Intelligent robot construction progress and quality linkage analysis system based on BIM

Abstract

The invention relates to the technical field of data processing, in particular to a BIM (Building Information Modeling)-based intelligent robot construction progress and quality linkage analysis system, which comprises a data acquisition module for acquiring robot sensor data and environment monitoring data, a semantic fusion module for generating structured data with semantic tags through a railway station building ontology library and outputting abnormal flag bits, the time sequence twinborn module generates a digital twinborn new version and a change vector based on an abnormal flag bit, the space matching module adopts an incremental learning mechanism to optimize a coordinate registration strategy, and the coupling analysis module establishes a quantitative evaluation model of the influence of the quality defect on the construction period and generates a risk early warning signal; the decision optimization module simulates a disposal scheme in a virtual environment to output a Pareto optimal solution set, and the execution tracing module generates a machine instruction set and constructs an auditable operation chain. Dynamic linkage analysis of construction progress and quality is realized through cooperation of multiple modules, and the problem of response delay is solved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a BIM-based intelligent robot construction progress and quality linkage analysis system. Background Technology

[0002] Building Information Modeling (BIM) accurately describes the geometric and physical properties of buildings digitally, laying the data foundation for intelligent robot construction. BIM is an abbreviation for Building Information Modeling. It involves digitally modeling the physical and functional characteristics of a building, including geometric information, spatial relationships, geographic information, and component attributes, providing a comprehensive digital foundation for intelligent construction. Intelligent robots rely on the 3D model information provided by BIM for path planning and task execution, such as automated masonry or assembly. With built-in sensors and algorithms, they achieve high-precision operations, reducing human intervention and improving construction quality and efficiency. This technological integration promotes the collaboration and predictability of the construction process because robots can dynamically analyze model changes and adjust their actions, thereby optimizing the overall workflow.

[0003] Existing technologies suffer from the following pain points: During the intelligent construction of railway stations, the lack of unified data interaction standards and dynamic correlation mechanisms between the BIM standard model, real-time robot construction data, and the schedule planning system prevents construction quality monitoring results from automatically triggering schedule adjustment instructions. For example, during the deep foundation pit construction of Station A, the support structure monitoring robot detected that the foundation settlement rate exceeded the safety threshold, but the settlement data was not immediately mapped to the structural stress parameter standards in the BIM model. Simultaneously, the schedule management system continued the earthwork excavation process as planned until manual verification revealed the quality anomaly, forcing subsequent electromechanical pre-embedded processes to be interrupted and reworked, delaying key milestones by up to 72 hours. Another typical scenario involves the welding of large-span steel beams in hub stations. The weld offset data collected in real-time by the welding robot needs to be manually compared with the seismic load-bearing design tolerances in the BIM model. The failure to automatically trigger an early warning mechanism resulted in accumulated deviations only being discovered during the curtain wall installation stage, causing the curved curtain wall to exceed the fitting standard and requiring overall demolition and reconstruction. Such data chain breaks directly cause a disconnect between schedule control and quality management, significantly increasing the risk of chain-like construction in railway station projects. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a BIM-based intelligent robot construction progress and quality linkage analysis system, which solves the technical problem of delayed progress and quality linkage response in intelligent construction of railway stations due to inconsistent cross-system data interaction standards and the lack of dynamic correlation mechanisms.

[0005] To solve the above-mentioned technical problems, the specific contents of the present invention are as follows: In a first aspect, the BIM-based intelligent robot construction progress and quality linkage analysis system provided by the present invention includes: The data acquisition module is used to acquire three-dimensional spatial coordinates, operation timestamps, construction process parameters, and temperature and humidity changes, structural deformation, and point cloud scan data captured by the intelligent robot's body sensors. The semantic fusion module is connected to the data acquisition module. It receives the three-dimensional spatial coordinates, operation timestamps, construction process parameters, temperature and humidity changes, structural deformation, and point cloud scanning data. It parses the data attributes through the railway station building ontology library, associates the sensor data with the quality indicators of specific components in the BIM model, generates structured data with semantic tags including component ID, indicator type, measured value, and threshold, and outputs anomaly flag bits. The temporal twin module is connected to the semantic fusion module. It receives the structured data with semantic tags and anomaly flags. When the anomaly flags indicate a quality anomaly, it injects a time dimension attribute into the BIM geometric model, generates a new version of the digital twin that records the sequence of component state changes, and outputs a change vector that represents the change of component state. The spatial matching module is connected to the temporal twin module. It receives the new version of the digital twin and the change vector. Based on the change vector and the new version of the digital twin, it optimizes the spatial matching strategy using an incremental learning mechanism, completes the registration of BIM component coordinates and robot coordinates, and outputs the component-equipment and data source mapping matrix. The coupling analysis module is connected to the spatial matching module, receives the component and equipment and data source mapping matrix, establishes a quantitative assessment model of the impact of quality defects on the project period, calculates the probability of subsequent process delays caused by quality defects and resource conflict detection results based on the quality anomaly data in the component and equipment and data source mapping matrix, and generates risk warning signals. The decision optimization module, connected to the coupled analysis module, receives the risk warning signal, simulates multiple disposal schemes in parallel in a virtual construction environment, evaluates the comprehensive impact of each disposal scheme on the total project duration and cost, and outputs the Pareto optimal solution set. The execution traceability module, connected to the decision optimization module, receives the Pareto optimal solution set, automatically generates a set of machine-executable instructions, adjusts the robot's process parameters and the logistics system's delivery priority, and constructs a digital thread to anchor key data of the disposal process to specific version nodes of the digital twin, forming an auditable operation chain, thereby solving the problem of delayed response in the linkage between progress and quality.

[0006] Furthermore, the data acquisition module of the BIM-based intelligent robot construction progress and quality linkage analysis system provided by the present invention includes: The robot sensor unit uses a fusion positioning system of lidar and inertial measurement unit to record the three-dimensional coordinates of the actuator at the working end in the world coordinate system in real time, and generates a spatial location data packet including a timestamp when each single operation is completed. The environmental monitoring unit independently deploys a micro-weather station to collect air temperature and humidity gradients at a first preset interval, and deploys a structural monitoring unit to capture strain fluctuations of the steel support frame at a second preset frequency. At the same time, a laser scanning station is deployed to emit a scanning beam through a rotating prism system and receive reflected signals to generate point cloud data frames. The spatiotemporal registration unit connects the robot sensor unit and the environmental monitoring unit, establishes a unified spatiotemporal reference axis, inputs the robot trajectory point cloud and the laser scanning point cloud into a feature matching algorithm, extracts feature points from the two types of point clouds for similarity calculation, eliminates coordinate system deviations between devices through an optimization algorithm, and outputs spatiotemporally aligned fused data as a snapshot of the construction status.

[0007] Furthermore, the semantic fusion module of the BIM-based intelligent robot construction progress and quality linkage analysis system provided by the present invention includes: The ontology library construction unit is based on the tree-structured semantic association rules that define component types, construction processes, and quality indicators according to railway station building engineering specifications. The dynamic mapping unit is connected to the ontology library construction unit. It receives three-dimensional spatial coordinates, operation timestamps, construction process parameters, temperature and humidity changes, structural deformation, and point cloud scan data from the data acquisition module. It queries the component IDs bound to the current operation robot, matches index items and thresholds according to the ontology library, and outputs structured data with component IDs, index types, measured values, thresholds, and anomaly flags. The data output unit is connected to the dynamic mapping unit and transmits the structured data to the time-series twin module.

[0008] Furthermore, the time-series twin module of the BIM-based intelligent robot construction progress and quality linkage analysis system provided by the present invention includes: The version generation unit imports the BIM model from the design phase as the initial version. After the process is completed, it generates a point cloud model through laser scanning, generates triangular mesh patches through surface reconstruction algorithm, compares the deviation vector between the design model and the measured model, generates a new version and records the change log. The prediction evolution unit, connected to the version generation unit, automatically starts the new version modeling process when the abnormal flag output by the semantic fusion module indicates that the displacement of the component continues to increase. It extracts the most recent scan data to establish a time series model, predicts future deformations, and generates a twin version containing the predicted state. The state chain management unit, connected to the prediction evolution unit, associates the version sequence as a timeline and supports historical state backtracking.

[0009] Furthermore, in the BIM-based intelligent robot construction progress and quality linkage analysis system provided by the present invention, the spatial matching module includes: The initial matching unit is rigidly aligned with the robot's GPS coordinates by the center coordinates of the BIM component, and the average positioning error is calculated. The learning optimization unit is connected to the initial matching unit, records the robot's typical operation path, and establishes a path feature library, which includes amplitude frequency and turning angle patterns. The prediction compensation unit, connected to the learning optimization unit, calls the current working area feature mode when the GPS signal is lost, combines the motion vector to predict the next position, calculates the coordinate offset compensation amount through point cloud feature points, and updates the spatial matching strategy of the spatial matching module.

[0010] Furthermore, the BIM-based intelligent robot construction progress and quality linkage analysis system provided by the present invention includes a coupling analysis module comprising: The input receiving unit receives the component and device mapping matrix and the data source mapping matrix, and identifies abnormal quality data. The historical retrieval unit is connected to the input receiving unit and retrieves the average repair time of similar quality anomaly cases in the stored historical data; The conflict detection unit is connected to the historical retrieval unit and detects resource conflicts in subsequent processes. The resource conflicts include conflicts between work group entry time and conflicts between equipment occupancy time. The probability calculation unit, connected to the conflict detection unit, calculates the critical path delay probability and generates a risk heat map to indicate the impact range.

[0011] Furthermore, the decision optimization module of the BIM-based intelligent robot construction progress and quality linkage analysis system provided by the present invention includes: The environment construction unit imports the current digital twin version generated by the time-series twin module and loads resource constraints, including machine scheduling rules and material supply conditions. The scheme simulation unit is connected to the environment construction unit and simulates multiple disposal schemes in parallel. The disposal schemes include a work stoppage detection scheme, a speed reduction operation scheme, and a backup team activation scheme. The virtual construction period and cost of each disposal scheme are output. The optimization screening unit is connected to the scheme simulation unit to evaluate the comprehensive impact of each disposal scheme on the total project duration and cost, eliminate inferior schemes, and generate a Pareto optimal solution set.

[0012] Furthermore, the execution traceability module of the BIM-based intelligent robot construction progress and quality linkage analysis system provided by the present invention includes: The instruction generation unit receives the Pareto optimal solution set, analyzes the resource requirements in the selected solution, and generates logistics system instructions and robot parameter adjustment instructions. An execution recording unit, connected to the instruction generation unit, records the instruction issuance time and execution confirmation time; The digital thread unit, connected to the execution record unit, anchors key data of the processing to a specific version node of the digital twin, and establishes an auditable operation chain including event ID, associated twin version number, decision scheme fingerprint, execution log and effect verification data.

[0013] Furthermore, the BIM-based intelligent robot construction progress and quality linkage analysis system provided by the present invention further includes a feedback closed-loop module connecting the execution traceability module and the data acquisition module. The feedback closed-loop module is used for: Receive new data acquired by the data acquisition module after the instruction is executed, the new data including quality recovery status data and progress data; The new data is sent to the semantic fusion module to generate structured data in a normal state; The time-series twin module is triggered to generate a new version record verification result; Update the spatial matching strategy of the spatial matching module and the decision model weight coefficients of the decision optimization module.

[0014] Furthermore, the BIM-based intelligent robot construction progress and quality linkage analysis system provided by the present invention also includes: The data flow between the data acquisition module, semantic fusion module, temporal twin module, spatial matching module, coupling analysis module, decision optimization module, and execution tracing module forms a closed loop: The data acquisition module outputs spatiotemporally aligned fused data as a snapshot of the construction status to the semantic fusion module; The semantic fusion module outputs structured data with semantic tags and anomaly flags to the temporal twin module; The temporal twin module outputs a new version of the digital twin and a change vector representing the change in the state of the component to the spatial matching module; The spatial matching module outputs a component-device and data source mapping matrix to the coupling analysis module; The coupling analysis module outputs a risk warning signal to the decision optimization module; The decision optimization module outputs the Pareto optimal solution set to the execution traceability module; The execution traceability module outputs the machine-executable instruction set and auditable operation chain, and feeds back the execution data to the data acquisition module and the update information to the spatial matching module, thus completing the system's self-evolution.

