[0078] (1) Analyze and process the quality control points in the intelligent BIM building twin to obtain the sensor layout strategy, and determine the installation position of the fixed sensors according to the sensor layout strategy to obtain the fixed sensor layout plan;

[0079] (2) Performing network topology design processing on the fixed sensor deployment scheme to obtain a low-power wide-area network structure, and performing communication protocol configuration processing on the low-power wide-area network structure to obtain a fixed sensor data transmission scheme;

[0080] (3) Assign data collection tasks to mobile collection devices to obtain mobile collection path planning, and perform data fusion processing on fixed sensor data transmission schemes and mobile collection path planning to obtain a comprehensive data collection strategy;

[0081] (4) Design the data storage structure for the comprehensive data collection strategy to obtain the real-time database architecture, and define the data interface of the real-time database architecture to obtain a real-time data collection system that includes a fixed sensor network and mobile collection devices.

[0082] Specifically, the quality control points in the intelligent BIM building twin are analyzed and processed to obtain the sensor layout strategy, and the fixed sensor installation position is determined according to the sensor layout strategy to obtain the fixed sensor layout plan: The deployment module 102 first uses the data analysis script developed in Python, which uses the pandas library to process the quality control point data exported by the BIM model. The key monitoring points are identified by the K-means clustering algorithm, and the appropriate sensor type is selected based on the preset decision tree algorithm. For example, for the concrete pouring process that requires precise control of temperature and humidity, the DHT22 digital temperature and humidity sensor is selected. The sensor layout strategy is implemented by a genetic algorithm optimization program, which considers factors such as sensor coverage, signal strength, and installation difficulty to calculate the best sensor layout plan. For example, BX120-50AA resistance strain gauges are installed at key structural nodes of high-rise buildings, such as columns, beams, and floor slabs, to monitor the stress state and deformation of these components in real time.

[0083] The fixed sensor deployment scheme is processed by network topology design to obtain a low-power wide-area network structure, and the low-power wide-area network structure is processed by communication protocol configuration to obtain a fixed sensor data transmission scheme: Deployment module 102 uses LoRaWAN technology to build a low-power wide-area network, and uses LoRa gateways (such as Multitech Conduit) and LoRa node modules (such as RFM95W) to build a wireless sensor network covering the entire construction site. The network topology design takes into account the number, location and data transmission bandwidth requirements of sensor nodes to ensure network reliability and scalability.

[0084] The communication protocol configuration uses the MQTT protocol, which is lightweight and efficient, and is suitable for resource-constrained sensor nodes. During the configuration process, data priority, transmission frequency, and encryption method are set to ensure secure and stable data transmission in complex environments. Data collection tasks are assigned to mobile collection devices to obtain mobile collection path planning, and data fusion processing is performed on fixed sensor data transmission schemes and mobile collection path planning to obtain a comprehensive data collection strategy:

[0085] Deployment module 102 develops a mobile device control program based on ROS (Robot Operating System) to automatically plan the mobile path and data collection tasks of the equipment. For example, a drone equipped with a high-precision laser scanner (such as Velodyne VLP-16) is used to regularly measure the excavation area and monitor surface deformation. Mobile collection path planning uses the A* algorithm combined with a dynamic obstacle avoidance strategy to plan the optimal path in real time. Data fusion processing uses time synchronization and spatial registration technology to organically combine the continuous monitoring data of fixed sensors with the point data of mobile devices to form a comprehensive data collection strategy.

[0086] The data storage structure design is performed on the comprehensive data collection strategy to obtain the real-time database architecture, and the data interface definition is performed on the real-time database architecture to obtain a real-time data collection system including a fixed sensor network and a mobile collection device: the deployment module 102 adopts a distributed message queue architecture based on Apache Kafka to ensure low-latency and highly reliable data transmission. The data storage uses a distributed storage system based on Apache Hadoop, and the real-time data processing uses the Apache Spark framework.

[0087] The data interface definition adopts RESTful API and is developed using the Spring Boot framework to ensure seamless integration of the database with other system components. The resulting real-time data collection system can integrate data from fixed sensor networks and mobile collection devices to provide all-round quality monitoring and data support for the construction process.

