A BIM-based intelligent quality assessment method and system for highways

Abstract

This invention relates to the field of dynamic modeling technology for quality assessment, specifically a BIM-based intelligent quality assessment method and system for highways. The method includes: establishing a BIM model using design drawings and specifications; preprocessing the BIM model to form a structured BIM dataset; collecting multi-source data during highway construction and, based on a timestamp synchronization mechanism and spatial positioning information, combining a BIM spatiotemporal dynamic mapping method based on component state phase space gating, associating the multi-source data with the BIM model to generate a time-series dataset; and establishing a dynamic update mechanism for the BIM model based on the time-series dataset to update the BIM model in a time-series manner, and introducing a construction process topology to constrain the time-series update process. This invention, by introducing a dynamic modeling mechanism, integrates multi-source construction data with the BIM model in real time and continuously corrects component states, effectively solving the problems of data lag and state misjudgment in traditional quality assessments.

Description

Technical Field

[0001] This invention relates to the field of dynamic modeling technology for quality assessment, and specifically to a method and system for intelligent quality assessment of highways using BIM. Background Technology

[0002] In traditional highway construction quality assessment, BIM models are usually used as static design carriers, only for visual briefings before construction or model archiving after completion. They lack the ability to deeply integrate with multi-source real-time data (such as measurement data, sensor data, inspection images, etc.) during construction and the ability to dynamically model them. This results in delayed updates to quality status, insufficient spatiotemporal mapping accuracy, and lack of logical constraints on processes, making it difficult to identify and classify quality risks in real time during construction. In addition, existing assessment methods mostly rely on manual sampling or post-event inspections, and cannot use time-series observation data to continuously deduce the evolution of component status, which can easily lead to misjudgment of status, delayed warnings, or even logical conflicts. Summary of the Invention

[0003] The purpose of this invention is to address the problems existing in the background technology by proposing a method and system for intelligent quality assessment of highway BIM.

[0004] The technical solution of this invention: A BIM-based intelligent quality assessment method for highways, comprising: S1. Establish a BIM model using design drawings and specification documents, preprocess the BIM model, and form a structured BIM dataset; S2. Collect multi-source data during the highway construction process, and based on the timestamp synchronization mechanism and spatial positioning information, combine the BIM spatiotemporal dynamic mapping method based on component state phase space gating to associate the multi-source data with the BIM model and generate a time series dataset. S3. Based on the time-series dataset, establish a dynamic update mechanism for the BIM model to update the BIM model in time. Introduce the construction process topology to constrain the time-series update process and generate a BIM evolution model that dynamically changes with the construction process. S4. Extract quality feature vectors based on the BIM evolution model, construct a quality assessment function based on the quality feature vectors, and obtain a comprehensive quality score; S5. Based on the comprehensive quality score, output the risk level through graded thresholds and conduct dynamic early warning of highway construction quality.

[0005] As a further improvement to this technical solution, in step S1, a BIM model is established using design drawings and specification documents, and the BIM model is preprocessed to form a structured BIM dataset, including the following steps: S1.1 Collect design drawings and specification documents, and construct a set of design parameters based on the design drawings. At the same time, a set of normative constraints is constructed based on the normative documents. ; S1.2, Based on the set of design parameters The model is layered according to component category; parametric modeling is used to define the geometric parameters, topological relationships, and attribute parameters of the components to generate the initial BIM model. ; S1.3, Initial BIM Model Preprocessing is performed to generate a BIM model. And based on BIM model The data is structured, converted into a unified data format, and component-level data records are generated. All component-level data records are then aggregated to form a structured BIM dataset.

[0006] As a further improvement to this technical solution, in step S2, based on the timestamp synchronization mechanism and spatial positioning information, combined with the BIM spatiotemporal dynamic mapping method based on component state phase space gating, multi-source data is associated with the BIM model, and a time-series dataset is generated, including the following steps: S2.1 Collect multi-source data during the highway construction process, establish a unified data access interface for data from different sources, encapsulate the multi-source data in a structured manner and record its source identifier, collection time and initial spatial information to form a multi-source raw dataset; S2.2 Constructing a continuous time axis using linear interpolation method Obtain a standardized timestamp Convert the acquisition time in step S2.1 into a standard timestamp. ; S2.3. Perform unified standardization processing on the initial spatial information in the multi-source data, obtain a unified three-dimensional spatial coordinate expression based on the pre-constructed spatial mapping function, and construct a standard spatial expression; S2.4 Based on standardized timestamps and standard spatial representation, a BIM spatiotemporal dynamic mapping method based on component state phase space gating is used to generate state-level component mapping output; S2.5. Bind the state-level component mapping output to the component ID in the BIM model, and extract the original observation feature vector from the multi-source data. Generate component-level data records ; S2.6. Using the component ID in the BIM model as the primary key, record component-level data according to time series. Organize the data to form a time-series dataset.

[0007] As a further improvement to this technical solution, in step S2.4, a BIM spatiotemporal dynamic mapping method based on component state phase space gating is used to generate a state-level component mapping output, including the following steps: Based on standard spatial representation, the component ID in the corresponding BIM model is located using spatial indexing methods, denoted as [ID]. Based on construction procedures and quality acceptance standards, the components in the BIM model are divided into discrete state sets. Based on standardized timestamps Constructing joint feature vectors with standard space representation ; Based on joint feature vectors Introducing a state phase space gating mechanism yields a set of candidate discrete states. Based on discrete state set State discrimination functions are constructed for different construction states, and the activation probabilities of each construction state are obtained by softmax normalization. ; Based on standard timestamps Mapping operators corresponding to each construction state are constructed using the standard spatial representation, and activation probabilities are combined. Generate state-level component mapping output.

[0008] As a further improvement to this technical solution, in step S3, a dynamic update mechanism for the BIM model is established to update the BIM model in a timely manner, and a construction sequence topology is introduced to constrain the time-series update process, generating a BIM evolution model that dynamically changes with the construction process, including the following steps: S3.1. Using the time-series dataset as input, for each component in the BIM model... Extract the corresponding time series data sequence, and extract the original observation feature vector from the time series data sequence. Combined with the state-level component mapping output, a component-level temporal state vector is constructed. ; S3.2, Based on component temporal state vector Dynamic evolution modeling of component states yields predicted states. Furthermore, a construction sequence topology is introduced to constrain the dynamic evolution modeling process of component states and to predict the states. Perform topological projection correction to generate the corrected predicted state. ; S3.3 Constructing the observation state based on component-level data records and state-level component mapping outputs at the same time. and the predicted state Perform deviation calculation to obtain the update increment. Based on update increment The component state is corrected to generate an updated component state vector; S3.4 Map the updated component state vector to the geometric and attribute parameter space of the BIM model to generate the component update function; S3.5. Based on the component update function, topological constraints and engineering logic constraints are introduced to generate a global BIM model; S3.6 Organize the global BIM model according to the time series to form a BIM model sequence that dynamically changes with the construction process. This BIM model sequence is the BIM evolution model, which is used to characterize the structural state evolution throughout the entire highway construction process.

