A method and system for engineering quality detection based on green building design

Abstract

The application discloses a kind of engineering quality detection method and system based on green building design, it is related to engineering quality detection technical field.This kind of engineering quality detection method and system based on green building design, including S1, to BIM structured data, real-time monitoring data, structure acceptance data and construction event and window data are pretreated;S2, based on BIM structured data is combined with BIM model to determine the qualified state of component and the quality grade of partition;S3, identify working condition switching anchor point and event, and segment division is carried out to working condition process;S4, characteristic clustering and performance state determination are carried out, and different working condition types are identified and labeled management is carried out;S5, through multidimensional label attribution and traceability determination, identify dominant factor and attribution assignment;S6, through multi-source time sequence item-by-item difference analysis, grade determination and hierarchical response are carried out.Solve the construction progress, BIM node and energy consumption environment data space-time mismatch, data asynchronous cause state interval disconnection, it is difficult to accurately attribute and stable diagnosis problem.

Description

Technical Field

[0001] This invention relates to the field of engineering quality testing technology, specifically to an engineering quality testing method and system based on green building design. Background Technology

[0002] In recent years, with the continuous advancement of green building and intelligent construction concepts, the engineering field has increasingly higher requirements for project quality, energy consumption, and environmental performance management. The construction industry is gradually shifting from traditional manual inspections and item-by-item testing to data-driven, information-based, and automated monitoring. Various environmental monitoring instruments, energy consumption data acquisition devices, and information management platforms are now widely used in the market to achieve real-time monitoring of building energy consumption, air quality, and the status of key processes. At the same time, relevant standards and specifications are driving the continuous improvement of engineering quality evaluation systems, making project management more scientific, systematic, and efficient throughout the entire process. Engineering managers and related technical personnel are increasingly emphasizing dynamic evaluation and risk control supported by multi-source data to ensure high-quality delivery and sustainable operation of construction projects.

[0003] For example, invention patent CN120031266A discloses a method and system for detecting the environmental quality of building projects, including an impact range module, an interfering population module, a health interference module, a living disturbance module, and an environmental quality module. This method first detects real-time environmental data of the building project, determines the environmental impact range of the building project based on the real-time environmental data, and identifies the interfering population by combining the environmental disturbance time and range; further, it obtains group information of the interfering population, and assesses the degree of health disturbance and living disturbance caused by the building project environment to the interfering population based on the real-time environmental data; finally, it comprehensively assesses the environmental quality of the building project by integrating the degree of health disturbance and living disturbance. This system realizes real-time dynamic assessment of the environmental impact of building projects, providing data support for environmental management and decision-making.

[0004] For example, invention patent CN115775116A discloses a BIM-based road and bridge engineering management method and system. The method includes: acquiring a BIM model of the road and bridge project and multiple construction site data; decomposing the BIM model based on the site data; extracting feature vectors from multiple decomposed BIM models using a convolutional neural network model as a filter; calculating the transition matrix between adjacent feature vectors to obtain multiple transition matrices; and calculating the global mean of each transition matrix to form a classification feature vector. Finally, a classifier processes the classification feature vectors and outputs a classification result indicating whether the road and bridge project construction progress is reasonable. The system structure includes a BIM model generation unit, a site data acquisition unit, a BIM model decomposition unit, a feature filtering unit, a transition unit, a mean-based unit, and a verification result generation unit, realizing intelligent evaluation and management of road and bridge project construction progress.

[0005] While existing building engineering management and environmental quality monitoring methods can collect and classify various types of data, in actual projects, there are often temporal and spatial mismatches and asynchronicities between construction progress, BIM nodes, and energy consumption and environmental data. This makes it difficult to accurately connect the working condition intervals, affecting the dynamic collaborative analysis of complex working conditions. Currently, the attribution and risk classification of abnormal working conditions mostly rely on single indicators or static judgments, lacking a multi-dimensional and systematic attribution mechanism and a closed-loop management throughout the entire process. This makes it difficult to achieve stable diagnosis and accurate response to engineering quality risks and environmental changes. The existing system has not yet established an intelligent decision-making framework that can fully link BIM structure, real-time monitoring, and event process data, making it difficult to meet the refined needs of high-standard projects such as green buildings in terms of quality, safety, and performance control.

[0006] Therefore, in order to address the above issues, there is an urgent need for an engineering quality testing method and system based on green building design. Summary of the Invention

[0007] Technical problems to be solved

[0008] To address the shortcomings of existing technologies, this invention provides an engineering quality inspection method and system based on green building design, which solves the problems of spatiotemporal mismatch between construction progress, BIM nodes and energy consumption and environmental data, and the disconnect between state intervals caused by asynchronous data, making it difficult to accurately attribute causes and make stable diagnosis.

[0009] Technical solution

[0010] To achieve the above objectives, this invention provides the following technical solution: an engineering quality inspection method based on green building design, comprising: S1, collecting BIM structured data, real-time monitoring data, and structural acceptance data to obtain construction events and window data; preprocessing the BIM structured data, real-time monitoring data, structural acceptance data, and construction events and window data; S2, determining the qualification status of components and the quality level of zones based on BIM structured data and BIM model; S3, identifying working condition switching anchor points and events in real-time monitoring data using a mutation detection algorithm and feature time-series clustering method, and segmenting the working condition process; S4, performing feature clustering and performance status discrimination on segmented working condition data and structural acceptance data, identifying different working condition types and managing them with labels; S5, for working condition anomaly types, identifying the dominant factors causing engineering quality fluctuations and assigning values ​​through multi-dimensional label attribution and source tracing discrimination; S6, determining the level and classifying responses based on BIM structured data, real-time monitoring data, and attribution value data through multi-source time-series item-by-item difference analysis.

[0011] Furthermore, BIM structured data, real-time monitoring data, and structural acceptance data are collected to obtain construction event and window data. The specific process for preprocessing the BIM structured data, real-time monitoring data, structural acceptance data, and construction event and window data is as follows: Collect BIM structured data, including: process node parameters, component attribute parameters, green performance design parameters, and spatial geometry and connection parameters; collect real-time monitoring data, including: electricity consumption data, indoor CO2 concentration, PM2.5 concentration, temperature and humidity, formaldehyde content, VOC content, and renewable energy utilization rate; collect structural acceptance data, including: heat transfer coefficient of exterior walls and roofs, thickness of door and window insulation layers, leakage test data of roof and exterior wall waterproofing layers, and concrete strength grade data; obtain construction event and window data, and establish a construction time database, including material arrival events, rework events, abnormal alarm events, and node changes. The system utilizes sliding windows for event data, historical time-series data, operating conditions, segmented sliding windows, and engineering unit windows. It employs a partition mapping relationship table, binding topology maps to equipment partitions, and a Hungarian allocation algorithm using the shortest path to match spatial partition mapping, unique process node numbers, and component numbers. Through multi-frequency timestamp resampling and linear interpolation, it standardizes and annotates the timestamps of real-time monitoring data, structural acceptance data, BIM model output data, and construction events, mapping different sampling frequencies and asynchronously reported data streams to a unified standard time-series axis. It performs multi-parameter correlation analysis and outlier identification on real-time monitoring data using MAD rules combined with multivariate covariance analysis and pre-feature clustering algorithms. Finally, it standardizes and normalizes BIM structured data, real-time monitoring data, construction events, and window data using distribution standardization and linear normalization algorithms, employing z-score standardization based on the training set mean and variance, combined with linear normalization.

[0012] Furthermore, the specific process for determining the qualification status of components and the quality level of zones based on BIM structured data and BIM model is as follows: Process node parameters, component attribute parameters, green performance design parameters, and spatial geometry and connection parameters are input into the BIM model. Based on component codes, the structured data is spatially aggregated and associated to form a data sample set with both structural and temporal characteristics. Using each component as a sample unit, a training sample set is constructed by combining historical time-series data with a sliding window. Quality indicators are modeled and learned. An integrated discriminant analysis-supervised learning algorithm is used to determine the qualification status of components, capturing the spatial correlation, temporal dependence, and multivariate dynamic relationships of the input parameters. A Bayesian optimization algorithm is used to adaptively adjust the model hyperparameters, optimizing the decision boundary and feature weights. After model training, based on the design and detection values ​​of the input samples, the following data is output: process node code values, zone quality grading values, energy consumption limits, air quality limits, importance level, three-dimensional coordinates of the geometric center, shortest path distance, structure number, process status, zone location, design parameters, and risk attribute data.

