An intelligent error detection method and system for special-shaped flow channels based on BIM data

Abstract

The application discloses a special-shaped runner intelligent error detection method and system based on BIM data, and relates to the technical field of computer-aided design.The method comprises the following steps: obtaining BIM model data and point cloud data of a special-shaped runner, performing semantic analysis and feature enhancement on the BIM model to generate semantic-enhanced BIM model data; extracting structural semantic features and assembly semantic features to construct a semantic feature correlation graph; based on the graph, dynamically aligning the point cloud and the semantic-enhanced BIM model to obtain semantic registration fusion data; then constructing a coupling error solving model, calculating three kinds of coupling errors respectively, binding category semantic labels, and forming a coupling error data set; further removing non-correlation error data, associating the non-correlation error data with structural semantic labels, and generating a heat map detection report containing error traceability links to complete error detection.The application can accurately detect the geometric and assembly errors of the special-shaped runner based on BIM data.

Description

A method and system for intelligent error detection of irregularly shaped flow channels based on BIM data Technical Field

[0001] This application relates to the field of computer-aided design technology, and in particular to an intelligent error detection method and system for irregular flow channels based on BIM data. Background Technology

[0002] In complex engineering projects such as modern large-scale public buildings, industrial plants, and data centers, HVAC, water supply and drainage, and process piping systems often contain a large number of non-standard, free-form, or complex irregularly shaped flow channel components. These components have extremely high requirements for airflow / water flow performance, spatial arrangement, and construction precision; their manufacturing and installation errors directly affect system energy efficiency, operational safety, and maintenance costs. With the widespread application of Building Information Modeling (BIM) technology, high-precision three-dimensional semantic models of irregularly shaped flow channels can be constructed during the design phase, providing a data foundation for subsequent construction quality control.

[0003] Currently, error detection for flow channel components mainly relies on traditional measurement methods, such as total stations, laser scanning combined with manual comparison, or general deviation analysis software (such as Geomagic and PolyWorks) based on point clouds and computer-aided design (CAD) models. Some projects have attempted to register on-site scanning data with BIM models and display deviation areas through visual chromatograms.

[0004] However, these methods are mostly geared towards regular pipe fittings (such as straight pipes, elbows, and tees), lacking a semantic understanding of the geometric features of irregularly shaped flow channels. Furthermore, error assessment typically relies on manually set thresholds based on experience, making it difficult to distinguish between design intent and construction deviations, and even more difficult to automatically identify semantic-level errors such as assembly misalignment and interface mismatch. Therefore, how to accurately detect geometric and assembly errors of irregularly shaped flow channels based on BIM data has become an urgent problem to be solved. Summary of the Invention

[0005] The purpose of this application is to provide a method and system for intelligent error detection of irregular flow channels based on BIM data, aiming to solve the technical problem of how to accurately detect the geometric and assembly errors of irregular flow channels based on BIM data.

[0006] To achieve the above objectives, this application proposes an intelligent error detection method for irregularly shaped flow channels based on BIM data, the method comprising:

[0007] The original BIM model data and multi-view time-series scan point cloud data of the irregular flow channel are obtained, and the original BIM model data is subjected to semantic parsing and feature enhancement to obtain semantically enhanced BIM model data.

[0008] Semantic features are extracted hierarchically from the semantically enhanced BIM model data, separating structural semantic features from assembly semantic features, and a semantic feature association map is constructed.

[0009] Based on the semantic feature association map, the multi-view temporal scan point cloud data after temporal denoising is dynamically aligned with the semantically enhanced BIM model data through a feature weight iterative registration algorithm to obtain semantic registration fusion data.

[0010] Based on the semantic registration and fusion data, a coupling error calculation model is constructed, and the surface topology error, assembly datum coupling error and flow channel conductivity error are calculated according to the coupling error calculation model to obtain a coupling error dataset with associated category semantic labels.

[0011] By using a semantic association threshold filtering mechanism, the coupling error dataset is checked for error coupling, and non-associative error data is removed to obtain the target coupling error data.

[0012] The target coupling error data is associated with the structural semantic tags of the semantically enhanced BIM model data to generate a heatmap detection report containing the error source tracing link, thus completing the error detection.

[0013] Furthermore, to achieve the above objectives, this application also proposes an intelligent error detection system for irregularly shaped flow channels based on BIM data, the system comprising:

[0014] The semantic enhancement module is used to acquire the original BIM model data and multi-view time-series scan point cloud data of the irregular flow channel, and to perform semantic parsing and feature enhancement on the original BIM model data to obtain semantically enhanced BIM model data.

[0015] The hierarchical extraction module is used to perform hierarchical extraction of semantic features from the semantically enhanced BIM model data, separate structural semantic features from assembly semantic features, and construct a semantic feature association map.

[0016] The iterative registration module is used to dynamically align the temporally denoised multi-view temporal scan point cloud data with the semantically enhanced BIM model data based on the semantic feature association map and through a feature weight iterative registration algorithm to obtain semantic registration fusion data.

[0017] The coupling error calculation module is used to construct a coupling error calculation model based on the semantic registration fusion data, and calculate the surface topology error, assembly datum coupling error and flow channel conductivity error according to the coupling error calculation model to obtain a coupling error dataset with associated category semantic labels.

[0018] The semantic association filtering module is used to perform error coupling verification on the coupling error dataset through a semantic association threshold filtering mechanism, remove non-associative error data, and obtain the target coupling error data.

[0019] The source tracing report generation module is used to associate the target coupling error data with the structural semantic tags of the semantically enhanced BIM model data, generate a heat map detection report containing the error source tracing link, and complete the error detection.

[0020] Furthermore, to achieve the above objectives, this application also proposes an intelligent error detection device for irregularly shaped flow channels based on BIM data. The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. The computer program is configured to implement the steps of the intelligent error detection method for irregularly shaped flow channels based on BIM data as described above.

[0021] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of the intelligent error detection method for irregular flow channels based on BIM data as described above.

[0022] One or more technical solutions proposed in this application have at least the following technical effects:

[0023] First, the original BIM model data and multi-view time-series scan point cloud data of the irregular flow channel are acquired. Semantic parsing and feature enhancement are performed on the original BIM model data to obtain semantically enhanced BIM model data, providing a semantically complete and geometrically accurate benchmark for subsequent analysis. Second, semantic features are extracted hierarchically from this semantically enhanced BIM model data, separating the structural semantic features describing the flow channel's morphology from the assembly semantic features describing the connection relationships. A semantic feature association map is constructed to link the two, providing clear semantic logical support for error analysis. Then, based on this map, a feature weight iterative registration algorithm is used to dynamically align the time-series denoised multi-view time-series scan point cloud data with the semantically enhanced BIM model data, resulting in semantic registration fusion. The data is then processed as follows: First, a coupling error calculation model is constructed based on the fused data to calculate surface topology errors, assembly datum coupling errors, and flow channel connectivity errors. Each type of error is then assigned a corresponding semantic label, forming a structured coupling error dataset. This enables simultaneous quantification and classification of multi-dimensional errors. Next, a semantic association threshold filtering mechanism is used to verify the coupling error dataset, eliminating non-associative error data that cannot be traced back through the semantic graph or has insufficient coupling strength. Target coupling error data with engineering authenticity is retained, effectively suppressing false alarms. Finally, the target coupling error data is associated with the structural semantic labels of the semantically enhanced BIM model data to generate a heatmap detection report containing error tracing links, completing the entire closed-loop detection process. This application can accurately detect geometric and assembly errors of irregularly shaped flow channels based on BIM data. Attached Figure Description

[0024] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0025] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 is a flowchart of the intelligent error detection method for irregular flow channels based on BIM data provided in Embodiment 1 of this application;

[0027] Figure 2 is a flowchart of the second embodiment of the intelligent error detection method for irregular flow channels based on BIM data in this application.

[0028] Figure 3 is a schematic diagram of the module structure of the intelligent error detection system for irregular flow channels based on BIM data according to an embodiment of this application;

[0029] Figure 4 is a schematic diagram of the hardware operating environment of the intelligent error detection method for irregular flow channels based on BIM data in the embodiments of this application.

[0030] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0031] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0032] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0033] It should be noted that the executing entity of this application embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device or error detection system capable of realizing the above functions. The following description uses an error detection system as an example to illustrate this embodiment and the subsequent embodiments.

[0034] Based on this, this application provides an intelligent error detection method for irregularly shaped flow channels based on BIM data. Referring to Figure 1, Figure 1 is a flowchart of the first embodiment of the intelligent error detection method for irregularly shaped flow channels based on BIM data of this application.

[0035] In this embodiment, the intelligent error detection method for irregular flow channels based on BIM data includes steps S10 to S60:

[0036] Step S10: Obtain the original BIM model data and multi-view time-series scan point cloud data of the irregular flow channel, and perform semantic parsing and feature enhancement on the original BIM model data to obtain semantically enhanced BIM model data.

[0037] Step S20: Extract semantic features hierarchically from the semantically enhanced BIM model data, separate structural semantic features from assembly semantic features, and construct a semantic feature association map;

[0038] Step S30: Based on the semantic feature association map, the multi-view temporal scan point cloud data after temporal denoising is dynamically aligned with the semantically enhanced BIM model data through a feature weight iterative registration algorithm to obtain semantic registration fusion data;

[0039] Step S40: Based on the semantic registration and fusion data, construct a coupling error calculation model, and calculate the surface topology error, assembly datum coupling error and flow channel conductivity error according to the coupling error calculation model to obtain a coupling error dataset of related category semantic labels;

[0040] Step S50: The coupling error dataset is checked for error coupling through a semantic association threshold filtering mechanism. Non-associative error data is removed to obtain the target coupling error data.

