Steel bar engineering quality detection method, system and equipment based on large model and medium

Abstract

The invention discloses a large-model-based steel bar engineering quality detection method, system and equipment and a medium, and relates to the technical field of data processing. The method comprises the following steps: acquiring vibration response data and three-dimensional point cloud data of a to-be-detected reinforcement project; inputting the vibration response data into a pre-trained vibration characteristic analysis model to obtain a stress distribution map and a stress transmission path, and performing segmentation processing on the three-dimensional point cloud data along the stress transmission direction to obtain segmented point cloud data; based on the segmented point cloud data, extracting morphological characteristics of each section of reinforcing steel bar in the to-be-detected reinforcing steel bar project, and establishing a stress-geometry coupled digital twinborn model according to the stress distribution map and the morphological characteristics; performing dynamic mapping on the digital twinborn model and a pre-calibrated BIM design model, and determining steel bar deformation parameters of the stress abnormal region; and based on the steel bar deformation parameters, generating a quality detection report of the to-be-detected steel bar project. By implementing the technical scheme provided by the invention, the accuracy of reinforcing steel bar engineering quality detection can be improved.

Description

Technical Field

[0001] This application relates to the field of data processing technology, specifically to a method, system, equipment, and medium for quality inspection of reinforced concrete engineering based on a large model. Background Technology

[0002] With the continuous expansion of construction projects and the increasing complexity of structural designs, steel reinforcement engineering, as a crucial component of building structures, directly impacts the overall safety performance of buildings through its construction quality. Therefore, steel reinforcement engineering quality inspection, as a key aspect of construction projects, has received widespread attention within the industry.

[0003] Currently, existing methods for quality inspection of reinforced concrete projects mainly include visual inspection, spot checks, ultrasonic testing, and magnetic induction testing. However, in practical applications, these methods often focus on the analysis of single data types, making it difficult to achieve a comprehensive evaluation of the reinforced concrete structure. This results in incomplete inspections and reduces the accuracy of quality inspection for reinforced concrete projects. Summary of the Invention

[0004] This application provides a method, system, equipment, and medium for inspecting the quality of reinforced concrete engineering based on a large model, which can improve the accuracy of inspecting the quality of reinforced concrete engineering.

[0005] Firstly, this application provides a method for quality inspection of reinforced concrete engineering based on a large model, including: Acquire vibration response data and 3D point cloud data of the steel reinforcement project to be inspected; The vibration response data is input into a pre-trained vibration feature analysis model to obtain a stress distribution map and stress transmission path. Based on the stress transmission path, the three-dimensional point cloud data is segmented along the stress transmission direction to obtain segmented point cloud data. Based on the segmented point cloud data, the morphological features of each segment of steel bar in the steel bar project to be inspected are extracted, and a stress-geometry coupled digital twin model is established according to the stress distribution map and morphological features. The digital twin model is dynamically mapped to the pre-calibrated BIM design model to determine the steel reinforcement deformation parameters in the stress anomaly area; Based on the steel bar deformation parameters, a quality inspection report for the steel bar project to be inspected is generated.

[0006] By adopting the above technical solution, vibration response data and 3D point cloud data of the steel reinforcement project to be inspected are obtained as multi-dimensional inputs. A pre-trained vibration feature analysis model is used to extract stress distribution maps and stress transmission paths. Based on the stress transmission direction, the 3D point cloud data is segmented, enabling segmented analysis of the steel reinforcement structure. Furthermore, by extracting the morphological characteristics of each steel segment and combining them with the stress distribution map to establish a stress-geometry coupled digital twin model, and then dynamically mapping and comparing it with the BIM design model, the deformation parameters of the steel reinforcement in areas of stress anomalies are accurately identified. Finally, a comprehensive quality inspection report is generated. This not only achieves a multi-dimensional evaluation of the structural performance of the steel reinforcement project, but also establishes a dynamic relationship between the physical steel reinforcement project and the virtual model through digital twin technology. This overcomes the problem that single detection methods in existing technologies cannot achieve comprehensive evaluation, achieving more comprehensive inspection and thus improving the accuracy of steel reinforcement quality inspection.

[0007] Optionally, a training sample library containing various types of reinforced concrete structures is constructed. Each training sample in the training sample library includes vibration response data, ground truth stress distribution data, and stress transmission path annotation data. Data augmentation processing is performed on the vibration response data in the training sample library to obtain an expanded training dataset. The expanded training dataset is input into a preset base model to obtain predicted stress distribution patterns and stress transmission paths. The loss function between the predicted stress distribution patterns and stress transmission paths and the corresponding ground truth data is calculated. Based on the loss function, the model parameters of the preset base model are iteratively optimized until the preset base model converges. The converged preset base model is used as a pre-trained vibration feature analysis model.