[0088] For example, in a large bridge construction project, the deployment module 102 arranges fiber Bragg grating strain sensors (such as os3600) at key stress points such as bridge piers and bridge towers, and arranges drones equipped with laser scanners (such as DJI Matrice 300RTK) to regularly monitor the progress of bridge deck construction. These data are transmitted to the central monitoring platform (based on Siemens MindSphere IoT operating system) via the OPC UA protocol, allowing construction managers to view the stress status and construction progress of each part of the bridge in real time, and promptly discover and deal with potential quality problems.

[0089] In a specific embodiment, the detection module 103 is specifically used to:

[0090] (1) Preprocess the data collected by the real-time data acquisition system to obtain cleaned and standardized data, and perform time series decomposition on the standardized data to obtain trend, seasonality, and random component data;

[0091] (2) Perform statistical feature extraction on trend, seasonality and random component data to obtain data feature vectors;

[0092] (3) Perform cluster analysis on the data feature vectors through clustering algorithms to obtain the classification results of the degree of data anomaly;

[0093] (4) Perform threshold judgment processing on the data anomaly classification results to obtain preliminary quality anomaly alerts;

[0094] (5) Perform contextual analysis on the preliminary quality abnormality alert to obtain the associated influencing factor data, and perform weight calculation on the associated influencing factor data to obtain the abnormality severity score;

[0095] (6) Prioritize the quality anomaly alarms according to the anomaly severity scores to obtain quality anomaly alarms and priority processing levels.

[0096] It should be noted that the data collected by the real-time data collection system is preprocessed to obtain cleaned standardized data, and the standardized data is decomposed into time series to obtain trend, seasonal and random component data: the detection module 103 uses Python's Pandas library for data cleaning, uses the moving median method to remove outliers, and uses the KNN (K-Nearest Neighbors) algorithm to interpolate missing data. For example, a threshold of ±3σ (standard deviation) is set for temperature data to identify outliers, and data points outside the range are removed or replaced with the average of neighboring valid values.

[0097] Standardization is implemented using the StandardScaler class of scikit-learn, which converts all features to a standard normal distribution with a mean of 0 and a variance of 1. Time series decomposition uses the seasonal_decompose function in the statsmodels library, using an additive model or a multiplicative model for decomposition. Statistical feature extraction is performed on trend, seasonal, and random component data to obtain data feature vectors:

[0098] The detection module 103 uses the NumPy library to calculate statistics such as mean, variance, skewness, kurtosis, etc. to form a feature vector. For example, for concrete strength data, extract its overall level (mean) and fluctuation range (variance) and other features. The data feature vector is clustered and analyzed by a clustering algorithm to obtain the data abnormality classification result:

[0099] The K-means clustering algorithm in the scikit-learn library is used in combination with the silhouette coefficient to automatically determine the optimal number of clusters. For example, the structural stress data of high-rise buildings are clustered to identify abnormal stress patterns. The data abnormality classification results are processed by threshold judgment to obtain preliminary quality abnormality alerts:

[0100] Use the NumPy library to perform matrix operations and quickly compare data with preset thresholds. For example, a column stress exceeding the design strength by 20% is considered abnormal. Perform contextual analysis on the initial quality abnormality alert to obtain the associated influencing factor data, and perform weight calculation on the associated influencing factor data to obtain the abnormality severity score:

[0101] Granger causality test is used to analyze the causal relationship between different time series. The weight calculation adopts the multivariate linear regression model and is implemented using the statsmodels library. For example, the weight of the influence of temperature and humidity on the abnormal strength of concrete is analyzed. The quality abnormality alarms are prioritized according to the abnormality severity score to obtain the quality abnormality alarms and priority processing levels:

[0102] A weighted scoring system is implemented using a custom function in Python. It takes into account factors such as the impact scope of the anomaly, potential consequences, and urgency of resolution, and calculates and sorts the comprehensive scores.