[0041] Step S60: Associate the target coupling error data with the structural semantic tags of the semantically enhanced BIM model data to generate a heatmap detection report containing the error source tracing link, thus completing the error detection.

[0042] It should be noted that irregular flow channels refer to pipe or channel components with non-standard geometric shapes, free-form surfaces, variable cross-sections, or complex topologies. These are commonly found in HVAC and industrial process systems, and their forms cannot be described using conventional pipe fittings (such as straight pipes and elbows). Raw BIM model data refers to the unprocessed 3D digital model of irregular flow channels directly exported from BIM, containing geometric information (such as vertices and faces), basic attributes (such as materials and dimensions), and some semantic information (such as component type and connection relationships). Multi-view temporal scan point cloud data refers to the set of spatial points on the surface of irregular flow channels collected from multiple observation angles and at different time points (temporal sequence) using equipment such as laser scanners at the construction or manufacturing site. Each point contains three-dimensional coordinates (x, y, z) to reflect the actual physical state. Semantic-enhanced BIM model data refers to an enhanced BIM model rich in semantic and geometric context information, formed by semantically parsing the raw BIM model data (such as identifying flow channel centerlines, interface flanges, support nodes, etc.) and enhancing its feature expression (such as adding curvature labels, flow direction vectors, assembly constraints, etc.). Structural semantic features refer to semantic information describing the geometric structure and functional attributes of irregular flow channels, such as the flow channel center path, cross-sectional variation law, curvature distribution, and wall thickness constraints, reflecting "what it is" and "how it is constructed". Assembly semantic features refer to semantic information describing the connection, positioning, and fit between irregular flow channels and other components, such as the flange interface coordinate system, bolt hole positions, alignment reference planes, and adjacent equipment numbers, reflecting "how it is installed" and "which it is associated with".

[0043] A semantic feature association graph is a data model that organizes structural and assembly semantic features in a graph structure. Nodes represent specific semantic elements, and edges represent logical, spatial, or assembly relationships between them, supporting subsequent intelligent registration and error analysis. The feature weight iterative registration algorithm is a registration method that integrates semantic feature importance assessment with the point cloud-model iterative closest point (ICP) concept. It first dynamically assigns registration weights to different regions based on the semantic feature association graph, and then achieves high-precision alignment through multiple rounds of iterative optimization. Semantic registration fusion data refers to a fused dataset generated after spatial alignment of multi-view temporal scan point cloud data and semantically enhanced BIM model data under semantic guidance. The correspondence between point clouds and models is established, and the original semantic labels are preserved. The coupled error solution model is a mathematical model that integrates multiple error calculation logics, simultaneously considering geometric deviations and assembly constraints, used to jointly solve different types of interrelated error indices. Topology error refers to the geometric deviation between the actual scanned surface of an irregular flow channel and the surface designed in the BIM, including local concavity / convexity, overall distortion, or topological connection errors (e.g., breaks instead of continuity). It is usually measured by the distance from a point to the surface. Assembly datum coupling error refers to the misalignment of the interface position or direction caused by the installation deviation of adjacent components, manifested as non-parallel flange faces, misaligned bolt holes, etc. This error not only involves individual flow channels but also couples to the state of their assembly partners.

[0044] Flow channel continuity error refers to semantic-level errors caused by manufacturing or installation deviations that obstruct the internal flow path, cause abrupt changes in cross-section, or deviate in direction, affecting the flow performance of the medium. For example, an excessively large bend angle can cause airflow separation (even with small geometric deviations, the function is still impaired). Category semantic labels are standardized semantic identifiers assigned to each type of error (e.g., "surface deformation," "flange misalignment," "conductivity obstruction"), used to map numerical error results to interpretable engineering problem categories. Coupled error datasets are structured datasets containing various error values ​​and their corresponding category semantic labels. Each data item records the error location, magnitude, type, and associated BIM component information. The semantic association threshold filtering mechanism uses logical rules and numerical thresholds set based on semantic feature association maps to determine whether an error is truly caused by design-construction coupling, rather than measurement noise or isolated deviations. Non-associative error data refers to error points or areas determined by the semantic association threshold filtering mechanism to have no logical association with the BIM semantic structure, such as spurious deviations caused by temporary occlusion, scanning noise, or reflections from irrelevant objects. Target coupling error data refers to real error data that has been filtered and retained, possessing clear semantic attribution and engineering significance, and can be used for subsequent analysis, early warning, or rework decisions. Error tracing chain refers to the complete path from the detected error point, tracing back through semantic tags to specific components, interfaces, or design parameters in the BIM model, realizing a traceable chain of "error → semantic tag → model location." Heatmap detection report is a visual result that overlays target coupling error data onto the BIM model in the form of color gradients, with red indicating high-error areas and blue indicating low-error areas, embedding error type tags and tracing information to facilitate engineers' rapid problem location.

[0045] Understandably, the error detection system first imports the original BIM model data of the irregular flow channel and acquires multi-view time-series scan point cloud data from the construction site at multiple time points and from different perspectives. Then, the system performs automated semantic parsing on the original BIM model data, identifying key components such as the flow channel centerline, interface flanges, and support nodes. Based on this, it injects enhanced features such as curvature variations, flow vectors, and assembly constraints to generate semantically enhanced BIM model data, improving the semantic perception capabilities for subsequent analysis. Secondly, the system performs layered processing on the semantically enhanced BIM model data: on the one hand, it extracts structural semantic features describing the flow channel's own geometry; on the other hand, it extracts assembly semantic features reflecting its connection relationships with other components. These two types of features are then organized into a semantic feature association graph through a graph structure, providing semantic guidance for subsequent accurate registration.

[0046] Then, the system first performs temporal denoising on the multi-view temporal scan point cloud data to eliminate dynamic interference and measurement noise. Next, based on the semantic feature association map, it dynamically assigns registration weights to different regions (e.g., assigning higher weights to interface regions). Using a feature weight iterative registration algorithm, it aligns the point cloud with the semantically enhanced BIM model data frame by frame, ultimately fusing them to generate spatially consistent and semantically aligned semantic registration fusion data. Following this fusion data, the system constructs a coupling error calculation model. This model considers both geometric shape and assembly logic, calculating the surface topology error between the irregular flow channel surface and the design curved surface, the assembly datum coupling error caused by deviations between adjacent components at the interface, and the flow channel connectivity error caused by path offset or abrupt cross-section changes. It automatically binds corresponding category semantic labels to each error result, forming a structured coupling error dataset.

[0047] Subsequently, the system invokes a preset semantic association threshold filtering mechanism. Based on the logical relationships and engineering tolerance rules in the semantic feature association graph, it performs coupling verification on each error in the coupling error dataset. If an error cannot be traced back to a valid BIM component or assembly relationship through a semantic path, it is determined to be non-associative error data and is removed, retaining only the target coupling error data with real engineering significance. Finally, the system accurately maps the target coupling error data to its corresponding structural semantic tags in the semantically enhanced BIM model data, establishing a complete error tracing link from the error point to the model component and then to the design parameters. The error magnitude is rendered onto the BIM model in the form of a color gradient, generating a heatmap detection report containing location, type, value, and tracing path, thereby completing high-precision, interpretable, and traceable intelligent error detection for irregularly shaped flow channels.

[0048] As an example, the steps of acquiring the original BIM model data and multi-view time-series scan point cloud data of the irregular flow channel, and performing semantic parsing and feature enhancement on the original BIM model data to obtain semantically enhanced BIM model data include: acquiring the original BIM model data and multi-view time-series scan point cloud data of the irregular flow channel; extracting flow channel structure tags, assembly relationship tags, and material attribute tags from the original BIM model data to obtain an initial semantic set; performing semantic completion on the initial semantic set to supplement flow channel topology association semantics and assembly benchmark semantics to obtain a complete semantic tag set; binding the complete semantic tag set to the original BIM model data point by point, performing semantically guided feature completion on the hidden areas of the model to obtain a semantically bound and completed model; and performing topological optimization on the semantically bound and completed model to obtain semantically enhanced BIM model data with deep fusion of semantic information and geometric features.

[0049] It should be noted that the flow channel structure label refers to the semantic identifier extracted from the original BIM model data to describe the geometric composition and functional form of the irregular flow channel itself, including structural elements such as centerline path, cross-section type, curved sections, variable diameter sections, and branch nodes. The assembly relationship label refers to the semantic identifier describing the connection method, interface type, and spatial fit between the irregular flow channel and other electromechanical components (such as equipment, valves, and adjacent pipe sections), such as "flange connected to fan outlet" or "associated with support and hanger fixing points." The material attribute label refers to the semantic information recording the material type and physical properties used in the irregular flow channel, such as "galvanized steel plate," "304 stainless steel," and "thermal conductivity W / (m·K)," used to support material-related judgments in subsequent performance or error analysis. The initial semantic set refers to the preliminary semantic label set automatically identified and summarized from the parameters, attributes, and naming rules of the original BIM model data using semantic extraction algorithms in Natural Language Processing (NLP). It includes flow channel structure labels, assembly relationship labels, and material attribute labels, but has not yet been completed or optimized. The semantic association of flow channel topology refers to the semantic information that describes the logical connection relationship between the components inside the irregular flow channel, such as the sequential connectivity of "inlet section → elbow section → straight pipe section → outlet flange", or the merging / diversion topology between multiple branch flow channels.