[0013] Furthermore, the specific process of applying the mutation detection algorithm and feature time-series clustering method to real-time monitoring data is as follows: Obtain the process node code values, construction event data, working condition sliding window, energy consumption data, indoor CO2 concentration, PM2.5 concentration, temperature and humidity, and zoning quality grading values; obtain the node code mean value using the working condition moving average algorithm; obtain the construction event signal value using the numerical mapping algorithm; obtain the energy consumption data mean value using the mean algorithm within the working condition sliding window, and calculate the energy consumption index mutation amount by calculating the difference between the energy consumption data and the mean energy consumption data; and obtain the energy consumption index mutation amount by analyzing the indoor CO2 concentration, PM2.5 concentration, and temperature and humidity... The abrupt changes of various environmental parameters are obtained through the window average difference method. The abrupt changes of various environmental parameters are then weighted and fused to obtain the abrupt changes of environmental green parameters. The difference between the process node code and the mean of the process node codes is calculated to obtain the process node change. The construction event signal value is extracted as the construction event impact value. The abrupt changes of energy consumption index and the abrupt changes of environmental green parameters are added together and then multiplied by the zoning quality classification value to obtain the energy consumption environmental risk weighted value. The process node change, the construction event impact value and the energy consumption environmental risk weighted value are added together and the absolute value is calculated as the segment discrimination value. Within the entire calculation window, the working condition sliding window discrimination value at each time point is traversed, and the maximum value is taken to obtain the dynamic working condition segment discrimination value.

[0014] Furthermore, the specific process of identifying operating condition switching anchor points and events, and segmenting the operating condition process, is as follows: Real-time comparison of dynamic operating condition segment discrimination values ​​and dynamic operating condition segment discrimination thresholds. The dynamic operating condition segment discrimination values ​​include a primary discrimination threshold and a secondary discrimination threshold. When the dynamic operating condition segment discrimination value is less than the secondary discrimination threshold, real-time monitoring data is created and incorporated into historical operating condition data without physical adjustments. When the dynamic operating condition segment discrimination value is greater than or equal to the secondary discrimination threshold but less than the primary discrimination threshold, it is determined as the starting point of a new operating condition segment. Segmentation anchor points are generated, segmented data is collected, and the detection frequency within this segment is increased. Sampling is added to the real-time monitoring data. After interpolation, removal, and smoothing of the segmented data, it enters the intelligent clustering discrimination module for operating condition performance calculation. When the dynamic operating condition segment discrimination value is greater than or equal to the primary discrimination threshold, it is determined as a risk operating condition point. The segmented real-time monitoring data is stored, and an abnormal operating condition database is constructed.

[0015] Furthermore, the specific process of performing feature clustering and performance status discrimination on segmented operating condition data, identifying different operating condition types, and managing them with labels is as follows: Acquiring data on energy consumption index fluctuations, indoor CO2 concentration, PM2.5 concentration, temperature and humidity, heat transfer coefficients of external walls and roofs, thickness of door and window insulation layers, leakage test data of roof and exterior wall waterproofing layers, concrete strength grade data, electricity consumption data, and historical operating condition data; obtaining the covariance of environmental green parameters for indoor CO2 concentration, PM2.5 concentration, and temperature and humidity using a covariance matrix algorithm; and determining the covariance of environmental green parameters for external walls and roofs, door and window insulation layer thickness, and roof and exterior wall waterproofing layer leakage test data. The quality index values ​​were obtained by using an arithmetic mean algorithm for the leakage test data of the external wall waterproofing layer and the concrete strength grade data; the rate of change of the quality index values ​​was obtained by using an adjacent time difference algorithm for the quality index value sequence; the Mahalanobis distance algorithm was used to calculate the covariance and mean of the electricity consumption data, indoor CO2 concentration, PM2.5 concentration, and temperature and humidity data with historical electricity consumption data, historical indoor CO2 concentration, historical PM2.5 concentration, and historical temperature and humidity to measure the degree of deviation and obtain the energy consumption-environment Mahalanobis distance; the product of the abrupt change in energy consumption index and the covariance of environmental green parameters was calculated to obtain the energy consumption-environment covariance. The same term is used; the difference between adjacent time points in the quality index value sequence is calculated, and the absolute value is calculated and added by one, which is used as the denominator of the quality index; the energy consumption and environment coordination term is divided by the denominator of the quality index to obtain the coordination value; the energy consumption and environment Mahalanobis distance is calculated, and the energy consumption and environment Mahalanobis distance is multiplied by the modulation coefficient, and the exponential function value is taken to obtain the modulation term; the coordination value is multiplied by the modulation term, the absolute value of the product is taken, and the maximum value is taken by traversing all time points within the segmented sliding window to obtain the segmented operating condition cluster value; the segmented operating condition cluster value and the segmented operating condition cluster threshold are compared in real time. The segmented operating condition cluster threshold includes the first-level cluster threshold, the second-level cluster threshold, and the third-level cluster threshold. Thresholds: When the cluster value of a segmented operating condition is less than the third-level cluster threshold, it is determined to be a normal operating condition and assigned a regular label. The regular label is assigned a value of 0.5 using the attribution label quantitative assignment method. When the cluster value of a segmented operating condition is greater than or equal to the third-level cluster threshold and less than the second-level cluster threshold, it is determined to be an energy consumption fluctuation condition and assigned a high energy consumption label. When the cluster value of a segmented operating condition is greater than or equal to the second-level cluster threshold and less than the first-level cluster threshold, it is determined to be an environmental anomaly condition and assigned a high environmental anomaly label. When the cluster value of a segmented operating condition is greater than or equal to the first-level cluster threshold, it is determined to be a complex interference condition and assigned a complex interference label.

[0016] Furthermore, for different types of abnormal operating conditions, the specific process of identifying and attributing the dominant factors causing fluctuations in project quality through multi-dimensional label attribution and source tracing is as follows: Combining the segmented operating condition labels with the structural number, process status, zoning location, design parameters, and risk attribute data output from the BIM model, the process aligns the detected energy consumption index mutations, environmental green parameter covariance, and quality index change rates within each segment. Simultaneously, it retrieves construction event data, equipment operation records, and external environmental monitoring information for the corresponding time period to determine the cause of the anomaly, prioritizing the determination of whether the segmented anomaly time coincides with the BIM model. If node and process status changes overlap, the cause is attributed to a change in the construction node. If there is no overlap, the synchronization between equipment operation records and sudden changes in physical quantities is assessed; if an anomaly exists, it is attributed to equipment malfunction. If the equipment is functioning normally, external environmental interference is further analyzed; if an anomaly exists, it is attributed to external disturbance. If multiple factors coexist or cannot be distinguished, the cause is attributed to multi-source complexity and unknown factors. A quantitative attribution labeling method is used to map the main cause category of the anomaly segment using digital factors. All attribution labels and corresponding segment numbers, real-time monitoring data, construction event data, and attribution process indexes are created and archived in the attribution database.

[0017] Furthermore, based on BIM structured data, real-time monitoring data, and attribution assignment data, the specific process of multi-source time-series item-by-item difference analysis is as follows: Energy consumption limits, air quality limits, compliance status, electricity consumption data, indoor CO2 concentration, PM2.5 concentration, anomaly attribution assignment, zoning quality grading values, importance level, geometric center 3D coordinates, and shortest path distance are obtained; the green performance detection deviation is obtained by using an item-by-item difference algorithm and weighted fusion for energy consumption limits, air quality limits, electricity consumption data, indoor CO2 concentration, and PM2.5 concentration; different risk scores are assigned to compliance status, zoning quality grading values, and importance level through interval mapping functions, and BIM risk coefficients are obtained through a weighted fusion algorithm; the green performance... The rate of change of the detection deviation is obtained by applying the adjacent time-difference algorithm to the time series of the detection deviation; the BIM spatial correlation factor is obtained by using the structural topology analysis algorithm based on the three-dimensional coordinates of the geometric center and the shortest path distance in space; the product of the green performance detection deviation and the attribution value of the abnormal working condition is calculated, and the result is added to the BIM risk coefficient to obtain the risk accumulation term; the adjacent time-difference of the green performance detection deviation time series is calculated, the absolute value is taken and one is added to obtain the performance denominator term; the risk accumulation term is divided by the performance denominator term to obtain the risk value term; the spatial structure correlation value is calculated and multiplied by the spatial adjustment coefficient, and the exponential function value is taken to obtain the spatial adjustment term; the risk value term is multiplied by the spatial adjustment term, the absolute value is taken, and the maximum value is taken by traversing all time points within the engineering unit window to obtain the engineering quality risk value.

[0018] Furthermore, the specific process of determining the risk level and implementing graded responses to drive closed-loop management of rectification, re-inspection, and archiving is as follows: Real-time comparison of engineering quality risk values ​​and engineering quality risk thresholds, including primary and secondary risk thresholds: When the engineering quality risk value is less than or equal to the secondary risk threshold, it is determined to be a green performance healthy zone, requiring only routine inspections and archiving of historical operating data; When the engineering quality risk value is greater than the secondary risk threshold but less than or equal to the primary risk threshold, it is determined to be a green performance concern zone, and encrypted testing tasks are pushed to conduct key spot checks and reviews of green indicators such as energy consumption, air quality, and renewable energy utilization rate; When the engineering quality risk value is greater than the primary risk threshold, it is determined to be a green performance risk zone, immediately triggering special rectification and re-inspection, generating an engineering quality risk report based on attribution database data and feeding it back to humans, listing the segment as a key monitoring object, and synchronously archiving all rectification records and re-inspection data to the abnormal operating condition database.