[0103] For example, in a subway tunnel construction project, the detection module 103 uses a HOBO U20L-04 water level recorder to monitor groundwater level changes. After data preprocessing, the Prophet library is used to perform time series prediction to determine whether the water level change is abnormal. If it is determined to be abnormal, the module analyzes relevant factors (such as rainfall data and drainage system sensor data obtained through the meteorological API). Finally, a rule-based expert system is used to make decisions, prioritize water level anomalies, and trigger immediate alarms through the MQTT protocol.

[0104] In a specific embodiment, the processing module 104 is specifically configured to:

[0105] (1) Perform 3D scene rendering processing on the intelligent BIM building twin to obtain a basic 3D visualization interface, and perform spatial positioning processing on quality abnormality alarms to obtain abnormal point location coordinate data;

[0106] (2) Marking the basic 3D visualization interface according to the coordinate data of the abnormal point position to obtain the abnormal point identification 3D interface, and color mapping the priority processing level to obtain the abnormal severity color scheme;

[0107] (3) Perform heat map processing on the abnormal point identification three-dimensional interface according to the color scheme of the abnormality severity, obtain the heat map superimposed three-dimensional interface, and perform correlation analysis on the quality abnormality alarm to obtain the relationship data between the abnormal points;

[0108] (4) Performing force-directed graph drawing processing on the heat map superimposed three-dimensional interface based on the relationship data between the abnormal points to obtain a force-directed graph superimposed three-dimensional interface;

[0109] (5) Perform interactive function design on the force-directed graph superimposed three-dimensional interface to obtain a dynamic three-dimensional visualization interface that includes a heat map and a force-directed graph.

[0110] Specifically, the intelligent BIM building twin is processed by 3D scene rendering to obtain a basic 3D visualization interface, and the quality abnormality alarm is processed by spatial positioning to obtain the coordinate data of the abnormal point position: the processing module 104 uses the Autodesk Forge API to render the 3D model and uses the WebGL technology to achieve high-performance 3D graphics display in the browser. At the same time, a custom plug-in based on Three.js is developed to convert 2D coordinates into precise positions in 3D space to achieve spatial positioning of abnormal points.

[0111] The basic three-dimensional visualization interface is marked according to the coordinate data of the abnormal point position to obtain the abnormal point identification three-dimensional interface, and the priority level is color mapped to obtain the abnormal severity color scheme: the processing module 104 uses the material system of Three.js to set the component color and mark the abnormal point. The color mapping uses the color scale function of the D3.js library to create a continuous color mapping from green to red to represent different priority levels.

[0112] According to the color scheme of abnormal severity, the three-dimensional interface of abnormal point identification is processed by heat map drawing to obtain the three-dimensional interface of heat map superposition, and the quality abnormality alarm is analyzed for correlation to obtain the relationship data between abnormal points: the heat map is realized by using WebGL shader technology, and the color intensity gradient effect is realized by custom fragment shader program. The correlation analysis uses a graph database (such as Neo4j) to store the relationship between abnormal points, and uses Cypher query language for complex relationship analysis.

[0113] According to the relationship data between the abnormal points, the heat map superimposed 3D interface is processed by force-directed graph drawing to obtain the force-directed graph superimposed 3D interface: The force-directed graph is implemented using the force-directed layout algorithm of the D3.js library, combined with WebGL technology for efficient rendering. A custom WebGL shader program is used to draw the connection lines, and the line thickness and color represent the strength of the association.

[0114] The interactive function design of the force-directed graph superimposed 3D interface is processed to obtain a dynamic 3D visualization interface including heat map and force-directed graph: the interactive function is realized by developing an orbit controller (OrbitControls) based on Three.js and a custom click event handler. The user interface is built using the React framework to realize interactive operations such as zooming, rotating, and clicking to view details.

[0115] For example, in a large office building project, processing module 104 first uses the Autodesk Forge API to render a 3D model of the entire building. For the substandard concrete strength points detected on the 15th floor, the system uses Three.js to add color marks at the corresponding locations: 3 seriously deviated points are marked in red, and 5 slightly deviated points are marked in yellow. A heat map is generated on the 15th floor using WebGL shader technology to show the problem concentration area. Correlation analysis uses Neo4j to store and analyze the temporal continuity between abnormal points, and finds that they may be related to the same batch of concrete. A force-directed graph is drawn using D3.js to connect related abnormal points, and the thickness of the line indicates the degree of correlation. Finally, the interactive interface built with React allows managers to zoom, rotate, click on abnormal points to view details, and call historical data for in-depth analysis through real-time connection with backend databases (such as MongoDB).