[0008] Optionally, the direction vector and node coordinates of the stress transfer path are obtained; a local coordinate system is established based on the node coordinates, and the three-dimensional point cloud data is mapped to the local coordinate system; density clustering analysis is performed on the mapped point cloud data along the direction vector to determine the segment boundaries of the point cloud data; and region growing processing is performed on the point cloud data between the segment boundaries to obtain segmented point cloud data.

[0009] Optionally, the normal vector and curvature feature value of each point in the point cloud data are calculated; the point with the smallest curvature feature value within each segment boundary is selected as the seed point; based on the normal vector of the seed point, the search radius and the normal vector angle threshold are calculated, wherein the search radius is proportional to the point cloud density, and the normal vector angle threshold is inversely proportional to the curvature feature value of the seed point; the neighboring points of the seed point are traversed, and the neighborhood points within the search radius and whose normal vector angle is less than the normal vector angle threshold are added to the current region in turn, and the average normal vector of the current region is updated. The updated average normal vector is used as the target search direction, and the traversal process is repeated until no new neighborhood points are added to the current region, thus obtaining the segmented point cloud data.

[0010] Optionally, the segmented point cloud data is analyzed in the main direction to determine the axial direction of each rebar segment; for each rebar, a sampling plane is established based on the axial direction, and cross-sectional point clouds of the rebar are obtained on the sampling plane at preset intervals; ellipse fitting is performed on the cross-sectional point clouds to obtain the cross-sectional parameters of the rebar; based on the cross-sectional parameters of the rebar, the curvature and torsion angle of the rebar are determined; the curvature and torsion angle are used as the morphological characteristics of the rebar.

[0011] Optionally, a geometric topology diagram of the rebar nodes in the rebar project to be inspected is constructed; based on the stress distribution map, the stress transfer weight between each rebar node is calculated; the stress transfer weight is superimposed on the geometric topology diagram to obtain a stress-geometric coupling diagram; based on the stress-geometric coupling diagram and the morphological features, a finite element mesh model of the rebar project to be inspected is constructed; stress field interpolation is performed on the finite element mesh model to generate a stress-geometric coupling digital twin model.

[0012] Optionally, a feature mapping matrix is ​​constructed between the digital twin model and the BIM design model; based on the feature mapping matrix, the deviation value between the measured coordinates and the design coordinates of the rebar nodes is determined; stress anomaly areas where the deviation value is greater than the deviation threshold are screened out; and the elastic deformation and plastic deformation of the rebars within the stress anomaly areas are extracted as rebar deformation parameters.

[0013] A second aspect of this application provides a large-model-based steel reinforcement engineering quality inspection system, the system comprising: The data acquisition module is used to acquire vibration response data and three-dimensional point cloud data of the steel reinforcement project to be inspected; The digital twin model building module is used to input the vibration response data into a pre-trained vibration feature analysis model to obtain a stress distribution map and stress transmission path. Based on the stress transmission path, the three-dimensional point cloud data is segmented along the stress transmission direction to obtain segmented point cloud data. Based on the segmented point cloud data, the morphological features of each segment of steel bar in the steel bar project to be inspected are extracted, and a stress-geometry coupled digital twin model is established according to the stress distribution map and morphological features. The steel reinforcement deformation parameter determination module is used to dynamically map the digital twin model with the pre-calibrated BIM design model to determine the steel reinforcement deformation parameters in the stress anomaly area. The quality inspection report generation module is used to generate a quality inspection report for the steel reinforcement project to be inspected based on the steel reinforcement deformation parameters.

[0014] A third aspect of this application provides an electronic device including a memory, a processor, and a program stored in the memory and executable on the processor, the program being loaded and executed by the processor to implement a method for inspecting the quality of reinforced concrete engineering based on a large model.

[0015] In a fourth aspect, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to implement a method for inspecting the quality of reinforced concrete engineering based on a large model.

[0016] In summary, one or more technical solutions provided in this application have at least the following technical effects or advantages: By adopting the above technical solution, vibration response data and 3D point cloud data of the steel reinforcement project to be inspected are obtained as multi-dimensional inputs. A pre-trained vibration feature analysis model is used to extract stress distribution maps and stress transmission paths. Based on the stress transmission direction, the 3D point cloud data is segmented, enabling segmented analysis of the steel reinforcement structure. Furthermore, by extracting the morphological characteristics of each steel segment and combining them with the stress distribution map to establish a stress-geometry coupled digital twin model, and then dynamically mapping and comparing it with the BIM design model, the deformation parameters of the steel reinforcement in areas of stress anomalies are accurately identified. Finally, a comprehensive quality inspection report is generated. This not only achieves a multi-dimensional evaluation of the structural performance of the steel reinforcement project, but also establishes a dynamic relationship between the physical steel reinforcement project and the virtual model through digital twin technology. This overcomes the problem that single detection methods in existing technologies cannot achieve comprehensive evaluation, achieving more comprehensive inspection and thus improving the accuracy of steel reinforcement quality inspection. Attached Figure Description

[0017] Figure 1 This is a schematic flowchart of a method for inspecting the quality of reinforced concrete engineering based on a large model, provided in an embodiment of this application. Figure 2 This is a structural schematic diagram of a steel reinforcement engineering quality inspection system based on a large model, provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0018] Explanation of reference numerals in the attached drawings: 300, electronic device; 301, processor; 302, communication bus; 303, user interface; 304, network interface; 305, memory. Detailed Implementation

[0019] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0020] In the description of the embodiments of this application, the words "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design that is described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design options. Rather, the use of the words "for example" or "for instance" is intended to present the relevant concepts in a specific manner.