[0050] Assembly datum semantics refers to the semantic information used to define the positioning and alignment of irregular flow channels during installation. This includes the datum coordinate system, alignment plane, bolt hole centerline, and interface normal vector, serving as a key reference for judging assembly errors. A complete semantic tag set refers to a set of semantic tags covering all dimensions of structure, connection, material, topology, and assembly, supplemented by a knowledge graph completion algorithm on topological association semantics and assembly datum semantics, based on the initial semantic set. This set possesses engineering integrity and logical consistency. Hidden areas of the model refer to geometric areas in the original BIM model data that lack semantic annotations due to modeling simplification, view occlusion, or non-explicit expression. Examples include pipe sections covered by insulation layers, interface parts embedded in walls, or internal flow channel surfaces not explicitly modeled. A semantically bound and completed model refers to an intermediate model generated by binding the complete semantic tag set to the original BIM model data point-by-point or face-by-face, and then using semantically guided reasoning (such as inferring missing parts based on adjacent known semantic areas) to complete the hidden areas of the model. Its geometric and semantic information is basically aligned and comprehensively covered.

[0051] Understandably, firstly, the error detection system retrieves the original BIM model data of the irregular flow channel from the semantic database of the project's BIM collaboration platform, and simultaneously controls the 3D laser scanning equipment deployed on-site to continuously collect point clouds at preset time intervals (e.g., every 0.5 seconds) from multiple fixed observation poses surrounding the flow channel, forming multi-view time-series scanning point cloud data covering the entire construction process to ensure that the true geometric state of the flow channel is captured at different installation stages. Secondly, the system calls a semantic extraction algorithm to perform structured parsing of the naming strings, parameter attribute tables, and family type information of each component in the original BIM model data, identifying and extracting flow channel structure tags (such as "variable diameter section", "90° bend", "main pipe branch node"), assembly relationship tags (such as "flange connected to AHU-01 outlet" "fixed with support SH-23"), and material attribute tags (such as "material: stainless steel 304, thermal conductivity: 16.2 W / (m·K)"), integrating these tags into an initial semantic set. Then, the system inputs the initial semantic set into a pre-trained electromechanical domain knowledge graph, and completes the missing high-level semantics through graph reasoning. Specifically, it derives the topological association semantics of the flow channel based on the connection order between flow channel segments (e.g., "inlet → elbow A → straight pipe B → tee C → outlet 1 / outlet 2"), and derives the assembly reference semantics based on the geometric constraints of the interface components (e.g., "flange end face is XY plane, bolt hole center circle diameter is xx mm, normal vector is along Z axis"), thereby generating a logically closed and engineering-usable complete semantic tag set.

[0052] Next, the system binds each semantic tag in the complete semantic tag set to its corresponding geometric element (vertices, edges, or faces) in the original BIM model data point by point. For hidden areas in the model that are not explicitly modeled or are occluded (such as wall penetrations and insulation wrapping areas), a semantically guided feature completion method is used. This method utilizes the semantic type, flow direction, and cross-sectional patterns of adjacent labeled areas to infer and fill in the missing semantic tags through interpolation or template matching, generating a semantically complete model without gaps. Finally, the system performs topology optimization on this semantically complete model: it calls a mesh simplification algorithm based on Quadric Error Metrics to merge redundant triangular faces and eliminate self-intersections and non-manifold edges while preserving key semantic boundaries (such as interface edges and curvature change lines). At the same time, it ensures that the deviation between the simplified geometry and the original design does not exceed a preset tolerance (e.g., 0.5 mm). The final output is a semantically enhanced BIM model data that is geometrically lightweight, topologically compliant, and deeply integrated with semantic information and geometric features, providing a reliable benchmark for subsequent high-precision error detection.

[0053] As an example, the steps of constructing a coupling error calculation model based on the semantic registration fusion data include: extracting structural semantic feature parameters, assembly semantic feature parameters, and point cloud fitting parameters from the semantic registration fusion data to obtain a model input dataset; constructing an initial model based on the model input dataset, the coupling correlation degree in the semantic feature association map, and the NSGA-III algorithm; setting constraints for the initial model to obtain a reference model, wherein the constraints include semantic label association accuracy, error calculation threshold, and data adaptation range; and using the least squares method to correct the parameters of the reference model to obtain the coupling error calculation model.

[0054] It should be noted that structural semantic feature parameters refer to numerical indicators extracted from semantic registration and fusion data that describe the geometric structure of the irregular flow channel, including centerline curvature, cross-sectional area change rate, wall thickness deviation, and surface normal continuity, used to characterize the consistency between its structural form and design intent. Assembly semantic feature parameters refer to quantitative indicators extracted from semantic registration and fusion data that reflect the connection status of the irregular flow channel with other components, such as flange surface flatness, bolt hole position deviation, interface coordinate system rotation angle error, and spacing between adjacent components, used to evaluate assembly alignment accuracy. Point cloud fitting parameters refer to the fitting quality indicators obtained after fitting multi-view time-series scan point cloud data onto the geometric surface of the BIM model, including the average distance from the point to the surface, standard deviation, residual distribution entropy, and local goodness of fit (e.g., R²). 2 The coefficient of variation (COP) measures the geometric fit between the measured data and the model. The model input dataset is a structured set composed of three types of data: structural semantic feature parameters, assembly semantic feature parameters, and point cloud fitting parameters. It serves as the original observation basis for the coupling error solution model. The coupling correlation degree refers to the numerical value of the correlation strength between different semantic feature nodes (such as a pipe bend and a downstream flange) in the semantic feature correlation graph due to topological connections or assembly constraints. It is used to quantify the degree of error propagation and mutual influence between the structural and assembly dimensions.

[0055] The NSGA-Ⅲ algorithm (Non-dominated Sorting Genetic Algorithm III) is an evolutionary computational method for high-dimensional multi-objective optimization. Here, it is used to search for Pareto optimal solutions among multiple conflicting error objectives (such as minimizing surface errors and assembly errors) to construct the parameter combination of the initial model. The initial model refers to an unconstrained mathematical model generated by the NSGA-Ⅲ algorithm based on the coupling correlation degree in the model input dataset and semantic feature association graph, containing multi-objective error solution functions and their preliminary parameter configurations. Constraints refer to the restrictive rules imposed to ensure the practicality of the model in engineering, including three core requirements: semantic label association accuracy, error solution threshold, and data adaptation range. Semantic label association accuracy means that when solving errors, the model must ensure that the matching accuracy between error points and their corresponding BIM component semantic labels is not lower than a preset value (e.g., 95%) to avoid misjudgments caused by semantic mismatches. Error threshold refers to the upper limit tolerance value set for different types of errors (e.g., surface topology error not exceeding 3mm, assembly datum coupling error not exceeding 1.5mm). Solution results exceeding this threshold will be marked as abnormal or invalid. Data adaptation range refers to the model's validity only for input data that meets specific geometric complexity, point cloud density (e.g., ≥50 points / cm²), and semantic integrity (e.g., label coverage ≥90%). Data exceeding this range will be excluded to ensure solution stability. Reference model refers to an intermediate model with engineering interpretability and physical rationality, formed by introducing the above constraints for feasibility screening and boundary definition based on the initial model. It serves as the basis for subsequent parameter fine-tuning.

[0056] Understandably, the error detection system first traverses each irregular flow channel component in the semantic registration and fusion data. For the fusion results after geometric and semantic alignment, it performs three types of parameter extraction operations: First, it calls the curvature analysis and section sampling modules to calculate the curvature values ​​of each node on the centerline, the area ratio of adjacent sections, and the wall thickness deviation, forming structural semantic feature parameters. Second, based on the bound assembly semantic tags, it parses the fitted plane of the interface flange and calculates its flatness and normal vector deflection angle, while comparing the Euclidean distance between the measured coordinates and design coordinates of the bolt holes to generate assembly semantic feature parameters. Third, it uses the nearest point projection results from the point cloud to the BIM surface to statistically analyze the residual mean, standard deviation, and local R-squared values ​​of all points. 2The goodness of fit and residual distribution entropy constitute the point cloud fitting parameters. These three types of parameters are then aligned and packaged according to component ID and timestamp to construct a complete model input dataset, providing high-dimensional observational basis for subsequent modeling. Next, the system reads the pre-stored coupling correlation matrix from the semantic feature association map (this matrix quantifies the mutual influence weights between structural features and assembly features through graph neural networks or expert rules), and inputs the model input dataset along with this matrix into the NSGA-Ⅲ algorithm framework. Within this framework, the system sets three optimization objective functions (corresponding to surface topology error, assembly datum coupling error, and flow channel conductivity error, respectively), runs a multi-generational genetic evolution process, and searches for a set of Pareto optimal solutions in the solution space that balance multi-objective performance and have a uniform distribution through non-dominated sorting and reference point guidance mechanisms. The solution with the highest comprehensive score is selected as the initial model's function structure and initial parameter values.

[0057] Next, the system imposes three hard constraints on the initial model: First, when backtracking error points, the associated BIM semantic tags must be consistent with the original annotations, and the overall matching accuracy must not be lower than the semantic tag association accuracy to prevent semantic drift; second, engineering tolerance limits are set for various types of errors, and exceeding these limits renders the solution invalid; third, the model is limited to high-quality inputs with a point cloud density ≥ 50 points / cm² and a semantic tag coverage ≥ 90%, thereby filtering low-quality scenarios, ensuring model robustness, and ultimately generating a reference model that meets engineering feasibility requirements. Finally, the system uses the least squares method to fine-tune the adjustable coefficients (such as error weighting factors and coupling attenuation coefficients) in the reference model: by constructing an objective function for the sum of squared errors between the observation residual vector and the model's predicted output, the optimal parameters are solved using gradient descent or QR decomposition, making the model output approximate the actual measurement deviation as closely as possible. After multiple iterations until convergence, a coupling error calculation model with accurate parameters, strong generalization ability, and the ability to simultaneously calculate multiple types of coupling errors is finally output.