[0116] In a specific embodiment, the decision module 105 is specifically configured to:

[0117] (1) Perform feature extraction on the quality abnormality information in the dynamic three-dimensional visualization interface to obtain an abnormal feature data set, and perform data normalization on the abnormal feature data set to obtain standardized abnormal feature data;

[0118] (2) The standardized abnormal feature data is classified and processed through the decision tree algorithm to obtain the abnormal type identification result;

[0119] (3) Based on the abnormal type identification results, similarity matching is performed on the historical rectification case library to obtain a set of candidate rectification measures. The feasibility of the candidate rectification measures is evaluated to obtain a preliminary rectification plan;

[0120] (4) Conduct resource demand analysis on the preliminary rectification plan to obtain a resource demand list, and optimize the scheduling of existing resources based on the resource demand list to obtain a resource allocation strategy;

[0121] (5) The preliminary rectification plan and resource allocation strategy are collaboratively optimized to obtain a target treatment plan that includes rectification measures and resource allocation strategy.

[0122] Specifically, feature extraction is performed on the quality abnormality information in the dynamic three-dimensional visualization interface to obtain an abnormal feature data set, and data normalization is performed on the abnormal feature data set to obtain standardized abnormal feature data: the decision module 105 uses the PCA (principal component analysis) algorithm in the scikit-learn library of Python to extract features, and combines a custom feature engineering function to capture domain-specific information. Data normalization is implemented using the StandardScaler class of scikit-learn to convert all features to a standard normal distribution with a mean of 0 and a variance of 1.

[0123] The standardized abnormal feature data is classified and processed through the decision tree algorithm to obtain the abnormal type identification result: the DecisionTreeClassifier class in the scikit-learn library is used to implement the decision tree algorithm, and the model complexity is controlled by setting parameters such as max_depth (maximum depth of the tree) and min_samples_split (minimum number of samples required for internal node re-division) to prevent overfitting.

[0124] According to the results of abnormal type identification, the historical rectification case library is matched by similarity to obtain a set of candidate rectification measures. The candidate rectification measure set is evaluated for feasibility to obtain a preliminary rectification plan: The historical rectification case library is implemented using MongoDB, which supports efficient document storage and retrieval. Similarity matching uses the cosine similarity algorithm (implemented by the cosine function of the scipy library). The feasibility assessment uses the multi-criteria decision analysis (MCDA) method, and uses Python's AHP (analytic hierarchy process) library to calculate the comprehensive score of each plan. The preliminary rectification plan is analyzed for resource requirements to obtain a resource requirement list, and the existing resources are optimized for scheduling based on the resource requirement list to obtain a resource allocation strategy: The resource requirement analysis uses Python's PuLP library to implement linear programming to solve the resource allocation problem, with the goal of minimizing cost or maximizing efficiency. Resource scheduling optimization uses a genetic algorithm, and uses Python's DEAP (distributed evolutionary algorithm framework) library for optimization calculations.

[0125] The preliminary rectification plan and resource allocation strategy are collaboratively optimized to obtain a target treatment plan that includes rectification measures and resource allocation strategies: Collaborative optimization uses a multi-objective optimization algorithm and is implemented using Python's DEAP library. It simultaneously considers multiple objectives such as the degree of quality improvement, resource utilization efficiency, and the impact on construction period to generate a Pareto optimal solution set.