[0009] As a further improvement to this technical solution, in step S3.2, a construction sequence topology is introduced to constrain the dynamic evolution modeling process of component states and to predict the states. Perform topological projection correction to generate the corrected predicted state. This includes the following steps: S3.21. Based on construction organization design and specifications, construct the directed process graph of BIM model components. ; S3.22, Activation probabilities obtained from step S2.4 Determine the construction state with the highest probability for the current BIM model component. ; S3.23, Based on the directed graph of the process For the construction state with the highest probability Perform process consistency assessment; S3.24. Based on the process consistency determination result, introduce a process-based directed graph. For the predicted state Perform constraint corrections to generate a topologically consistent state vector. ; S3.25, Based on topologically consistent state vectors Based on the process consistency judgment results, the predicted status is... Perform selective corrections to generate corrected predicted states. .

[0010] As a further improvement to this technical solution, S3.24 introduces a process-oriented directed graph. For the predicted state Perform constraint corrections to generate a topologically consistent state vector. This includes the following steps: Based on the directed graph of the process For the construction state with the highest probability Extract the set of transitional states that satisfy the process constraints. Based on discrete state set Calculate discrete state Embedded representation and according to the embedded representation and predicted state Generate embedding distances and based on the process directed graph. Calculate the construction state with the highest probability. To discrete state The topological distance is calculated based on the embedding distance and the topological distance, and a process topological penalty coefficient is introduced. Generate a weighted cost function; based on the weighted cost function, generate a topologically consistent state vector through discrete projection.

[0011] As a further improvement to this technical solution, in step S4, quality feature vectors are extracted based on the BIM evolution model, and a quality assessment function is constructed based on the quality feature vectors to obtain a comprehensive quality score, including the following steps: S4.1 Extracting quality observation vectors based on the BIM evolution model; S4.2, Based on the set of normative constraints in S1 Extract construction quality control parameters, construct a set of quality indicators, calculate the deviation of each quality indicator, and generate a quality deviation vector. S4.3. Based on the quality observation vector and the quality deviation vector, generate the quality feature vector and normalize the quality feature vector; S4.4 Construct a multi-index evaluation function based on the normalized quality feature vector to generate a quality score; S4.5. Weighted summation of the quality scores of all components to generate a comprehensive quality score.

[0012] As a further improvement to this technical solution, in step S5, the risk level is output based on the comprehensive quality score through a graded threshold, and dynamic early warning of highway construction quality is performed, including the following steps: The comprehensive quality scores are organized in chronological order to form a quality score time series. Based on the quality score time series, the risk assessment is performed on the comprehensive quality score at each moment using preset grading thresholds to generate a risk level. Based on the correspondence between the risk level and preset early warning triggering rules, the corresponding level of construction quality early warning information is output for dynamic early warning of construction quality.

[0013] On the other hand, the present invention provides a highway BIM intelligent quality assessment system, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned highway BIM intelligent quality assessment method.

[0014] Compared with the prior art, the above-mentioned technical solution of the present invention has the following beneficial technical effects: by introducing a dynamic modeling mechanism, multi-source construction data and BIM model are integrated in real time and the component status is continuously corrected, which effectively solves the problems of data lag and status misjudgment in traditional quality assessment; at the same time, by using state phase space gating and process topology constraints, the robustness of spatiotemporal mapping and engineering logic consistency are significantly improved, thereby realizing accurate and dynamic assessment and graded early warning of highway construction quality. Attached Figure Description

[0015] Figure 1 This is a flowchart of the overall method of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0017] Example 1: Please refer to Figure 1 As shown, this embodiment provides a BIM-based intelligent quality assessment method for highways, including the following steps: S1. Establish a BIM model using design drawings and specification documents, preprocess the BIM model, and form a structured BIM dataset; In this embodiment, a BIM model is established using design drawings and specification documents. The BIM model is then preprocessed to form a structured BIM dataset, including the following steps: S1.1 Collect design drawings and specification documents, and construct a set of design parameters based on the design drawings (such as CAD drawings, route longitudinal profiles, and cross profiles). (Including route geometric parameters (horizontal curves, vertical curves), structural dimensional parameters (thickness, width, elevation), and construction layering information), while also constructing a set of specification constraints based on specification documents. (e.g., compaction standards, flatness index limits, and allowable deviation ranges for structures); specifically, by constructing a unified data access process, design drawings and specification documents are systematically collected: First, engineering-related design drawings are acquired, including CAD drawings, route longitudinal profiles, cross-sections, and structural details, and their formats are parsed (e.g., DWG / DXF to structured data); simultaneously, specification documents corresponding to the project are collected (e.g., construction specifications, quality acceptance standards, control index descriptions, etc.), and the text content is structured (e.g., index clause extraction and parameterized expression); subsequently, data cleaning and field standardization are performed on the design drawings and specification documents respectively, extracting key design parameters (e.g., geometric dimensions, line parameters) and specification constraint parameters (e.g., allowable deviation ranges, quality control thresholds), and establishing a set of design parameters. With normative constraint set (After completing the structured analysis of the design drawings and specification documents, the design drawing data is first analyzed at the element level and mapped to fields to identify and extract key design parameters, including route alignment parameters (horizontal curve radius, transition curve length, longitudinal slope and vertical curve parameters), structural geometric dimensions (thickness, width, elevation, cross slope, etc.), and construction layer information. These are then categorized and indexed based on component IDs to form parameter entries with spatial relationships. Simultaneously, the specification documents are analyzed at the clause level. Control indicators corresponding to the above design parameters are identified through rule matching or semantic extraction, and specification constraint parameters are extracted, including allowable deviation ranges (such as elevation deviation).) Thickness deviation , This is the absolute value of the allowable elevation deviation. (absolute value of allowable thickness deviation), quality control threshold (such as compaction degree) Flatness , This represents the minimum threshold for compaction. Flatness index The maximum allowable value of the International Flatness Index (IFI) and acceptance criteria are established, and a mapping relationship between the index and component categories or process status is created. Subsequently, the extracted design parameters and specification parameters undergo unified dimensional conversion and standardized coding to eliminate unit differences and inconsistent expressions. Finally, the design parameter sets are organized according to the structure of "component ID, design parameter type, and parameter value". And the set of structural organization specification constraints according to "component category or process, control index type, and specification constraint parameters". ); S1.2, Based on the set of design parameters Layered modeling is performed according to component categories (at least including roadbed, pavement, bridges, culverts, and ancillary facilities). (Components are road and ancillary structural components in highway engineering, representing the basic engineering object units modeled in the BIM model of highway engineering). Parametric modeling is used to define the geometric parameters, topological relationships, and attribute parameters of the components to generate the initial BIM model. Specifically, based on the set of design parameters. First, semantic recognition is performed on the engineering object information obtained from the analysis of design drawings. Based on pre-defined engineering component classification rules, components are categorized into roadbeds, pavements, bridges, culverts, and ancillary facilities. On this basis, a hierarchical modeling framework is established for different component categories, organized according to a hierarchical relationship of "overall structure layer, substructure layer, and component unit layer," where each layer maintains spatial connectivity and dependencies through topological constraints. Subsequently, a parametric modeling method is employed to... Geometric parameters (such as line type, elevation, thickness, and width) are used as driving variables to construct the geometric parameter representation of each component. Simultaneously, the connection relationships between components (such as adjacency, support, and containment topological relationships) and their attribute parameter information (such as material type, design strength grade, and construction process identifiers) are defined. Finally, the aforementioned geometric parameters, topological relationships, and attribute information are uniformly encapsulated to form an initial BIM model with complete semantic information and spatial structural representation. This provides a foundational model for subsequent spatiotemporal data mapping and dynamic evolution modeling; among which, the initial BIM model The system is organized using a four-layered architecture: geometry, topology, attributes, and semantics. At the bottom layer (geometric layer), the three-dimensional geometric parameters of each road engineering component (such as roadbed, pavement, bridges, culverts, etc.) are constructed parametrically, forming the foundation for spatial representation. At the topology layer, the connections, adjacencies, and dependencies between components (such as hierarchical structural relationships and spatial continuity) are defined, constructing an overall structural network. At the attribute layer, design parameter information (such as dimensions, materials, design indicators, etc.) and a unique component identifier are bound to each component, enabling component-level data management. At the semantic layer, components are assigned category labels and engineering semantic information (such as component type, sub-project to which it belongs, construction stage attributes, etc.), thus forming a unified model architecture with complete spatial structure, engineering logic, and information expression capabilities. S1.3, Initial BIM Model Preprocessing is performed to generate a BIM model. And based on BIM model The initial BIM model is structured, converted into a unified data format (such as JSON / relational tables / graph structure data), generates component-level data records, and aggregates all component-level data records to form a structured BIM dataset. Specifically, this involves: [the process of] the initial BIM model... During preprocessing, model normalization and cleaning operations are first performed, including geometric consistency checks (such as duplicate component removal, topological break repair, and coordinate system unification) and attribute integrity checks (missing field completion and outlier correction), resulting in a normalized model. Subsequently, lightweighting and standardization processes are performed, reducing redundant geometric complexity through mesh simplification or parameter trimming, and standardizing component attribute fields according to a unified data dictionary (such as field naming, unit unification, and enumeration value normalization), generating the preprocessed BIM model. Based on this, The process involves a structured representation, breaking down each component into a unified data structure consisting of "component ID, geometric representation (such as parametric dimensions or simplified mesh features), topological relationships (adjacency / connection / containment relationships), attribute fields (materials, design parameters, etc.), and semantic tags (component category, sub-project)". This structure is then converted into standard data formats (such as JSON objects, relational data tables, or graph structure node and edge representations) according to application requirements. Finally, component-level data records are generated using the component ID as the primary key, and all component-level data records are summarized and indexed to form a queryable and computable structured BIM dataset, providing a unified data foundation for subsequent spatiotemporal mapping and dynamic modeling.