[0058] As an example, the steps of constructing an initial model based on the model input dataset, the coupling correlation degree in the semantic feature association graph, and the NSGA-III algorithm include: fusing the model input dataset and the coupling correlation degree in the semantic feature association graph according to the corresponding feature dimensions to obtain target input data; importing the target input data into the NSGA-III algorithm with the target accuracy of solving surface topology error, assembly datum coupling error, and flow channel conductivity error; iteratively adjusting the correlation weights of each feature parameter in the target input data through the NSGA-III algorithm to obtain target parameter weights; and building a process framework for data input, weight allocation, and error calculation based on the target input data and the target parameter weights to obtain the initial model.

[0059] It should be noted that the target input data refers to the new dataset formed by dimensional alignment and weighted fusion of the structural semantic feature parameters, assembly semantic feature parameters, and point cloud fitting parameters from the model input dataset according to their corresponding coupling correlation in the semantic feature correlation map. Each feature dimension is supplemented with information on its correlation strength with other features in error propagation, serving as the optimization input basis for the NSGA-III algorithm. The target parameter weights refer to a set of optimal weight coefficients obtained through iterative search during the multi-objective optimization process of the NSGA-III algorithm. These weights are used to quantify the relative contributions of the structural semantic feature parameters, assembly semantic feature parameters, and point cloud fitting parameters in solving surface topology errors, assembly datum coupling errors, and flow channel continuity errors. This weighting ensures that the overall solution accuracy of the three types of errors reaches Pareto optimality. The process framework refers to an executable computational structure consisting of a data input module, a weight allocation module, and an error calculation module connected sequentially. The data input module receives the target input data, the weight allocation module dynamically assigns weights to each feature parameter according to the target parameter weights, and the error calculation module calls the corresponding error calculation sub-functions (such as surface deviation integral, interface pose error decomposition, and flow section connectivity verification) based on the weighted parameters, ultimately outputting three types of coupled error results. The framework itself is not a single mathematical formula, but an error calculation process with determined parameters (i.e., fixed weights), a closed logical loop, and reproducible, which is the specific implementation of the initial model.

[0060] Understandably, the error detection system first aligns the dimensions of each structural semantic feature parameter, assembly semantic feature parameter, and point cloud fitting parameter in the model input dataset according to their corresponding coupling correlation degree in the semantic feature correlation map. For example, the coupling correlation degree value (e.g., 0.78) between the "bend curvature" parameter and its downstream "flange interface flatness" parameter is used as a fusion coefficient to perform weighted concatenation or tensor concatenation on the original parameters, generating a high-dimensional vector containing semantic interaction information. Finally, the data of all components and time frames are integrated to form the target input data. This is done so that the subsequent optimization process can explicitly perceive the engineering coupling relationship between different features and avoid isolated processing that leads to error calculation distortion. Secondly, the system imports the target input data into the NSGA-Ⅲ algorithm and explicitly sets three optimization objectives: minimizing the deviation between the predicted and measured values ​​of surface topology error, minimizing the residual of assembly datum coupling error, and maximizing the recognition accuracy of flow channel conductivity error. During algorithm operation, the system continuously adjusts the correlation weight of each feature parameter in error calculation through multiple generations of evolutionary iterations (for example, increasing the weight of the "interface normal vector" in the assembly error objective, or reducing the influence of point cloud fitting parameters in low signal-to-noise ratio regions). It retains diverse high-quality solutions using non-dominated sorting and guides the population to converge towards high-precision regions through preset reference points. Finally, it outputs a set of target parameter weights that optimize the comprehensive performance of the three types of errors. The purpose of this is to automatically balance the sensitivity of various errors under multi-objective conflict and avoid subjective bias caused by manual parameter tuning. Finally, based on the determined target input data structure and the converged target parameter weights, the system constructs a fixed three-layer computational process framework: the first layer receives the target input data, the second layer performs linear or nonlinear weighting on each feature parameter according to the target parameter weights, and the third layer performs parallel calculations on surface topology errors, assembly datum coupling errors, and flow channel conductivity errors, respectively, and outputs three types of error results. This process framework is the initial model, and its structure and parameters are fixed, which can be directly used for subsequent constraint screening and fine-tuning, thereby ensuring that the error solution process has both a multi-objective optimization basis and engineering feasibility.

[0061] As an example, the step of calculating the surface topology error, assembly datum coupling error, and flow channel conductivity error according to the coupling error calculation model to obtain a coupling error dataset with associated category semantic labels includes: extracting target parameter weights from the coupling error calculation model, wherein the target parameter weights include structural semantic feature weights, assembly semantic feature weights, and point cloud fitting parameter weights; calculating the surface topology error according to the structural semantic feature parameters, the structural semantic feature weights, and preset standard structural parameters; calculating the assembly datum coupling error according to the assembly semantic feature parameters, the assembly semantic feature weights, and preset standard assembly parameters; calculating the flow channel conductivity error according to the point cloud fitting parameters, the point cloud fitting parameter weights, and preset standard conductivity parameters; and binding the surface topology error, the assembly datum coupling error, and the flow channel conductivity error with corresponding category semantic labels according to the semantic feature association map to obtain a coupling error dataset with associated semantic labels.

[0062] It should be noted that the structural semantic feature weights (denoted as...) ), assembly semantic feature weights (denoted as ) and the weights of the point cloud fitting parameters (denoted as ) ) refers to the relative importance coefficients of the three types of characteristic parameters determined iteratively by the NSGA-III algorithm during multi-objective optimization in the error calculation, satisfying the normalization constraint ( + + =1), and its allocation ratio is consistent with the structure-assembly-geometric coupling strength reflected in the semantic feature association map, which is used to dynamically adjust the contribution of various errors to the final result.

[0063] Preset standard structural parameters (denoted as) ( ) refers to the set of reference values ​​defined in the design phase to describe the ideal geometry of irregular flow channels, including standard centerline path, theoretical cross-sectional dimensions, design curvature distribution and wall thickness, etc., which serve as a reference for calculating surface topology errors.

[0064] Preset standard assembly parameters (denoted as) ( ) refers to the ideal state parameters of the assembly interface that are explicitly defined in the BIM model, including the standard plane equation of the flange end face, the design coordinates of the bolt holes, the direction of the interface normal vector, the alignment reference coordinate system, etc., which are used to measure the deviation between the actual installation position and the design intent.

[0065] Preset standard conduction parameters (denoted as) ( ) refers to the flow performance benchmark indicators required to ensure the normal function of the flow channel, such as minimum flow cross-sectional area, maximum allowable flow direction deviation angle, and continuous unobstructed centerline connectivity threshold, which are used to determine whether there is functional blockage or serious deviation in the flow channel.

[0066] Understandably, surface topology errors The calculation formula is as follows:

[0067]

[0068] in, It refers to the structural semantic feature parameters in semantic registration and fusion data.

[0069] Assembly datum coupling error The calculation formula is as follows:

[0070]

[0071] in, It refers to the assembly semantic feature parameters in semantic registration and fusion data.

[0072] Flow channel continuity error The calculation formula is as follows:

[0073]

[0074] in, It refers to the point cloud fitting parameters in semantic registration and fusion data.

[0075] Finally, the system traverses the semantic feature association graph and, based on the predefined semantic mapping rules in the graph—such as "curvature anomaly node → category semantic label 'surface deformation'", "flange normal deviation node → 'interface misalignment'", and "section abrupt change path → 'conduction obstruction'"—calculates the semantic features associated with each item. , , Automatically bind the corresponding category semantic labels, and package them together with their spatial coordinates, component IDs, and error values ​​to generate a coupled error dataset with a unified structure, clear semantics, and can be directly used for subsequent filtering and visualization.

[0076] As an example, the step of performing error coupling verification on the coupling error dataset through a semantic association threshold filtering mechanism, removing non-associative error data, and obtaining target coupling error data includes: obtaining semantic association threshold parameters, which include structural semantic association threshold, assembly semantic association threshold, and coupling association threshold; comparing the structural semantic association degree of each error data in the coupling error dataset with the structural semantic association threshold, and the assembly semantic association degree with the assembly semantic association threshold, retaining error data whose two association degrees are both higher than the corresponding thresholds, to obtain associated error data; performing coupling analysis on the associated error data according to the semantic feature association map, identifying single errors and coupled errors, and labeling error coupling links and coupling association degrees; removing non-associative error data and weakly associated error data from the associated error data to obtain target coupling error data, wherein the non-associative error data does not contain the error coupling links, and the coupling association degree of the weakly associated error data is lower than the coupling association threshold.