[0015] Secondly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements a BIM-based intelligent robot construction progress and quality linkage analysis system.

[0016] Beneficial effects of this invention; The BIM-based intelligent robot construction progress and quality linkage analysis system provided by this invention effectively solves the technical problem of delayed response in the linkage between progress and quality in the intelligent construction of railway stations by constructing a closed-loop data flow and dynamic correlation mechanism. The system first acquires multi-source heterogeneous sensor data through a data acquisition module and outputs spatiotemporally aligned fused data through a spatiotemporal registration unit, providing a unified benchmark for subsequent processing. The semantic fusion module uses a railway station domain ontology library to parse data attributes, generate structured data with semantic tags, and output anomaly flags, establishing a machine-understandable association between the original monitoring data and BIM quality indicators. The temporal twin module injects time dimension attributes when the anomaly flags indicate quality anomalies, generating a new version of the digital twin and a change vector, supporting historical backtracking and future prediction. The spatial matching module employs… The incremental learning mechanism optimizes the matching strategy, completes the registration of BIM component coordinates with robot coordinates, and outputs a mapping matrix of components, equipment, and data sources. The coupling analysis module establishes a quantitative evaluation model based on the quality anomaly data in the mapping matrix, calculates the probability of subsequent process delays and resource conflict detection results, and generates risk warning signals. The decision optimization module simulates multiple disposal schemes in parallel in a virtual environment, outputs a Pareto optimal solution set to provide multi-objective optimization choices. The execution traceability module automatically generates a set of machine-executable instructions and constructs an auditable operation chain, while realizing system self-evolution through a feedback closed-loop module. The entire system forms a closed loop through data flow between modules, realizing unified data interaction standards and establishing a dynamic correlation mechanism, thereby achieving real-time linkage response between progress and quality, eliminating response delays, and improving construction efficiency and quality control levels. Attached Figure Description

[0017] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on the drawings without creative effort.

[0018] Figure 1 The system architecture diagram of the BIM-based intelligent robot construction progress and quality linkage analysis system provided in the embodiments of the present invention is shown. Detailed Implementation

[0019] To make the technical solution of the present invention clearer, the present invention will be clearly and completely described below with reference to specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. 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. The present invention provided by various embodiments will be described in detail below with reference to the accompanying drawings. To better understand the purpose of the present invention, the present invention will be described in further detail below.

[0020] Firstly, please refer to Figure 1 The present invention provides a BIM-based intelligent robot construction progress and quality linkage analysis system, comprising: The data acquisition module is used to acquire three-dimensional spatial coordinates, operation timestamps, construction process parameters, and temperature and humidity changes, structural deformation, and point cloud scan data captured by the intelligent robot's body sensors. The data acquisition module integrates multiple sensor units and data processing units to collect and initially fuse multi-source heterogeneous data during the intelligent robot's construction process. The robot's sensor unit has built-in lidar and inertial measurement units, and uses fusion positioning technology to acquire the three-dimensional spatial coordinates of the actuator in the world coordinate system in real time. It also automatically generates a spatial position data packet with a precise timestamp after each operation, thereby recording the robot's motion trajectory and operation sequence.

[0021] The environmental monitoring unit is independently deployed in the construction area, including a micro-weather station that collects air temperature and humidity gradient data at fixed intervals, a structural monitoring unit that captures strain fluctuations of the steel support frame at a specific frequency to reflect the amount of structural deformation, and a laser scanning station that emits scanning beams through a rotating prism system and receives reflected signals to generate high-density point cloud scanning data, thereby comprehensively monitoring construction environmental parameters and structural status.

[0022] The spatiotemporal registration unit connects the robot sensor unit and the environmental monitoring unit, establishing a unified spatiotemporal reference axis based on the time of the Global Navigation Satellite System. This unit inputs the robot trajectory point cloud and the laser scan point cloud into a feature descriptor-based matching algorithm, extracts corner and edge feature points from the two types of point clouds for similarity calculation, eliminates coordinate system deviations between devices through an iterative nearest-point optimization algorithm, and outputs spatiotemporally aligned fused data as a snapshot of the construction status, providing standardized input for subsequent modules.

[0023] The data flow logic between the units is as follows: the robot sensor unit and the environmental monitoring unit collect raw data in parallel; the spatiotemporal registration unit receives this data and performs spatiotemporal alignment processing; finally, the fused data is output. This design ensures the consistency of multi-source data in time and space, providing a reliable data foundation for subsequent semantic parsing and digital twin construction.

[0024] The semantic fusion module is connected to the data acquisition module. It receives the three-dimensional spatial coordinates, operation timestamps, construction process parameters, temperature and humidity changes, structural deformation, and point cloud scanning data. It parses the data attributes through the railway station building ontology library, associates the sensor data with the quality indicators of specific components in the BIM model, generates structured data with semantic tags including component ID, indicator type, measured value, and threshold, and outputs anomaly flag bits. The semantic fusion module achieves semantic transformation of raw monitoring data into quality information through a multi-level processing flow. The module receives six types of monitoring data from the data acquisition module, including three-dimensional spatial coordinates, operation timestamps, construction process parameters, temperature and humidity changes, structural deformation, and point cloud scan data. This data is input into the dynamic mapping unit through a unified data interface to maintain spatiotemporal consistency during data acquisition.

[0025] The ontology library construction unit establishes a machine-readable semantic knowledge network based on railway station building engineering specifications. This unit defines the inheritance relationship between component types and construction methods, constructing a tree structure of quality indicators, where leaf nodes include indicator name, unit of measurement, design threshold, and testing method attributes. Semantic association rules explicitly define the mapping relationship between construction techniques and quality indicators, forming a domain-specific quality evaluation system.

[0026] The dynamic mapping unit employs a multi-level matching algorithm to process the input data. The unit first parses the construction process parameters to determine the current work sequence, and then retrieves the associated component set from the semantic network based on the sequence type. For point cloud scan data, the unit calls a surface fitting algorithm to calculate the actual geometric parameters of the components and matches them with the corresponding quality index items in the semantic network. When the monitored value of structural deformation exceeds the design threshold, the unit generates an anomaly flag and records the deviation value. The output data uses a standardized structure including component ID, index type, measured value, and threshold fields.

[0027] The data output unit performs lightweight serialization processing on structured data, adding a timestamp and data source information to the data packet header. The unit transmits the data packets to the time-series twin module via a message broker, employing a verification mechanism to ensure data integrity during transmission. An exception flag is embedded as an independent field at the end of the data packet, providing a basis for subsequent modules to determine the quality status.

[0028] Each processing unit executes sequentially within a pipeline architecture: the ontology building unit provides static knowledge base support, the dynamic mapping unit completes real-time data parsing and quality evaluation, and the data output unit is responsible for data encapsulation and transmission. This design ensures that the semantic parsing process is synchronized with the data acquisition rhythm, providing high-quality input for the construction of the digital twin.

[0029] The temporal twin module is connected to the semantic fusion module. It receives the structured data with semantic tags and anomaly flags. When the anomaly flags indicate a quality anomaly, it injects a time dimension attribute into the BIM geometric model, generates a new version of the digital twin that records the sequence of component state changes, and outputs a change vector that represents the change of component state. The temporal twin module achieves time-dimensional expansion and version-based management of construction status through multi-stage processing. The module receives structured data with semantic tags and anomaly flags output by the semantic fusion module. The structured data includes component ID, index type, measured value and threshold information, and the anomaly flags indicate the component quality status.

[0030] When the anomaly flag indicates a quality anomaly, the module activates the version generation process. The process first injects a time dimension attribute into the static BIM geometric model. The time dimension attribute is derived from the work timestamp in the structured data. The injection process binds the timestamp to the spatial coordinates of the component, constructing a four-dimensional spatiotemporal coordinate system.

[0031] Based on a four-dimensional spatiotemporal coordinate system, the module generates a new version of a digital twin that records the sequence of component state changes. The new version generates a change vector, including parameters such as position offset and deformation, by comparing the geometric deviation between the current measured state and historical versions. This change vector quantifies the magnitude and direction of the component's state change from the previous version to the current version.

[0032] The version generation unit performs the following operations: imports the BIM model from the design phase as the initial version; acquires laser-scanned point cloud data after the completion of a new process; converts the point cloud into a measured model represented by triangular mesh patches using a surface reconstruction algorithm; performs spatial registration with the design model; and calculates the Euclidean distance deviation between corresponding vertices. When the deviation exceeds a preset threshold, a new version number is generated, and the deviation value, timestamp, and affected area are recorded in the change log.

[0033] Each processing step forms a logical closed loop: anomaly flags trigger the injection of the time dimension, the time dimension supports the recording of state transition sequences, state transitions generate a new version twin, and the new version outputs a change vector. The change vector, as the core data for quantifying component state changes, provides dynamically updated geometric feature inputs for the subsequent spatial matching module.

[0034] The spatial matching module is connected to the temporal twin module. It receives the new version of the digital twin and the change vector. Based on the change vector and the new version of the digital twin, it optimizes the spatial matching strategy using an incremental learning mechanism, completes the registration of BIM component coordinates and robot coordinates, and outputs the component-equipment and data source mapping matrix. The spatial matching module achieves dynamic matching between the building information model and the physical space through a progressive optimization process. The module receives the new version of the digital twin and change vectors output by the temporal twin module. The new version of the digital twin includes the updated geometric attributes of the components, and the change vectors represent the changes in the position and shape of the components.

[0035] The initial matching unit first performs basic spatial registration. This unit reads the component center coordinates from the new version of the digital twin and aligns them with the robot's global positioning system coordinates using a rigid transformation. The alignment process uses the least squares method to calculate the rotation and translation matrix. After the transformation, the Euclidean distance of each matching point is calculated, and the average positioning error value is output. The average positioning error is used as a benchmark parameter for the initial matching accuracy and input to subsequent units.

[0036] The learning optimization unit initiates path feature analysis based on the average positioning error. The unit continuously collects pose data of the robot's actual operating path, extracts path amplitude frequency features through Fourier transform, calculates path curvature using differential geometry methods, and statistically analyzes the distribution of turning angles. The analysis results are normalized to form a path feature library, which stores feature patterns by dividing the construction area into grid cells.

[0037] The predictive compensation unit enables dynamic spatial correction. When the GPS signal strength falls below the operating threshold, the unit indexes the path feature library based on the robot's current position and calls upon the corresponding grid's regional feature pattern. Combining the motion vector output in real time by the inertial measurement unit, a kinematic model is used to predict the theoretical coordinates for the next moment. Simultaneously, structural feature points in the environmental point cloud are scanned, and the offset between the actual and predicted coordinates is calculated using a point cloud registration algorithm to generate real-time coordinate offset compensation.

[0038] The module employs an incremental learning mechanism to update the spatial matching strategy. The offset compensation amount generated during each prediction and compensation process is recorded in the strategy database, and the spatial transformation parameters are dynamically corrected using a weighted average algorithm. The final output is a mapping matrix between components, equipment, and data sources. This matrix records the correspondence between component numbers, construction equipment identifiers, and data acquisition sources in a two-dimensional table format, establishing a traceability path for equipment-level quality data for coupled analysis.

[0039] The coupling analysis module is connected to the spatial matching module, receives the component and equipment and data source mapping matrix, establishes a quantitative assessment model of the impact of quality defects on the project period, calculates the probability of subsequent process delays caused by quality defects and resource conflict detection results based on the quality anomaly data in the component and equipment and data source mapping matrix, and generates risk warning signals. The coupling analysis module uses a multi-step analysis process to quantitatively assess the impact of quality defects on construction progress. The module receives a component-equipment and data source mapping matrix output from the spatial matching module. This matrix records the ternary correspondence between component number, construction equipment identifier, and data acquisition source in a two-dimensional relational table format.

[0040] The input receiving unit parses the structured quality data in the mapping matrix and uses a rule engine to identify quality anomaly records. The identification process judges the status based on the threshold range of quality indicators. When the measured value of a component continuously exceeds the design threshold, it is marked as a valid quality anomaly event, and feature vectors of the abnormal component location, deviation magnitude, and detection timestamp are extracted.