[0126] For example, in a large residential construction project, the decision module 105 extracts abnormal information of the exterior wall insulation layer from the dynamic three-dimensional visualization interface, including spatial location, severity, duration, etc. After dimensionality reduction using the PCA algorithm, the abnormalities are classified into "material performance not meeting the standard" and "construction process problems" through the decision tree. The system uses the ElasticSearch full-text search engine to quickly match historical cases and proposes two measures: replacing insulation materials and improving construction processes. Resource demand analysis uses the PuLP library for linear programming, considering constraints such as manpower, materials, and equipment. Finally, the multi-objective optimization algorithm implemented by the DEAP library generates an overall plan including local repairs, process improvements, and full-staff training. Use the API of project management software (such as Microsoft Project) to generate a detailed implementation plan, including a work breakdown structure (WBS) and a Gantt chart. The entire plan is visualized on a three-dimensional interface through Autodesk Forge Viewer, which is easy for managers to understand and execute.

[0127] In a specific embodiment, the analysis module 106 is specifically configured to:

[0128] (1) Decompose the corrective measures and resource allocation strategies in the target treatment plan in the time dimension to obtain a time series data set, and perform seasonal adjustment on the time series data set to obtain deseasonalized time series data;

[0129] (2) The ARIMA algorithm is used to perform trend forecasting on the deseasonalized time series data to obtain quality parameter trend forecast data. The confidence interval of the quality parameter trend forecast data is calculated to obtain trend forecast results with uncertainty ranges.

[0130] (3) Conduct correlation analysis on various measures and strategies in the target treatment plan to obtain a correlation coefficient matrix between parameters;

[0131] (4) The correlation coefficient matrix between parameters is reduced in dimension through principal component analysis to obtain a set of key influencing factors;

[0132] (5) Sort the key influencing factors by importance to obtain parameter influence weight data;

[0133] (6) Comprehensively analyze and process the trend prediction results with uncertainty ranges and parameter impact weight data to obtain a dynamic quality assessment report containing trend prediction data and parameter impact weights.

[0134] Specifically, the time dimension decomposition processing is performed on the corrective measures and resource allocation strategies in the target treatment plan to obtain a time series data set, and the time series data set is seasonally adjusted to obtain deseasonalized time series data: the analysis module 106 uses the Python pandas library to decompose the time dimension and arrange the corrective measures and resource allocation strategies by time point. The seasonal adjustment uses the seasonal_decompose function in the statsmodels library, and uses an additive model or a multiplicative model for decomposition to remove cyclical fluctuations.

[0135] The ARIMA algorithm is used to process the trend forecast of the deseasonalized time series data, and the quality parameter trend forecast data is obtained. The confidence interval calculation is performed on the quality parameter trend forecast data to obtain the trend forecast result with uncertainty range: the ARIMA class in the statsmodels library is used to implement trend forecasting, and the grid search method is used to automatically select the optimal p, d, and q parameters. The confidence interval calculation uses the get_forecast method of the ARIMA model, and the 95% confidence level is set to obtain the upper and lower bounds of the forecast value.

[0136] And conduct correlation analysis on various measures and strategies in the target treatment plan to obtain the correlation coefficient matrix between parameters: use the corr function of the pandas library to calculate the Pearson correlation coefficient and obtain the correlation coefficient matrix between parameters.

[0137] The correlation coefficient matrix between parameters is reduced in dimension through principal component analysis to obtain a set of key influencing factors: the PCA class in the scikit-learn library is used for principal component analysis, and the number of principal components retained is controlled by setting the n_components parameter.

[0138] The key influencing factors are ranked by importance to obtain parameter influence weight data: the ranking is based on the variance contribution rate of the principal component or the factor loading, and the calculation and ranking are realized using the NumPy library.

[0139] Comprehensively analyze and process the trend forecast results with uncertainty range and parameter impact weight data to obtain a dynamic quality assessment report containing trend forecast data and parameter impact weights: Use Python's reportlab library to automatically create a PDF document containing charts and text descriptions. The visualization part uses matplotlib and seaborn libraries to create time series forecast graphs, correlation heat maps, and principal component contribution rate bar graphs.