[0077] It should be noted that the structural semantic association threshold is a preset minimum correlation value (e.g., 0.65) used to determine whether an error is effectively associated with its corresponding flow channel structural semantic features. Only when the actual calculated structural semantic association degree is not lower than this threshold is the error considered to truly reflect structural deviation rather than noise. The assembly semantic association threshold is a preset lower limit value (e.g., 0.70) used to determine the validity of the association between errors and assembly semantic features, used to filter out false assembly deviations caused by scan occlusion or isolated components. The coupling association threshold is a critical value (e.g., 0.60) defined in the semantic feature association map to distinguish between strong and weak coupling relationships. Interactions between multiple features below this value are considered negligible in engineering. The structural semantic association degree refers to the quantitative association strength between an error data point obtained from the semantic feature association map and its corresponding flow channel structural semantic features (such as bends or diameter variations), reflecting the reliability of the error's attribution at the structural level. Assembly semantic correlation refers to the numerical value of the correlation strength between a certain error data and its associated assembly semantic features (such as flange interface and support point) in the semantic feature correlation map, which is used to measure whether it truly originates from the deviation in the assembly process.

[0078] Associated error data refers to the error data in the coupled error dataset that simultaneously satisfies the structural semantic association degree ≥ the structural semantic association threshold and the assembly semantic association degree ≥ the assembly semantic association threshold, that is, the effective error initially identified as having dual semantic support. A single error refers to an isolated error term caused by only a single factor (such as local geometric deformation or independent assembly offset) and not forming a propagation or interaction relationship with other types of errors through the semantic feature association map. Coupled error refers to a composite error caused by the interaction of structural deformation and assembly deviation, or the superposition of error transmission of multiple components, and its generation mechanism involves the联动 influence of two or more semantic feature nodes. An error coupling link refers to a causal or associative path formed in the semantic feature association map that starts from an error point and traces back to multiple related semantic feature nodes through edge connections (such as "abnormal elbow curvature → causes downstream flange angle deflection → leads to misalignment of adjacent equipment interfaces"). Non-associated error data refers to the error terms in the associated error data that do not identify any error coupling links after coupled analysis, that is, they cannot be traced back to the multi-feature interaction relationship through the semantic map, usually isolated or mis-matched results. Weakly associated error data refers to error terms that, although there are error coupling links, their overall coupling association degree (i.e., the comprehensive measure of the weights of each edge on the link) is lower than the preset coupling association threshold, indicating that their coupling effect is not significant in engineering and should be removed to avoid interfering with the diagnosis of the core problem. <0000!89> It should be noted that there may be an error in the original text "

[0079] ", and it is translated as "<0000!89>" here according to the original. You may need to check and correct it if necessary.Understandably, firstly, the error detection system reads three preset semantic association threshold parameters from the local configuration library. These thresholds are determined jointly by historical engineering data statistics and expert experience, and are used to define the semantic credibility boundary of valid errors. Secondly, the system traverses each error record in the coupled error dataset, extracts its structural semantic association degree and assembly semantic association degree already labeled in the semantic feature association map, and compares them numerically with the corresponding structural semantic association threshold and assembly semantic association threshold, respectively. Only when both association degrees are strictly greater than their respective thresholds is the record retained as associated error data—the purpose of this is to ensure that the retained errors have both reliable structural attribution and assembly context, avoiding misjudgment due to single-point noise or semantic mismatch. Then, the system performs graph-driven coupling analysis on all associated error data: using depth-first search (DFS), it starts from the semantic node corresponding to the error in the semantic feature association graph to detect whether there is a path leading to other types of semantic nodes (such as from "abnormal pipe curvature" to "downstream flange angle offset"). If so, it is determined to be a coupled error, and the weighted average of the weights of the entire error coupling link and its path is recorded as the coupling correlation degree. If not, it is marked as a single error. Finally, the system removes two types of data: non-associated error data and weakly associated error data, retaining only error terms with clear multi-feature interaction mechanisms and qualified coupling strength, thus obtaining target coupled error data that focuses on real engineering problems and can support accurate rework decisions.

[0080] As an example, the step of associating the target coupling error data with the structural semantic tags of the semantically enhanced BIM model data to generate a heatmap detection report containing error tracing links, and completing error detection, includes: associating the error type corresponding to the target coupling error data with the structural semantic tags of the semantically enhanced BIM model data; taking each associated target coupling error data as a starting point, traversing backwards according to the labeled structural association relationship based on the semantic feature association map to determine the error tracing links; converting the target coupling error data into heatmap values ​​and superimposing them onto the corresponding areas of the semantically enhanced BIM model data according to feature positions to obtain an error heatmap; fusing the error heatmap, the tracing links, and the semantic tags, and labeling the error type, source, and transmission information to obtain a heatmap detection report, thus completing error detection.

[0081] It should be noted that structural semantic tags refer to standardized semantic identifiers that describe the ontological structural attributes of each geometric region (such as bends, diameter change zones, branch nodes, etc.) in the semantically enhanced BIM model data, and are used to uniquely identify the functional and morphological role of that region in the flow channel system. Structural relationships refer to the topological connections or geometric dependencies between different structural semantic tags defined in the semantic feature association map (e.g., "upstream straight pipe section → connection → 90° bend → connection → downstream diameter change section"), used to characterize the logical order and spatial transmission path between components within the flow channel. Thermal values ​​refer to the color intensity values ​​obtained by converting the error values ​​(e.g., surface topology error of 3.2 mm) in the target coupling error data through a preset mapping function (e.g., linear normalization or piecewise coloring rules), used in visualization to represent the severity of the error using gradient colors such as red-yellow-green. Feature location refers to the specific geometric region coordinates or patch indexes corresponding to the target coupling error data in the semantically enhanced BIM model data. This is determined by the semantic binding relationship established during error detection, ensuring that heat values ​​are accurately superimposed onto the physical location of the model. The error heatmap is a color-coded 3D visualization model generated after rendering all heat values ​​onto the surface of the semantically enhanced BIM model according to their corresponding feature locations. Warmer colors (e.g., red) indicate larger errors, while cooler colors (e.g., blue) indicate smaller errors, visually presenting the spatial distribution of errors.

[0082] Understandably, the error detection system first performs precise matching and association between each record in the target coupling error data (containing error value, category, and component ID) and the structural semantic tag with the same component ID in the semantically enhanced BIM model data. Specifically, it searches for the corresponding geometric entity in the attribute table of the BIM model using the component's unique identifier and writes the bound structural semantic tag (such as "variable diameter segment_01") into the metadata field of the error record, thereby establishing a direct mapping from error to model semantic unit. This is done to ensure that subsequent visualization and tracing can accurately locate the specific part in the design model. Secondly, the system takes each associated error record as a starting point and performs a reverse traversal in the semantic feature association graph: starting from the structural semantic label node corresponding to the error, it recursively backtracks along the pre-stored "upstream connection" edges in the graph (e.g., edge attributes marked as "flow from" or "predecessor component"). Each time a parent node is visited, it is added to the path list until the root node without upstream connection (e.g., the main inlet section) is reached, thereby generating a complete error tracing link (e.g., "branch flange → elbow section → main duct diameter change section"). This process adopts the DFS strategy and limits the maximum backtracking depth (e.g., no more than 5 hops) to avoid infinite loops and focus on engineering-related transmission paths. The goal is to reveal the possible propagation source of the error rather than viewing the current deviation in isolation.

[0083] Then, the system substitutes the error value (e.g., 2.8 mm) of each target coupling error data into a preset thermal mapping function (e.g., piecewise linear mapping: 0–1 mm → blue, 1–3 mm → yellow, >3 mm → red) to calculate the corresponding thermal value (i.e., color RGBA encoding). Based on the feature location stored in the error record (i.e., the corresponding patch index set or spatial bounding box coordinates in the BIM model), the color value is directly written into the visualization attributes of the corresponding geometric patch in the semantically enhanced BIM model data, thereby generating a full-model-covered error thermal map. Finally, the system uses the error thermal map as a base map and overlays an interactive information layer on it: for each error point area, text labels are automatically embedded, clearly marking three items—error type (e.g., "flow channel continuity error"), source (i.e., the last node in the tracing link, such as "main duct diameter change section"), and transmission information (i.e., the complete link string). Finally, it outputs a thermal map detection report integrating 3D coloring, semantic annotation, and causal tracing, completing high-precision, interpretable, and actionable error detection.

[0084] This embodiment provides an intelligent error detection method for irregularly shaped flow channels based on BIM data. First, it acquires the original BIM model data and multi-view time-series scan point cloud data of the irregularly shaped flow channel. Then, it performs semantic parsing and feature enhancement on the original BIM model data to obtain semantically enhanced BIM model data, thus providing a semantically complete and geometrically accurate benchmark for subsequent analysis. Second, it performs hierarchical semantic feature extraction on the semantically enhanced BIM model data, separating the structural semantic features describing the flow channel's morphology from the assembly semantic features describing the connection relationships, and constructs a semantic feature association map between the two, providing clear semantic logical support for error analysis. Finally, based on this map, it uses a feature weight iterative registration algorithm to perform further analysis on the time-series denoised multi-view time-series scan point cloud data and the semantically enhanced BIM model data. Dynamic alignment is performed to obtain semantic registration fusion data. Next, a coupling error calculation model is constructed based on this fusion data to calculate surface topology errors, assembly datum coupling errors, and channel connectivity errors, respectively. Each type of error is then assigned a corresponding semantic label, forming a structured coupling error dataset, enabling simultaneous quantification and classification of multi-dimensional errors. Subsequently, a semantic association threshold filtering mechanism is used to verify the coupling error dataset, eliminating non-associative error data that cannot be traced back through the semantic graph or has insufficient coupling strength, while retaining target coupling error data with engineering authenticity, effectively suppressing false alarms. Finally, the target coupling error data is associated with the structural semantic labels of the semantically enhanced BIM model data to generate a heatmap detection report containing error tracing links, completing the entire closed-loop detection process. This embodiment can accurately detect geometric and assembly errors of irregularly shaped channels based on BIM data.