[0021] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple systems means two or more systems, and multiple screen terminals means two or more screen terminals. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0022] This application provides a method for inspecting the quality of reinforced concrete engineering based on a large model. In one embodiment, please refer to... Figure 1 , Figure 1 This is a flowchart illustrating the large-model-based steel reinforcement engineering quality inspection method provided in this application embodiment. This method can be implemented using a computer program, which can be integrated into an application or run as a standalone utility application. The method can also be implemented using a microcontroller and can run on a large-model-based steel reinforcement engineering quality inspection system based on the von Neumann architecture. Specifically, the method may include the following steps: Step 101: Obtain vibration response data and three-dimensional point cloud data of the steel reinforcement project to be tested.

[0023] Among them, the steel reinforcement projects to be inspected refer to the steel reinforcement structural systems in building projects that require quality inspection, including but not limited to the steel reinforcement configuration in beams, columns, slabs and other components of buildings, including steel reinforcement components with different functions such as main bars, stirrups, and distribution bars, as well as the connection nodes and anchorages between them. These steel reinforcement structures may be in the acceptance stage after construction is completed, or they may be existing structures that require quality assessment during use, and their quality status is directly related to the overall structural safety of the building.

[0024] Vibration response data refers to structural vibration signals collected by an array of vibration sensors deployed at key nodes of a reinforced steel structure under external environmental or artificial excitation. This data includes acceleration time history data, displacement response data, and strain response data. These data reflect the dynamic characteristics of the reinforced steel structure, containing dynamic parameters such as the structure's natural frequencies, damping ratio, and mode shapes. Analysis of this data allows for the assessment of the overall stiffness distribution and local stress state of the structure, providing fundamental data support for subsequent stress distribution map generation and stress transfer path analysis.

[0025] Three-dimensional point cloud data refers to a dataset of spatial geometric information obtained by using 3D laser scanning equipment to scan a steel reinforcement project from all angles. It represents the geometric features of the structure in the form of a massive number of discrete points, with each data point containing spatial three-dimensional coordinates, reflection intensity values, and normal vector information. This data accurately records the spatial location, geometric dimensions, and surface features of the steel reinforcement components. Through point cloud data processing, it is possible to accurately measure geometric parameters such as the actual arrangement position, spacing, and protective layer thickness of the steel reinforcement, as well as to detect and evaluate quality issues such as steel reinforcement deformation and displacement.

[0026] Specifically, to comprehensively assess the quality of the reinforced concrete structure, it is necessary to simultaneously acquire vibration response data reflecting the structural dynamic characteristics and three-dimensional point cloud data characterizing the geometric morphology. In practical implementation, a multi-point vibration sensor array is first deployed at key nodes of the reinforced concrete structure. The sensor placement is optimized based on the stress characteristics of the reinforced concrete structure and the importance of the nodes. High-precision vibration sensors are used to collect vibration response signals of the structure under environmental and artificial excitation, with a sampling frequency set to 1000Hz to ensure the capture of high-frequency vibration characteristics of the reinforced concrete structure. The vibration response data acquisition duration is no less than 10 minutes to ensure data stability and representativeness. Simultaneously, a three-dimensional laser scanner is used to perform a 3D scan of the reinforced concrete structure from all angles. The scanner's spatial resolution is set to 0.5mm. During the scanning process, a multi-site joint measurement method is used to eliminate occlusion effects and ensure the acquisition of complete three-dimensional point cloud data. During point cloud data acquisition, a unified spatial coordinate system is established by setting target points to achieve accurate registration of point cloud data from different measurement stations. The acquired vibration response data includes information such as the structure's natural frequencies, modal characteristics, and dynamic response, reflecting the overall stiffness and local stress distribution of the reinforced concrete structure. Meanwhile, the three-dimensional point cloud data records in detail the spatial location, geometric dimensions, and deformation state of the reinforced concrete members. This collaborative acquisition method using multi-source data enables both dynamic evaluation of structural performance and geometric analysis of member quality, providing a complete data foundation for subsequent comprehensive analysis and evaluation. Practice has shown that this data acquisition scheme effectively obtains comprehensive information on reinforced concrete engineering, with stable and reliable data quality. The spatial accuracy and temporal resolution both meet the requirements of subsequent analysis and processing, laying a solid data foundation for accurate assessment of the quality of reinforced concrete engineering projects.