[0041] The historical retrieval unit performs similar case matching based on the feature vectors of quality anomaly events. The unit calculates the cosine similarity between the current event features and the case features in the historical database, filters out the case set with similarity higher than a set threshold, extracts the repair time data from the cases, calculates the arithmetic mean, and generates the average repair time of similar quality anomaly cases.

[0042] The conflict detection unit performs resource conflict scanning based on the average repair time. The unit accesses the process logic relation library of the schedule planning system to locate critical subsequent processes affected by abnormal components. Based on the average repair time, it extrapolates the changes in process time windows and performs time window overlap analysis with the shift entry plans in the resource scheduling library to detect shift entry time conflict events. Simultaneously, it compares the data with the equipment occupancy plan to identify equipment occupancy time conflict events, generating a resource conflict report including conflict type and duration.

[0043] The probability calculation unit performs risk quantification modeling. The unit integrates resource conflict reports and critical path models, employing a Monte Carlo simulation algorithm to iteratively calculate the probability of critical path delays. The simulation process considers variables such as repair time volatility and resource adjustment flexibility, outputting probability distribution data to drive the risk heatmap generation module. This ultimately generates a red, yellow, and green risk heatmap covering the construction area, with contour lines indicating the boundaries of different risk levels.

[0044] The decision optimization module, connected to the coupled analysis module, receives the risk warning signal, simulates multiple disposal schemes in parallel in a virtual construction environment, evaluates the comprehensive impact of each disposal scheme on the total project duration and cost, and outputs the Pareto optimal solution set. The decision optimization module achieves scientific decision-making on disposal plans through virtual simulation and multi-objective optimization. The module receives risk warning signals output by the coupled analysis module, which include predicted data on resource conflict types and impact ranges. The environment construction unit first imports the current digital twin version generated by the temporal twin module, which fully reflects the spatial geometric attributes and physical parameters of the construction scenario. Simultaneously, it loads a resource constraint library, where machine scheduling rules define the maximum continuous working time and switching cooldown time of equipment, and material supply conditions specify key building material inventory thresholds and procurement lead times, constructing a virtual construction environment with a complete constraint system.

[0045] The simulation unit executes discrete event simulations of three handling strategies in parallel within a virtual environment. The shutdown inspection scheme simulates pausing the current process and then initiating a non-destructive testing process, calculating the quality re-inspection time and the cost of idle resources during downtime. The speed-down operation scheme reduces the robot's operating speed by a preset ratio, extrapolating the cascading impact of increased unit working hours on subsequent processes and calculating the incremental costs of labor and equipment usage. The backup shift activation scheme simulates a multi-shift collaborative scenario after utilizing reserve human resources, quantifying the time-compression effect and additional management costs brought about by parallel operations. Each simulation outputs the virtual total project duration and comprehensive cost value for the corresponding scheme, forming the original solution set.

[0046] The optimization and screening unit employs a multi-objective optimization algorithm to process the original solution set. The unit establishes a two-dimensional evaluation coordinate system of schedule and cost, mapping the simulation results of each scheme to discrete points in the coordinate system. A non-dominated sorting algorithm is used to identify the Pareto front solution set, automatically eliminating inferior schemes whose schedule and cost both exceed those of other schemes. An inferior scheme is defined as a non-boundary scheme that fails to achieve optimality in both schedule and cost dimensions. The final output is a Pareto optimal solution set, which includes multiple non-inferior solutions and their schedule and cost data, providing decision-makers with a choice of schemes that balance efficiency and economy.

[0047] The execution traceability module, connected to the decision optimization module, receives the Pareto optimal solution set, automatically generates a set of machine-executable instructions, adjusts the robot's process parameters and the logistics system's delivery priority, and constructs a digital thread to anchor key data of the disposal process to specific version nodes of the digital twin, forming an auditable operation chain, thereby solving the problem of delayed response in the linkage between progress and quality.

[0048] The execution traceability module achieves precise conversion from decision-making schemes to physical execution through instruction conversion and data anchoring mechanisms. The module receives the Pareto optimal solution set output by the decision optimization module, which includes multiple non-dominated solutions and their time and cost data. The instruction generation unit parses the resource requirement details of the selected scheme, including the required material codes, quantities, and target warehouse location information, and generates warehousing and allocation instructions for the logistics system. Simultaneously, it extracts robot operation parameter adjustment items, generating a robot control instruction set including speed adjustment coefficients and path replanning parameters, forming a complete set of machine-executable instructions.

[0049] The execution recording unit uses the time source provided by the time synchronization server as a reference to record the precise timestamp of each instruction issued by the instruction generation unit. When the logistics management system and the robot control system return an instruction reception confirmation signal, the unit captures and records the execution confirmation time, forming an execution event log that includes the instruction type, issuance time, and confirmation time.

[0050] The digital thread unit constructs an auditable operation chain based on a cryptographic hash chain. The unit extracts the current activity version number of the time-series twin module as an anchor point and encapsulates key data from the handling process into data blocks. A unique event identifier is written to the header of each data block, and the body embeds the associated twin version number, decision scheme digest hash value, execution event log, and quality review data. The quality review data originates from the anomaly flag state change records of the semantic fusion module. The operation chain establishes a bidirectional association with the digital twin nodes through a version number index, supporting the retrieval of the complete handling trajectory by event identifier.

[0051] The workflow of each unit forms a logical closed loop: the instruction generation unit converts decision-making schemes into machine instructions, the execution record unit establishes instruction lifecycle tracking, and the digital thread unit completes full-process data anchoring. The operation chain uses blockchain storage technology to ensure immutability, and event identifiers serve as retrieval keys to support full-link traceability, ultimately solving the problem of delayed response in the linkage between progress and quality.

[0052] The BIM-based intelligent robot construction progress and quality linkage analysis system achieves linkage analysis of progress and quality through the collaborative work of multiple modules. The data acquisition module first acquires the three-dimensional spatial coordinates, operation timestamps, construction process parameters collected by the intelligent robot's sensors, and temperature and humidity changes, structural deformation, and point cloud scan data captured by the environmental monitoring unit. The data acquisition module includes a robot sensor unit, an environmental monitoring unit, and a spatiotemporal registration unit. The robot sensor unit uses a fusion positioning system of lidar and inertial measurement unit to record the three-dimensional coordinates of the actuator in the world coordinate system in real time, and generates a spatial location data packet including a timestamp after each operation. The environmental monitoring unit independently deploys a micro-weather station to collect air temperature and humidity gradients at preset intervals, and deploys a structural monitoring unit to capture strain fluctuations of the steel support frame at a preset frequency. Simultaneously, a laser scanning station is deployed to emit a scanning beam through a rotating prism system and receive reflected signals to generate point cloud data frames. The spatiotemporal registration unit connects the robot sensor unit and the environmental monitoring unit, establishes a unified spatiotemporal reference axis, inputs the robot trajectory point cloud and the laser scanning point cloud into the feature matching algorithm, extracts the feature points in the two types of point clouds for similarity calculation, eliminates the coordinate system deviation between devices through the optimization algorithm, and outputs spatiotemporally aligned fused data as a snapshot of the construction status.

[0053] The semantic fusion module connects to the data acquisition module, receiving 3D spatial coordinates, operation timestamps, construction process parameters, temperature and humidity changes, structural deformation, and point cloud scan data. The semantic fusion module includes an ontology library construction unit, a dynamic mapping unit, and a data output unit. The ontology library construction unit defines a tree-structured semantic association rule based on railway station building engineering specifications, defining component types, construction methods, and quality indicators. The dynamic mapping unit connects to the ontology library construction unit, queries the component ID bound to the currently operating robot, matches indicator items and thresholds according to the ontology library, and outputs structured data with component ID, indicator type, measured value, threshold, and anomaly flag. The data output unit connects to the dynamic mapping unit, transmitting the structured data to the temporal twin module. This process associates sensor data with the quality indicators of specific components in the BIM model, generating structured data with semantic tags and outputting anomaly flags.

[0054] The temporal twin module connects to the semantic fusion module, receiving structured data with semantic tags and anomaly flags. When the temporal twin module detects anomaly flags indicating quality anomalies, it injects time-dimensional attributes into the BIM geometric model, generates a new version of the digital twin recording the sequence of component state changes, and outputs a change vector representing the component state changes. The temporal twin module includes a version generation unit, a prediction evolution unit, and a state chain management unit. The version generation unit imports the BIM model from the design phase as the initial version. After the process is completed, it generates a point cloud model through laser scanning, generates triangular mesh patches through a surface reconstruction algorithm, compares the deviation vectors between the design model and the measured model, generates a new version, and records the change log. The prediction evolution unit connects to the version generation unit. When the anomaly flags output by the semantic fusion module indicate that the component displacement is continuously increasing, it automatically starts the new version modeling process, extracts the most recent scan data to build a time series model, predicts future deformations, and generates a twin version containing the predicted state. The state chain management unit connects to the prediction evolution unit, associating the version sequence as a timeline, supporting historical state backtracking.

[0055] The spatial matching module connects to the temporal twin module, receiving new versions of the digital twin and change vectors. Based on the change vectors and the new version of the digital twin, the spatial matching module uses an incremental learning mechanism to optimize the spatial matching strategy, complete the registration of BIM component coordinates with robot coordinates, and output a mapping matrix of components, equipment, and data sources. The spatial matching module includes an initial matching unit, a learning optimization unit, and a prediction compensation unit. The initial matching unit rigidly aligns the center coordinates of the BIM component with the robot's GPS coordinates and calculates the average positioning error. The learning optimization unit connects to the initial matching unit, records typical robot operation paths, and establishes a path feature library, which includes amplitude frequency and turning angle patterns. The prediction compensation unit connects to the learning optimization unit. When the GPS signal is lost, it calls the current operation area feature pattern, combines it with the motion vector to predict the next position, and calculates the coordinate offset compensation amount through point cloud feature points to update the spatial matching strategy of the spatial matching module.

[0056] The coupling analysis module connects to the spatial matching module, receiving the component-equipment and data source mapping matrix. The coupling analysis module establishes a quantitative assessment model of the impact of quality defects on the project schedule. Based on quality anomaly data in the component-equipment and data source mapping matrix, it calculates the probability of subsequent process delays caused by quality anomalies and resource conflict detection results, generating risk warning signals. The coupling analysis module includes an input receiving unit, a historical retrieval unit, a conflict detection unit, and a probability calculation unit. The input receiving unit receives the component-equipment and data source mapping matrix and identifies quality anomaly data. The historical retrieval unit connects to the input receiving unit and retrieves the average repair time of similar quality anomaly cases from stored historical data. The conflict detection unit connects to the historical retrieval unit and detects resource conflicts in subsequent processes, including conflicts in team arrival time and equipment occupancy time. The probability calculation unit connects to the conflict detection unit, calculates the critical path delay probability, and generates a risk heatmap to indicate the impact range.

[0057] The decision optimization module connects to the coupling analysis module and receives risk warning signals. In a virtual construction environment, the decision optimization module simulates multiple disposal plans in parallel, evaluates the comprehensive impact of each plan on the total project duration and cost, and outputs a Pareto optimal solution set. The decision optimization module includes an environment construction unit, a plan simulation unit, and an optimization screening unit. The environment construction unit imports the current digital twin version generated by the time-series twin module and loads resource constraints, including machinery scheduling rules and material supply conditions. The plan simulation unit connects to the environment construction unit and simulates multiple disposal plans in parallel, including shutdown and inspection plans, speed-up operation plans, and plans to activate backup work teams, outputting the virtual project duration and cost for each plan. The optimization screening unit connects to the plan simulation unit, evaluates the comprehensive impact of each plan on the total project duration and cost, eliminates suboptimal plans, and generates a Pareto optimal solution set.

[0058] The execution traceability module connects to the decision optimization module and receives the Pareto optimal solution set. It automatically generates a set of executable machine instructions, adjusts robot process parameters and logistics system delivery priorities, and constructs a digital thread to anchor key data from the disposal process to specific version nodes of the digital twin, forming an auditable operation chain. The execution traceability module includes an instruction generation unit, an execution recording unit, and a digital thread unit. The instruction generation unit receives the Pareto optimal solution set, analyzes the resource requirements of the selected solution, and generates logistics system instructions and robot parameter adjustment instructions. The execution recording unit connects to the instruction generation unit and records the instruction issuance time and execution confirmation time. The digital thread unit connects to the execution recording unit, anchors key data from the disposal process to specific version nodes of the digital twin, and establishes an auditable operation chain including event ID, associated twin version number, decision solution fingerprint, execution log, and effect verification data.