[0019] The second aspect of this invention provides an engineering quality inspection system based on green building design, including a data acquisition and preprocessing module for acquiring BIM structured data, real-time monitoring data, and structural acceptance data, and obtaining construction event and window data; preprocessing the BIM structured data, real-time monitoring data, structural acceptance data, and construction event and window data; a quality index extraction and compliance determination module for determining the compliance status of components and the quality level of zones based on BIM structured data and BIM model; and a dynamic working condition segmentation and event detection module for processing real-time monitoring data through mutation detection algorithms and feature temporal clustering. The method identifies anchor points and events for work condition switching and segments the work condition process; the intelligent clustering and discrimination module for work condition performance is used to perform feature clustering and performance status discrimination on segmented work condition data and structural acceptance data, identify different work condition types and manage them with labels; the work condition anomaly attribution and tracing module is used to identify the dominant factors causing fluctuations in project quality and assign values ​​based on the type of work condition anomaly through multi-dimensional label attribution and tracing discrimination; the quality risk classification and closed-loop management module is used to determine the level and classify the response based on BIM structured data, real-time monitoring data and attribution assignment data through multi-source time series item-by-item difference analysis.

[0020] Beneficial effects

[0021] The present invention has the following beneficial effects:

[0022] (1) The present invention adopts segmented working condition identification and event synchronous retrieval, which can capture working condition changes and key events in a timely manner, thereby achieving the effect of dynamic segmentation of complex working conditions and accurate locking of abnormal time periods, effectively solving the problems of untimely working condition identification and inaccurate anomaly tracking in the prior art.

[0023] (2) This invention integrates green performance testing, attribution labeling and spatial risk factor assessment to drive dynamic risk response throughout the process, thereby achieving the effect of intelligent risk classification and closed-loop management of rectification, effectively solving the problems of crude risk identification and non-closed-loop rectification tracking in the prior art.

[0024] (3) This invention integrates multi-dimensional index attribution and label assignment methods to classify and quantify the main causes of abnormal working conditions, thereby achieving the effect of quantitative attribution of multi-source abnormal working conditions, effectively solving the problems of single attribution method and lack of quantitative mechanism in the existing technology.

[0025] (4) This invention achieves efficient linkage of multiple types of data such as BIM model, construction progress and energy consumption environment through multi-source data collaborative fusion and dynamic spatiotemporal alignment, thereby realizing accurate mapping of the working conditions of the whole process and the effect of data collaboration, effectively solving the problems of spatiotemporal mismatch of multi-source data and disconnection of state intervals in the existing technology.

[0026] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0027] Figure 1 This is a flowchart of an engineering quality inspection method based on green building design according to the present invention;

[0028] Figure 2 This is a structural diagram of an engineering quality inspection system based on green building design according to the present invention;

[0029] Figure 3 This is a comparison trend chart of process node parameters in this invention;

[0030] Figure 4 This is a flowchart of the working condition determination and tag assignment process of the present invention. Detailed Implementation

[0031] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] Please see Figures 1-4This invention provides a technical solution: an engineering quality inspection method based on green building design, comprising: S1, collecting BIM structured data, real-time monitoring data, and structural acceptance data to obtain construction events and window data; preprocessing the BIM structured data, real-time monitoring data, structural acceptance data, and construction events and window data; S2, determining the qualification status of components and the quality level of zones based on BIM structured data and BIM model; S3, identifying working condition switching anchor points and events in real-time monitoring data using mutation detection algorithms and feature time-series clustering methods, and segmenting the working condition process; S4, performing feature clustering and performance status discrimination on segmented working condition data and structural acceptance data, identifying different working condition types and managing them with labels; S5, for working condition anomaly types, identifying the dominant factors causing engineering quality fluctuations and assigning values ​​through multi-dimensional label attribution and source tracing discrimination; S6, determining the level and classifying responses based on BIM structured data, real-time monitoring data, and attribution value data through multi-source time-series item-by-item difference analysis.

[0033] Specifically, the process involves collecting BIM structured data, real-time monitoring data, and structural acceptance data, as well as acquiring construction event and window data. The preprocessing of this data includes: collecting BIM structured data, which includes: process node parameters, component attribute parameters, green performance design parameters, and spatial geometry and connection parameters. Process node parameters include: unique ID, name, type, spatial partition mapping, and temporal attributes. Component attribute parameters include: quality grade, testing standards, component number, target parameter range, node type, stress function attributes, importance level, and design redundancy coefficient. Green performance design parameters include: design energy consumption limits, sub-item energy consumption, green target descriptions, indoor CO2 concentration, PM2.5 concentration, formaldehyde content, VOC content target limits and related design parameters, renewable energy utilization target values, and related equipment parameters. Spatial geometry and connection parameters include: three-dimensional coordinates, dimensions, spatial boundary points, voxel bounding boxes, structural component connection information, adjacency matrix, path dependency sequence, and shortest path distance.

[0034] Real-time monitoring data is collected, including: electricity consumption data, indoor CO2 concentration, PM2.5 concentration, temperature and humidity, formaldehyde content, VOC content, and renewable energy utilization rate. Electricity consumption data and temperature and humidity are sampled at a 1-2 minute interval, indoor CO2 concentration and PM2.5 concentration are collected in groups of 4-6 minutes, and formaldehyde content, VOC content, and renewable energy utilization rate are collected every 10-12 minutes. All sensors support zero-point drift calibration and periodic self-test to ensure the timeliness, continuity, and traceability of the data.

[0035] Structural acceptance data is collected, including: heat transfer coefficient of exterior walls and roof, thickness of insulation layer of doors and windows, leakage test data of waterproof layer of roof and exterior walls, and concrete strength grade data. All structural acceptance data are collected regularly by testing equipment. The testing cycle for heat transfer coefficient and insulation layer thickness is no less than once per quarter. The frequency of leakage test and concrete strength grade test is dynamically adjusted according to the construction status and zone.

[0036] Acquire construction event and window data, and establish a construction time database. The construction event and window data include material arrival events, rework events, abnormal alarm events, node change event data, historical time series data sliding windows, working condition sliding windows, segmented sliding windows, and engineering unit windows, to support subsequent time series analysis and process traceability.

[0037] The Hungarian allocation algorithm, which binds the topology map of the equipment partitions to the partition mapping table and the shortest path, constructs a cost matrix based on the shortest path value between the process node and the component number. A simplified matrix is ​​obtained through row and column reduction, and all zero elements are covered with the fewest possible lines. If the number of covered lines is less than the matrix dimension, the matrix is ​​adjusted until the number of covered lines equals the dimension. The optimal allocation scheme is determined based on the position of the zero elements in the simplified matrix. Spatial partition mapping, unique process node numbers, and component numbers are matched one-to-one to generate a partition mapping table and verify number consistency. Using multi-frequency timestamp resampling and linear interpolation methods, the timestamps of real-time monitoring data, structural acceptance data, BIM model output data, and construction events are standardized and annotated. Different sampling frequencies and asynchronously reported data streams are uniformly mapped to a standard timeline. Data streams are resampled at standard intervals of 1-2 minutes, and missing points are filled using linear interpolation. Data from different sources are aligned and all have original frequency and alignment status labels. The MAD rule is utilized, and the formula is... Using multiples of the absolute deviation of the median as the outlier threshold, and combining multivariate covariance analysis (MAD), statistical distances are constructed by calculating the eigenvalues ​​and eigenvectors of the multi-parameter covariance matrix. Observations exceeding the MAD threshold and whose Mahalanobis distances significantly deviate from the main distribution are identified as multivariate outliers. Combined with a pre-feature clustering algorithm, multi-parameter correlation analysis and outlier identification are performed on real-time monitoring data. All outlier results are associated with the source of the original parameters and the sampling time, facilitating subsequent source tracing. Through distribution standardization and linear normalization algorithms, z-score standardization is performed using the mean and variance of the training set. Combined with linear normalization, BIM structured data, real-time monitoring data, and construction event and window data are standardized and normalized. The standardization of all parameters strictly uses the mean and standard deviation of the training set. Normalization maps all indicators to the 0-1 interval, ensuring that the data distribution is consistent and the feature space is comparable during the modeling and inference stages.

[0038] This implementation plan achieves systematic aggregation and dynamic verification of various core indicators by uniformly collecting and organizing multi-source engineering data, including BIM structured data, real-time monitoring data, structural acceptance data, and construction event data, combined with an automated parameter preprocessing mechanism. After processes such as spatial mapping, temporal alignment, and multi-dimensional anomaly identification, the data is efficiently integrated into a standardized data system. All key parameters are processed through normalization and standardization algorithms to ensure the consistency of data characteristics, the comparability of distribution, and the traceability of analysis. The overall process not only improves the efficiency and utilization value of multi-source heterogeneous data fusion but also provides a high-quality, structured, and automated data foundation for subsequent risk assessment, work condition segmentation, and full-process quality traceability, strongly supporting the achievement of intelligent diagnosis, dynamic decision-making, and refined management goals in complex engineering scenarios.