[0140] For example, in a large commercial complex construction project, the analysis module 106 processes measures such as improving the concrete pouring process and strengthening the quality control of steel bar binding. The pandas library is used to decompose these measures according to the implementation time point to obtain the time series data of the expected value of daily quality parameters and resource input. After seasonal adjustment, the ARIMA (2,1,1) model is used to predict the development trend of concrete strength in the next 30 days, and the 95% confidence interval is calculated to be ±2MPa. The correlation analysis uses the corr function of pandas and finds that the concrete strength is strongly positively correlated with the curing time and ambient temperature, and weakly negatively correlated with the pouring speed. The PCA class of scikit-learn is used for dimensionality reduction to extract the three main influencing factors of curing conditions, raw material quality and construction technology. The NumPy library is used to calculate and sort the influence weights, and the curing condition weight is 0.45, the raw material quality weight is 0.35, and the construction technology weight is 0.20. Finally, the reportlab library is used to generate a dynamic quality assessment report, which includes the trend prediction chart drawn by matplotlib and the correlation heat map generated by seaborn. The report recommends focusing on maintenance condition control and conducting a comprehensive quality inspection after the 15th day to provide the project management team with precise quality management strategy guidance.

[0141] The above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. Although the present invention has been described in detail with reference to the embodiments, a person skilled in the art should understand that the specific implementation modes of the present invention can still be modified or replaced by equivalents, and any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A BIM-based intelligent supervision system for building construction quality information, characterized in that: include: The modeling module is used to perform BIM modeling on the architectural engineering design drawings and pre-collected building specification data to obtain an intelligent BIM building twin that includes quality parameters and quality control points; A deployment module, used to deploy a field data acquisition system for the intelligent BIM building twin to obtain a real-time data acquisition system including a fixed sensor network and mobile acquisition devices; A detection module, used to perform abnormality detection processing on the data collected by the real-time data acquisition system to obtain quality abnormality alarms and priority processing levels; A processing module, used for performing multi-dimensional visualization processing on the quality abnormality alarm and the priority processing level to obtain a dynamic three-dimensional visualization interface including a heat map and a force-directed map; A decision-making module, used to perform intelligent decision-making analysis on the quality abnormality information in the dynamic three-dimensional visualization interface to obtain a target processing solution including corrective measures and resource allocation strategies; The analysis module is used to perform time series analysis and multivariate correlation analysis on the target treatment scheme to obtain a dynamic quality assessment report containing trend prediction data and parameter impact weights.

2. The BIM-based intelligent construction quality information supervision system according to claim 1 is characterized in that: The modeling module is specifically used for: Performing digital conversion processing on the architectural engineering design drawings to obtain building structure digital data, and performing parameterization processing on the building structure digital data to obtain initial BIM data containing geometric information and attribute information; Performing quality parameter implantation processing on the initial BIM data to obtain parametric BIM data including material strength, dimensional error and construction process requirements, and performing quality standard association processing on the parametric BIM data according to the pre-collected building specification data to obtain standardized BIM data with a quality assessment benchmark; Performing quality control point setting processing on the standardized BIM data to obtain monitoring point layout data including key structural nodes and construction quality sensitive areas, and performing data association processing on the monitoring point layout data to obtain data collection rules corresponding to each quality parameter; The historical project data is analyzed and processed by a deep learning algorithm to obtain quality risk prediction data, and the standardized BIM data is labeled with risk areas based on the quality risk prediction data to obtain an intelligent BIM building twin that contains high-risk area identification.

3. The BIM-based intelligent construction quality information supervision system according to claim 1 is characterized in that: The deployment module is specifically used for: Analyze and process the quality control points in the intelligent BIM building twin to obtain a sensor layout strategy, and determine the installation positions of fixed sensors according to the sensor layout strategy to obtain a fixed sensor layout plan; Performing network topology design processing on the fixed sensor arrangement scheme to obtain a low-power wide-area network structure, and performing communication protocol configuration processing on the low-power wide-area network structure to obtain a fixed sensor data transmission scheme; Performing data collection task allocation processing on the mobile collection device to obtain the mobile collection path planning, and performing data fusion processing on the fixed sensor data transmission scheme and the mobile collection path planning to obtain a comprehensive data collection strategy; The comprehensive data collection strategy is processed by data storage structure design to obtain a real-time database architecture, and the real-time database architecture is processed by data interface definition to obtain a real-time data collection system including a fixed sensor network and a mobile collection device.