[0018] S2. Collect multi-source data during the construction process, and based on the timestamp synchronization mechanism and spatial positioning information, combine the BIM spatiotemporal dynamic mapping method based on component state phase space gating to associate the multi-source data with the BIM model and generate a time series dataset. In this embodiment, based on the timestamp synchronization mechanism and spatial positioning information, combined with the BIM spatiotemporal dynamic mapping method based on component state phase space gating, multi-source data is associated with the BIM model, and a time-series dataset is generated, including the following steps: S2.1 Collect multi-source data during highway construction, including at least 3D laser scanning point clouds, construction progress data (such as process completion time, construction logs), measurement data (such as elevation, flatness), sensor data (temperature, strain, compaction, etc.), and inspection image data (images / videos). Establish a unified data access interface for data from different sources, and structurally encapsulate the multi-source data, recording its source identifier, acquisition time, and initial spatial information (initial spatial information includes at least GPS geographic coordinates, RTK engineering measurement coordinates, and linear reference coordinates based on mileage markers), forming a multi-source raw dataset. Specifically, deploy corresponding data acquisition methods for different data types, including obtaining construction progress data (process completion time, construction records, etc.) from the construction management system or log system, collecting measurement data (elevation, flatness, etc.) through total stations / laser measurement equipment, and utilizing on-site sensor networks. Physical data (temperature, strain, compaction, etc.) and image data (images / videos) are collected through inspection equipment (drones, mobile terminals). A unified data access interface is then designed to adapt protocols and convert formats for data from different sources, achieving standardized access to multi-source data. Each data entry is assigned a source identifier (e.g., device ID, system source identifier). Based on this, the collected data is structured and encapsulated into a standard data structure containing "data content, source identifier, collection time, and spatial information." The collection time is uniformly recorded as a raw timestamp, and the spatial information retains multiple coordinate representations, including GPS geographic coordinates, RTK high-precision engineering coordinates, and linear reference coordinates based on mileage markers. Finally, the encapsulated data is uniformly stored and indexed to form a multi-source raw dataset, providing fundamental data support for subsequent time synchronization, spatial mapping, and component association. S2.2 Constructing a continuous time axis using linear interpolation method ( (Indexing discrete-time observation points) to obtain standardized timestamps Convert the acquisition time (i.e., the original timestamp) in step S2.1 into a standard timestamp. Specifically, this involves: firstly, analyzing the original timestamps collected during the construction process. A time reference unification process is performed, establishing a standard time reference system based on a pre-defined time reference system (such as UTC or a unified engineering clock). Subsequently, the original time series is sorted and interval analyzed to identify time gaps and non-uniform distribution characteristics between adjacent acquisition time points, and discrete-time observation sequences are constructed accordingly. Based on this, linear interpolation methods are used to reconstruct the non-uniform time sampling points, connecting two adjacent known time points... and Between these points, interpolation calculations are performed on missing or irregular time points based on the proportional relationship of time changes, thereby generating a uniformly distributed continuous time axis. Finally, the original timestamp Mapped to the nearest standard time scale on this continuous time axis, and through a time mapping function. Perform a normalization transformation to obtain a standardized timestamp. In the formula, As the starting point of the continuous time axis, This is the end point of the continuous time axis. This is the lower limit of the original time. This represents the original time limit. S2.3. Standardize the initial spatial information from multi-source data using a pre-built spatial mapping function. (Converting any local coordinate or measurement coordinate system into a unified project 3D coordinate system) yields a unified 3D spatial coordinate representation. And construct a standard spatial representation Specifically, when standardizing the initial spatial information from multi-source data, the spatial coordinates from different sources are first analyzed and their types identified, including GPS geographic coordinates, RTK engineering measurement coordinates, and linear reference coordinates based on mileage markers, and corresponding coordinate representations are established for each. Then, based on a pre-constructed spatial mapping function... ( This is a collection of raw spatial information, including GPS coordinates, RTK measurement coordinates, and mileage marker coordinates. GPS geographic coordinate data, For RTK engineering measurement coordinates, Linear reference coordinates for mileage station numbers. This indicates that geographic coordinates are transformed to the engineering coordinate system through coordinate rotation. This indicates that GPS geographic coordinates are transformed to the BIM engineering coordinate system through coordinate rotation and translation transformations. This process involves transforming RTK engineering survey coordinates to the BIM engineering coordinate system through coordinate adjustment and benchmark unification correction. A unified model is constructed to model the transformation relationships between different coordinate systems. This includes projecting GPS coordinates onto the engineering coordinate system using a transformation model from the geographic coordinate system; performing benchmark adjustment and coordinate correction on the RTK survey coordinates; and mapping mileage markers to 3D spatial trajectory coordinates using a road centerline parametric function, thus achieving unified alignment between different spatial representations. After coordinate transformation, all spatial data undergoes scale unification, orientation correction, and outlier removal to eliminate measurement errors and system biases. Finally, the processed spatial information is uniformly represented in standard 3D spatial coordinates. And through standardized mapping functions Initial spatial information Convert to standard space representation (Standardized mapping function) It is a processing operator used to uniformly convert multi-source heterogeneous spatial information into standard BIM engineering coordinate representation. Its input is initial spatial information. The output is in standard space representation. Its function is to convert the original spatial information After completing coordinate system one, scale standardization and error correction, it is mapped to the three-dimensional spatial representation under the unified BIM engineering coordinate system, thereby forming computable spatial data under the unified engineering coordinate system, providing a consistent spatial basis for subsequent spatiotemporal mapping and component correlation analysis; S2.4 Based on standardized timestamps and standard spatial representation, a BIM spatiotemporal dynamic mapping method based on component state phase space gating is used to generate state-level component mapping output; In this embodiment, the BIM spatiotemporal dynamic mapping method based on component state phase space gating addresses the specific problem of accurately and dynamically establishing spatiotemporal correlations and identifying the current construction state of components between multi-source heterogeneous data (such as measurement data, sensor data, inspection images, etc.) and components in the BIM model during highway construction. Existing technologies typically only perform simple matching based on spatial location (such as nearest neighbor or spatial inclusion queries), ignoring the temporal evolution of construction states and the uncertainties in multi-source observations, which can easily lead to misjudgment of states or discontinuous mapping. This invention constructs a joint feature vector (timestamp and spatial coordinates) and introduces a state phase space gating mechanism: on the one hand, it utilizes... State embedding maps discrete construction states to a continuous feature space. On the other hand, it performs soft assignment of each state based on the discriminant function and softmax probability, thereby achieving robust handling of noise, missing data, and spatial bias, and can output a weighted fusion continuous state mapping result. This method has stronger noise resistance and fault tolerance, and can adapt to the actual situation of inconsistent data quality at the construction site. It can probabilistically represent the state of components, providing richer confidence information for subsequent dynamic evolution modeling. Through state embedding and gating mechanism, it implicitly utilizes the temporal constraints of construction procedures, avoiding state jumps that violate construction logic, thereby significantly improving the accuracy and engineering rationality of spatiotemporal mapping. The method of generating state-level component mapping output by using BIM spatiotemporal dynamic mapping based on component state phase space gating includes the following steps: According to standard spatial expression The component ID in the corresponding BIM model is located by matching and locating it using spatial indexing methods (such as nearest neighbor search) according to standard spatial representation. (Already unified to the project's 3D coordinate system), firstly, a spatial index structure (such as an R-tree) is established for all components in the BIM model, and the bounding box (AABB) or precise geometric boundary of each component is registered in the index; then, for a given... The index is used to quickly filter out a set of candidate components that may contain the spatial point; then, the