[0027] Step 102: Input the vibration response data into the pre-trained vibration feature analysis model to obtain the stress distribution map and stress transmission path. Based on the stress transmission path, the three-dimensional point cloud data is segmented along the stress transmission direction to obtain segmented point cloud data.

[0028] The pre-trained vibration feature analysis model refers to a deep learning model optimized with a large number of training samples. These training samples include known vibration response data, ground truth stress distribution data, and stress propagation path annotation data. The model employs a multi-layer neural network structure, including a temporal feature extraction layer, a spatial correlation analysis layer, and a stress mapping layer. It can transform the input vibration response data into stress distribution and stress propagation characteristics of the reinforcing steel nodes. Through loss function optimization and parameter iteration, the model achieves an accurate mapping from vibration features to stress features.

[0029] A stress distribution map is a visual representation of the stress state of steel reinforcement nodes calculated by a pre-trained vibration characteristic analysis model. It displays the stress distribution at each node in a steel reinforcement structure in the form of a heat map. The map includes stress values, stress levels, and spatial distribution information. Different color intensities visually reflect the relative magnitude of the stress at the nodes, and it can be used to identify stress concentration areas and potential structural weaknesses.

[0030] The stress transfer path refers to the main channels and directions through which stress is transferred between nodes in a reinforced steel structure, characterizing the flow characteristics of internal force distribution after the structure is subjected to stress. It consists of a series of directional path segments, each containing a transfer direction vector and node coordinate information, reflecting the stress transfer intensity and spatial relationship between adjacent nodes, and is an important basis for guiding the segmentation of point cloud data.

[0031] Segmented point cloud data refers to the dataset obtained by segmenting the original 3D point cloud data according to the stress transfer path. Each segment corresponds to an independent region along the stress transfer direction. This data preserves the geometric features of the original point cloud while reflecting the stress transfer characteristics of the structure, facilitating subsequent extraction and analysis of the rebar morphology. Each segment contains complete spatial coordinate information and point cloud density distribution characteristics.

[0032] Specifically, to accurately assess the stress state of the reinforced concrete structure and achieve reasonable segmentation of the point cloud data, this step employs a pre-trained vibration feature analysis model to conduct in-depth analysis of the vibration response data. In practice, the collected vibration response data is first pre-processed, including signal denoising, outlier removal, and data standardization to improve data quality. The pre-processed vibration response data is then input into the pre-trained vibration feature analysis model according to its time-series characteristics. This model employs a deep learning architecture, including a time-series feature extraction layer, a spatial correlation analysis layer, and a stress mapping layer. Through time-frequency feature analysis of the vibration response data, combined with the topological relationship of the reinforced concrete structure, the model calculates a stress distribution map of the reinforced concrete nodes. This map visually displays the stress magnitude and distribution at each node in the form of a heatmap. Simultaneously, based on the stress transfer relationship between nodes, the model generates stress transfer paths characterizing the direction and intensity of stress transfer. After obtaining the stress transfer paths, they are used as a guide for segmenting the point cloud data. Specifically, by calculating the direction vector of the stress transfer path and the coordinates of key nodes, a local coordinate system consistent with the stress transfer direction is established. Based on this, the 3D point cloud data is mapped onto this local coordinate system, and density clustering analysis is performed along the stress transfer direction to determine the segment boundary positions of the point cloud data. For the point cloud data between the determined segment boundaries, a region growing algorithm is used for segmentation processing, thereby obtaining segmented point cloud data adapted to the stress transfer characteristics. Practice shows that this point cloud segmentation method based on stress transfer characteristics can effectively identify the weak stress areas of reinforced concrete structures. The obtained segmented point cloud data not only maintains the integrity of the original geometric features but also reflects the stress transfer law of the structure, providing a reasonable data foundation for subsequent morphological feature extraction and digital twin modeling. In addition, by combining the stress distribution map and the segmented point cloud analysis, the stress-deformation coupling effect of the reinforced concrete structure can be intuitively reflected, improving the accuracy and reliability of quality inspection.

[0033] Based on the above embodiments, as an optional embodiment, in step 102: the three-dimensional point cloud data is segmented along the stress transmission direction based on the stress transmission path to obtain segmented point cloud data. This step may further include the following steps: Step 201: Obtain the direction vector and node coordinates of the stress transfer path; establish a local coordinate system based on the node coordinates, and map the 3D point cloud data to the local coordinate system.