[0085] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in the first embodiment can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to Figure 2, which is a flowchart illustrating the second embodiment of the intelligent error detection method for irregularly shaped flow channels based on BIM data. Step S20 of the intelligent error detection method for irregularly shaped flow channels based on BIM data includes steps S21 to S25:

[0086] Step S21: Based on the feature similarity between the structural semantic labels and the assembly semantic labels in the semantically enhanced BIM model data, the structural semantic layer and the assembly semantic layer are separated by the K-means algorithm.

[0087] Step S22: Extract structural semantic features from the structural semantic layer and label the structural relationships. The structural semantic features include the topology of the flow channel surface, curvature distribution, cross-sectional profile, flow channel axis orientation, and the boundaries of the concave and convex regions of the surface.

[0088] Step S23: Extract assembly semantic features from the assembly semantic layer and record assembly constraint relationships. The assembly semantic features include assembly reference surface, connection hole position, gap size, assembly positioning pin hole and component mating surface parameters.

[0089] Step S24: Based on the structural semantic features, the structural association relationships, the assembly semantic features, and the assembly constraint relationships, construct a semantic feature association matrix and label the coupling association degree between each feature;

[0090] Step S25: Visualize the semantic feature association matrix to generate a semantic feature association map containing coupling degree annotations.

[0091] It should be noted that assembly semantic tags refer to semantic markers in semantically enhanced BIM model data used to identify the connection, positioning, or mating relationships between irregular flow channels and other components, such as "flange connected to fan outlet" or "fixed with support brackets," reflecting their assembly role in the system. Feature similarity refers to the cosine similarity or Euclidean distance between structural semantic tags and assembly semantic tags in vector space (such as semantic vectors generated through embedding the model), used to measure the semantic or functional closeness of the two types of tags, serving as the basis for K-means clustering. The structural semantic layer refers to the subset of semantic tags separated from the semantically enhanced BIM model data using the K-means algorithm, mainly containing descriptions of the geometric and morphological features of the flow channel itself, constituting the exclusive scope for subsequent structural feature extraction. The assembly semantic layer refers to the subset of semantic tags separated using the K-means algorithm, containing concentrated information on interfaces, connections, and installation constraints, used to define the boundaries of assembly feature extraction. The topology of a flow channel surface refers to the connectivity and overall shape logic of the outer surface of an irregularly shaped flow channel, including whether it is closed, whether it has branches, the number of holes, and the adjacency relationship between surface patches, reflecting its overall geometric complexity. Curvature distribution refers to the numerical variation law of the principal curvature or Gaussian curvature at various points on the flow channel surface, used to characterize the spatial distribution characteristics of curved, flat, or saddle-shaped regions. Cross-sectional profile refers to the shape and dimensional parameters (such as circle, rectangle, ellipse, or free curve) of the cross-sectional boundary perpendicular to the flow channel axis, which may dynamically adjust with changes in flow direction. The flow channel axis orientation refers to the directional trend of the reference path (i.e., the centerline) running through the center of the flow channel in three-dimensional space, usually represented by a series of control points or spline curves, used to define the overall layout of the flow channel.

[0092] The boundary of a curved surface's concave-convex region refers to the dividing line between concave and convex areas on the flow channel surface, defined by changes in the sign of curvature (e.g., from positive to negative) or abrupt changes in normal. This boundary is used to identify local geometric anomalies or areas with manufacturing difficulties. Assembly constraints refer to the geometric or logical rules that restrict the degrees of freedom of components during assembly, such as "flange faces must be coplanar," "pin holes must be coaxial," and "gap must not exceed 2 mm," used to guide error judgment. An assembly reference surface refers to a specific plane used as a positioning reference during installation (e.g., flange end face or equipment interface face). Its spatial orientation (position and direction) is the core basis for evaluating assembly deviations. Connection hole positions refer to the center coordinates and arrangement patterns (e.g., circular distribution, rectangular array) of holes through which bolts, rivets, or other fasteners pass, and are key elements for judging alignment accuracy. Gap size refers to the actual empty distance between two adjacent assembled parts, usually specified in the design as a maximum allowable value (e.g., ≤1.5 mm). Exceeding this tolerance may affect sealing performance or structural stability. Assembly positioning pin holes refer to guide holes or pin hole pairs used for precise positioning. Their coaxiality and positional accuracy directly affect assembly repeatability and alignment performance. Component mating surface parameters refer to the geometric matching indices between two contacting assembly surfaces, including mating area, average gap, maximum local gap, and normal deviation, used to quantify contact quality. A semantic feature correlation matrix is ​​a two-dimensional matrix constructed using structural and assembly semantic features as row and column indices. Each element value represents the coupling correlation degree between the corresponding feature pair (typically ranging from 0 to 1), used to quantify the interaction strength of different semantic features in error propagation or functional impact. Coupling degree annotation refers to the quantitative numerical label assigned to the mutual influence or correlation strength between any two semantic features (such as "abnormal pipe curvature" and "downstream flange angle offset") in the semantic feature correlation matrix or its visualization. This value is typically a real number between 0 and 1, reflecting the tightness of coupling between the two at the geometric, functional, or assembly levels, and is directly derived from a comprehensive evaluation of structural correlation relationships, assembly constraint relationships, and engineering experience rules.

[0093] Understandably, the system first converts each structural semantic label and assembly semantic label in the semantically enhanced BIM model data into a fixed-dimensional semantic vector using a pre-trained text embedding model (such as BERT). It then calculates the cosine similarity between any two label vectors as the feature similarity. Next, using all label vectors as input, it sets the number of clusters K=2 and runs the K-means algorithm for iterative clustering. Elements closer to the cluster center ("morphological description") are assigned to the structural semantic layer, and elements closer to the cluster center ("connection description") are assigned to the assembly semantic layer. This divides the extraction range of the two types of semantic features. This is done to avoid human classification bias and ensure that subsequent feature extraction is focused and mutually exclusive. Secondly, the system traverses each component in the structural semantic layer and calls the geometric analysis module to extract its flow channel surface topology (determining connected branches through the adjacency graph of patches), curvature distribution (calculating principal curvature and sampling statistics), cross-sectional profile (fitting boundary curves by equidistant slices along the axis), flow channel axis orientation (extracting centerline control points through skeletonization algorithm), and surface concave and convex region boundaries (detecting zero-crossing lines based on Gaussian curvature sign change). At the same time, based on the upstream and downstream order of the components in the flow channel and the shared boundary relationship, it marks structural association relationships such as "upstream → downstream" or "main trunk → branch".

[0094] Then, the system traverses each interface or connection node in the assembly semantic layer, extracting its assembly reference plane (fitting the flange end face point cloud to obtain the plane equation), connection hole position (identifying the hole center coordinates and distribution pattern), gap size (calculating the minimum distance between adjacent surface point clouds), assembly positioning pin hole (detecting the cylindrical hole axis and matching the design position), and component mating surface parameters (calculating the contact surface overlap area and average normal deviation). It also reads and records the corresponding assembly constraint relationships (such as "gap ≤ 1.5 mm" and "pin hole coaxiality tolerance 0.1 mm") from BIM attributes or rule bases. Next, the system constructs an N×N semantic feature association matrix using all extracted structural and assembly semantic features as nodes. For each pair of features (such as "abnormal bend curvature" and "downstream flange angle offset"), if they are spatially adjacent, have assembly references in BIM, or have a co-occurrence frequency higher than a threshold in historical data, a high coupling association degree (between 0 and 1) is assigned; otherwise, it is set to a low value or zero. Finally, the system calls Graphviz's dot layout engine to convert the matrix into a directed graph: structural semantic features are represented by blue circular nodes, assembly semantic features are represented by orange square nodes, edges are colored according to their coupling degree, and specific values ​​are marked on the edges, generating a semantic feature association graph with clear hierarchy, well-defined paths, and interactive viewing, which is used to support subsequent high-precision error registration and tracing.

[0095] As an example, the step of dynamically aligning the temporally denoised multi-view temporal scan point cloud data with the semantically enhanced BIM model data using a feature weight iterative registration algorithm based on the semantic feature association map to obtain semantic registration fusion data includes: performing temporal smoothing denoising on the multi-view temporal scan point cloud data to obtain temporally stable point cloud data; assigning a first initial weight to the structural semantic features and a second initial weight to the assembly semantic features based on the coupling correlation degree of the semantic feature association map; performing preliminary registration of the temporally stable point cloud data with the semantically enhanced BIM model data based on the first initial weight and the second initial weight, and calculating the registration deviation value of each semantic feature; iteratively adjusting the first initial weight and the second initial weight based on the registration deviation value, and increasing the weight ratio of features whose registration deviation value exceeds a preset deviation value range; and fusing the registered data and semantic labels when all the registration deviation values ​​are within a preset accuracy range to obtain semantic registration fusion data.