[0039] Specifically, the process of determining the qualification status and zoning quality level of components based on BIM structured data and BIM model is as follows: Work process node parameters, component attribute parameters, green performance design parameters, and spatial geometry and connection parameters are input into the BIM model. All parameters are imported in batches in a standardized format to ensure consistency in data type, coding standards, and spatial coordinate system. Based on component coding, the structured data is spatially aggregated and associated to form a data sample set with both structural and temporal characteristics. During the aggregation process, spatial zoning labels and temporal indexes are generated to achieve a dual mapping of structural attributes and temporal evolution at the sample level. Using each component as a sample unit, a training sample set is constructed using a sliding window of historical temporal data. An integrated discriminant analysis-supervised learning method is employed to model and learn the quality indicators and qualification status, capturing the spatial correlation, temporal dependence features, and multivariate dynamic relationships of the input parameters. Multiple discriminant base classifiers are introduced, and weighted ensemble is used to enhance the modeling ability for complex feature relationships, fully exploring the spatial adjacency and temporal dynamic evolution between components. Information, combined with Bayesian optimization algorithms, enables a framework that uses a Gaussian process as a surrogate model and the desired improvement as the acquisition function. Hyperparameters range from a learning rate of 1e-5 to 1e-1, L1 and L2 regularization coefficients of 1e-6 to 1e-1, and a maximum tree model depth of 3 to 15. The termination condition is set to allow for adaptive adjustment of the model's hyperparameters after 50 consecutive iterations on the validation set, where the objective function improvement is less than 1e-4 or the maximum number of iterations (200) is reached. This allows for adaptive adjustment of the model's hyperparameters, optimizing the decision boundary and feature weights. During hyperparameter tuning, key parameter combinations such as the learning rate and regularization terms are iterated, and feature importance ranking is dynamically adjusted based on validation set performance. After model training, based on the design and detection values ​​of the input samples, the system outputs process node codes, partition quality grading values, energy consumption limits, air quality limits, importance levels, three-dimensional coordinates of the geometric center, shortest path distance, structure number, process status, partition location, design parameters, and risk attribute data. The output results are grouped and archived by spatial partition, process node, and component number for easy subsequent risk analysis and decision-making.

[0040] This implementation plan, by aggregating and deeply integrating BIM structured data into the BIM model, achieves the collection of key attributes, design parameters, and temporal characteristics of various components and spatial zoning. Combining advanced discriminant analysis and supervised learning algorithms, it models and intelligently judges component qualification and zoning quality levels, effectively capturing the complex relationships between structure, space, and time, and adaptively optimizing the decision model based on multi-source data. Ultimately, the process outputs decision data covering multiple aspects such as process nodes, zoning levels, energy consumption and air quality limits, structural spatial attributes, and risk indicators, significantly improving the automation, intelligence, and scientific level of engineering quality management, and providing efficient and reliable data and model support for subsequent risk analysis, refined operation and maintenance, and intelligent control.

[0041] Specifically, the process of applying mutation detection algorithms and feature-based temporal clustering methods to real-time monitoring data is as follows: Obtain process node code values, construction event data, working condition sliding windows, energy consumption data, indoor CO2 concentration, PM2.5 concentration, temperature and humidity, and zoning quality grading values. All data types are associated with corresponding spatial zoning attributes to ensure the relevance of data input. The process node code values ​​are then processed using a working condition moving average algorithm to obtain the node code mean. Moving average processing helps smooth short-term fluctuations and highlights the phased change trends of the node codes. Finally, the construction event data is processed using a numerical mapping algorithm to obtain the construction event signal values. Numerical mapping enables direct comparability of different types of events for subsequent risk analysis; the mean value of energy consumption data within the sliding window of the operating condition is obtained through the mean algorithm, and the difference between the energy consumption data and the mean value is calculated to obtain the mutation amount of energy consumption index, reflecting the abnormal deviation of the current window stage relative to the baseline consumption, which is conducive to the detection of abnormal energy use; the mutation amounts of various environmental parameters such as indoor CO2 concentration, PM2.5 concentration, temperature and humidity are obtained through the window average difference method, and the mutation amounts of various environmental parameters are obtained through weighted fusion to obtain the mutation amount of environmental green parameters, which helps to comprehensively assess the impact of environmental quality fluctuations on the operating condition.

[0042] The difference between the process node code and the mean of the process node codes is calculated to obtain the process node change, which can intuitively reveal the abnormal switching of node status and the transition of process stages. Construction event signal values ​​are extracted as the construction event impact, reflecting the actual interference of construction activities or anomalies on the current working condition. The energy consumption index mutation and the environmental green parameter mutation are added together and then multiplied by the zoning quality grading value to obtain the energy consumption and environmental risk weighted quantity, reflecting the significant impact of energy consumption and environmental anomalies on zoning quality. The process node change, construction event impact, and energy consumption and environmental risk weighted quantity are added together and their absolute values ​​are calculated as the segmented discrimination quantity, facilitating the quantitative integration of risk contributions from multiple dimensions. Within the entire calculation window, the working condition sliding window discrimination quantity is traversed at each moment, and the maximum value is taken to obtain the dynamic working condition segmented discrimination value. The specific calculation formula is as follows:

[0043] ;

[0044] In the formula, It represents the dynamic operating condition segment discrimination value, which is used to quantify the comprehensive variation amplitude of multi-source risk factors within the operating condition window. It is the main control indicator for discriminating segments and anomalies. The numerical code of the process node reflects the spatial status and coding characteristics of the current process progress in the BIM model, providing a structural basis for segmented aggregation; It represents the average value of node codes, reflecting the stable level of process nodes in the current time segment, which is convenient for capturing sudden changes; This represents the signal value of a construction event, measuring the degree of impact of on-site construction events on operating conditions. It indicates the sudden change in energy consumption indicators, reflecting the deviation of real-time energy consumption data from the benchmark average, which helps to identify energy consumption anomalies and abnormal operating conditions. It represents the abrupt changes in environmental green parameters, comprehensively depicting the dynamic fluctuation characteristics of environmental quality; This represents the zoning quality grading value, assigned based on the BIM zoning quality evaluation, and is used to adjust the weight and influence of risk in different spatial zones.

[0045] In this embodiment, Table 1 is a data table for process node and dynamic working condition discrimination. It records in detail the key discrimination parameters of process node code, mean node code, node change, energy consumption index mutation, zoning quality grading value, segment discrimination value, and dynamic working condition segment discrimination value at different times. The environmental green parameter mutation and construction event signal value auxiliary parameters not mentioned in the table have been uniformly fixed at 0.00 to eliminate their interference with the main discrimination process and ensure the comparability of multiple parameters at different times. For example, at time t=3, the process node code is 0.31, the node change is 0.07, the energy consumption index mutation is 0.16, and the segment discrimination value reaches 0.206. The dynamic working condition segment discrimination value is significantly higher than at other times, reflecting that the working condition has undergone obvious segmentation variation at this time. This table is used to systematically quantify and compare the engineering quality risk characteristics of each working condition time period, providing basic data support for subsequent dynamic segmentation and risk control decisions.

[0046] Table 1. Process Node and Dynamic Operating Condition Judgment Data Table

[0047] 1 0.20 0.19 0.01 0.03 0.80 0.03 0.04 2 0.23 0.21 0.02 0.08 0.82 0.09 0.09 3 0.31 0.24 0.07 0.16 0.85 0.21 0.21 4 0.28 0.25 0.03 0.05 0.84 0.07 0.12 5 0.20 0.22 -0.02 0.01 0.81 0.01 0.03 6 0.45 0.29 0.16 0.28 0.90 0.28 0.29 7 0.40 0.35 0.05 0.09 0.86 0.088 0.10

[0048] like Figure 3 The chart shows the trend of process node parameters. Combined with Table 1, it can intuitively reflect the dynamic change trend of each main control parameter under different time windows. In the chart, the blue curve and bubbles represent the change of process node code over time, yellow represents the sudden change in energy consumption index, and red represents the segmentation discrimination value. It can be seen that the node code, energy consumption sudden change, and segmentation discrimination value corresponding to the 6th window all reach high values, indicating that this stage is a typical area of ​​sudden changes in operating conditions and concentrated risks. Conversely, the parameter values ​​in the 5th window are all low, indicating that the operating conditions are relatively stable. The trend chart intuitively shows the temporal fluctuation characteristics of process nodes, energy consumption, and segmentation risk values, which helps to quickly locate abnormal periods of operating conditions and provides a visual decision-making basis for dynamic segmentation and accurate risk response.