[0034] Specifically, to establish a point cloud analysis coordinate system adapted to the stress transfer characteristics, it is necessary to first extract direction vectors and node coordinate information from the stress transfer path. In practice, the direction vector Vi = (vix, viy, viz) reflecting the spatial stress transfer direction is calculated by the positional difference between adjacent nodes in the stress transfer path. Simultaneously, the three-dimensional coordinates Pi = (xi, yi, zi) of each node in the global coordinate system are recorded. Based on the acquired node coordinates, a principal direction analysis algorithm is used to establish a local coordinate system: the principal direction vector of stress transfer is normalized and used as the z-axis of the local coordinate system. The local x-axis direction is obtained by performing a cross product with the y-axis of the global coordinate system, and then the y-axis direction is obtained by performing a cross product between the z-axis and x-axis, thus constructing a complete local coordinate system transformation matrix T. Finally, the original three-dimensional point cloud data set {P} is mapped from the global coordinate system to the newly established local coordinate system using the coordinate transformation matrix T, achieving spatial alignment of the point cloud data along the stress transfer direction. This coordinate transformation method based on stress transfer characteristics provides a reasonable spatial reference basis for subsequent point cloud segmentation processing.

[0035] Step 202: Perform density clustering analysis on the mapped point cloud data along the direction vector to determine the segment boundaries of the point cloud data.

[0036] Specifically, to achieve reasonable segmentation of point cloud data, an improved DBSCAN density clustering algorithm is used to analyze the mapped point cloud data along the stress transfer direction. First, clustering parameters are determined based on the density distribution characteristics of the point cloud data: the search radius ε is set to 2.5 times the average point spacing, and the minimum number of points, MinPts, is determined to be 50 points based on the minimum feature size of the reinforcing steel member. During algorithm execution, an unvisited point p is selected, and the number of points n in its ε-neighborhood is calculated. When n > MinPts, p is marked as a core point and assigned a cluster label, while its neighboring points are added to the processing queue. The neighborhood search and label assignment process is repeated for each point in the queue until the queue is empty, completing the generation of one cluster. Unvisited points are then processed, ultimately resulting in multiple point cloud clusters. By analyzing the point density changes and spatial distribution characteristics between adjacent clusters, the segmentation boundary coordinate set {Bi} of the point cloud data is determined. This density-based segmentation method fully considers the spatial distribution characteristics of the point cloud data and can accurately identify the natural segmentation positions of the reinforcing steel member.

[0037] Step 203: Perform region growing processing on the point cloud data between the segment boundaries to obtain segmented point cloud data.

[0038] Specifically, to obtain accurate segmented point cloud data, region growing processing is performed on the point cloud data between the segment boundaries determined in step 202. First, principal component analysis is used to calculate the normal vector ni and curvature eigenvalue ci of each point pi in the point cloud. The normal vector is obtained by analyzing the eigenvectors of the covariance matrix of the point's K nearest neighbors (K=30), and the curvature eigenvalue is calculated by fitting a local quadratic surface. Within each segment boundary, the top 5% of points with the smallest curvature eigenvalues ​​are selected as the initial seed point set S. For each seed point s, its search radius Rs=βd (where d is the local point cloud density, β is the density coefficient, and the value is 1.5), and the normal vector angle threshold θs=α / cs (where α is the scaling factor, and the value is 0.8). Region growing iteration is performed: the point set {p} within the Rs neighborhood of the seed point s is searched, and the angle θ between each point and the average normal vector of the current region is calculated. If θ<θs, the point is added to the current region, and the average normal vector of the region is updated. Repeat the process until no new points are added, thus completing the region growing for one segment and obtaining complete segmented point cloud data. This adaptive region growing method ensures both the geometric integrity of the segments and achieves accurate segmentation based on local features.

[0039] Based on the above embodiments, as an optional embodiment, step 203, which involves performing region growing processing on the point cloud data between segment boundaries to obtain segmented point cloud data, may further include the following steps: Step 213: Calculate the normal vector and curvature feature value of each point in the point cloud data; select the point with the smallest curvature feature value within each segment boundary as the seed point.

[0040] Specifically, to accurately identify local geometric features in point cloud data, it is necessary to calculate the normal vector and curvature eigenvalue of each point. In practice, for each point pi in the point cloud data, its K nearest neighbor set (K=30) is first determined, and the covariance matrix M of the local neighborhood is constructed. By performing eigenvalue decomposition on the covariance matrix M, three eigenvalues ​​λ1≥λ2≥λ3 and corresponding eigenvectors v1, v2, and v3 are obtained. The eigenvector v3 corresponding to the smallest eigenvalue λ3 is the normal vector ni of point pi. The curvature eigenvalue ci of point pi is calculated using the formula ci=λ3 / (λ1+λ2+λ3). After obtaining the normal vector and curvature eigenvalue of each point, the points within each segment boundary are sorted according to their curvature eigenvalues, and the top 5% of points with the smallest curvature eigenvalues ​​are selected as the seed point set S. These points are usually located in smooth areas on the surface of the reinforcing steel, making them suitable as starting points for region growth.

[0041] Step 223: Based on the normal vector of the seed point, calculate the search radius and the normal vector angle threshold, where the search radius is directly proportional to the point cloud density and the normal vector angle threshold is inversely proportional to the curvature feature value of the seed point.