precise geometry (such as triangular meshes or solid representations) of each candidate component is examined one by one, and the point containment test is used to determine the exact geometry. Whether it falls inside a component or within its allowable adjacent tolerance range; if multiple components are hit simultaneously (such as overlapping areas), the component corresponding to the lowest level or the nearest construction step is selected according to engineering logic; finally, a unique component ID is output. (Completing the matching and positioning from spatial representation to BIM components), denoted as Based on construction procedures and quality acceptance standards, the components in the BIM model are divided into discrete state sets. ( For the first component A specific construction status. For example (highways): For the unconstructed or initial state, (for the completion of roadbed construction), based on standardized timestamps Constructing joint feature vectors with standard space representation (Used to characterize the spatiotemporal observation state of a component at a specific construction stage, where,) This is the normalized standard timestamp (using Min-Max normalization). This is the normalized standard space representation (using Min-Max normalization); Based on joint feature vectors Introducing a state phase space gating mechanism yields a set of candidate discrete states. (First, in the discrete state set) An internal state phase space gating mechanism is introduced to initially screen all candidate construction states based on the topological constraints of construction procedures and the spatiotemporal consistency condition, eliminating states that do not match the current spatiotemporal observation or violate the procedure sequence constraints, thus obtaining a set of candidate discrete states. (When initially screening all candidate construction states based on the topological constraints of construction procedures and the spatiotemporal consistency conditions, firstly, according to the procedure sequence rules in the construction organization design, establish the sequential logical constraint relationship between construction states, and determine the allowable evolution stage range of the current component in the observed time series; then, the joint feature vectors are...) The standard spatiotemporal characteristics of each candidate construction state (first based on the construction schedule (such as Gantt chart or network plan) for each construction state) start time With end time Perform analysis to determine its standard time window interval. Or the standard duration; then, based on the station range, structural segmentation information, and spatial distribution of BIM model components in the design drawings, combined with the geometric model of the line centerline, the work section corresponding to each construction state is mapped to the three-dimensional engineering coordinate system to form its standard spatial area; on this basis, the process constraints and quality control requirements in the construction specifications are introduced to make consistency corrections and boundary constraint adjustments to the time window and spatial area, eliminating unreasonable or conflicting time and space ranges; finally, the time characteristics and spatial characteristics after unified verification are combined to form the standard time and space characteristics of each construction state for matching, where the time dimension is determined by the current Consistency checks are performed to verify whether the timestamp falls within the standard construction time interval or allowable duration range corresponding to the construction state. Spatial dimension constraints are determined by calculating the spatial distance deviation between the current spatial coordinates and the design location or standard construction area corresponding to the construction state. Based on this, candidate states that do not meet the construction sequence logic (i.e., violate the preset procedure order rules), time window constraints, or spatial deviation threshold constraints are eliminated, retaining only construction states that satisfy consistency conditions in procedure logic, temporal evolution, and spatial distribution. This ultimately yields a subset of candidate states that matches the current spatiotemporal observations, based on a discrete state set. Construct state discrimination functions for different construction states. Used to measure the current joint feature vector With the The degree of matching between each construction state in the phase space is determined, and the activation probability of each construction state is obtained by softmax normalization. ;in, (In the formula, Discrete state Embedded vector, This is a state embedding function used to map discrete construction process states into continuous vector representations. It maps state indices to continuous vectors using a learnable embedding matrix. For vector concatenation operators, For state discrimination weights, For transpose operation, For state discrimination bias, and (Set using expert experience method) to quantify the probability distribution of components belonging to various construction states under the current spatiotemporal observation, thereby providing a basis for state weighted fusion output and determination of the construction state with the highest probability; Based on standard timestamps Mapping operators corresponding to each construction state are constructed using standard spatial representation. ( , For the first The mapping weight matrix for the nth construction state represents the weight matrix for the nth construction state. Under various construction conditions, this set of parameters characterizes the influence of temporal and spatial features on the mapping result. Essentially, it is a learnable or empirically defined linear transformation matrix used to weight and project the input standard timestamps and standard spatial representations into a state-related feature space. This achieves differentiated responses and expressions of spatiotemporal information to the mapping output under different construction conditions. The input is a standardized timestamp. Standard spatial expression The concatenated joint vector outputs a vector in the state embedding space, with the output dimension being... Choose 16 to 32 (to match the dimension of the subsequent state vector). No. One construction status bias item and All parameters are preset or learnable, and are combined with activation probability. Generate state-level component mapping output for each component and each observation time. State-level component mapping output Its core function is to connect the originally discrete construction states (such as "subgrade construction completed" and "pavement base construction in progress") with continuous spatiotemporal observations (timestamps). Standard spatial expression The components are fused into a continuous, differentiable vectorized representation through probabilistic weighting, thus providing a unified, smooth, and probabilistically rich representation of component states for subsequent dynamic evolution modeling and quality assessment. This output retains the contribution of each state to the current spatiotemporal observation (through activation probabilities). ), and also utilize mapping operators By embedding the geometric and physical properties of different states (such as elevation shift and thickness variation trend) into the same vector space, the model can still obtain a continuous, reasonable and numerically applicable state expression even when the state boundary is ambiguous or the observation data is noisy, which significantly improves the robustness and engineering applicability of spatiotemporal dynamic mapping. S2.5. Bind the state-level component mapping output to the component ID in the BIM model, and extract the original observation feature vector from the multi-source data. Specifically, based on the component ID and standardized timestamp, all data records corresponding to the component and the specified time are retrieved from multi-source data, including measurement data (such as elevation difference and flatness value), sensor data (such as temperature, strain, and compaction degree), construction progress data (such as process completion markers), and inspection image data (feature vectors extracted from images using methods such as SIFT, HOG, or deep learning networks). Then, different types of data undergo dimensional unification and missing value processing. For example, continuous measurement values ​​are Z-score standardized, discrete process markers are one-hot encoded, and image features are L2 normalized. Finally, all processed data are concatenated into a fixed-length one-dimensional vector according to a preset dimensional order (such as geometric measurement first, then physical sensing, and finally process and image processing), serving as the original observation feature vector of the component at that time. Generate component-level data records. ; S2.6. Using the component ID in the BIM model as the primary key, record component-level data according to time series. Organize the data to form a time-series dataset (where each record also contains the original observation feature vector). Mapping output with state-level components Specifically, using the component ID in the BIM model as the unique primary key, all component-level data records are first processed. Grouping and aggregating the data involves dividing the observation data from multiple time points corresponding to the same component ID into independent data sequences; then, standardized timestamps are used to... Based on the sorting criteria, the record set of each component is arranged in ascending order of time to form a time-series data chain that evolves continuously according to the construction process. On this basis, structural consistency verification and missing data completion processing are performed on the data at different time steps to ensure the integrity and comparability of the time series. Finally, the ordered data sequences of each component are indexed and organized according to component ID and time series structure to construct a time-series dataset for component-level evolution analysis, thereby realizing continuous tracking and unified management of the state, spatial and semantic information of a single component throughout the entire construction process.