[0059] The system also includes a feedback loop module, connecting the execution traceability module and the data acquisition module. The feedback loop module receives new data collected by the data acquisition module after instruction execution; this new data includes quality recovery status data and progress data. The feedback loop module sends the new data to the semantic fusion module to generate structured data for the normal state. The feedback loop module triggers the time-series twin module to generate a new version record verification result. The feedback loop module updates the spatial matching strategy of the spatial matching module and the decision model weight coefficients of the decision optimization module. The data flow of the entire system forms a closed loop. The data acquisition module outputs spatiotemporally aligned fused data to the semantic fusion module. The semantic fusion module outputs structured data with semantic tags and anomaly flags to the temporal twin module. The temporal twin module outputs the new version of the digital twin and change vectors to the spatial matching module. The spatial matching module outputs the component-device and data source mapping matrix to the coupling analysis module. The coupling analysis module outputs risk warning signals to the decision optimization module. The decision optimization module outputs the Pareto optimal solution set to the execution traceability module. The execution traceability module outputs the machine-executable instruction set and auditable operation chain, and feeds back the execution data to the data acquisition module and the update information to the spatial matching module, thus completing the system's self-evolution.

[0060] Specifically, the data acquisition module of the BIM-based intelligent robot construction progress and quality linkage analysis system provided by the present invention includes: The robot sensor unit uses a fusion positioning system of lidar and inertial measurement unit to record the three-dimensional coordinates of the actuator at the working end in the world coordinate system in real time, and generates a spatial location data packet including a timestamp when each single operation is completed. The environmental monitoring unit independently deploys a micro-weather station to collect air temperature and humidity gradients at a first preset interval, and deploys a structural monitoring unit to capture strain fluctuations of the steel support frame at a second preset frequency. At the same time, a laser scanning station is deployed to emit a scanning beam through a rotating prism system and receive reflected signals to generate point cloud data frames. The spatiotemporal registration unit connects the robot sensor unit and the environmental monitoring unit, establishes a unified spatiotemporal reference axis, inputs the robot trajectory point cloud and the laser scanning point cloud into a feature matching algorithm, extracts feature points from the two types of point clouds for similarity calculation, eliminates coordinate system deviations between devices through an optimization algorithm, and outputs spatiotemporally aligned fused data as a snapshot of the construction status.

[0061] The data acquisition module constructs a complete digital image of the construction status through the collaborative acquisition and fusion processing of multi-source heterogeneous sensors. The robot sensor unit adopts a positioning scheme combining lidar and inertial measurement unit. The lidar acquires high-precision distance information of the surrounding environment by emitting laser beams and receiving reflected signals, while the inertial measurement unit continuously tracks the angular velocity and linear acceleration of the actuator at the working end through gyroscopes and accelerometers. The two types of data are input into a Kalman filter algorithm for fusion and calculation, and the three-dimensional coordinates of the actuator at the working end in the world coordinate system are output in real time. After each operation is completed, the system automatically generates a spatial location data packet including an international standard timestamp with millisecond-level accuracy, providing a time reference for subsequent time series analysis.

[0062] The environmental monitoring unit independently deploys multiple dedicated sensors to form a distributed monitoring network. Micro-weather stations collect air temperature and humidity gradient data at different altitudes within the construction area according to preset sampling intervals, and the monitoring data is transmitted to the central processor via a wireless transmission module. The structural monitoring unit uses a Wheatstone bridge circuit composed of resistance strain gauges to collect micro-strain fluctuation data of the steel support frame at a specific frequency; the strain data is converted into digital signals via an analog-to-digital converter. The laser scanning station emits a scanning beam through a high-speed rotating prism system, calculates the distance value based on the time difference of the received reflected signals, and generates a high-density point cloud data frame, which includes millions of three-dimensional coordinate points and their reflection intensity information.

[0063] The spatiotemporal registration unit establishes a unified spatiotemporal reference axis based on the BeiDou Navigation Satellite System time. This unit receives robot trajectory point clouds from the robot sensor unit and laser scan point clouds from the environmental monitoring unit, inputting the two types of point cloud data into a feature descriptor-based matching algorithm. The algorithm first extracts salient feature points such as corners and edges from the two types of point clouds, then calculates the similarity matrix between feature points, and finally eliminates coordinate system deviations between different devices through an iterative nearest-point optimization algorithm. The output is a set of spatiotemporally aligned fused data, which serves as a complete snapshot representing the current construction status and provides standardized input for subsequent modules.

[0064] Specifically, the semantic fusion module of the BIM-based intelligent robot construction progress and quality linkage analysis system provided by the present invention includes: The ontology library construction unit is based on the tree-structured semantic association rules that define component types, construction processes, and quality indicators according to railway station building engineering specifications. The dynamic mapping unit is connected to the ontology library construction unit. It receives three-dimensional spatial coordinates, operation timestamps, construction process parameters, temperature and humidity changes, structural deformation, and point cloud scan data from the data acquisition module. It queries the component IDs bound to the current operation robot, matches index items and thresholds according to the ontology library, and outputs structured data with component IDs, index types, measured values, thresholds, and anomaly flags. The data output unit is connected to the dynamic mapping unit and transmits the structured data to the time-series twin module.

[0065] The semantic fusion module achieves precise mapping from construction status to quality indicators through semantic parsing and structured transformation of multi-source heterogeneous data. The ontology library construction unit establishes a tree-like semantic network with component type as the root node, based on the design specifications and acceptance standards for railway station construction projects. This semantic network includes process and method sub-nodes and quality indicator leaf nodes. The quality indicator nodes are pre-defined with design threshold ranges, units of measurement, and testing method attributes, forming a machine-interpretable quality evaluation knowledge system. Semantic association rules explicitly define the binding relationship between components and processes; for example, a strong association is established between the "steel column installation" process and the "verticality deviation" indicator.

[0066] The dynamic mapping unit receives six types of real-time monitoring data from the data acquisition module, including three-dimensional spatial coordinates, operation timestamps, construction process parameters, temperature and humidity changes, structural deformation, and point cloud scan data. The system analyzes the current execution process type through the robot's operation log and retrieves the bound component set in the semantic network based on the process type. Taking point cloud scan data as an example, the unit calls a surface fitting algorithm to calculate the actual axis coordinates of the steel beam, matches the "axis offset" index under the "steel beam installation" node in the semantic network, and compares the calculation result with a preset threshold. When the monitored value of structural deformation exceeds the beam deflection threshold, the dynamic mapping unit automatically generates an anomaly flag and outputs a structured data record including component ID, index type, measured value, and threshold.

[0067] The data output unit encapsulates structured data using a lightweight data exchange format, and the encapsulated data packets retain the node encoding system in the semantic network. Each data packet header includes a timestamp and data source identifier, while the main body arranges indicator records by component ID index. Anomaly flags are embedded as a separate field at the end of each record. The data output unit transmits the structured data to the time-series twin module via a message queue. A checksum mechanism ensures data integrity during transmission, providing standardized input for the construction of the digital twin.

[0068] Specifically, the time-series twin module of the BIM-based intelligent robot construction progress and quality linkage analysis system provided by the present invention includes: The version generation unit imports the BIM model from the design phase as the initial version. After the process is completed, it generates a point cloud model through laser scanning, generates triangular mesh patches through surface reconstruction algorithm, compares the deviation vector between the design model and the measured model, generates a new version and records the change log. The prediction evolution unit, connected to the version generation unit, automatically starts the new version modeling process when the abnormal flag output by the semantic fusion module indicates that the displacement of the component continues to increase. It extracts the most recent scan data to establish a time series model, predicts future deformations, and generates a twin version containing the predicted state. The state chain management unit, connected to the prediction evolution unit, associates the version sequence as a timeline and supports historical state backtracking.

[0069] The temporal twin module achieves full lifecycle mapping of the construction process by constructing a digital twin that can trace historical states and predict future evolution. The version generation unit imports the BIM model from the design phase as initial version zero, which includes component geometric attributes and design parameters. Upon completion of a specific process, the module calls upon the point cloud model generated by laser scanning and uses a Poisson surface reconstruction algorithm to transform the discrete point cloud into continuous triangular mesh patches, generating a measured physical model. Spatial registration technology aligns the design model with the measured model, calculates the Euclidean distance between corresponding vertices to form a deviation vector, and generates a new version when the deviation vector exceeds a preset tolerance, recording a change log including the deviation value, timestamp, and scope of impact.

[0070] The predictive evolution unit monitors the anomaly flags output by the semantic fusion module in real time. When the anomaly flags indicate a monotonically increasing displacement of the same component for three consecutive sampling periods, the unit automatically activates the prediction process. The process extracts the most recent ten laser scan data to form a time series, uses an autoregressive integral moving average model to analyze the deformation rate, and extrapolates the deformation values ​​for the next six hours. The prediction results are overlaid onto the current twin version in the form of a semi-transparent heatmap, generating a new version twin with a fused prediction state. The new version identifies the prediction confidence level.

[0071] The state chain management unit assigns a globally unique version number to each twin version, which includes a timestamp and a version type identifier. The unit uses a doubly linked list data structure to associate the version sequence, with each list node storing a pointer to the version file and pointers to adjacent versions. When a user initiates a history rewind request, the unit locates the target version node based on the timeline index, loads the corresponding version data, and highlights the change vector between versions, supporting filtering of component state evolution paths by time range. The version rewind function renders the component state change process frame-by-frame according to the time sequence recorded in the logs.

[0072] Specifically, the spatial matching module of the BIM-based intelligent robot construction progress and quality linkage analysis system provided by the present invention includes: The initial matching unit is rigidly aligned with the robot's GPS coordinates by the center coordinates of the BIM component, and the average positioning error is calculated. The learning optimization unit is connected to the initial matching unit, records the robot's typical operation path, and establishes a path feature library, which includes amplitude frequency and turning angle patterns. The prediction compensation unit, connected to the learning optimization unit, calls the current working area feature mode when the GPS signal is lost, combines the motion vector to predict the next position, calculates the coordinate offset compensation amount through point cloud feature points, and updates the spatial matching strategy of the spatial matching module.

[0073] The spatial matching module achieves accurate mapping between the Building Information Model (BIM) and physical space through a multi-stage optimization mechanism. The initial matching unit establishes a rigid transformation matrix between the center coordinates of BIM components and the robot's GPS coordinates, and uses the least squares method to solve for the optimal rotation and translation parameters. After transformation, the average Euclidean distance of each component's matching points is calculated as the average positioning error, and this error value is recorded in the matching report for subsequent optimization reference.

[0074] The learning optimization unit continuously collects pose data during robot operations, including time-series position and attitude information. The unit performs Fourier transform analysis on typical work paths to extract path amplitude-frequency features; it also calculates path curvature changes using differential geometry methods and statistically analyzes the histogram of turning angle distribution. After normalization, the feature data is stored in a path feature library. This library is indexed spatially according to a grid of the construction area, with each grid cell associated with a specific amplitude-frequency pattern and turning angle pattern.

[0075] The prediction and compensation unit monitors the GPS signal strength in real time. When the signal strength falls below the reliable communication threshold, the unit retrieves regional feature patterns from the path feature library based on the robot's current grid index. Combining the motion vector output by the inertial measurement unit, a Kalman filter predictor is used to estimate the theoretical position for the next sampling period. Simultaneously, stable feature points such as column corners and wall edges are extracted from the environmental point cloud, and the coordinate offset between the actual and predicted positions is calculated using an iterative nearest-point algorithm. Finally, the offset compensation is injected into the spatial matching strategy database to generate an updated spatial matching strategy for use in the next round of matching.

[0076] Specifically, the BIM-based intelligent robot construction progress and quality linkage analysis system provided by this invention includes a coupling analysis module comprising: The input receiving unit receives the component and device mapping matrix and the data source mapping matrix, and identifies abnormal quality data. The historical retrieval unit is connected to the input receiving unit and retrieves the average repair time of similar quality anomaly cases in the stored historical data; The conflict detection unit is connected to the historical retrieval unit and detects resource conflicts in subsequent processes. The resource conflicts include conflicts between work group entry time and conflicts between equipment occupancy time. The probability calculation unit, connected to the conflict detection unit, calculates the critical path delay probability and generates a risk heat map to indicate the impact range.