[0042] Specifically, to ensure the rationality and controllability of the region growth process, search parameters need to be dynamically calculated based on the characteristics of the seed points. For each seed point s in the seed point set S, the point cloud density ρs is first calculated in its local neighborhood, defined as the number of points per unit volume. The search radius Rs is set using the formula Rs=α·ρs, where α is a scaling factor with a value range of [1.2, 1.8], and a larger point cloud density corresponds to a larger search radius. The normal vector angle threshold θs is determined using the formula θs=β / cs, where β is an adjustment coefficient with a value range of [0.6, 1.0], and cs is the curvature characteristic value of the seed point. This parameter setting method ensures that a larger angle threshold is used in flat regions (low curvature), while a smaller angle threshold is used in regions with drastic curvature changes, thus adapting to changes in local geometric features.

[0043] Step 233: Traverse the neighboring points of the seed point, and add the neighboring points within the search radius that have a normal vector angle less than the normal vector angle threshold to the current region, and update the average normal vector of the current region.

[0044] Specifically, to achieve gradual expansion of the region, it is necessary to traverse the neighborhood of the seed point based on the set search parameters. For the seed point s, the KD-tree structure is used to quickly find its neighborhood point set {p} within the search radius Rs. For each neighborhood point pi, the angle θi = arccos(ni·ns) between its normal vector ni and the normal vector ns of the current seed point is calculated. If θi is less than the calculated normal vector angle threshold θs, point pi is added to the current region and simultaneously added to the candidate seed point queue Q. After each new point is added, the average normal vector navg of the current region R is updated using the formula navg = (n·|R|+ni) / (|R|+1), where |R| is the number of points contained in the current region. This dynamic update mechanism ensures a smooth transition of the normal vector during the region growth process.

[0045] Step 243: Using the updated average normal vector as the target search direction, repeat the traversal process until no new neighborhood points are added to the current region, and obtain segmented point cloud data.

[0046] Specifically, to ensure the continuity and integrity of the region growth, iterative expansion based on the updated average normal vector is required. The updated average normal vector `navg` is used as the new target search direction, and points are continuously retrieved from the candidate seed point queue `Q` for the aforementioned traversal process. For each point in queue `Q`, the neighborhood search, angle determination, and region update operations are repeatedly performed. When no new points are added to the current region in a given iteration, the region growth is considered complete. The resulting segmented point cloud data exhibits good geometric continuity and boundary integrity, with points within each segment having a similar normal vector distribution, effectively preserving the local geometric features of the reinforcing steel structure.

[0047] Based on the above embodiments, as an optional embodiment, the training process of the pre-trained vibration feature analysis model specifically includes: Step 204: Construct a training sample library containing various types of reinforced concrete structures. Each training sample in the training sample library includes vibration response data, stress distribution true value data, and stress transfer path annotation data.

[0048] Specifically, to ensure the vibration characteristic analysis model has good generalization ability and adaptability, it is necessary to construct a training sample library containing various typical reinforced concrete structure types. In practice, data samples are collected under different structural types (including frames, shear walls, beam-slab systems, etc.) and different load conditions (including static, dynamic, fatigue, etc.) through a combination of finite element simulation and actual engineering testing. Each training sample contains three types of data: multi-point vibration response time history data with a sampling frequency of 1000Hz, true value data of nodal stress distribution calibrated by strain gauges and displacement sensors, and stress transfer path annotation data confirmed by experts. The training sample library contains a total of 10,000 valid samples, covering more than 95% of the common reinforced concrete structure load scenarios, providing a sufficient data foundation for model training.

[0049] Step 205: Perform data augmentation on the vibration response data in the training sample library to obtain an expanded training dataset; input the expanded training dataset into the preset base model to obtain the predicted stress distribution spectrum and stress transmission path.

[0050] Specifically, to improve the robustness and anti-interference ability of the model, data augmentation was performed on the vibration response data in the training sample library. Five augmentation methods were employed: 1) adding Gaussian white noise (signal-to-noise ratio in the range of 20-40 dB) to simulate sensor noise; 2) performing random time-scale stretching (stretch factor between 0.8 and 1.2) to simulate sampling frequency fluctuations; 3) implementing random amplitude scaling (scaling factor between 0.9 and 1.1) to simulate sensor sensitivity changes; 4) performing random time segment truncation (duration between 30% and 70% of the original data) to increase data diversity; and 5) generating combined working condition data through signal superposition. After data augmentation, the number of training samples was expanded to 50,000 sets. The expanded training dataset was input into a pre-defined deep neural network base model. This model adopts an encoder-decoder architecture, including a temporal feature extraction module, a spatial feature association module, and a stress mapping module, outputting predicted stress distribution maps and stress transmission paths.