[0019] S3. Based on the time-series dataset, establish a dynamic update mechanism for the BIM model to update the BIM model in time. Introduce the construction process topology to constrain the time-series update process and generate a BIM evolution model that dynamically changes with the construction process. In this embodiment, a dynamic update mechanism for the BIM model is established to perform time-series updates on the BIM model. A construction sequence topology is introduced to constrain the time-series update process, generating a BIM evolution model that dynamically changes with the construction process. This includes the following steps: S3.1. Using the time-series dataset as input, for each component in the BIM model... Extract its corresponding time series data sequence ( The length of the time-series data sequence is given. When using a time-series dataset as input, the structured time-series dataset is first grouped and retrieved using the component ID in the BIM model as the index key, grouping components belonging to the same component. All component-level data records were filtered from the time-series dataset; then, based on standardized timestamps... The selected results are sorted and serialized by time to eliminate time disorder caused by different data sources, and missing time nodes are filled in by interpolation or neighboring states. Based on this, the component observation records arranged in ascending time order are organized into an ordered time series data sequence. Extracting raw observation feature vectors from time-series data sequences. And combine state-level component mapping output Construct component-level temporal state vectors ( The feature encoding function (such as a multilayer perceptron, using a three-layer fully connected network, with the input layer dimension being the original observed feature vector) is used. The dimension and state-level component mapping output The sum of dimensions (original observed feature vectors) The dimension is , The dimension is The total input dimension is The hidden layer contains two fully connected layers with 128 and 64 neurons respectively. The activation function for both layers is ReLU. The output layer dimension is a predefined component-level temporal state vector. Dimensions The output layer does not use an activation function; the state transition matrix... Implemented as a learnable linear layer with an input dimension of (in External driving variables The dimension is 8), and the output dimension is... No activation function; mapping operator The weight matrix in Implemented as a fully connected layer, with an input dimension of 2 (the dimension resulting from concatenating the standardized timestamp and the standardized spatial representation), and an output dimension of... Bias term For learnable parameters, no activation function is used; embedding matrix The dimension is ,in The total number of discrete states. The above network structures are all implemented using standard linear layers in the PyTorch framework (parameters are trained end-to-end using mean squared error as the loss function from a time-series dataset generated during construction). This is used to fuse component observation data with state mapping results, generating a unified temporal state representation of the component. Essentially, it is a feature transformation function that linearly or nonlinearly fuses multi-source component observation data with state-level mapping results and maps them to a unified temporal state vector. Indicates component index, (representing time); where, for the first observation time... Define the initial component state vector Generation method: based on encoding function Mapping ,in , The first record from the time series dataset This state serves as the starting point for the dynamic update mechanism, used to predict the next time step. The state of ). From Begin by performing a recursive update following steps S3.2–S3.4; S3.2, Based on component temporal state vector Dynamic evolution modeling of component states yields predicted states. ( Externally driven variables (such as construction procedures, environmental conditions, and equipment status) are derived from extracting the current procedure identifier from construction logs, acquiring environmental and equipment status data from the on-site sensor network, and normalizing (e.g., using Min-Max normalization) and vectorizing and encoding (one-hot encoding) these data to form a... External driving vector for dimension matching The state transition matrix (input is the component-level temporal state vector) (dimension) (e.g., 32) with external driving variables (dimension) (e.g., 4-10, including process identification, ambient temperature, construction equipment status, etc.) spliced ​​together. A dimensional vector, the output of which is a predicted state vector, with dimensions equal to or greater than the given dimensions. Same (i.e.) ( ), which is a learnable linear weight matrix. (These are bias terms, preset or learnable parameters), and a construction sequence topology is introduced to constrain the dynamic evolution modeling process of component states and to predict the states. Perform topological projection correction to generate the corrected predicted state. ; In this embodiment, a construction sequence topology is introduced to constrain the dynamic evolution modeling process of component states and to predict the states. Topology projection correction addresses the specific problem of process logic violations (such as jumping to surface layer construction before completing base layer construction or state reversal) that occur when relying solely on sensor or measurement data to predict component states during highway construction. This invention explicitly incorporates topological constraints of the directed graph of construction processes and corrects predicted states that do not conform to process logic to the correct path through projection operators. This ensures the engineering rationality and standard consistency of the state evolution sequence, while also taking into account the flexibility of data-driven approaches. It effectively avoids non-physical state jumps caused by noise or outliers in pure data methods, thereby improving the credibility of quality assessment and early warning. This involves introducing a construction sequence topology to constrain the dynamic evolution modeling process of component states and to predict the states. Perform topological projection correction to generate the corrected predicted state. This includes the following steps: S3.21. Based on construction organization design and specifications, construct the directed process graph of BIM model components. , This refers to the sequence of processes (directed edges), for example: road surface structure. , The construction status at the grassroots level. This refers to the construction state of the subbase or foundation layer. This is the surface layer construction state; S3.22, Activation probabilities obtained from step S2.4 Determine the construction state with the highest probability for the current BIM model component. (Select the construction state with the highest probability from the multi-state probability distribution as the deterministic state representation of the current component, which is used for subsequent process determination and state evolution modeling.) S3.23, Based on the directed graph of the process For the construction state with the highest probability Perform process consistency judgment: (In the formula, This is the result of the process consistency judgment. This indicates that the process topology constraints are satisfied. This indicates that the process topology constraints are not met. For components In the next moment The construction state with the highest probability; specifically: the predicted state obtained from step S3.2. Mapping to discrete states yields the construction state with the highest probability. The mapping method is: calculate Embedded representations of each discrete state The Euclidean distance is used to determine the state with the smallest distance. Then calculate the process consistency judgment result: if , ,otherwise ; The purpose of process consistency determination is to align the current construction state with the highest probability obtained through data-driven analysis with the permissible process paths constrained by the construction organization design and specifications. This determines whether the state conforms to the preset construction logic sequence, avoiding process logic deviations or unreasonable state identification caused by relying solely on probabilistic inference. The core function of this determination process is to provide a constraint basis for subsequent state evolution and prediction correction by identifying whether the current state transition falls within the directed process graph. Within the allowed edge set, structured constraints are implemented on the evolution path of the construction process, thereby ensuring the rationality and feasibility of the process of component state evolution in the time dimension in the BIM model; S3.24, Based on the process consistency judgment result (when a state transition that does not meet the construction process topology constraints is detected ( When introducing a process-based directed graph, For the predicted state Perform constraint corrections to generate a topologically consistent state vector. This is used to ensure that the evolution path strictly follows the topological constraints of the construction process; Furthermore, regarding the predicted state Perform constraint corrections to generate a topologically consistent state vector. This includes the following steps: Based on the directed graph of the process For the construction state with the highest probability Extract the set of transitional states that satisfy the process constraints. (Procedure constraints refer to the permissible rules for state transitions defined by the construction organization design and specifications, specifically embodied in a directed process graph) The set of directed edges in : Only if there exists a path from the current state Pointing to state The directed edges (i.e.) Only when ) is it allowed to be from ^Transfer to It also allows remaining in the current state ( This disallows actions that violate construction logic, such as skipping steps (e.g., proceeding directly to the surface layer without completing the base layer), reversing steps, or skipping intermediate processes, based on a discrete state set. Calculate discrete state Embedded representation (The embedding representation is obtained by multiplying one-hot encoding with the learnable embedding matrix) First, for each discrete state Assign a unique index and construct a dimension that represents the total number of states. One-hot encoded vector , of which only in the first One position is 1 and the rest are 0; then a learnable embedding matrix is ​​defined. Its dimensions are ( For embedded dimensions, (total number of states), multiply the one-hot encoded vector by the embedding matrix (i.e., The result is one A dimensional vector, which represents the discrete state. Embedded representation ), and according to the embedded representation and predicted state Generate embedding distance (Used to characterize the degree of deviation in feature space between the predicted state and the candidate state), and based on the directed graph of the process. Calculate the construction state with the highest probability. To discrete state Topological distance (Based on process directed graph) First, each construction state node in the graph is considered a vertex, and each directed edge represents the sequence of operations and is assigned a weight of 1 (i.e., the transition cost of each operation is 1). Then, the construction state with the highest probability is selected. Starting node, discrete state Using the destination as the starting point, a breadth-first search (BFS) algorithm is used to traverse the directed graph. If the target node can be reached from the starting node along the directed edges, then the topological distance is calculated. It equals the number of edges traversed by the shortest path (i.e., the minimum number of steps required); if there is no directed path (such as in discrete states) If the process is unreachable or violates the process sequence, then set a topology distance. A penalty constant that is much larger than the reasonable path length. (For example (thus effectively penalizing unreachable state transitions in the projection cost function); based on embedding distance and topological distance And introduce process topology penalty coefficient. Generate weighted cost function Topologically consistent state vectors are generated through discrete projection based on a weighted cost function. The output is a continuous vector (embedded representation), with dimensions equal to... The same applies; it can be directly used for subsequent updates and deviation calculations. S3.25, Based on topologically consistent state vectors Based on the process consistency judgment results, the predicted status is... Perform selective corrections to generate corrected predicted states. Based on the consistency determination results of the construction process, only when the predicted state violates the directed graph constraints of the construction process (i.e., Only when the state evolution strictly follows the allowed process path is the original predicted state replaced by the state corrected by topological projection. This preserves the original information of the data-driven prediction while avoiding changes to the allowed state transition. This selective correction mechanism not only corrects illegal process jumps (such as skipping necessary processes) caused by sensor noise, missing data, or model errors, but also maintains a sensitive response to the normal construction process, achieving a balance between engineering logic constraints and data-driven flexibility. S3.3 Based on the component-level data records and state-level component mapping outputs at the same time, construct the component-level temporal state vector according to step S3.1. Construct the observation state in the same way and the predicted state Perform deviation calculation to obtain the update increment. (Used to characterize the deviation between the predicted state and the actual observed state, and to correct the prediction results accordingly, thereby improving the accuracy and consistency of component state evolution); based on update increments The component state is corrected to generate an updated component state vector. ( (To update the gain coefficient), used to fuse predicted state and actual observation information, and reduce model prediction error through closed-loop correction, thereby improving the accuracy and robustness of component state evolution; S3.4. Update the component state vector Mapping the geometric and attribute parameters to the BIM model's space generates component update functions: In the formula, For components In the The constantly updated BIM sub-model For components In the BIM sub-models for each construction phase For component update operators (component update operators) The execution logic is as follows: First, the updated component state vector is processed. Semantic groupings are used to split the data into geometric sub-vectors (corresponding to elevation offset, thickness correction, alignment adjustment, etc.), physical sub-vectors (corresponding to compaction degree, strength, construction completion degree, etc.), and semantic sub-vectors (corresponding to construction status, etc.). Each sub-vector is converted into correction values ​​corresponding to the BIM model parameters through a learnable linear mapping layer: geometric parameter correction values ​​and attribute parameter correction values. Then, the current component model is used as the basis for further processing. Based on the geometric and attribute parameters, corresponding correction amounts are superimposed to obtain preliminary update results. On this basis, a set of standard constraints is introduced to perform boundary checks and consistency corrections on the updated geometric and attribute parameters to avoid parameter out-of-bounds errors or structural conflicts. Finally, the corrected parameters are written back to the corresponding components in the BIM model to generate updated component sub-models. Specifically: first, the component state vector... Semantic decoding and dimensional decomposition are performed, dividing the data into geometrically related sub-vectors (such as elevation offset, thickness correction, and alignment adjustment) and attribute-related sub-vectors (such as compaction degree, strength, and construction completion degree). Then, a mapping function is established from state parameters to BIM model parameters. The geometrically related sub-vectors are applied to the geometric control points, dimensional parameters, and spatial coordinates of the components through a parameter-driven approach, enabling dynamic adjustment of the component's shape. Simultaneously, the attribute-related sub-vectors are mapped to the component attribute fields to update material properties, quality indicators, and construction status. Based on this, standard constraints are introduced to perform boundary checks and rationality corrections on the update results, avoiding parameter out-of-bounds errors or structural conflicts. Finally, a unified component update operator is used. The mapped geometric and attribute parameters are written back to the corresponding BIM model sub-parameters. This enables dynamic updating of the component-level model; S3.5. Based on the component update function, topological constraints and engineering logic constraints are introduced to generate a global BIM model: In the formula, Represents the component model variables to be optimized. This is the evolution sequence of the BIM model. To standardize the constraint penalty function, representing the degree of deviation between the BIM model and the standard parameter set, For quality, geometric or attribute indicators, The penalty weight for the indicator, For components The A set of actual parameter values ​​(such as elevation, thickness, compaction degree, etc.). To standardize the standard values, Tolerance (allowable deviation range); S3.6 Organize the global BIM model according to the time sequence to form a BIM model sequence that dynamically changes with the construction process. ( (The total number of discrete time steps in the construction process), this BIM model sequence is the BIM evolution model, used to characterize the structural state evolution throughout the entire highway construction process.