[0096] It should be noted that time-stable point cloud data refers to the point cloud sequence obtained after applying the Kalman filter algorithm to the original multi-view time-series scan point cloud data for time-series smoothing and denoising. Its dynamic jitter, measurement noise, and outliers are effectively suppressed, and the geometric consistency between adjacent frames is significantly improved, making it suitable for high-precision registration. The first initial weight refers to the initial importance coefficient assigned to structural semantic features (such as flow channel surfaces, axes, etc.) in the preliminary registration stage, set to 0.55, used to weight the matching contribution of structural regions in the registration error function. The second initial weight refers to the initial importance coefficient assigned to assembly semantic features (such as flange faces, hole positions, etc.) concurrently, set to 0.45, reflecting the relative priority of assembly areas in the overall registration. The registration deviation value of each semantic feature refers to the mean distance or RMS error between the point cloud in the structural semantic feature region (such as the surface of a bend) and the corresponding geometric element in the BIM model, calculated separately after preliminary registration, used to quantify the alignment quality of different semantic regions. The preset deviation range refers to the threshold interval for judging whether a certain type of semantic feature is poorly registered, set to ±0.1 mm. If the registration deviation value of a feature exceeds this range, the region is considered not fully aligned, and its weight in the next round of registration needs to be increased. The preset weight percentage refers to the percentage increase in weight automatically added by the system when the registration deviation value of a certain type of semantic feature exceeds the limit (e.g., an increase of 0.05 in each iteration), to strengthen the constraint of that region in subsequent registrations and encourage the algorithm to prioritize optimizing high-deviation regions. The preset accuracy range refers to the criterion for registration convergence, set to ±0.05 mm. Only when the registration deviation values ​​of all semantic features fall within this range is the registration considered to have reached the required engineering accuracy, the iteration can be terminated, and the result can be output.

[0097] Understandably, firstly, the system inputs multi-view temporal scan point cloud data frame by frame into a Kalman filter, using the point cloud state of the previous frame to predict the point coordinates of the current frame, and performs weighted correction based on actual observations. This effectively suppresses high-frequency noise caused by equipment jitter or environmental interference during laser scanning, outputting geometrically continuous and inter-frame stable temporal point cloud data. This is done to eliminate the interference of random errors on subsequent high-precision registration. Secondly, the system reads the coupling correlation degree between structural semantic features and assembly semantic features from the semantic feature association map, and initializes the weight allocation accordingly—assigning the first initial weight to structural semantic features (such as curved pipe surfaces and axis areas), and the second initial weight to assembly semantic features (such as flange surfaces and hole areas). Then, a weighted ICP objective function is constructed using these two weights. During registration, the point-to-point error of different semantic regions is multiplied by its corresponding weight, and preliminary rigid body transformation registration is performed. The average distance from the structural semantic feature region and the assembly semantic feature region to the BIM model surface is calculated to obtain the registration deviation value of each semantic feature. Then, the system checks whether each registration deviation value exceeds the preset deviation range. If a certain feature exceeds the limit (e.g., the deviation of the assembly area is 0.13 mm), a fixed preset weight percentage is added to its corresponding weight, and the first and second initial weights are updated using the gradient descent algorithm. The ICP registration is then rerun with the new weights, and this process is iterated repeatedly to dynamically increase the constraint strength of high-deviation areas in the objective function, thereby guiding the registration to converge towards the critical area. Finally, when the registration deviation values ​​of all semantic features are less than or equal to the preset accuracy range, the system stops iterating, merges the finally aligned temporally stable point cloud data with the semantically enhanced BIM model data in a unified coordinate system, and retains the original structural semantic labels and assembly semantic labels of each geometric region, generating semantic registration fusion data that combines millimeter-level geometric alignment accuracy with complete engineering semantics.

[0098] This embodiment first uses the K-means algorithm to automatically cluster the labels into structural semantic layers and assembly semantic layers based on the feature similarity between structural semantic labels and assembly semantic labels in the semantically enhanced BIM model data, achieving clear separation of the two semantic types and avoiding errors from manual classification. Secondly, it extracts the topology, curvature distribution, cross-sectional contour, flow channel axis direction, and boundaries of concave and convex areas of the flow channel surface from the structural semantic layer, and labels their upstream and downstream or branch structural relationships to accurately depict the geometric characteristics of the flow channel body and its inherent connection logic. Then, it extracts the assembly reference surface, connection hole positions, gap size, assembly positioning pin holes, and component mating surface parameters from the assembly semantic layer, and records the corresponding assembly constraint relationships to provide a basis for quantifying the interface installation status. Next, it constructs a semantic feature association matrix based on the above structural semantic features, structural relationships, assembly semantic features, and assembly constraint relationships, and labels the coupling correlation degree for each pair of features to explicitly express the mutual influence strength between different features. Finally, it uses the Graphviz visualization algorithm to generate a semantic feature association map containing coupling degree labels, intuitively presenting feature types, levels, and association paths.

[0099] This application also provides a BIM data-based intelligent error detection system for irregularly shaped flow channels. Referring to Figure 3, the BIM data-based intelligent error detection system for irregularly shaped flow channels includes:

[0100] The semantic enhancement module 10 is used to acquire the original BIM model data and multi-view time-series scan point cloud data of the irregular flow channel, and to perform semantic parsing and feature enhancement on the original BIM model data to obtain semantically enhanced BIM model data.

[0101] The hierarchical extraction module 20 is used to perform hierarchical extraction of semantic features on the semantically enhanced BIM model data, separate structural semantic features from assembly semantic features, and construct a semantic feature association map.

[0102] The iterative registration module 30 is used to dynamically align the temporally denoised multi-view temporal scan point cloud data with the semantically enhanced BIM model data based on the semantic feature association map and through a feature weight iterative registration algorithm to obtain semantic registration fusion data.

[0103] The coupling error calculation module 40 is used to construct a coupling error calculation model based on the semantic registration fusion data, and calculate the surface topology error, assembly datum coupling error and flow channel conductivity error respectively according to the coupling error calculation model to obtain a coupling error dataset with associated category semantic labels.

[0104] The semantic association filtering module 50 is used to perform error coupling verification on the coupling error dataset through a semantic association threshold filtering mechanism, remove non-associative error data, and obtain target coupling error data.

[0105] The source tracing report generation module 60 is used to associate the target coupling error data with the structural semantic tags of the semantically enhanced BIM model data, generate a heat map detection report containing the error source tracing link, and complete the error detection.

[0106] The intelligent error detection system for irregularly shaped flow channels based on BIM data provided in this application, employing the intelligent error detection method for irregularly shaped flow channels based on BIM data in the above embodiments, can solve the technical problem of how to accurately detect the geometric and assembly errors of irregularly shaped flow channels based on BIM data. Compared with the prior art, the beneficial effects of the intelligent error detection system for irregularly shaped flow channels based on BIM data provided in this application are the same as those of the intelligent error detection method for irregularly shaped flow channels based on BIM data provided in the above embodiments, and other technical features of the intelligent error detection system for irregularly shaped flow channels based on BIM data are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0107] This application provides a BIM data-based intelligent error detection device for irregular flow channels. The device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the BIM data-based intelligent error detection method for irregular flow channels in Embodiment 1 above.

[0108] As shown in Figure 4, the intelligent error detection device for irregularly shaped flow channels based on BIM data may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to the program stored in ROM (Read Only Memory) 1002 or the program loaded from storage device 1003 into RAM (Random Access Memory) 1004. RAM 1004 also stores various programs and data required for the operation of the intelligent error detection device for irregularly shaped flow channels based on BIM data. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via bus 1005. Input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, LCDs (Liquid Crystal Displays), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the BIM data-based intelligent error detection device for irregularly shaped flow channels to wirelessly or wiredly communicate with other devices to exchange data. Although the figure shows a BIM data-based intelligent error detection device for irregularly shaped flow channels with various systems, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented alternatively.

[0109] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0110] The intelligent error detection device for irregularly shaped flow channels based on BIM data provided in this application, employing the intelligent error detection method for irregularly shaped flow channels based on BIM data in the above embodiments, can solve the technical problem of how to accurately detect the geometric and assembly errors of irregularly shaped flow channels based on BIM data. Compared with the prior art, the beneficial effects of the intelligent error detection device for irregularly shaped flow channels based on BIM data provided in this application are the same as those of the intelligent error detection method for irregularly shaped flow channels based on BIM data provided in the above embodiments, and other technical features in this intelligent error detection device for irregularly shaped flow channels based on BIM data are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0111] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0112] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, which are used to execute the intelligent error detection method for irregular flow channels based on BIM data in the above embodiments.

[0113] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, RAM (Random Access Memory), ROM (Read Only Memory), EPROM (Erasable Programmable Read Only Memory or Flash Memory), optical fibers, CD-ROM (CD-Read Only Memory), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0114] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by the BIM data-based intelligent error detection device for irregularly shaped flow channels, the device performs the following: acquires the original BIM model data and multi-view time-series scan point cloud data of the irregularly shaped flow channels; performs semantic parsing and feature enhancement on the original BIM model data to obtain semantically enhanced BIM model data; performs hierarchical extraction of semantic features on the semantically enhanced BIM model data, separates structural semantic features and assembly semantic features, and constructs a semantic feature association map; based on the semantic feature association map, it uses a feature weight iterative registration algorithm to denoise the multi-view time-series scan point cloud data. The sequential scan point cloud data is dynamically aligned with the semantically enhanced BIM model data to obtain semantic registration and fusion data. Based on the semantic registration and fusion data, a coupling error calculation model is constructed, and the surface topology error, assembly datum coupling error, and flow channel conductivity error are calculated according to the coupling error calculation model to obtain a coupling error dataset with associated category semantic labels. Through a semantic association threshold filtering mechanism, the coupling error dataset is checked for error coupling, and non-associated error data is eliminated to obtain target coupling error data. The target coupling error data is associated with the structural semantic labels of the semantically enhanced BIM model data to generate a heatmap detection report containing error tracing links, thus completing error detection.