[0049] This implementation scheme integrates mutation detection algorithms and feature-based temporal clustering to achieve dynamic segmentation and intelligent feature extraction of multi-source monitoring data. Based on parameters such as process node codes, construction events, energy consumption, and environmental indicators, it identifies abnormal transitions, risk accumulation, and state mutations during the working process, achieving highly sensitive identification of working condition segments and comprehensive analysis of multi-dimensional risk signals. The parameter processing workflow not only eliminates interference from different types of auxiliary parameters on the main process but also enhances the temporal sensitivity and regional difference response of risk features through mechanisms such as sliding windows and weighted fusion. Finally, the relevant discrimination results and trend analyses can be intuitively displayed in tables and graphs, providing solid data support and visualization for engineering quality risk control, anomaly tracing, and segmented intelligent decision-making, significantly improving the ability for refined diagnosis and dynamic management in complex working conditions.

[0050] Specifically, the process of identifying operating condition switching anchor points and events and segmenting the operating condition process is as follows: real-time comparison of dynamic operating condition segment discrimination values ​​and dynamic operating condition segment discrimination thresholds. The dynamic operating condition segment discrimination values ​​include a primary discrimination threshold and a secondary discrimination threshold. When the dynamic operating condition segment discrimination value is less than the secondary discrimination threshold, the real-time monitoring data is created and included in the historical operating condition data, which facilitates subsequent traceability and data review without the need for physical adjustments.

[0051] When the dynamic operating condition segment discrimination value is greater than or equal to the secondary discrimination threshold and less than the primary discrimination threshold, it is determined as the starting point of a new operating condition segment, a segment anchor point is generated, segment data is collected and the detection frequency within this segment is increased, the spatial location information and data window number of the segment anchor point are recorded synchronously, and all processing processes retain processing logs, supporting fine-grained tracing of abnormal samples and cluster feature tracking, adding sampling inspection for real-time monitoring data, and performing interpolation, removal and smoothing processing on segment data before entering the intelligent cluster discrimination module for operating condition performance calculation.

[0052] When the dynamic operating condition segment discrimination value is greater than or equal to the first-level discrimination threshold, it is judged as a risk operating condition point. The segment real-time monitoring data is stored and an abnormal operating condition database is built. The data is then pushed to relevant personnel to realize real-time closed-loop response and multi-level early warning archiving for abnormal operating conditions.

[0053] In this implementation plan, by comparing the dynamic operating condition segment discrimination value with multi-level thresholds in real time, the identification of operating condition switching anchor points and key events, as well as the dynamic segmentation of operating conditions throughout the entire process, are realized. Relying on segment anchor point positioning, segment data collection, anomaly handling logs, and feature clustering analysis, the system can efficiently perceive the phased changes in operating conditions, quickly locate risk segments, and establish a complete data traceability chain for each segment. For risky operating conditions, anomaly alarms and multi-level early warnings can be triggered simultaneously, supporting rapid manual response and management closed loop. This step greatly improves the intelligent identification capability of operating condition switching and anomaly risks during engineering operation, strengthens the systematicness and transparency of segment data management, and lays a solid data and logical foundation for subsequent quality traceability, risk control, and intelligent operation and maintenance decision-making.

[0054] Specifically, the process of performing feature clustering and performance status discrimination on segmented operating condition data, identifying different operating condition types, and managing them with labels involves: acquiring data on energy consumption index mutations, indoor CO2 concentration, PM2.5 concentration, temperature and humidity, heat transfer coefficients of exterior walls and roofs, thickness of door and window insulation layers, leakage test data of roof and exterior wall waterproofing layers, concrete strength grade data, electricity consumption data, and historical operating condition data to ensure that input parameters comprehensively cover both operating condition and structural feature dimensions; obtaining the covariance of environmental green parameters for indoor CO2 concentration, PM2.5 concentration, and temperature and humidity using a covariance matrix algorithm; covariance matrix analysis can reveal the correlation and coupling strength between different environmental indicators, facilitating the identification of multivariate linkage characteristics of environmental risks; and obtaining the covariance of environmental green parameters for exterior walls and roofs, door and window insulation layer thickness, roof and exterior wall waterproofing layer leakage test data, concrete strength grade data, electricity consumption data, and historical operating condition data. The quality index values ​​are obtained by using an arithmetic mean algorithm for the leakage test data of the external wall waterproofing layer and the concrete strength grade data. This can effectively reflect the overall level of structural quality parameters under a specified zone or operating condition window. The rate of change of quality index values ​​is obtained by using an adjacent time difference algorithm for the quality index value sequence. This can capture the gradual fluctuations and sudden anomalies in structural performance and help to dynamically monitor the quality health of the zone. The Mahalanobis distance algorithm is used for the electricity consumption data, indoor CO2 concentration, PM2.5 concentration, and temperature and humidity. The Mahalanobis distance is calculated with the covariance and mean of historical electricity consumption data, historical indoor CO2 concentration, historical PM2.5 concentration, and historical temperature and humidity data to measure the degree of deviation and obtain the energy consumption environment Mahalanobis distance. This can quantify the overall deviation of the current operating condition from the historical steady-state range and identify the collaborative risks of multi-source anomalies.

[0055] The product of the abrupt change in energy consumption indicators and the covariance of environmental green parameters is calculated to obtain the energy consumption-environment synergy term, which reflects the superimposed effect of energy consumption anomalies and environmental linkage risks, and is particularly sensitive to risk identification under complex operating conditions. The difference between adjacent time points in the quality indicator value sequence is calculated, and the absolute value is calculated and incremented by one as the denominator of the quality indicator, used to adjust the normalization strength of the synergy term and avoid amplified bias caused by drastic fluctuations in quality indicators. The synergy value is obtained by dividing the energy consumption-environment synergy term by the denominator of the quality indicator, reflecting the joint dynamic influence of multidimensional risk factors. The modulation term is obtained by multiplying the energy consumption-environment Mahalanobis distance by the modulation coefficient and taking the exponential function value. Magnifying or reducing the energy consumption-environment Mahalanobis distance and inputting it into the exponential function can nonlinearly enhance the response sensitivity to extreme anomalies, achieving exponentially sensitive discrimination of operating conditions far from historical distribution ranges. The absolute value of the synergy value multiplied by the modulation term is taken, and the maximum value is obtained by traversing all time points within the segmented sliding window. The specific calculation formula is as follows:

[0056] ;

[0057] In the formula, The segmented operating condition cluster value is used to quantitatively characterize the synergistic extreme value of multi-source risks and operating condition abrupt changes within the sliding window. It is a key basis for segmented operating condition classification and risk identification. It indicates the sudden change in energy consumption indicators, reflects the difference between real-time consumption and the baseline value, captures abnormal energy consumption signals, and responds sensitively to changes in energy consumption. It represents the covariance of environmental green parameters, reflecting the dynamic linkage and abnormal diffusion trend of multiple environmental factors; These represent quality index values, reflecting the overall quality level of each segment. It indicates the rate of change of quality indicators, reveals the real-time fluctuations and sudden changes in structural performance parameters, and helps in dynamic risk monitoring; It represents the Mahalanobis distance between energy consumption and environmental parameters, which measures the overall deviation of current energy consumption and environmental parameters from historical steady-state distributions. Combined with modulation coefficients and exponential mapping, it is used to amplify the nonlinear effects of extreme abnormal operating conditions. This represents the modulation coefficient, which is obtained from historical operating data using the Fisher discriminant analysis algorithm. The value ranges from 0.1 to 0.5.

[0058] like Figure 4 The diagram shows the flowchart for determining the working condition and assigning labels. It compares the segmented working condition cluster values ​​and the segmented working condition cluster thresholds in real time. The segmented working condition cluster thresholds include first-level cluster thresholds, second-level cluster thresholds, and third-level cluster thresholds, which are used to classify and identify the severity of the working condition, ensuring the accuracy and operability of risk segmentation.

[0059] When the cluster value of segmented working conditions is less than the three-level clustering threshold, it is judged as a normal working condition and assigned a regular label, indicating that the risk characteristics within this period are all within the normal fluctuation range and no significant abnormalities are observed. The regular label is assigned a value of 0.5 by the attribution label quantitative assignment method.

[0060] When the cluster value of a segmented operating condition is greater than or equal to the third-level clustering threshold but less than the second-level clustering threshold, it is judged as an energy consumption fluctuation situation and is labeled as high energy consumption. This reflects that the energy consumption data in the operating condition stage has a short-term fluctuation but has not exceeded the warning limit, and energy consumption management needs to be paid attention to.

[0061] When the cluster value of segmented operating conditions is greater than or equal to the second-level clustering threshold but less than the first-level clustering threshold, it is judged as an environmental anomaly and is assigned a high environmental anomaly label, indicating that the environmental parameters in this interval have fluctuated significantly and have a phased impact on green performance.