[0020] S4. Extract quality feature vectors based on the BIM evolution model, construct a quality assessment function based on the quality feature vectors, and obtain a comprehensive quality score; In this embodiment, a quality feature vector is extracted based on the BIM evolution model, and a quality assessment function is constructed based on the quality feature vector to obtain a comprehensive quality score, including the following steps: S4.1 Extracting Quality Observation Vectors Based on BIM Evolution Model ( Geometric quality characteristics (elevation deviation, thickness deviation, flatness, etc.) These are physical quality characteristics (compaction degree, strength, temperature uniformity, etc.). Semantic features (construction status, process completion, etc.); specifically: firstly, based on time steps BIM model For input, position the component In the BIM evolution model, the geometric and attribute parameter nodes are analyzed to obtain raw information by parsing their underlying BIM data structures (such as the geometric representation of IFC objects, attribute sets Pset, and semantic tags). Then, geometric quality characteristics are calculated from the geometric parameters, including elevation deviation, section thickness difference, line center offset obtained by comparing design values ​​with actual model coordinates, and flatness error obtained based on point cloud or mesh fitting. Simultaneously, physical quality characteristics such as compaction degree, material strength, and temperature field uniformity are extracted from component attribute fields, and semantic features are obtained by combining construction logs or process status tags. Based on this, data from different sources are scaled and normalized to map heterogeneous indicators to the same numerical space. Finally, the normalized geometric features, physical quality features, and semantic features are fused according to component and time indices to form a quality observation vector. ; S4.2, Based on the set of normative constraints in S1 Extract construction quality control parameters, construct a set of quality indicators, calculate the deviation of each quality indicator, and generate a quality deviation vector. Specifically: First from The analysis of construction quality control parameters should include at least the geometric tolerance range (such as elevation error limit, thickness tolerance, and flatness index); from The system filters entries whose "Control Index Type" field contains geometric keywords such as "Elevation Deviation," "Thickness Deviation," "Smoothness IRI," "Width Deviation," and "Cross Slope Deviation." Then, it analyzes the numerical and symbolic representations in the "Specification Constraint Parameters" field, as well as physical performance control thresholds (such as lower compaction limit, strength grade requirements, and temperature uniformity range). The process involves filtering entries whose "Control Index Type" field contains physical performance keywords such as "compaction degree," "strength grade," "temperature uniformity," "deflection value," and "elastic modulus," then parsing the comparison operators and target values ​​in the "Specification Constraint Parameters" field. Finally, the aforementioned specification parameters are compared with the component at time... quality observation vector The corresponding indicators are aligned and matched one by one to form a quality indicator comparison relationship. On this basis, the deviation between the actual observed value and the standard value of each quality indicator is calculated, and expressed quantitatively, for example, using difference or relative deviation. The part that exceeds the allowable range is explicitly marked. Finally, the deviation values ​​of all quality indicators are structured and vectorized according to geometric, physical and semantic dimensions to generate a quality deviation vector. ; S4.3, Based on quality observation vector and quality deviation vector Generate quality feature vectors and the quality feature vector Perform normalization; S4.4 Construct a multi-index evaluation function based on the normalized quality feature vector to generate a quality score. (In the formula, This represents the total number of quality indicator dimensions used for quality assessment. For quality indicator index, As the indicator weight, For the first The evaluation mapping function for each indicator uses the Sigmoid function, and the normalized eigenvalues ​​are used as the basis for the evaluation mapping function. To avoid saturation, the region is adjusted to the sensitive range of the Sigmoid function (e.g., [-3, 3]) using a linear transformation. (The normalized quality eigenvectors). S4.5. Weighted summation of the quality scores for all components to generate a comprehensive quality score. (In the formula, The total number of components. (Component weights).