[0051] Step 206: Calculate the loss function between the predicted stress distribution spectrum and stress transmission path and the corresponding true data; iteratively optimize the model parameters of the preset base model based on the loss function until the preset base model converges, and use the converged preset base model as the pre-trained vibration feature analysis model.

[0052] Specifically, to achieve accurate model training, a composite loss function was designed to evaluate the accuracy of the model's prediction results. The loss function consists of three parts: 1) stress distribution loss Ls, which uses mean squared error to measure the difference between the predicted stress value and the true value; 2) transmission path loss Lp, which uses an improved Hausdorff distance to evaluate the geometric similarity between the predicted path and the labeled path; and 3) topological consistency loss Lt, which calculates the structural consistency between the predicted path and the labeled path using a graph matching algorithm. The total loss function is L = α·Ls + β·Lp + γ·Lt, where the weighting coefficients α = 0.4, β = 0.4, and γ = 0.2. Based on the calculated loss function values, the Adam optimizer is used to iteratively update the parameters of the base model. The initial learning rate is set to 0.001, and a cosine annealing strategy is used for dynamic adjustment. When the change in the validation set loss function value is less than 1% over 10 consecutive epochs, the model is considered to have converged, and the converged model is saved as a pre-trained vibration feature analysis model. After testing, the model achieved an average accuracy of 92% and 88% in stress prediction and path recognition tasks, respectively, demonstrating its practicality.

[0053] Step 103: Based on segmented point cloud data, extract the morphological features of each segment of steel bar in the steel bar project to be inspected, and establish a stress-geometry coupled digital twin model according to the stress distribution map and morphological features.

[0054] Morphological characteristics refer to a set of quantitative parameters used to characterize the geometric deformation state of reinforced steel members, specifically including two core indicators: curvature and torsion angle. Curvature κ is calculated by analyzing the curvature change of the steel reinforcement centerline, characterizing the degree to which the reinforced steel member deviates from a straight line; the torsion angle ψ is determined by the angular change between adjacent cross sections, reflecting the amount of rotational deformation of the reinforced steel member around its axial direction. These two parameters together constitute a morphological characteristic vector, which can comprehensively describe the deformation state of the steel reinforcement in space and is an important basis for evaluating the construction quality of reinforced steel. By analyzing these characteristic parameters, the location and degree of abnormal deformation of the steel reinforcement can be accurately identified.

[0055] A stress-geometry coupled digital twin model refers to a digital representation model that integrates the geometric information, mechanical properties, and deformation behavior of reinforced concrete engineering. This model is constructed based on the geometric topology graph G=(V,E) and stress transfer weights W, and is discretized through finite element mesh generation. The model includes dynamically updated material parameters and deformation parameters based on morphological features, enabling a two-way mapping between stress state and geometric deformation. This coupled model not only reflects the static geometric characteristics of reinforced concrete engineering but also simulates the deformation response during dynamic stress processes, providing a digital platform for the quality assessment and performance prediction of reinforced concrete engineering.

[0056] Specifically, to achieve accurate digital characterization and mechanical performance evaluation of reinforced concrete engineering, it is necessary to extract the morphological features of the reinforced concrete based on segmented point cloud data and establish a stress-geometry coupled digital twin model. In the implementation process, firstly, the principal direction of each segment of point cloud data is analyzed, and the axial direction vector v of the reinforced concrete is determined using the PCA algorithm. Based on this axial vector, a vertical sampling plane is established, and a sampling section is set every 50 mm along the axial direction of the reinforced concrete to obtain the cross-sectional point cloud set {Pi}. For each cross-sectional point cloud, the RANSAC ellipse fitting algorithm is used to process the data, obtaining the geometric parameters of the section, including the major axis a, minor axis b, and inclination angle θ. By analyzing the variation relationship of parameters between adjacent sections, the local curvature κ=|d²r / ds²| (where r is the centerline position vector and s is the arc length parameter) and torsion angle ψ=dθ / ds of the reinforced concrete are calculated. The obtained curvature and torsion angle are used as the morphological features of the reinforced concrete, and a morphological feature vector F=(κ, ψ) is established.

[0057] Subsequently, a geometric topology graph G=(V, E) of the rebar nodes is constructed, where the vertex set V represents the rebar nodes and the edge set E represents the connection relationships between nodes. Based on the pre-obtained stress distribution map, the stress transfer weight w between nodes is calculated, and the weight value is determined by the functional relationship between stress gradient and transfer distance. The stress transfer weight is superimposed on the geometric topology graph to form a stress-geometry coupling graph G'=(V, E, W). Based on this coupling graph and morphological features, a tetrahedral meshing algorithm is used to construct a finite element mesh model M for the rebar engineering. In the mesh model, the material parameters of the elements are dynamically adjusted according to the local stress state, while the deformation parameters are corrected based on the morphological features. Through a stress field interpolation algorithm, discrete stress data is mapped to a continuous stress field distribution, ultimately generating a digital twin model integrating geometric, mechanical, and deformation information.