[0077] The coupling analysis module achieves a quantitative assessment of the impact of quality defects on construction progress through multi-dimensional correlation analysis. The input receiving unit parses the structured quality data in the component and equipment mapping matrix and the data source, and uses a rule-based pattern recognition engine to filter abnormal records. The identification process judges the abnormal state of the data based on the threshold range of quality indicators. When the measured value continuously exceeds the threshold, it is marked as a valid quality abnormality event, and the location of the abnormal component, the magnitude of the deviation, and the detection timestamp are extracted.

[0078] The historical retrieval unit generates retrieval vectors based on the component type and deviation magnitude of quality anomaly events. These retrieval vectors are then matched with cosine similarity vectors of cases stored in the historical database, filtering out cases with similarity scores exceeding a set threshold. Repair time data is extracted from the matched cases, and an arithmetic mean is calculated to generate average repair time data for similar quality anomaly cases, providing historical data for project schedule prediction.

[0079] After receiving the average repair time data, the conflict detection unit activates resource conflict scanning. The unit accesses the process logic relationship database of the schedule planning system to locate the critical subsequent processes affected by the current abnormal component. Based on the average repair time, it extrapolates changes in process time windows and performs time window overlap detection with the team entry plans in the resource scheduling database to identify team entry time conflict events. Simultaneously, it verifies the equipment occupancy plan table, discovers equipment occupancy time conflict events, and generates a resource conflict report including the conflicting parties and conflict duration.

[0080] The probability calculation unit combines resource conflict reports with the critical path model to perform Monte Carlo simulations. The simulation process considers variables such as repair time volatility and resource adjustment flexibility, iteratively calculating the frequency of critical path delays. The calculation results are converted into probability distribution curves, which are used to generate a risk heatmap covering the construction area. The heatmap uses a red, yellow, and green gradient to indicate the risk level of different spatial areas and uses contour lines to mark the boundaries of the risk impact range, visually demonstrating the cascading schedule risks caused by quality anomalies.

[0081] Specifically, the decision optimization module of the BIM-based intelligent robot construction progress and quality linkage analysis system provided by the present invention includes: The environment construction unit imports the current digital twin version generated by the time-series twin module and loads resource constraints, including machine scheduling rules and material supply conditions. The scheme simulation unit is connected to the environment construction unit and simulates multiple disposal schemes in parallel. The disposal schemes include a work stoppage detection scheme, a speed reduction operation scheme, and a backup team activation scheme. The virtual construction period and cost of each disposal scheme are output. The optimization screening unit is connected to the scheme simulation unit to evaluate the comprehensive impact of each disposal scheme on the total project duration and cost, eliminate inferior schemes, and generate a Pareto optimal solution set.

[0082] The decision optimization module optimizes construction plans under quality anomalies by simulating multi-dimensional handling schemes and selecting the optimal solution in a digital environment. The environment construction unit obtains the latest digital twin version from the temporal twin module, which includes the spatial geometric attributes and physical parameters of the current construction status. Simultaneously, it loads the project resource constraint library, where machine scheduling rules define restrictions such as equipment switching cool-down time and maximum continuous working duration; material supply conditions specify the minimum inventory threshold and procurement lead time for key building materials, forming a complete virtual simulation constraint framework.

[0083] The simulation unit performs discrete event simulations based on a constraint framework, executing three handling strategies in parallel. The shutdown inspection scheme simulates pausing the current process and then initiating a non-destructive testing process, calculating the quality review time and resource idleness losses during downtime. The speed-down operation scheme proportionally reduces the robot's operating speed, extrapolating the cascading impact of extended unit working hours on subsequent processes and calculating the incremental costs of labor and equipment usage. The backup shift activation scheme simulates a multi-shift collaborative scenario after utilizing reserve human resources, evaluating the time-compression effect and additional management costs brought about by parallel operations. Each simulation outputs the virtual total project duration and comprehensive cost value corresponding to the chosen scheme.

[0084] An optimized screening unit establishes a two-dimensional evaluation coordinate system based on schedule and cost, mapping the simulation results of each scheme to discrete points in the coordinate system. A non-dominated sorting algorithm is used to identify the Pareto front solution set, eliminating inferior schemes that exceed other schemes in both schedule and cost. An inferior scheme is defined as one that fails to achieve Pareto optimality in both schedule and cost dimensions. The final Pareto optimal solution set includes multiple non-inferior schemes and their schedule and cost data, providing decision-makers with a multi-objective balance selection space.

[0085] Specifically, the BIM-based intelligent robot construction progress and quality linkage analysis system provided by this invention includes an execution traceability module comprising: The instruction generation unit receives the Pareto optimal solution set, analyzes the resource requirements in the selected solution, and generates logistics system instructions and robot parameter adjustment instructions. An execution recording unit, connected to the instruction generation unit, records the instruction issuance time and execution confirmation time; The digital thread unit, connected to the execution record unit, anchors key data of the processing to a specific version node of the digital twin, and establishes an auditable operation chain including event ID, associated twin version number, decision scheme fingerprint, execution log and effect verification data.

[0086] The execution traceability module achieves precise implementation of decision-making schemes through instruction automation and operation chain traceability. The instruction generation unit receives the Pareto optimal solution set output by the decision optimization module and analyzes the resource requirement details of the selected disposal scheme. The resource requirement details include the required material codes, quantities, and target warehouse location information, generating warehousing and allocation instructions for the logistics system; at the same time, it extracts robot operation parameter adjustment items and generates a robot control instruction set including speed adjustment coefficients and path replanning parameters.

[0087] The execution recording unit uses the time source provided by the time synchronization server as a reference to record the precise timestamp of the instructions issued by the instruction generation unit. When the logistics management system and the robot control system return an instruction reception confirmation signal, the unit captures and records the execution confirmation time, forming an execution event log that includes the instruction type, issuance time, and confirmation time.

[0088] The digital thread unit constructs an auditable operation chain based on a cryptographic hash chain. The unit extracts the current activity version number of the time-series twin module as an anchor point and encapsulates key data from the handling process into data blocks. A unique event identifier is written to the header of each data block, and the body embeds the associated twin version number, decision scheme digest hash value, execution event log, and quality review data. The quality review data originates from the anomaly flag state change records of the semantic fusion module. The operation chain establishes a bidirectional association with the digital twin nodes through a version number index, supporting the retrieval of the complete handling trajectory by event identifier.

[0089] Specifically, the BIM-based intelligent robot construction progress and quality linkage analysis system provided by the present invention further includes a feedback closed-loop module connecting the execution traceability module and the data acquisition module. The feedback closed-loop module is used for: Receive new data acquired by the data acquisition module after the instruction is executed, the new data including quality recovery status data and progress data; The new data is sent to the semantic fusion module to generate structured data in a normal state; The time-series twin module is triggered to generate a new version record verification result; Update the spatial matching strategy of the spatial matching module and the decision model weight coefficients of the decision optimization module.

[0090] The feedback closed-loop module achieves system adaptive optimization through a multi-level feedback mechanism. The module receives new datasets collected by the data acquisition module after instruction execution. These new datasets include quality recovery status data and progress data. The quality recovery status data originates from the component deformation monitoring values ​​re-collected by the structural monitoring unit, while the progress data includes the deviation between the actual completion timestamps of the processes and the planned times. These data reflect the actual effectiveness of the remediation plan.

[0091] The feedback loop module transmits the new dataset to the semantic fusion module for standardization. The semantic fusion module uses the railway station domain ontology to parse the quality recovery status data. When the monitored values ​​remain within the design threshold range, it generates structured data indicating a normal state. This structured data retains the component ID and indicator type fields, updates the measured values ​​to the latest monitored values, and identifies the abnormal status as a normal status code. Progress data is converted into standardized progress deviation records.

[0092] The feedback loop module sends a new version creation command to the time-series twin module. The time-series twin module obtains the current digital twin version number and creates a verification version based on the structured data generated by the semantic fusion module. The new version records the deviation between the quality recovery status data and the predicted variables, and adds a verification entry for the treatment plan's effectiveness to the change log. The version number uses the naming convention of "V + timestamp + verification identifier".

[0093] The feedback loop module initiates the strategy update process. For the spatial matching module, it extracts the deviation matrix between the robot's actual trajectory point cloud and the predicted trajectory in the new dataset, and adjusts the compensation parameters in the spatial matching strategy using the backpropagation algorithm. For the decision optimization module, it calculates the error rate between the actual project duration and cost of the disposal plan and the predicted value, and updates the weight coefficients of the decision model using the gradient descent method. The updated strategy parameters take effect immediately and are used in subsequent decision-making processes.

[0094] Specifically, the BIM-based intelligent robot construction progress and quality linkage analysis system provided by the present invention further includes: The data flow between the data acquisition module, semantic fusion module, temporal twin module, spatial matching module, coupling analysis module, decision optimization module, and execution tracing module forms a closed loop: The data acquisition module outputs spatiotemporally aligned fused data as a snapshot of the construction status to the semantic fusion module; The semantic fusion module outputs structured data with semantic tags and anomaly flags to the temporal twin module; The temporal twin module outputs a new version of the digital twin and a change vector representing the change in the state of the component to the spatial matching module; The spatial matching module outputs a component-device and data source mapping matrix to the coupling analysis module; The coupling analysis module outputs a risk warning signal to the decision optimization module; The decision optimization module outputs the Pareto optimal solution set to the execution traceability module; The execution traceability module outputs the machine-executable instruction set and auditable operation chain, and feeds back the execution data to the data acquisition module and the update information to the spatial matching module, thus completing the system's self-evolution.

[0095] The BIM-based intelligent robot construction progress and quality linkage analysis system achieves dynamic optimization through closed-loop data flow. The data acquisition module outputs the spatiotemporally aligned fused data generated by the spatiotemporal registration unit as a construction status snapshot to the semantic fusion module. This snapshot integrates robot trajectory and environmental monitoring data to form a complete picture of the construction status under a unified spatiotemporal reference.

[0096] The semantic fusion module processes construction status snapshots based on semantic rules defined by the ontology library building units, outputting structured data with semantic tags, including component IDs, indicator types, measured values, thresholds, and anomaly flags. This structured data and anomaly flags are then transmitted to the time-series twin module. Semantic tags enable a computable correlation between the original monitoring data and BIM quality indicators.

[0097] The temporal twin module generates a new version of the digital twin based on structured data and anomaly flags. This new version records the component's state transition sequence and outputs a change vector characterizing the component's state changes. The change vector and the new version of the digital twin are input into the spatial matching module. The change vector quantifies the geometric deviation between the component's actual state and its design state.

[0098] The spatial matching module employs an incremental learning mechanism to process the new version of the change vector and digital twin, and after optimization, outputs a mapping matrix of components, equipment, and data sources to the coupling analysis module. This mapping matrix establishes the correlation between quality anomaly data and construction equipment and data acquisition sources, providing a data foundation for risk analysis.

[0099] The coupling analysis module parses the quality anomaly data in the mapping matrix, generates risk warning signals that identify potential cascading risks, and outputs them to the decision optimization module. The risk warning signals include predictions of resource conflict types and their impact scope.

[0100] After receiving the risk warning signal, the decision optimization module outputs the Pareto optimal solution set to the execution traceability module. The solution set includes multiple time-cost balanced solutions. The Pareto optimal solution set is generated through virtual simulation and inferior solutions are eliminated.

[0101] The execution tracking module transforms the Pareto optimal solution set into a machine-executable instruction set and generates an auditable operation chain. Simultaneously, it feeds back execution data to the data acquisition module and update information to the spatial matching module. Execution data includes the robot's actual operational parameters, and update information includes spatial matching strategy corrections. These two types of feedback data drive the system's self-evolution.

[0102] Secondly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements a BIM-based intelligent robot construction progress and quality linkage analysis system.