[0021] S5. Based on the comprehensive quality score, output the risk level through the graded threshold and conduct dynamic early warning of highway construction quality; In this embodiment, the risk level is output, and dynamic early warning of highway construction quality is provided, including the following steps: Comprehensive quality score The data is organized chronologically to form a quality score time series. Based on this time series, a risk assessment is performed on the comprehensive quality score at each moment using a preset grading threshold, generating a risk level (where the preset grading threshold is...). ( This serves as the dividing line between high risk and medium-high risk. This serves as the dividing line between medium-high risk and medium-low risk. In this embodiment, the threshold between low-to-medium risk and safety level is used as the dividing line. It is 60. It is 75. Given a score of 90 (out of 100), the scores are first sorted according to their numerical values ​​to construct a risk grading interval system, resulting in four consecutive intervals: , , and Subsequently, a comprehensive quality score was calculated for each moment. Perform interval matching determination, when Falling into the first interval ( When it falls into the first interval, it is judged as a high-risk level; when it falls into the second interval, it is judged as a high-risk level. When it falls into the third interval, it is judged as a medium-to-high risk level. When it falls into the fourth interval, it is judged as a medium-low risk level. When the risk level is determined to be "high risk," a safety level is established, thus achieving threshold-based segmentation of construction quality risk classification. Based on the correspondence between risk levels and preset early warning trigger rules, corresponding level of construction quality early warning information is output for dynamic early warning of construction quality. (Preset early warning trigger rule: When the risk level is "high risk," a level one early warning is triggered, immediately sent to the project manager and supervising engineer via a pop-up window in the BIM model interface and SMS.) This is also reflected on the digital construction platform (a collaborative management information system integrating BIM models, construction progress, quality monitoring, and early warning notifications, used to achieve digitalization, visualization, and dynamic monitoring of the construction process). (For control purposes) Components are highlighted in red. When the risk level is "medium-high," a level two warning is triggered, sent to the construction team leader via system notification and email, and the problem area is flashed in orange in the BIM model, requiring a rectification plan to be submitted within 24 hours. When the risk level is "medium-low," a level three warning is triggered, recorded only in the system log and highlighted in yellow on the quality dashboard, requiring daily monitoring by quality inspectors. When the risk level is "safe," no warning is triggered, only a green indicator. All warning information includes the component ID, timestamp, quality score, and deviation details, thus achieving dynamic early warning and closed-loop management of construction quality.