[0062] When the cluster value of a segmented operating condition is greater than or equal to the first-level clustering threshold, it is determined to be a complex interference operating condition and is assigned a complex interference label. This indicates that multiple risk factors within the segment are synchronously abnormal, the operating condition interference characteristics are prominent, and key monitoring and intervention are required.

[0063] In this implementation plan, this step achieves intelligent identification and hierarchical labeling management of operating condition types through the integrated processing of multi-dimensional monitoring parameters and feature clustering of segmented operating condition data. Relying on the collaborative analysis of core indicators such as environment, structure, and energy consumption, it can accurately capture operating condition fluctuations, abnormal trends, and complex disturbances, enabling dynamic risk identification and classification for each segment. The clustering threshold grading mechanism ensures the scientific and operational nature of operating condition differentiation, while the labeling system enhances the precision of risk management and the traceability of engineering events. This step provides a quantifiable and traceable multi-dimensional operating condition risk assessment system for engineering operation and maintenance, laying a solid foundation for dynamic diagnosis, anomaly early warning, and decision support in high-requirement scenarios such as green buildings.

[0064] Specifically, for different types of abnormal operating conditions, the process of identifying and attributing the dominant factors causing fluctuations in project quality through multi-dimensional label attribution and source tracing is as follows: Combining segmented operating condition labels with the structural number, process status, partition location, design parameters, and risk attribute data output from the BIM model, the process aligns the detected energy consumption index mutations, environmental green parameter covariance, and quality index change rates within each segment, ensuring that all key control parameters are correlated one-to-one at the spatial, temporal, and process information levels. This enhances the integrity of the data link for anomaly attribution. Simultaneously, construction event data, equipment operation records, and external environmental monitoring information for the corresponding time period are retrieved for anomaly attribution determination. Priority is given to determining whether the segmented anomaly time overlaps with changes in BIM nodes and process status. If so, it is attributed to a construction node change. If there is no overlap, the synchronization between equipment operation records and physical quantity mutations is assessed; if anomalies exist, they are attributed to equipment anomalies. If the equipment is normal, further analysis of external environmental interference is conducted; if anomalies exist, they are attributed to external disturbances. If multiple factors coexist or cannot be distinguished, the cause is attributed to multi-source composite and unknown factors. The attribution rule uses prior logic to determine priority, and can also calculate the weight allocation based on the magnitude of change of each main factor indicator. If the magnitude of each factor's influence is similar, joint probability attribution is used. After the multi-source composite and unknown categories are identified, the weights and probability distributions of all suspicious factors are recorded to facilitate subsequent manual review and secondary attribution. A quantitative attribution label assignment method is adopted, and the main factor categories of the anomaly segment are mapped through digital factors. The construction node change label is assigned a value of 0.8, the equipment anomaly label is assigned a value of 1.5, the external disturbance label is assigned a value of 1.2, and the multi-source composite and unknown label is assigned a value of 1.8. The numerical assignment is based on the degree of impact of each type of risk on the stability of the project. The assignment results can be directly used for subsequent risk quantification and classification strategy formulation. All attribution labels and corresponding segment numbers, real-time monitoring data, construction event data, and attribution process indexes are created and archived in the attribution database to realize full-process traceability, graded response, and closed-loop management of risk responsibility for anomalies.

[0065] In this implementation plan, multi-dimensional label attribution and source tracing can accurately identify the dominant factors causing fluctuations in engineering quality within the spatiotemporal link of segmented working condition data. This enables the orderly hierarchical attribution and digital assignment of various anomalies, combined with BIM model output, real-time monitoring, and event recording. All working condition anomalies can be fully correlated with master control parameters and process information, effectively improving the completeness of data traceability and the accuracy of attribution judgment. The attribution results are optimized through prior logic and joint probability mechanisms, which not only enhances the attribution discrimination in multi-source composite or complex anomaly situations, but also allows attribution labels and quantitative values ​​to be directly used for subsequent risk classification, response strategy formulation, and closed-loop responsibility management. This step significantly improves the ability to locate, interpret, and respond to engineering anomalies, providing solid data and decision support for the intelligent management and full life-cycle quality traceability of green high-standard projects.

[0066] Specifically, based on BIM structured data, real-time monitoring data, and attribution assignment data, the process of multi-source time-series item-by-item difference analysis is as follows: Energy consumption limits, air quality limits, electricity consumption data, compliance status, indoor CO2 concentration, PM2.5 concentration, abnormal operating condition attribution assignments, zonal quality grading values, importance levels, geometric center three-dimensional coordinates, and shortest path distance are obtained to ensure that the data covers all dimensions of green performance, structural grading, and spatial distribution, which is beneficial for multi-angle risk assessment. The green performance detection deviation is obtained by using an item-by-item difference algorithm and weighted fusion for energy consumption limits, air quality limits, electricity consumption data, indoor CO2 concentration, and PM2.5 concentration. The difference algorithm highlights the relationship between each parameter and the target setting. The deviation of the calculation is weighted and fused to enhance the sensitivity of the comprehensive green performance evaluation. Different risk scores are assigned to the qualification status, zoning quality grading value and importance level through interval mapping function. The BIM risk coefficient is obtained through weighted fusion algorithm. The interval mapping function reflects the risk impact of different levels and attributes in a quantitative way, which is convenient for risk classification management. The time series of green performance detection deviation is obtained by applying the adjacent time difference algorithm to obtain the rate of change of detection deviation, which can effectively capture the dynamic fluctuation of green performance indicators and help identify sudden anomalies. The three-dimensional coordinates of the geometric center and the shortest path distance in space are obtained by structural topology analysis algorithm to obtain BIM spatial correlation factors, revealing the spatial dependence and risk diffusion path between zoning.

[0067] The product of the green performance detection deviation and the attribution value of the abnormal working conditions is calculated. The result is added to the BIM risk coefficient to obtain the risk accumulation term, which comprehensively reflects the superimposed effect of multiple factors such as green deviation, abnormal attribution, and structural classification risk. The difference between adjacent time-series green performance detection deviation is calculated, and the absolute value is incremented by one to obtain the performance denominator term, which is used to normalize the risk amplitude and suppress the deviation amplification caused by extreme values. The risk accumulation term is divided by the performance denominator term to obtain the risk value term, which quantitatively measures the comprehensive risk level of a single window period. The spatial structure correlation value is calculated and multiplied by the spatial adjustment coefficient, and the exponential function value is taken to obtain the spatial adjustment term, which significantly amplifies the risk performance of areas with high structural correlation and easy risk diffusion through exponential mapping. The risk value term is multiplied by the spatial adjustment term, and the absolute value is taken. The maximum value is then taken after traversing all time points within the engineering unit window to obtain the engineering quality risk value. The specific calculation formula is as follows:

[0068] ;

[0069] In the formula, It represents the engineering quality risk value, quantitatively characterizes the comprehensive extreme value after the superposition of multiple risk factors, and is the main quantitative indicator for engineering health and risk warning; It indicates the deviation in green performance testing and sensitively reveals fluctuations in green performance. This indicates the assignment of abnormal operating conditions to different factors, which assigns weights based on the dominant abnormality types in segmented operating conditions to quantify the actual contribution of each abnormal factor to the overall risk. This represents the BIM risk coefficient, reflecting the fundamental moderating effect of zoning structural attributes on risk. It indicates the rate of change of detection deviation, captures the dynamic changes in the degree of green performance detection deviation over time, and reflects the sensitive areas of drastic fluctuations and sudden changes in indicators; It represents the spatial structure correlation value, which measures the potential diffusion capacity and dependence of risks in spatial intervals; This represents the spatial adjustment coefficient, which is obtained by using the maximum grading accuracy optimization algorithm on historical operating data. The value ranges from 0.1 to 0.5.

[0070] This implementation plan utilizes multi-source time-series item-by-item difference analysis based on BIM structured data, real-time monitoring data, and attribution assignment data to organically integrate multi-dimensional risk factors related to green performance, structural classification, and spatial topology. Various parameters work synergistically at the spatial, temporal, and attribution levels, effectively enhancing the sensitivity to capture the health status and abnormal risks of the project. Combined with weighted fusion, interval mapping, structural topology analysis, and dynamic normalization, it not only quantitatively reflects the extreme risk values ​​of each zone of the project during real-time operation but also accurately identifies sudden anomalies and potential diffusion paths. The final output of the project quality risk value possesses multi-source heterogeneous data support, dynamic threshold discrimination, and spatial risk adjustment capabilities, providing a comprehensive and reliable data foundation and decision-making basis for risk warning, strategy decision-making, and anomaly closed-loop management during project operation and maintenance.

[0071] Specifically, the process of determining the risk level and implementing graded responses to drive closed-loop management of rectification, re-inspection, and archiving involves: real-time comparison of engineering quality risk values ​​and engineering quality risk thresholds, including primary and secondary risk thresholds, dynamically adjusted to adapt to different project types and control requirements.