[0103] This invention addresses the technical problem of delayed progress and quality linkage response in intelligent railway station construction by constructing an integrated system architecture to achieve unified cross-system data interaction standards and establish a dynamic correlation mechanism. The system first acquires multi-source heterogeneous data through a data acquisition module, including 3D spatial coordinates, operation timestamps, construction process parameters collected by the intelligent robot's sensors, and temperature and humidity changes, structural deformation, and point cloud scan data captured by the environmental monitoring unit. A unified spatiotemporal reference axis is established through a spatiotemporal registration unit, outputting spatiotemporally aligned fused data as a snapshot of the construction status, laying the foundation for data interaction standards. The semantic fusion module connects to the data acquisition module, using a railway station domain ontology library to parse data attributes, associating sensor data with quality indicators of specific components in the BIM model, generating structured data with semantic tags, and outputting anomaly flags. This endows the original data with semantic consistency, enabling machines to understand and process data from different systems. The temporal twin module connects to the semantic fusion module. When an anomaly flag indicates a quality anomaly, it injects a time dimension attribute into the BIM geometric model, generating a new version of the digital twin that records the sequence of component state changes. It also outputs a change vector representing the component's state change, creating a dynamically evolving digital representation that supports historical retrospection and future prediction. The spatial matching module connects to the temporal twin module. Based on the change vector and the new version of the digital twin, it uses an incremental learning mechanism to optimize the spatial matching strategy, completing the registration of BIM component coordinates with robot coordinates. It outputs a mapping matrix of components, equipment, and data sources, achieving real-time and accurate mapping between the physical space and the digital model. The coupling analysis module connects to the spatial matching module, establishing a quantitative assessment model of the impact of quality defects on the project schedule. Based on the quality anomaly data in the component, equipment, and data source mapping matrix, it calculates the probability of subsequent process delays caused by quality anomalies and the results of resource conflict detection, generating risk warning signals that directly link quality status and schedule impact. The decision optimization module, connected to the coupling analysis module, receives risk warning signals, simulates multiple disposal plans in parallel within a virtual construction environment, evaluates the comprehensive impact of each plan on the total project duration and cost, and outputs a Pareto optimal solution set, providing data-driven optimization decisions. The execution traceability module, connected to the decision optimization module, receives the Pareto optimal solution set, automatically generates a set of machine-executable instructions, adjusts robot process parameters and logistics system delivery priorities, and constructs a digital thread to anchor key data of the disposal process to specific version nodes of the digital twin, forming an auditable operation chain to achieve decision execution with full traceability. The feedback closed-loop module, connected to the execution traceability module and the data acquisition module, receives new data collected after instruction execution, including quality recovery status data and progress data, and feeds it back to the semantic fusion module to generate structured data for the normal state. This triggers the temporal twin module to generate a new version record verification result, updates the spatial matching strategy of the spatial matching module and the decision model weight coefficients of the decision optimization module, achieving system self-evolution.The entire system's data flow forms a closed loop. From the data acquisition module, spatiotemporally aligned fused data is output to the semantic fusion module. The semantic fusion module outputs structured data with semantic tags and anomaly flags to the temporal twin module. The temporal twin module outputs the new version of the digital twin and change vectors to the spatial matching module. The spatial matching module outputs the component-device and data source mapping matrix to the coupling analysis module. The coupling analysis module outputs risk warning signals to the decision optimization module. The decision optimization module outputs the Pareto optimal solution set to the execution traceability module. The execution traceability module outputs the machine-executable instruction set and auditable operation chain, and feeds back the execution data to the data acquisition module and the update information to the spatial matching module. This closed-loop design enables smooth data interaction between different systems based on a unified standard, and achieves real-time linkage response of progress and quality through a dynamic correlation mechanism, eliminating response lag.

[0104] The specific implementation process of the BIM-based intelligent robot construction progress and quality linkage analysis system of this invention is as follows. In the intelligent construction scenario of railway station buildings, the data acquisition module starts working after the system is started. The robot sensor unit uses the fusion positioning of lidar and inertial measurement unit to record the three-dimensional coordinates of the actuator in the world coordinate system in real time, and generates a spatial location data packet including a timestamp when each operation is completed. The environmental monitoring unit independently deploys micro-weather stations to collect air temperature and humidity gradients at preset intervals, the structural monitoring unit captures the strain fluctuations of the steel support frame at preset frequencies, and the laser scanning station emits scanning beams through a rotating prism system and receives reflected signals to generate point cloud data frames. The spatiotemporal registration unit connects the robot sensor unit and the environmental monitoring unit, establishes a unified spatiotemporal reference axis, inputs the robot trajectory point cloud and the laser scanning point cloud into the feature matching algorithm, extracts feature points from the two types of point clouds for similarity calculation, eliminates coordinate system deviations between equipment through optimization algorithms, and outputs spatiotemporally aligned fused data as a construction status snapshot to be transmitted to the semantic fusion module.

[0105] After receiving the spatiotemporally aligned fused data, the semantic fusion module parses it using the ontology library construction unit, which uses a tree-structured semantic association rule based on component types, construction methods, and quality indicators defined in railway station building engineering specifications. The dynamic mapping unit queries the component identifiers bound to the currently operating robot, matches indicator items and thresholds according to the ontology library, and outputs structured data containing component identifiers, indicator types, measured values, thresholds, and anomaly flags. The data output unit transmits the structured data to the temporal twin module. When the strain fluctuation of the steel support frame exceeds the design threshold, the anomaly flag is set and the deviation value is recorded.

[0106] The temporal twin module receives structured data with semantic tags and anomaly flags. When the anomaly flags indicate a quality anomaly, the processing flow is initiated. The version generation unit imports the building information model from the design phase as the initial version, generates a point cloud model through laser scanning, and generates triangular mesh patches through a surface reconstruction algorithm. After comparing the deviation vectors between the design model and the measured model, a new version is generated and a change log is recorded. The prediction evolution unit automatically initiates a new version modeling process when the component displacement continues to increase. It extracts the most recent scan data to build a time series model to predict future deformations and generates a twin version containing the predicted state. The state chain management unit associates the version sequence as a timeline to support historical state backtracking and outputs the new version of the digital twin and the change vector representing the component state change to the spatial matching module.

[0107] The spatial matching module, based on the new version of change vectors and digital twins, employs an incremental learning mechanism to optimize the matching strategy. The initial matching unit calculates the average positioning error by rigidly aligning the component center coordinates from the Building Information Model (BIM) with the robot's GPS coordinates. The learning optimization unit records typical robot operation paths to establish a path feature library including amplitude frequency and turning angle patterns. When the GPS signal is lost, the prediction compensation unit calls upon the current operation area feature patterns, combines motion vectors to predict the next position, and calculates the coordinate offset compensation amount using point cloud feature points. After updating the spatial matching strategy, it outputs the component-equipment and data source mapping matrix to the coupling analysis module.

[0108] The coupling analysis module receives the component-equipment and data source mapping matrix and establishes a quantitative assessment model of the impact of quality defects on the project schedule. The input receiving unit identifies quality anomaly data, and the historical retrieval unit retrieves the average repair time of similar quality anomaly cases from stored historical data. The conflict detection unit detects resource conflicts in subsequent processes, including conflicts in team arrival time and equipment occupancy time. The probability calculation unit calculates the critical path delay probability and generates a risk heatmap to indicate the impact range, outputting a risk warning signal to the decision optimization module.

[0109] After receiving a risk warning signal, the decision optimization module simulates multiple disposal plans in parallel within the virtual construction environment. The environment construction unit imports the current digital twin version generated by the time-series twin module and loads resource constraints, including machinery scheduling rules and material supply conditions. The scheme simulation unit simulates the shutdown detection plan, the speed reduction operation plan, and the backup team activation plan in parallel, outputting the virtual duration and cost of each disposal plan. The optimization screening unit evaluates the comprehensive impact of each disposal plan on the total duration and cost, eliminates inferior plans, and generates a Pareto optimal solution set, which is then output to the execution traceability module.

[0110] After receiving the Pareto optimal solution set, the execution traceability module automatically generates a set of machine-executable instructions. The instruction generation unit parses the resource requirements in the selected solution and generates logistics system instructions and robot parameter adjustment instructions. The execution recording unit records the instruction issuance time and execution confirmation time. The digital thread unit anchors key data of the disposal process to specific version nodes of the digital twin, establishing an auditable operational chain including event identifiers, associated twin version numbers, decision solution fingerprints, execution logs, and effect verification data. Simultaneously, the system receives new data collected by the data acquisition module after instruction execution, including quality recovery status data and progress data, through the feedback closed-loop module. This new data is sent to the semantic fusion module to generate structured data for the normal state, triggering the temporal twin module to generate new version record verification results, and updating the spatial matching strategy of the spatial matching module and the decision model weight coefficients of the decision optimization module, completing the system's self-evolutionary closed loop.

[0111] The technical features of this invention are explained below: The quantitative assessment model is built by analyzing the cascading impact of quality defects on construction progress. Based on historical data statistics and machine learning algorithms, the model takes as input the quality anomaly data in the component and equipment and data source mapping matrix, and outputs the probability of delay in subsequent processes and the results of resource conflict detection. When the model is applied, it first identifies quality anomaly events, retrieves the average repair time of similar cases, detects resource conflicts, and finally calculates the probability of critical path delay, thereby generating risk warning signals and providing data support for decision-making.

[0112] The time series model is used to predict the future deformation of components. The model is built based on historical scanning data and uses an autoregressive integral moving average algorithm to analyze the time series data and capture deformation trends and seasonal changes. When the model is applied, when the anomaly flag output by the semantic fusion module indicates that the component displacement is continuously increasing, the prediction process is automatically started, the most recent scanning data is extracted to build a time series model, the predicted deformation is output, and a digital twin version containing the predicted state is generated to support the forward-looking management of construction status.

[0113] A digital twin is constructed by injecting time-dimensional attributes into a BIM geometric model. The model is built starting with the BIM model from the design phase. A point cloud model is generated through laser scanning, and triangular mesh patches are generated using a surface reconstruction algorithm. The deviation vector between the design model and the measured model is compared, and a change log is recorded. When the model is applied, the version generation unit creates a new version, the prediction evolution unit adds predicted states, and the state chain management unit associates the version sequence to form a timeline, supporting historical state backtracking and future state simulation, providing a dynamic digital representation for the construction process.

[0114] The spatial matching strategy is built by optimizing the registration of BIM component coordinates and robot coordinates through an incremental learning mechanism. The strategy is built based on the positioning error calculated by the initial matching unit, the learning optimization unit's path feature library, and the predicted compensation unit's back-calculated coordinate offset compensation amount. When the strategy is applied, the matching parameters are dynamically adjusted according to the change vector and the new version of the digital twin. When the GPS signal is lost, the path feature mode is called to predict the position, and the component, equipment, and data source mapping matrix is ​​output to achieve accurate alignment between the physical space and the digital model.

[0115] The decision-making model is constructed by simulating multiple disposal options in parallel in a virtual construction environment. The model is built based on the current version of the digital twin and resource constraints, such as machine scheduling rules and material supply conditions. Discrete event simulation technology is used to evaluate the impact of each option on the project schedule and cost. When the model is applied, the scheme simulation unit runs options such as shutdown detection, deceleration operation and activation of standby teams, and the optimization and screening unit generates Pareto optimal solution set through non-dominated sorting algorithm, providing decision-makers with multi-objective optimization choices.

[0116] The path feature library is built by recording feature data of typical robot operation paths. The library is built based on pose data collected by the learning optimization unit, using Fourier transform to extract amplitude and frequency features, differential geometry to calculate steering angle patterns, and normalizing and storing the data. When the library is used, if it is necessary to optimize the spatial matching strategy or predict the robot position, the regional feature patterns in the path feature library are called, and combined with motion vectors for prediction and compensation, thereby improving the accuracy and robustness of spatial matching.

[0117] Semantic association rules are constructed using a tree structure based on the definition of component types, processes, and quality indicators in railway station building engineering specifications. The rules are built in the ontology library construction unit, which clarifies the binding relationship between components and processes and the threshold of quality indicators. When the rules are applied, the dynamic mapping unit queries the component identifiers bound to the current working robot, matches the indicator items and thresholds according to the rules, and outputs structured data with semantic tags, realizing machine-resolvable mapping of sensor data to BIM quality indicators.