[0115] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including LAN (Local Area Network) or WAN (Wide Area Network)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0116] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described intelligent error detection method for irregularly shaped flow channels based on BIM data. This method can solve the technical problem of how to accurately detect the geometric and assembly errors of irregularly shaped flow channels based on BIM data. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the intelligent error detection method for irregularly shaped flow channels based on BIM data provided in the above embodiments, and will not be repeated here.

[0117] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for intelligent error detection of irregularly shaped flow channels based on BIM data, characterized in that, The method includes: acquiring original BIM model data and multi-view time-series scan point cloud data of irregular flow channels; performing semantic parsing and feature enhancement on the original BIM model data to obtain semantically enhanced BIM model data; performing hierarchical semantic feature extraction on the semantically enhanced BIM model data, separating structural semantic features and assembly semantic features, and constructing a semantic feature association map; based on the semantic feature association map, dynamically aligning the time-series denoised multi-view time-series scan point cloud data with the semantically enhanced BIM model data using a feature weight iterative registration algorithm to obtain semantic registration fusion data. Based on the semantic registration and fusion data, a coupling error calculation model is constructed, and the surface topology error, assembly datum coupling error, and flow channel conductivity error are calculated according to the coupling error calculation model to obtain a coupling error dataset with associated category semantic tags. Through a semantic association threshold filtering mechanism, the coupling error dataset is checked for error coupling, and non-associated error data is eliminated to obtain target coupling error data. The target coupling error data is associated with the structural semantic tags of the semantically enhanced BIM model data to generate a heatmap detection report containing error tracing links, thus completing the error detection.

2. The method as described in claim 1, characterized in that, The steps of extracting semantic features hierarchically from the semantically enhanced BIM model data, separating structural semantic features from assembly semantic features, and constructing a semantic feature association map include: separating the structural semantic layer and the assembly semantic layer using the K-means algorithm based on the feature similarity between the structural semantic labels and assembly semantic labels in the semantically enhanced BIM model data; extracting structural semantic features from the structural semantic layer and labeling structural relationships, the structural semantic features including the topology of the flow channel surface, curvature distribution, cross-sectional contour, flow channel axis direction, and the boundary of the concave and convex regions of the surface; extracting assembly semantic features from the assembly semantic layer and recording assembly constraint relationships, the assembly semantic features including assembly reference surfaces, connection holes, gap sizes, assembly positioning pin holes, and component mating surface parameters; constructing a semantic feature association matrix based on the structural semantic features, the structural relationships, the assembly semantic features, and the assembly constraint relationships, and labeling the coupling degree between each feature; and visualizing the semantic feature association matrix to generate a semantic feature association map containing coupling degree labels.

3. The method as described in claim 2, characterized in that, The step of dynamically aligning the temporally denoised multi-view temporal scan point cloud data with the semantically enhanced BIM model data using a feature weight iterative registration algorithm based on the semantic feature association map to obtain semantic registration fusion data includes: performing temporal smoothing and denoising on the multi-view temporal scan point cloud data to obtain temporally stable point cloud data; assigning a first initial weight to the structural semantic features and a second initial weight to the assembly semantic features based on the coupling correlation degree of the semantic feature association map; performing preliminary registration of the temporally stable point cloud data with the semantically enhanced BIM model data based on the first initial weight and the second initial weight, and calculating the registration deviation value of each semantic feature; iteratively adjusting the first initial weight and the second initial weight based on the registration deviation value, and increasing the weight ratio of features whose registration deviation value exceeds a preset deviation value range; and fusing the registered data and semantic labels when all the registration deviation values ​​are within a preset accuracy range to obtain semantic registration fusion data.

4. The method as described in claim 1, characterized in that, The steps for constructing a coupling error calculation model based on the semantic registration fusion data include: extracting structural semantic feature parameters, assembly semantic feature parameters, and point cloud fitting parameters from the semantic registration fusion data to obtain a model input dataset; constructing an initial model based on the model input dataset, the coupling correlation degree in the semantic feature association map, and the NSGA-III algorithm; setting constraints for the initial model to obtain a reference model, wherein the constraints include semantic label association accuracy, error calculation threshold, and data adaptation range; and using the least squares method to correct the parameters of the reference model to obtain the coupling error calculation model.

5. The method as described in claim 4, characterized in that, The steps for constructing an initial model based on the model input dataset, the coupling correlation degree in the semantic feature association graph, and the NSGA-III algorithm include: fusing the model input dataset and the coupling correlation degree in the semantic feature association graph according to the corresponding feature dimensions to obtain target input data; importing the target input data into the NSGA-III algorithm with the target accuracy of solving surface topology error, assembly datum coupling error, and flow channel conductivity error; iteratively adjusting the correlation weights of each feature parameter in the target input data through the NSGA-III algorithm to obtain target parameter weights; and building a process framework for data input, weight allocation, and error calculation based on the target input data and the target parameter weights to obtain the initial model.

6. The method as described in claim 4, characterized in that, The step of calculating the surface topology error, assembly datum coupling error, and flow channel conductivity error according to the coupling error calculation model to obtain a coupling error dataset with associated category semantic labels includes: extracting target parameter weights from the coupling error calculation model, wherein the target parameter weights include structural semantic feature weights, assembly semantic feature weights, and point cloud fitting parameter weights; calculating the surface topology error according to the structural semantic feature parameters, the structural semantic feature weights, and preset standard structural parameters; calculating the assembly datum coupling error according to the assembly semantic feature parameters, the assembly semantic feature weights, and preset standard assembly parameters; calculating the flow channel conductivity error according to the point cloud fitting parameters, the point cloud fitting parameter weights, and preset standard conductivity parameters; and binding the surface topology error, the assembly datum coupling error, and the flow channel conductivity error with corresponding category semantic labels according to the semantic feature association map to obtain a coupling error dataset with associated semantic labels.

7. The method as described in claim 1, characterized in that, The step of performing error coupling verification on the coupling error dataset through a semantic association threshold filtering mechanism, removing non-associative error data, and obtaining target coupling error data includes: obtaining semantic association threshold parameters, which include structural semantic association threshold, assembly semantic association threshold, and coupling association threshold; comparing the structural semantic association degree of each error data in the coupling error dataset with the structural semantic association threshold, and the assembly semantic association degree with the assembly semantic association threshold, retaining error data whose two association degrees are both higher than the corresponding thresholds, to obtain associated error data; performing coupling analysis on the associated error data according to the semantic feature association map, identifying single errors and coupled errors, and labeling error coupling links and coupling association degrees; removing non-associative error data and weakly associated error data from the associated error data to obtain target coupling error data, wherein the non-associative error data does not contain the error coupling links, and the coupling association degree of the weakly associated error data is lower than the coupling association threshold.

8. The method as described in claim 1, characterized in that, The steps of acquiring the original BIM model data and multi-view time-series scan point cloud data of the irregular flow channel, and performing semantic parsing and feature enhancement on the original BIM model data to obtain semantically enhanced BIM model data include: acquiring the original BIM model data and multi-view time-series scan point cloud data of the irregular flow channel; extracting flow channel structure tags, assembly relationship tags, and material attribute tags from the original BIM model data to obtain an initial semantic set; performing semantic completion on the initial semantic set to supplement flow channel topology association semantics and assembly benchmark semantics to obtain a complete semantic tag set; binding the complete semantic tag set to the original BIM model data point by point, performing semantically guided feature completion on the hidden areas of the model to obtain a semantically bound and completed model; and performing topological optimization on the semantically bound and completed model to obtain semantically enhanced BIM model data with deep fusion of semantic information and geometric features.

9. The method according to any one of claims 1 to 8, characterized in that, The steps of associating the target coupling error data with the structural semantic tags of the semantically enhanced BIM model data to generate a heatmap detection report containing error tracing links, and completing error detection, include: associating the error type corresponding to the target coupling error data with the structural semantic tags of the semantically enhanced BIM model data; taking each associated target coupling error data as a starting point, traversing backwards according to the labeled structural association relationship based on the semantic feature association map to determine the error tracing links; converting the target coupling error data into heatmap values ​​and superimposing them onto the corresponding areas of the semantically enhanced BIM model data according to feature positions to obtain an error heatmap; fusing the error heatmap, the tracing links, and the semantic tags, and labeling the error type, source, and transmission information to obtain a heatmap detection report, thus completing error detection.

10. A smart error detection system for irregularly shaped flow channels based on BIM data, characterized in that, The system includes: a semantic enhancement module, used to acquire original BIM model data and multi-view time-series scan point cloud data of irregular flow channels, and to perform semantic parsing and feature enhancement on the original BIM model data to obtain semantically enhanced BIM model data; a hierarchical extraction module, used to perform hierarchical extraction of semantic features from the semantically enhanced BIM model data, separating structural semantic features and assembly semantic features, and constructing a semantic feature association map; and an iterative registration module, used to dynamically align the time-series denoised multi-view time-series scan point cloud data with the semantically enhanced BIM model data based on the semantic feature association map using a feature weight iterative registration algorithm to obtain semantically registered fused data; and coupling... The coupling error calculation module is used to construct a coupling error calculation model based on the semantic registration and fusion data, and calculate the surface topology error, assembly datum coupling error, and flow channel continuity error according to the coupling error calculation model to obtain a coupling error dataset with associated category semantic tags; the semantic association filtering module is used to perform error coupling verification on the coupling error dataset through a semantic association threshold filtering mechanism, eliminate non-associated error data, and obtain the target coupling error data; the source tracing report generation module is used to associate the target coupling error data with the structural semantic tags of the semantically enhanced BIM model data, generate a heat map detection report containing the error source tracing link, and complete the error detection.

Images