[0072] When the engineering quality risk value is less than or equal to the secondary risk threshold, it is judged as a green performance health zone. All green and quality indicators are stable and no special intervention is required. Only daily inspections and historical working condition data archiving are needed to facilitate data accumulation and subsequent traceability.

[0073] When the engineering quality risk value is greater than the secondary risk threshold but less than or equal to the primary risk threshold, it is identified as a green performance concern area. Enhanced monitoring tasks are then pushed to conduct key spot checks and reviews on energy consumption, air quality, and renewable energy utilization rates. If the indicators show fluctuations or critical anomalies, the monitoring frequency and data review efforts need to be increased to prevent potential risks from developing further.

[0074] When the engineering quality risk value exceeds the first-level risk threshold, it is identified as a green performance risk zone, and a special rectification and re-inspection are immediately triggered. An engineering quality risk report is generated based on the attribution database data and fed back to the human team. The segment is listed as a key monitoring target, and all rectification records and re-inspection data are synchronously archived to the abnormal working condition database to ensure full-process recording, problem tracing, and closed-loop rectification of high-risk segments, forming a dynamic risk control chain.

[0075] In this implementation plan, the level determination and tiered response mechanism in this step greatly enhances the sensitivity and accuracy of engineering quality control, enabling risk signals to be dynamically captured and responded to in a tiered manner. Through data-driven archiving and closed-loop rectification throughout the entire process, the operation and maintenance team can track changes in operating conditions and risk evolution in a timely manner, ensuring that high-risk sections are quickly addressed and closely monitored. This step effectively reduces the accumulation of potential quality hazards, improves rectification efficiency and traceability transparency, and provides full life-cycle risk controllability and management resilience for green buildings and high-standard projects. It not only strengthens the intelligent level of engineering operation and maintenance but also lays a solid data and decision-making foundation for subsequent optimization strategy formulation, quality improvement, and the continuous achievement of green goals.

[0076] like Figure 2 As shown, the second aspect of this invention provides an engineering quality inspection system based on green building design, including a data acquisition and preprocessing module for acquiring BIM structured data, real-time monitoring data, and structural acceptance data, and obtaining construction event and window data; preprocessing the BIM structured data, real-time monitoring data, structural acceptance data, and construction event and window data; a quality index extraction and qualification determination module for determining the qualification status of components and the quality level of zones based on BIM structured data and BIM model; and a dynamic working condition segmentation and event detection module for processing real-time monitoring data using a mutation detection algorithm and feature time series. Clustering methods identify anchor points and events for work condition switching, and segment the work condition process; the intelligent clustering and discrimination module for work condition performance is used to perform feature clustering and performance status discrimination on segmented work condition data and structural acceptance data, identify different work condition types, and manage them with labels; the work condition anomaly attribution and tracing module is used to identify the dominant factors causing fluctuations in project quality and assign values ​​based on the type of work condition anomaly through multi-dimensional label attribution and tracing discrimination; the quality risk classification and closed-loop management module is used to determine the level and classify the response based on BIM structured data, real-time monitoring data, and attribution value data through multi-source time-series item-by-item difference analysis.

[0077] This implementation plan achieves full-process, multi-dimensional, and intelligent detection and management of green building project quality. It effectively integrates various data types, including BIM modeling, on-site multi-source monitoring, structural acceptance, and event archiving. Through modular division of labor, it ensures efficient collection and intelligent preprocessing of raw data while enabling component-level quality qualification and zoning quality grading, significantly improving the granularity and scientific rigor of detection. Dynamic operating condition segmentation, event detection, and intelligent clustering of operating condition performance enable rapid location of operating condition transitions, identification of abnormal events, and multi-dimensional tagging management, achieving precise characterization and differentiated control of operating conditions. The attribution and tracing module effectively supports the identification and quantification of complex risk causes, providing data support for precise rectification and risk response. The hierarchical risk assessment and closed-loop management mechanism ensures full-process traceability, timely intervention, and dynamic archiving of quality issues, significantly improving project risk management capabilities and green operation and maintenance efficiency. The system provides a solid intelligent foundation and decision-making guarantee for the quality safety, energy conservation, carbon reduction, and sustainable operation and maintenance of high-standard green building projects.

[0078] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0079] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. An engineering quality inspection method based on green building design, characterized in that, Includes the following steps: S1: Collect BIM structured data, real-time monitoring data, and structural acceptance data; obtain construction event and window data; preprocess the BIM structured data, real-time monitoring data, structural acceptance data, and construction event and window data. S2, based on BIM structured data and BIM model, determines the qualification status of components and the quality level of zones. The specific process is as follows: Input process node parameters, component attribute parameters, green performance design parameters, spatial geometry and connection parameters into the BIM model. Based on component codes, the structured data is spatially aggregated and associated to form a data sample set with both structural and temporal characteristics. Taking each component as a sample unit, a training sample set is constructed by combining historical time-series data with a sliding window. The quality indicators are modeled and learned. An integrated discriminant analysis method is used to supervise the learning algorithm to determine the qualification status of components, capturing the spatial correlation, temporal dependence features and multivariate dynamic relationships of the input parameters. Combined with the Bayesian optimization algorithm, the model hyperparameters are adaptively adjusted to optimize the decision boundary and feature weights. After the model training is completed, based on the design value and detection value features of the input samples, the following data are output: process node code value, zone quality level value, energy consumption limit, air quality limit, importance level, three-dimensional coordinates of geometric center, shortest path distance, structure number, process status, zone location, design parameters and risk attribute data. By binding the partition mapping relationship table with the equipment partition topology map and the shortest path Hungarian allocation algorithm, the spatial partition mapping, the unique number of the process node and the component number are matched one by one. S3, using a mutation detection algorithm and a feature-based temporal clustering method, the real-time monitoring data is used to identify working condition switching anchor points and events, and to segment the working condition process. Specifically, the process of using the mutation detection algorithm and feature-based temporal clustering method on the real-time monitoring data includes: acquiring process node code values, construction event data, working condition sliding windows, energy consumption data, indoor CO2 concentration, PM2.5 concentration, temperature and humidity, and zoning quality grading values; obtaining the node code mean value using a working condition moving average algorithm; obtaining the construction event signal value using a numerical mapping algorithm; obtaining the average energy consumption data within the working condition sliding window using a mean algorithm; calculating the difference between the energy consumption data and the average energy consumption data to obtain the energy consumption index mutation amount; obtaining the mutation amounts of various environmental parameters for indoor CO2 concentration, PM2.5 concentration, and temperature and humidity using a window average difference method; and obtaining the environmental green parameter mutation amount through weighted fusion of the various environmental parameter mutation amounts. The process node change is obtained by calculating the difference between the process node code and the mean of the process node codes; the construction event signal value is extracted as the construction event impact value; the energy consumption index mutation value and the environmental green parameter mutation value are added together and then multiplied by the zoning quality classification value to obtain the energy consumption environmental risk weighted value; the process node change value, the construction event impact value and the energy consumption environmental risk weighted value are added together and the absolute value is calculated as the segmentation discrimination value; within the entire calculation window, the working condition sliding window discrimination value at each moment is traversed, and the maximum value is taken to obtain the dynamic working condition segmentation discrimination value; the specific process of identifying the working condition switching anchor point and event and segmenting the working condition process is as follows: the dynamic working condition segmentation discrimination value and the dynamic working condition segmentation discrimination threshold are compared in real time. The dynamic working condition segmentation discrimination threshold includes a first-level discrimination threshold and a second-level discrimination threshold: when the dynamic working condition segmentation discrimination value is less than the second-level discrimination threshold, the real-time monitoring data is created and included in the historical working condition data without physical adjustment; When the dynamic operating condition segment discrimination value is greater than or equal to the secondary discrimination threshold and less than the primary discrimination threshold, it is determined as the starting point of a new operating condition segment, a segment anchor point is generated, segment data is collected and the detection frequency within this segment is increased, sampling inspection is added for real-time monitoring data, and the segment data is interpolated, removed, and smoothed before entering the calculation of S4; when the dynamic operating condition segment discrimination value is greater than or equal to the primary discrimination threshold, it is determined as a risk operating condition point, the segment real-time monitoring data is stored and an abnormal operating condition database is built; S4. Feature clustering and performance status assessment are performed on segmented operating condition data and structural acceptance data to identify different operating condition types and manage them with labels. The specific process involves: acquiring data on energy consumption index fluctuations, indoor CO2 concentration, PM2.5 concentration, temperature and humidity, heat transfer coefficients of exterior walls and roofs, thickness of door and window insulation layers, leakage test data of roof and exterior wall waterproofing layers, concrete strength grade data, electricity consumption data, and historical operating condition data; indoor CO2 concentration, PM2.5 concentration, and temperature and humidity are assessed through a collaborative process. The difference matrix algorithm is used to obtain the covariance of environmental green parameters; the arithmetic mean algorithm is used to obtain the quality index values ​​of the heat transfer coefficient of the exterior wall and roof, the thickness of the insulation layer of the doors and windows, the leakage test data of the waterproof layer of the roof and exterior wall, and the concrete strength grade data; the rate of change of the quality index values ​​is obtained by the difference algorithm between adjacent time points; the Mahalanobis distance algorithm is used to calculate the Mahalanobis distance between the electricity consumption data, indoor CO2 concentration, PM2.5 concentration, and temperature and humidity data and the covariance and mean of historical electricity consumption data, historical indoor CO2 concentration, historical PM2.5 concentration, and historical temperature and humidity, to measure the degree of deviation, and to obtain the energy consumption environment Mahalanobis distance; The product of the energy consumption index mutation and the covariance of the environmental green parameters is calculated to obtain the energy consumption-environment synergy term. The difference between adjacent time points in the quality index value sequence is calculated, and the absolute value is incremented by one to serve as the denominator of the quality index. The synergy term is divided by the denominator of the quality index to obtain the synergy value. The Mahalanobis distance between energy consumption and environment is calculated, multiplied by the modulation coefficient, and the exponential function value is obtained to obtain the modulation term. The synergy value is multiplied by the modulation term, and the absolute value of the product is taken. The maximum value is obtained by iterating through all time points within the segmented sliding window to obtain the segmented operating condition cluster value. The segmented operating condition cluster value and the segmented operating condition cluster threshold are compared in real time. The segmented operating condition cluster threshold includes a first-level cluster threshold, a second-level cluster threshold, and a third-level cluster threshold. When the segmented operating condition cluster value is less than the third-level cluster threshold, it is determined to be a normal operating condition and assigned a regular label. The regular label is assigned a value of 0.5 using a quantitative attribution label assignment method. When the segmented operating condition cluster value is greater than or equal to the third-level clustering threshold and less than the second-level clustering threshold, it is judged as an energy consumption fluctuation situation and is labeled as high energy consumption; when the segmented operating condition cluster value is greater than or equal to the second-level clustering threshold and less than the first-level clustering threshold, it is judged as an environmental anomaly situation and is labeled as high environmental anomaly; when the segmented operating condition cluster value is greater than or equal to the first-level clustering threshold, it is judged as a complex interference situation and is labeled as complex interference. S5, for abnormal working conditions, identifies and assigns values ​​to the dominant factors causing fluctuations in project quality through multi-dimensional label attribution and source tracing. Specifically, it combines segmented working condition labels with structural numbers, process statuses, zoning locations, design parameters, and risk attribute data output from the BIM model. It aligns the detected abrupt changes in energy consumption indicators, covariance of environmental green parameters, and rate of change in quality indicators within each segment. Simultaneously, it retrieves construction event data, equipment operation records, and external environmental monitoring information for the corresponding time period to determine the cause of the anomalies, prioritizing whether the segmented anomaly time coincides with a BIM node. If there is overlap with changes in process status, it is attributed to a change in construction node; if there is no overlap, the synchronization between equipment operation records and sudden changes in physical quantities is assessed, and if an anomaly exists, it is attributed to equipment anomaly; if the equipment is normal, external environmental interference is further analyzed, and if an anomaly exists, it is attributed to external disturbance; if multiple factors coexist and cannot be distinguished, it is attributed to multi-source composite and unknown factors. A quantitative attribution labeling method is used to map the main cause category of the anomaly segment through digital factors; all attribution labels and corresponding segment numbers, real-time monitoring data, construction event data, and attribution process indexes are created and archived in the attribution database. S6, based on BIM structured data, real-time monitoring data, and attribution data, performs level determination and graded response through multi-source time-series item-by-item difference analysis.