[0118] The risk heatmap is constructed by calculating the probability of critical path delay and visualizing the scope of impact. The graph is built based on the Monte Carlo simulation output of the probability calculation unit, which transforms the probability distribution data into a color gradient, with red indicating high risk and green indicating low risk. When the graph is applied, a heatmap covering the construction area is generated, and the risk level boundaries are marked by contour lines, which intuitively shows the schedule risks caused by quality anomalies, helping on-site managers to quickly identify and respond to them.

[0119] The Pareto optimal solution set is constructed by screening disposal schemes through a multi-objective optimization algorithm. The set is built based on the simulation results of the decision optimization module, identifies non-dominated solutions in the two-dimensional evaluation coordinate system of schedule and cost, and eliminates inferior schemes. When the set is applied, it outputs multiple non-inferior solution schemes and their schedule and cost data, providing decision-makers with trade-offs and choices, and achieving balanced optimization of the scheme in terms of schedule and cost.

[0120] Digital threads are constructed by anchoring key data of the disposal process to specific version nodes of a digital twin. The thread construction is based on cryptographic hash chain technology and includes event identifiers, associated twin version numbers, decision scheme fingerprints, execution logs, and effect verification data. When the thread is applied, the execution traceability module records the time of instruction issuance and execution confirmation, establishes an auditable operation chain, supports full-process traceability and verification, and achieves transparency and reliability of decision execution.

[0121] A feature matching algorithm is applied to the spatiotemporal registration unit to align robot trajectory point clouds with laser scan point clouds. This algorithm extracts corner and edge feature points from both types of point clouds, calculates the similarity matrix between feature points, and eliminates coordinate system deviations between devices through an iterative nearest-point optimization algorithm. This achieves spatial alignment of multi-source point cloud data, providing fused data in a unified coordinate system for subsequent processing.

[0122] The rule engine algorithm is applied to the dynamic mapping unit to identify abnormal quality data. Based on a predefined threshold range for quality indicators, the engine judges the status of sensor measurements. When a value continuously exceeds the threshold, it is marked as a valid abnormal event, and feature vectors of the location and magnitude of deviation of the abnormal component are extracted to achieve automated determination of quality status.

[0123] The surface reconstruction algorithm is applied to the version generation unit to convert point cloud data into a geometric model. This algorithm uses the Poisson surface reconstruction method to transform discrete point clouds into continuous triangular mesh patches, generating a high-precision measured model. By comparing this model with the design model, a deviation vector is generated, providing geometric change data for digital twin version updates.

[0124] Time series forecasting algorithms are applied to predict evolutionary units to analyze component deformation trends. This algorithm uses an autoregressive integral moving average model to process historical scan data, establishes a time series model of deformation variables, and predicts deformation variable values ​​for specific future periods, providing data support for forward-looking decision-making.

[0125] Incremental learning algorithms are applied to the spatial matching module to optimize coordinate registration strategies. This algorithm continuously records robot operation path features to build a path feature library, dynamically updates matching parameters when new data is input, and gradually improves the registration accuracy between BIM component coordinates and robot coordinates, adapting to environmental changes.

[0126] Monte Carlo simulation algorithms are applied to probabilistic calculation units to assess the risk of project delays. This algorithm simulates multiple possible scenarios using random sampling, iteratively calculates the probability of delays on the critical path, generates probability distribution curves, quantifies the potential impact of quality anomalies on the project schedule, and supports risk warning.

[0127] The cosine similarity algorithm is applied to the historical retrieval unit to match similar quality cases. This algorithm calculates the cosine of the angle between the feature vector of the current abnormal event and the feature vectors of historical cases, filtering historical cases with similarity higher than a threshold, thus providing a reference for estimating repair time.

[0128] The non-dominated sorting algorithm is applied to the optimization screening unit for multi-objective optimization decision-making. This algorithm hierarchically sorts the set of disposal options, identifies the Pareto optimal solution set, eliminates inferior solutions that are comprehensively outperformed by other solutions, and provides a comprehensive optimization solution that balances time and cost.

[0129] Cryptographic hash algorithms are applied to digital thread units to construct auditable operation chains. This algorithm generates digital fingerprints for key data in the processing, links data blocks chronologically through a hash chain, ensuring the integrity and immutability of operation records and supporting full-process traceability.

[0130] Gradient descent is applied to the feedback closed-loop module to update the parameters of the decision model. This algorithm calculates the error gradient between the actual effect of the treatment plan and the predicted value, iteratively adjusting the weight coefficients of the decision model to gradually bring the model output closer to the actual result, thereby improving the accuracy of subsequent decisions.

[0131] Spatiotemporally aligned fused data is a snapshot of the construction status formed by integrating multi-source sensor data through a unified spatiotemporal reference axis. This data is generated by processing robot trajectory point clouds and laser scan point clouds using a feature matching algorithm. The algorithm extracts corner and edge feature points from the two types of point clouds to calculate a similarity matrix. Through iterative optimization, it eliminates coordinate system deviations between devices and outputs a comprehensive dataset with a unified spatiotemporal reference system, providing a standardized input basis for subsequent modules.

[0132] Structured data with semantic tags is a machine-understandable format generated from raw monitoring data after parsing by a domain ontology library. The data includes component ID, indicator type, measured value, and threshold fields. Through semantic association rules defined in railway station building engineering specifications, sensor data is mapped to specific quality indicators in the BIM model, transforming physical quantities such as temperature and deformation into standardized quality parameters with engineering significance, thus providing semantic consistency assurance for the construction of digital twins.

[0133] The anomaly flag is a binary indicator used to identify the quality status. When the sensor's measured value continuously exceeds the design threshold, the semantic fusion module automatically generates this flag. Its status change triggers the version generation process of the timing twin module, and at the same time, it serves as the start signal for the quality anomaly event, driving the subsequent prediction and decision-making process, forming an instantaneous conversion mechanism from quality status to system response.

[0134] A change vector is a set of geometric parameters that quantifies changes in the state of a component. The vector is generated by comparing the geometric deviations between the new and historical versions of the digital twin, including parameters such as position offset and deformation. This provides the spatial matching module with a dynamically updated basis for coordinate correction, enabling continuous calibration between the BIM model and the physical space.

[0135] The component-equipment-data source mapping matrix is ​​a two-dimensional relational table that establishes traceability relationships for quality data. The matrix records the correspondence between component numbers, construction equipment identifiers, and data acquisition sources. When a quality anomaly is detected, the matrix can be used to quickly locate the relevant equipment and data source, providing an equipment-level quality data association path for coupled analysis.

[0136] Resource conflict detection is a technical method for identifying the competition status of construction resources through time window overlap analysis. This method, based on the extrapolation of changes in process time windows, compares the results with the resource scheduling plan over a time dimension to identify conflicts in team arrival times and equipment occupancy times. It generates a diagnostic report including conflict type and duration, providing input for quantifying schedule risks.

[0137] The Pareto optimal solution set is the set of non-dominated solutions output by multi-objective optimization algorithms. The set includes multiple time-cost balanced solutions, each of which cannot be further optimized in either the time or cost dimension without compromising the other, providing decision-makers with a scientific multi-objective trade-off selection space.

[0138] The auditable operational chain is a full-process traceability data structure built on a cryptographic hash chain. The chain structure anchors key data in the handling process to digital twin version nodes, including fields such as event ID, decision fingerprint, and execution log. It supports retrieving the complete handling trajectory by timeline or event identifier, achieving transparency and traceability of the operation process.

[0139] The version number of a digital twin is a unique code that identifies the sequence of state changes of the twin. The version number adopts a composite structure of "version type + timestamp + identifier", and links each version node through a doubly linked list. It supports tracing back to historical states in chronological order or jumping to a specific version, providing a traceable digital image of the construction process.

[0140] The feedback loop mechanism is a control strategy that enables system self-optimization through data feedback. The mechanism feeds back quality recovery data and progress data after instruction execution to the front-end module, triggering semantic data updates, version generation, and strategy adjustments. This forms a closed-loop control system encompassing perception, decision-making, execution, and optimization, enabling the system to continuously learn and adapt.

[0141] The unified spatiotemporal reference axis is a multi-source data synchronization framework based on the time of the BeiDou Navigation Satellite System. The framework provides millisecond-level time synchronization for all sensors through a timing server, establishes a unified spatial reference by combining it with the world coordinate system, and enables the comparability and fusion of data collected by different devices, laying the foundation for cross-system data interaction.

[0142] The path feature library is a feature database that stores typical robot operation modes. The library structure is divided into grids according to the construction area. Each grid cell is associated with motion feature patterns such as amplitude frequency and turning angle. When GPS signal is lost, the feature patterns are called to predict the robot trajectory, providing redundant positioning support for spatial matching.

[0143] A risk heat map is a temperature map that visualizes the spatial distribution of quality risks. It uses a red, yellow, and green color gradient to represent risk levels and contour lines to mark the boundaries of the affected area, intuitively showing the cascading effects of quality anomalies on the construction area and helping on-site managers quickly locate high-risk areas.

[0144] A virtual construction environment is a simulation platform that integrates digital twins and resource constraints. The environment loads the current twin version and constraint parameters such as machine scheduling rules and material supply conditions. It simulates the execution process of handling solutions using discrete event simulation technology, outputting virtual construction period and cost data, providing a high-fidelity testing environment for decision-making.

[0145] The digital thread unit is a data pipeline connecting the physical world and the digital space. The unit encapsulates key data flows such as execution instructions, sensor readings, and decision-making schemes in a structured manner, and establishes a bidirectional association with digital twin nodes through version numbers, forming a continuous data chain that runs through physical construction and digital mapping.

[0146] Incremental learning is a method for continuous optimization of spatial matching strategies. The mechanism records each coordinate compensation amount and updates the transformation parameters by weighted averaging, allowing the spatial matching accuracy to gradually improve with the construction process. This adapts to systematic errors caused by equipment wear, environmental changes, etc., ensuring the timeliness and accuracy of the matching strategy.

[0147] Non-dominated sorting algorithms are solution set selection methods in multi-objective optimization. By comparing the dominance relationships between solution vectors, the algorithm divides the solution set into different frontier levels, automatically eliminating inferior solutions that are comprehensively surpassed by other solutions, retaining Pareto optimal solutions, and achieving the overall superiority of the output solution.

[0148] Anchoring technology for auditable operation chains is a method of binding critical data to a twin version. This technology uses cryptographic hash values ​​to associate data blocks with version nodes; any data tampering will cause a change in the hash value, thus ensuring the authenticity and integrity of operation records and meeting the data reliability requirements of engineering audits.

[0149] The feature matching algorithm in the spatiotemporal registration unit includes: extracting corner and edge feature points from the robot trajectory point cloud and laser scan point cloud; calculating the similarity matrix between feature points; determining the correspondence between feature points based on the similarity matrix; and applying an iterative nearest-point optimization algorithm to calculate the optimal rigid body transformation matrix based on the correspondence between feature points. When calculating the similarity matrix, the matching threshold is adjusted according to the point cloud density and noise level.

[0150] The data parsing in the dynamic mapping unit includes: normalizing the input data; extracting component identifiers from the normalized data; querying the ontology library to obtain the quality index items and thresholds corresponding to the component identifiers; calculating the deviation rate using the measured values ​​and thresholds; and generating an anomaly flag when the deviation rate continuously exceeds the preset threshold.

[0151] The surface reconstruction algorithm in this generation unit includes: converting point cloud data into an octree structure; calculating the point cloud normal field from the octree structure; using the normal field to solve the Poisson equation to generate implicit function surfaces; and applying the moving cube algorithm to extract triangular mesh patches from the implicit function surfaces. The time series prediction includes: employing an autoregressive integral moving average model; automatically selecting model parameters using information criteria; fitting historical deformation data using maximum likelihood estimation; and predicting future deformation based on the fitted model.

[0152] The incremental learning mechanism includes: analyzing the amplitude and frequency characteristics of the robot's work path using Fourier transform; calculating the path curvature and statistically analyzing the turning angle distribution using differential geometry methods; and establishing a path feature library based on the amplitude and frequency characteristics, path curvature, and turning angle distribution. When the GPS signal is lost, the system retrieves the regional feature pattern of the current work area from the path feature library; predicts the robot's position by combining the Kalman filter algorithm and the regional feature pattern; and calculates the coordinate offset compensation between the actual and predicted positions using the iterative nearest point algorithm.