4. The BIM-based intelligent construction quality information supervision system according to claim 1 is characterized in that: The detection module is specifically used for: Preprocessing the data collected by the real-time data acquisition system to obtain cleaned standardized data, and performing time series decomposition processing on the standardized data to obtain trend, seasonality and random component data; Performing statistical feature extraction processing on the trend, seasonality and random component data to obtain a data feature vector; Performing cluster analysis on the data feature vectors by using a clustering algorithm to obtain a classification result of the degree of data anomaly; Performing threshold judgment processing on the data abnormality degree classification results to obtain a preliminary quality abnormality alarm; Performing context association analysis on the preliminary quality abnormality alarm to obtain associated influencing factor data, and performing weight calculation on the associated influencing factor data to obtain an abnormality severity score; The quality abnormality alarms are prioritized according to the abnormality severity scores to obtain quality abnormality alarms and priority processing levels.

5. The BIM-based intelligent construction quality information supervision system according to claim 1 is characterized in that: The processing module is specifically used for: Performing three-dimensional scene rendering processing on the intelligent BIM building twin to obtain a basic three-dimensional visualization interface, and performing spatial positioning processing on the quality abnormality alarm to obtain abnormal point position coordinate data; Marking the basic three-dimensional visualization interface according to the abnormal point position coordinate data to obtain an abnormal point identification three-dimensional interface, and color mapping the priority processing level to obtain an abnormal severity color scheme; Performing heat map processing on the abnormal point identification three-dimensional interface according to the abnormal severity color scheme to obtain a heat map superimposed three-dimensional interface, and performing correlation analysis processing on the quality abnormality alarm to obtain relationship data between abnormal points; Performing force-directed graph drawing processing on the heat map superimposed three-dimensional interface according to the relationship data between the abnormal points to obtain a force-directed graph superimposed three-dimensional interface; The force-directed graph and the three-dimensional interface are superimposed and interactively designed to obtain a dynamic three-dimensional visualization interface including a heat map and a force-directed graph.

6. The BIM-based intelligent construction quality information supervision system according to claim 1 is characterized in that: The decision module is specifically used for: Performing feature extraction processing on the quality abnormality information in the dynamic three-dimensional visualization interface to obtain an abnormal feature data set, and performing data normalization processing on the abnormal feature data set to obtain standardized abnormal feature data; Classify the standardized abnormal feature data using a decision tree algorithm to obtain an abnormal type identification result; Performing similarity matching processing on the historical rectification case library according to the abnormal type identification result to obtain a candidate rectification measure set, performing feasibility assessment processing on the candidate rectification measure set to obtain a preliminary rectification plan; Performing resource demand analysis on the preliminary rectification plan to obtain a resource demand list, and performing scheduling optimization on existing resources according to the resource demand list to obtain a resource allocation strategy; The preliminary rectification plan and the resource allocation strategy are collaboratively optimized to obtain a target processing plan including rectification measures and resource allocation strategy.

7. The BIM-based intelligent construction quality information supervision system according to claim 6 is characterized in that: The analysis module is specifically used for: Performing time dimension decomposition processing on the rectification measures and resource allocation strategies in the target processing plan to obtain a time series data set, and performing seasonal adjustment processing on the time series data set to obtain deseasonalized time series data; Performing trend forecasting processing on the de-seasonalized time series data by using the ARIMA algorithm to obtain quality parameter trend forecasting data, and performing confidence interval calculation processing on the quality parameter trend forecasting data to obtain a trend forecasting result with an uncertainty range; And carry out correlation analysis on various measures and strategies in the target treatment plan to obtain a correlation coefficient matrix between parameters; Performing dimensionality reduction processing on the correlation coefficient matrix between the parameters through principal component analysis to obtain a set of key influencing factors; Sorting the key influencing factors according to their importance to obtain parameter influencing weight data; The trend prediction results with uncertainty ranges and parameter impact weight data are comprehensively analyzed and processed to obtain a dynamic quality assessment report including trend prediction data and parameter impact weights.

Images