2. The engineering quality inspection method based on green building design according to claim 1, characterized in that: The specific process for preprocessing the BIM structured data, real-time monitoring data, and structural acceptance data, as well as construction event and window data, is as follows: Collecting BIM structured data, which includes: process node parameters, component attribute parameters, green performance design parameters, and spatial geometry and connection parameters; Collecting real-time monitoring data, which includes: electricity consumption data, indoor CO2 concentration, PM2.5 concentration, temperature and humidity, formaldehyde content, VOC content, and renewable energy utilization rate; Collecting structural acceptance data, which includes: heat transfer coefficient of exterior walls and roofs, thickness of door and window insulation layers, leakage test data of roof and exterior wall waterproofing layers, and concrete strength grade data. Construction event and window data are acquired to establish a construction time database. This data includes material arrival events, rework events, anomaly alarm events, node change events, historical time-series sliding windows, work condition sliding windows, segmented sliding windows, and engineering unit windows. Through multi-frequency timestamp resampling and linear interpolation methods, the timestamps of real-time monitoring data, structural acceptance data, BIM model output data, and construction events are standardized and annotated, mapping different sampling frequencies and asynchronously reported data streams to a unified standard time-series axis. Using MAD rules combined with multivariate covariance analysis and pre-feature clustering algorithms, multi-parameter correlation analysis and outlier identification are performed on real-time monitoring data. Distribution standardization and linear normalization algorithms are used, employing z-score standardization based on the training set mean and variance, combined with linear normalization to standardize and normalize BIM structured data, real-time monitoring data, and construction event and window data.

3. The engineering quality inspection method based on green building design according to claim 1, characterized in that: The specific process of multi-source time-series item-by-item difference analysis based on BIM structured data, real-time monitoring data, and attribution assignment data is as follows: Energy consumption limits, air quality limits, compliance status, electricity consumption data, indoor CO2 concentration, PM2.5 concentration, anomaly attribution assignment, zoning quality grading values, importance level, geometric center 3D coordinates, and shortest path distance are obtained; the green performance detection deviation is obtained by item-by-item difference algorithm and weighted fusion for energy consumption limits, air quality limits, electricity consumption data, indoor CO2 concentration, and PM2.5 concentration; different risk scores are assigned to compliance status, zoning quality grading values, and importance level through interval mapping functions, and BIM risk coefficients are obtained through weighted fusion algorithm; the rate of change of detection deviation is obtained by applying an adjacent time-series difference algorithm to the green performance detection deviation; and BIM spatial correlation factors are obtained from the geometric center 3D coordinates and shortest path distance through structural topology analysis algorithm. Calculate the product of the green performance detection deviation and the attribution value of the abnormal working conditions, add the result to the BIM risk coefficient, and obtain the risk accumulation item; Calculate the time series difference between adjacent time points of the green performance detection deviation, take the absolute value and add one to obtain the denominator of the performance; Divide the risk accumulation term by the performance denominator term to obtain the risk value term; Calculate the spatial structure correlation value, multiply it by the spatial adjustment coefficient, and take the exponential function value to obtain the spatial adjustment term; multiply the risk value term by the spatial adjustment term, take the absolute value, traverse all times within the engineering unit window, and take the maximum value to obtain the engineering quality risk value.

4. The engineering quality inspection method based on green building design according to claim 1, characterized in that: The specific process of determining the level and responding in a graded manner is as follows: Real-time comparison of the engineering quality risk value and the engineering quality risk threshold, which includes a first-level risk threshold and a second-level risk threshold: When the engineering quality risk value is less than or equal to the second-level risk threshold, it is determined to be a green performance health zone, requiring only daily inspection and archiving of historical working condition data. When the engineering quality risk value is greater than the secondary risk threshold but less than or equal to the primary risk threshold, it is identified as a green performance concern area, and encrypted detection tasks are pushed to conduct key spot checks and reviews on energy consumption, air quality and renewable energy utilization rate. When the engineering quality risk value exceeds the first-level risk threshold, it is identified as a green performance risk zone, and a special rectification and re-inspection are immediately triggered. An engineering quality risk report is generated based on the attribution database data and fed back to the human team. The segment is listed as a key monitoring object, and all rectification records and re-inspection data are synchronously archived to the abnormal working condition database.

5. An engineering quality inspection system based on green building design, employing an engineering quality inspection method based on green building design as described in any one of claims 1-4, comprising: The data acquisition and preprocessing module is used to acquire BIM structured data, real-time monitoring data, and structural acceptance data, and to obtain construction events and window data. Preprocess BIM structured data, real-time monitoring data, structural acceptance data, construction events and window data; The quality indicator extraction and conformity determination module is used to determine the conformity status of components and the quality level of zones based on BIM structured data and BIM model. The dynamic operating condition segmentation and event detection module is used to identify operating condition switching anchor points and events from real-time monitoring data through mutation detection algorithms and feature time series clustering methods, and to segment the operating condition process. The intelligent clustering and discrimination module for working condition performance is used to perform feature clustering and performance status discrimination on segmented working condition data and structural acceptance data, identify different working condition types and manage them with labels. The working condition anomaly attribution and tracing module is used to identify the dominant factors causing fluctuations in engineering quality and assign them values ​​based on the type of working condition anomaly through multi-dimensional label attribution and tracing. The quality risk classification and closed-loop management module is used to determine the level and classify the response based on BIM structured data, real-time monitoring data and attribution data through multi-source time series item-by-item difference analysis.

Images