[0153] The Monte Carlo simulation in the probability calculation unit includes: taking a quality anomaly event as the starting point of the simulation; extracting repair time data of similar cases from the historical database to generate random repair time samples; extracting conflict parameters from the resource scheduling library to generate resource conflict parameter samples; performing multiple iterative simulations using the repair time samples and conflict parameter samples; calculating the critical path delay time in each iteration; applying the kernel density estimation method to process the iteration results to generate probability distribution curves; and generating a risk heatmap based on the probability distribution curves.

[0154] The multi-objective optimization algorithm in the optimization screening unit includes: encoding the disposal plan into a two-dimensional vector of schedule and cost to form an initial population; performing crossover and mutation operations on the initial population to generate a offspring population; merging the offspring population with the parent population to form a mixed population; applying a fast non-dominated sorting algorithm to the mixed population to divide it into non-dominated levels; calculating the crowding distance between individuals within the same non-dominated level; screening the Pareto optimal solution set based on the non-dominated level and the crowding distance; and introducing resource continuity constraints and mechanical scheduling rule constraints during the screening process to eliminate individuals that violate the constraints.

[0155] The construction of an auditable operation chain in the digital thread unit includes: applying the SHA-256 algorithm to calculate the hash value of the decision scheme to generate a digital fingerprint; combining and encapsulating the event identifier, twin version number, execution log, and digital fingerprint into a data block; calculating the hash value of the data block as a chain pointer; linking the chain pointer with the hash value of the previous data block to form a hash chain structure; obtaining the Coordinated Universal Time (UTC) timestamp from the time server and writing it into the data block header; and verifying the digital signature of the timestamp to ensure the credibility of the time source.

[0156] The model update in the feedback closed-loop module includes: calculating the error value between the actual effect data and the predicted effect data of the treatment plan; calculating the gradient of the loss function based on the error value; updating the weight coefficients of the decision model using the gradient descent algorithm; extracting the deviation matrix between the actual motion trajectory point cloud and the predicted motion trajectory point cloud of the robot; calculating the partial derivative of the deviation matrix with respect to the spatial matching compensation parameters; and updating the compensation parameters in the spatial matching strategy using the backpropagation algorithm.

[0157] The innovation of this invention lies in constructing a construction progress and quality linkage analysis system based on the deep integration of BIM and intelligent robots. This invention achieves dynamic correlation and self-evolution of cross-domain data through multi-module collaboration and closed-loop data flow. Specifically, the system is as follows: the data acquisition module generates spatiotemporally aligned construction status snapshots through multi-source sensor fusion and spatiotemporal registration units; the semantic fusion module parses the raw data into structured data with semantic tags based on the railway station domain ontology library and outputs anomaly flags; the temporal twin module generates a new version of the digital twin and change vector recording the component state change sequence based on the anomaly flags; the spatial matching module optimizes the registration strategy between BIM components and robot coordinates using an incremental learning mechanism; the coupling analysis module establishes a quantitative assessment model of the impact of quality defects on the construction period and generates risk warning signals; the decision optimization module simulates multiple disposal schemes in parallel in a virtual environment and outputs a Pareto optimal solution set; the execution traceability module automatically generates machine instruction sets and constructs an auditable operation chain; and the feedback closed-loop module drives the system parameters to self-adjust through data feedback. This invention solves the problem of response delay caused by data silos between BIM models, real-time robot data and the schedule planning system by standardizing data interfaces between modules and dynamic association mechanisms, and realizes real-time linkage analysis and closed-loop control of construction progress and quality.

Claims

1. A BIM-based intelligent robot construction progress and quality linkage analysis system, characterized in that, include: The data acquisition module is used to acquire three-dimensional spatial coordinates, operation timestamps, construction process parameters, and temperature and humidity changes, structural deformation, and point cloud scan data captured by the intelligent robot's body sensors. The semantic fusion module is connected to the data acquisition module. It receives the three-dimensional spatial coordinates, operation timestamps, construction process parameters, temperature and humidity changes, structural deformation, and point cloud scanning data. It parses the data attributes through the railway station building ontology library, associates the sensor data with the quality indicators of specific components in the BIM model, generates structured data with semantic tags including component ID, indicator type, measured value, and threshold, and outputs anomaly flag bits. The temporal twin module is connected to the semantic fusion module. It receives the structured data with semantic tags and anomaly flags. When the anomaly flags indicate a quality anomaly, it injects a time dimension attribute into the BIM geometric model, generates a new version of the digital twin that records the sequence of component state changes, and outputs a change vector that represents the change of component state. The spatial matching module is connected to the temporal twin module. It receives the new version of the digital twin and the change vector. Based on the change vector and the new version of the digital twin, it optimizes the spatial matching strategy using an incremental learning mechanism, completes the registration of BIM component coordinates and robot coordinates, and outputs the component-equipment and data source mapping matrix. The coupling analysis module is connected to the spatial matching module, receives the component and equipment and data source mapping matrix, establishes a quantitative assessment model of the impact of quality defects on the project period, calculates the probability of subsequent process delays caused by quality defects and resource conflict detection results based on the quality anomaly data in the component and equipment and data source mapping matrix, and generates risk warning signals. The decision optimization module, connected to the coupled analysis module, receives the risk warning signal, simulates multiple disposal schemes in parallel in a virtual construction environment, evaluates the comprehensive impact of each disposal scheme on the total project duration and cost, and outputs the Pareto optimal solution set. The execution traceability module is connected to the decision optimization module, receives the Pareto optimal solution set, automatically generates a set of machine-executable instructions, adjusts the robot's process parameters and the logistics system's delivery priority, and constructs a digital thread to anchor key data of the disposal process to a specific version node of the digital twin, forming an auditable operation chain.

2. The BIM-based intelligent robot construction progress and quality linkage analysis system according to claim 1, characterized in that, The data acquisition module includes: The robot sensor unit uses a fusion positioning system of lidar and inertial measurement unit to record the three-dimensional coordinates of the actuator at the working end in the world coordinate system in real time, and generates a spatial location data packet including a timestamp when each single operation is completed. The environmental monitoring unit independently deploys a micro-weather station to collect air temperature and humidity gradients at a first preset interval, and deploys a structural monitoring unit to capture strain fluctuations of the steel support frame at a second preset frequency. At the same time, a laser scanning station is deployed to emit a scanning beam through a rotating prism system and receive reflected signals to generate point cloud data frames. The spatiotemporal registration unit connects the robot sensor unit and the environmental monitoring unit, establishes a unified spatiotemporal reference axis, inputs the robot trajectory point cloud and the laser scanning point cloud into a feature matching algorithm, extracts feature points from the two types of point clouds for similarity calculation, eliminates coordinate system deviations between devices through an optimization algorithm, and outputs spatiotemporally aligned fused data as a snapshot of the construction status.

3. The BIM-based intelligent robot construction progress and quality linkage analysis system according to claim 2, characterized in that, The semantic fusion module includes: The ontology library construction unit is based on the tree-structured semantic association rules that define component types, construction processes, and quality indicators according to railway station building engineering specifications. The dynamic mapping unit is connected to the ontology library construction unit. It receives three-dimensional spatial coordinates, operation timestamps, construction process parameters, temperature and humidity changes, structural deformation, and point cloud scan data from the data acquisition module. It queries the component IDs bound to the current operation robot, matches index items and thresholds according to the ontology library, and outputs structured data with component IDs, index types, measured values, thresholds, and anomaly flags. The data output unit is connected to the dynamic mapping unit and transmits the structured data to the time-series twin module.

4. The BIM-based intelligent robot construction progress and quality linkage analysis system according to claim 3, characterized in that, The time-series twin module includes: The version generation unit imports the BIM model from the design phase as the initial version. After the process is completed, it generates a point cloud model through laser scanning, generates triangular mesh patches through surface reconstruction algorithm, compares the deviation vector between the design model and the measured model, generates a new version and records the change log. The prediction evolution unit, connected to the version generation unit, automatically starts the new version modeling process when the abnormal flag output by the semantic fusion module indicates that the displacement of the component continues to increase. It extracts the most recent scan data to establish a time series model, predicts future deformations, and generates a twin version containing the predicted state. The state chain management unit, connected to the prediction evolution unit, associates the version sequence as a timeline and supports historical state backtracking.

5. The BIM-based intelligent robot construction progress and quality linkage analysis system according to claim 4, characterized in that, The spatial matching module includes: The initial matching unit is rigidly aligned with the robot's GPS coordinates by the center coordinates of the BIM component, and the average positioning error is calculated. The learning optimization unit is connected to the initial matching unit, records the robot's typical operation path, and establishes a path feature library, which includes amplitude frequency and turning angle patterns. The prediction compensation unit, connected to the learning optimization unit, calls the current working area feature mode when the GPS signal is lost, combines the motion vector to predict the next position, calculates the coordinate offset compensation amount through point cloud feature points, and updates the spatial matching strategy of the spatial matching module.

6. The BIM-based intelligent robot construction progress and quality linkage analysis system according to claim 5, characterized in that, The coupling analysis module includes: The input receiving unit receives the component-device and data source mapping matrix and identifies abnormal quality data. The historical retrieval unit is connected to the input receiving unit and retrieves the average repair time of similar quality anomaly cases in the stored historical data. The conflict detection unit is connected to the historical retrieval unit and detects resource conflicts in subsequent processes. The resource conflicts include conflicts between work group entry time and conflicts between equipment occupancy time. The probability calculation unit, connected to the conflict detection unit, calculates the critical path delay probability and generates a risk heat map to indicate the impact range.

7. The BIM-based intelligent robot construction progress and quality linkage analysis system according to claim 6, characterized in that, The decision optimization module includes: The environment construction unit imports the current digital twin version generated by the time-series twin module and loads resource constraints, including machine scheduling rules and material supply conditions. The scheme simulation unit is connected to the environment construction unit and simulates multiple disposal schemes in parallel. The disposal schemes include a work stoppage detection scheme, a speed reduction operation scheme, and a backup team activation scheme. The virtual construction period and cost of each disposal scheme are output. The optimization screening unit is connected to the scheme simulation unit to evaluate the comprehensive impact of each disposal scheme on the total project duration and cost, eliminate inferior schemes, and generate a Pareto optimal solution set.

8. The BIM-based intelligent robot construction progress and quality linkage analysis system according to claim 7, characterized in that, The execution traceability module includes: The instruction generation unit receives the Pareto optimal solution set, analyzes the resource requirements in the selected solution, and generates logistics system instructions and robot parameter adjustment instructions. An execution recording unit, connected to the instruction generation unit, records the instruction issuance time and execution confirmation time; The digital thread unit, connected to the execution record unit, anchors key data of the processing to a specific version node of the digital twin, and establishes an auditable operation chain including event ID, associated twin version number, decision scheme fingerprint, execution log and effect verification data.

9. The BIM-based intelligent robot construction progress and quality linkage analysis system according to claim 8, characterized in that, The system further includes a feedback closed-loop module, connecting the execution tracing module and the data acquisition module. The feedback closed-loop module is used for: Receive new data acquired by the data acquisition module after the instruction is executed, the new data including quality recovery status data and progress data; The new data is sent to the semantic fusion module to generate structured data in a normal state; The time-series twin module is triggered to generate a new version record verification result; Update the spatial matching strategy of the spatial matching module and the decision model weight coefficients of the decision optimization module.

10. The BIM-based intelligent robot construction progress and quality linkage analysis system according to claim 9, characterized in that, Also includes: The data acquisition module outputs spatiotemporally aligned fused data as a snapshot of the construction status to the semantic fusion module; The semantic fusion module outputs structured data with semantic tags and anomaly flags to the temporal twin module; The temporal twin module outputs a new version of the digital twin and a change vector representing the change in the state of the component to the spatial matching module; The spatial matching module outputs a component-device and data source mapping matrix to the coupling analysis module; The coupling analysis module outputs a risk warning signal to the decision optimization module; The decision optimization module outputs the Pareto optimal solution set to the execution traceability module; The execution traceability module outputs the machine-executable instruction set and auditable operation chain, and feeds back the execution data to the data acquisition module and the update information to the spatial matching module.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the BIM-based intelligent robot construction progress and quality linkage analysis system according to any one of claims 1 to 10.

Images