[0022] Example 2: This example provides a highway BIM intelligent quality assessment system, including a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement the highway BIM intelligent quality assessment method described in Example 1 above.

[0023] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A BIM-based intelligent quality assessment method for highways, characterized in that, include: S1. Establish a BIM model using design drawings and specification documents, preprocess the BIM model, and form a structured BIM dataset; S2. Collect multi-source data during the highway construction process, and based on the timestamp synchronization mechanism and spatial positioning information, combine the BIM spatiotemporal dynamic mapping method based on component state phase space gating to associate the multi-source data with the BIM model and generate a time series dataset. S3. Based on the time-series dataset, establish a dynamic update mechanism for the BIM model to update the BIM model in time. Introduce the construction process topology to constrain the time-series update process and generate a BIM evolution model that dynamically changes with the construction process. S4. Extract quality feature vectors based on the BIM evolution model, construct a quality assessment function based on the quality feature vectors, and obtain a comprehensive quality score; S5. Based on the comprehensive quality score, output the risk level through graded thresholds and conduct dynamic early warning of highway construction quality.

2. The intelligent quality assessment method for highway BIM according to claim 1, characterized in that, In step S1, a BIM model is established using design drawings and specification documents. The BIM model is then preprocessed to form a structured BIM dataset, including the following steps: S1.1 Collect design drawings and specification documents, and construct a set of design parameters based on the design drawings. At the same time, a set of normative constraints is constructed based on the normative documents. ; S1.2, Based on the set of design parameters The model is layered according to component categories; parametric modeling is used to define the geometric parameters, topological relationships, and attribute parameters of the components to generate the initial BIM model. ; S1.3, Initial BIM Model Preprocessing is performed to generate a BIM model. And based on BIM model The data is structured, converted into a unified data format, and component-level data records are generated. All component-level data records are then aggregated to form a structured BIM dataset.

3. The intelligent quality assessment method for highway BIM according to claim 1, characterized in that, In step S2, based on the timestamp synchronization mechanism and spatial positioning information, combined with the BIM spatiotemporal dynamic mapping method based on component state phase space gating, multi-source data is associated with the BIM model, and a time-series dataset is generated, including the following steps: S2.1 Collect multi-source data during the highway construction process, establish a unified data access interface for data from different sources, encapsulate the multi-source data in a structured manner and record its source identifier, collection time and initial spatial information to form a multi-source raw dataset; S2.2 Constructing a continuous time axis using linear interpolation method Obtain a standardized timestamp Convert the acquisition time in step S2.1 into a standard timestamp. ; S2.

3. Perform unified standardization processing on the initial spatial information in the multi-source data, obtain a unified three-dimensional spatial coordinate expression based on the pre-constructed spatial mapping function, and construct a standard spatial expression; S2.4 Based on standardized timestamps and standard spatial representation, a BIM spatiotemporal dynamic mapping method based on component state phase space gating is used to generate state-level component mapping output; S2.

5. Bind the state-level component mapping output to the component ID in the BIM model, and extract the original observation feature vector from the multi-source data. Generate component-level data records ; S2.

6. Using the component ID in the BIM model as the primary key, record component-level data according to time series. Organize the data to form a time-series dataset.

4. The intelligent quality assessment method for highway BIM according to claim 3, characterized in that, In step S2.4, the state-level component mapping output is generated using the BIM spatiotemporal dynamic mapping method based on component state phase space gating, including the following steps: Based on standard spatial representation, the component ID in the corresponding BIM model is located using spatial indexing methods, denoted as [ID]. Based on construction procedures and quality acceptance standards, the components in the BIM model are divided into discrete state sets. Based on standardized timestamps Constructing joint feature vectors with standard space representation ; Based on joint feature vectors Introducing a state phase space gating mechanism yields a set of candidate discrete states. Based on discrete state set State discrimination functions are constructed for different construction states, and the activation probabilities of each construction state are obtained by softmax normalization. ; Based on standard timestamps Mapping operators corresponding to each construction state are constructed using the standard spatial representation, and activation probabilities are combined. Generate state-level component mapping output.

5. The intelligent quality assessment method for highway BIM according to claim 1, characterized in that, In step S3, a dynamic update mechanism for the BIM model is established to update the BIM model over time. A construction sequence topology is introduced to constrain the time-series update process, generating a dynamically changing BIM evolution model that evolves with the construction process. This includes the following steps: S3.

1. Using the time-series dataset as input, for each component in the BIM model... Extract the corresponding time series data sequence, and extract the original observation feature vector from the time series data sequence. Combined with the state-level component mapping output, a component-level temporal state vector is constructed. ; S3.2, Based on component temporal state vector Dynamic evolution modeling of component states yields predicted states. Furthermore, a construction sequence topology is introduced to constrain the dynamic evolution modeling process of component states and to predict the states. Perform topological projection correction to generate the corrected predicted state. ; S3.3 Constructing the observation state based on component-level data records and state-level component mapping outputs at the same time. and the predicted state Perform deviation calculation to obtain the update increment. Based on update increment The component state is corrected to generate an updated component state vector; S3.4 Map the updated component state vector to the geometric and attribute parameter space of the BIM model to generate the component update function; S3.

5. Based on the component update function, topological constraints and engineering logic constraints are introduced to generate a global BIM model; S3.6 Organize the global BIM model according to the time series to form a BIM model sequence that dynamically changes with the construction process. This BIM model sequence is the BIM evolution model, which is used to characterize the structural state evolution throughout the entire highway construction process.

6. The intelligent quality assessment method for highway BIM according to claim 5, characterized in that, In step S3.2, a construction sequence topology is introduced to constrain the dynamic evolution modeling process of component states and to predict the states. Perform topological projection correction to generate the corrected predicted state. This includes the following steps: S3.

21. Based on construction organization design and specifications, construct the directed process graph of BIM model components. ; S3.22, Activation probabilities obtained from step S2.4 Determine the construction state with the highest probability for the current BIM model component. ; S3.23, Based on the directed graph of the process For the construction state with the highest probability Perform process consistency assessment; S3.

24. Based on the process consistency determination result, introduce a process-based directed graph. For the predicted state Perform constraint corrections to generate a topologically consistent state vector. ; S3.25, Based on topologically consistent state vectors Based on the process consistency judgment results, the predicted status is... Perform selective corrections to generate corrected predicted states. .

7. The intelligent quality assessment method for highway BIM according to claim 6, characterized in that, In S3.24, a directed graph based on process steps is introduced. For the predicted state Perform constraint corrections to generate a topologically consistent state vector. This includes the following steps: Based on the directed graph of the process For the construction state with the highest probability Extract the set of transitional states that satisfy the process constraints. Based on discrete state set Calculate discrete state Embedded representation and according to the embedded representation and predicted state Generate embedding distances and based on the process directed graph. Calculate the construction state with the highest probability. To discrete state The topological distance is calculated based on the embedding distance and the topological distance, and a process topological penalty coefficient is introduced. Generate a weighted cost function; based on the weighted cost function, generate a topologically consistent state vector through discrete projection.

8. The intelligent quality assessment method for highway BIM according to claim 1, characterized in that, In step S4, quality feature vectors are extracted based on the BIM evolution model, and a quality assessment function is constructed based on the quality feature vectors to obtain a comprehensive quality score. This includes the following steps: S4.1 Extracting quality observation vectors based on the BIM evolution model; S4.2, Based on the set of normative constraints in S1 Extract construction quality control parameters, construct a set of quality indicators, calculate the deviation of each quality indicator, and generate a quality deviation vector; S4.

3. Based on the quality observation vector and the quality deviation vector, generate the quality feature vector and normalize the quality feature vector; S4.4 Construct a multi-index evaluation function based on the normalized quality feature vector to generate a quality score; S4.

5. Weighted summation of the quality scores of all components to generate a comprehensive quality score.

9. The intelligent quality assessment method for highway BIM according to claim 1, characterized in that, In S5, the risk level is output based on the comprehensive quality score through a graded threshold, and dynamic early warning of highway construction quality is carried out, including the following steps: The comprehensive quality scores are organized in chronological order to form a quality score time series. Based on the quality score time series, the risk assessment is performed on the comprehensive quality score at each moment using preset grading thresholds to generate a risk level. Based on the correspondence between the risk level and preset early warning triggering rules, the corresponding level of construction quality early warning information is output for dynamic early warning of construction quality.

10. A highway BIM intelligent quality assessment system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: The processor executes a computer program to implement the highway BIM intelligent quality assessment method as described in any one of claims 1-9.

Images