[0058] This digital twin modeling method, based on multi-source data fusion, not only achieves high-precision reconstruction of the geometric morphology of reinforced concrete structures but also establishes a coupling relationship between stress distribution and geometric deformation, accurately reflecting the stress state and deformation characteristics of reinforced concrete structures. Through the digital twin model, the performance status of reinforced concrete structures can be monitored and predicted in real time, providing reliable data support for quality assessment.

[0059] Based on the above embodiments, as an optional embodiment, step 103: extracting the morphological features of each segment of steel bar in the steel bar project to be inspected based on segmented point cloud data, this step may further include the following steps: Step 301: Perform main direction analysis on the segmented point cloud data to determine the axial direction of each rebar segment; for each rebar, establish a sampling plane based on the axial direction, and obtain the cross-sectional point cloud of the rebar at a preset interval on the sampling plane.

[0060] Specifically, to accurately extract the geometric features of the reinforcing bars, it is first necessary to determine the spatial orientation of the reinforcing bars and obtain their cross-sectional information. In practice, a covariance matrix C is constructed for each segment of point cloud data: C = (1 / n)∑(pi-p̄)(pi-p̄)T, where pi is the coordinate of a point in the point cloud, p̄ is the coordinate of the center point of the point cloud, and n is the number of points. By performing eigenvalue decomposition on the covariance matrix C, three eigenvalues ​​λ1≥λ2≥λ3 and their corresponding eigenvectors v1, v2, and v3 are obtained. The eigenvector v1 corresponding to the largest eigenvalue λ1 is the axial direction vector of the reinforcing bar. Based on this axial vector, a vertical sampling plane π: n·x+d=0 is established, where n is the plane normal vector (parallel to v1), and d is the distance from the plane to the origin. A sampling plane is set every 50mm along the axial direction of the reinforcing bar, and point cloud data within a range of ±2.5mm is extracted from each sampling plane to form a cross-sectional point cloud set {Si}. This cross-sectional sampling method based on the main direction can accurately obtain the cross-sectional morphological characteristics of the reinforcing bars, providing a reliable data foundation for subsequent shape analysis.

[0061] Step 302: Perform ellipse fitting on the cross-sectional point cloud to obtain the cross-sectional parameters of the reinforcing bars.

[0062] Specifically, to obtain the geometric parameters of the rebar cross-section, ellipse fitting analysis is performed on the collected cross-sectional point cloud. An improved RANSAC ellipse fitting algorithm is employed, which effectively handles cross-sectional data containing noise and outliers. In practice, five points are randomly selected from the cross-sectional point cloud as initial samples, and an ellipse equation is constructed based on the quadratic curve equation ax² + bxy + cy² + dx + ey + f = 0. The equation parameters are solved using the least squares method to obtain the initial ellipse model. The distance from all points to this ellipse is calculated, and points with a distance less than a threshold ε (0.5 mm) are marked as inliers. This process is repeated 1000 times, and the ellipse model with the most inliers is selected as the optimal fitting result. Cross-sectional parameters, including the major axis a, minor axis b, and the angle θ between the principal axis and the reference direction, are extracted from the optimal ellipse model. This RANSAC-based ellipse fitting method has strong noise resistance and can accurately obtain the cross-sectional geometric features of the rebar.

[0063] Step 303: Based on the cross-sectional parameters of the reinforcing bar, determine the bending degree and torsion angle of the reinforcing bar; use the bending degree and torsion angle as the morphological characteristics of the reinforcing bar.

[0064] Specifically, to characterize the spatial deformation features of reinforcing bars, it is necessary to calculate the curvature and torsion angle of the bars based on cross-sectional parameters. For the curvature calculation, the spatial centerline of the reinforcing bar is first constructed using the coordinates of the center points of adjacent cross-sections. A continuous centerline curve r(s) is obtained using cubic spline interpolation, where s is the arc length parameter along the centerline. The curvature κ is calculated using the formula κ=|d²r / ds²|, reflecting the degree to which the reinforcing bar deviates from a straight line. For the torsion angle calculation, the change in the direction angle θ of the principal axis of the ellipse between adjacent cross-sections is analyzed, and the torsion angle is calculated using the formula ψ=dθ / ds, representing the amount of rotational deformation of the reinforcing bar around the axial direction. The calculated curvature κ and torsion angle ψ are combined to form a morphological feature vector F=(κ, ψ), which is used to characterize the spatial deformation state of the reinforcing bar. This feature extraction method based on differential geometry can quantitatively describe the deformation characteristics of the reinforcing bar, providing reliable data for subsequent quality assessment.

[0065] Based on the above embodiments, as an optional embodiment, step 103: establishing a stress-geometry coupled digital twin model based on the stress distribution spectrum and morphological characteristics may further include the following steps: Step 304: Construct a geometric topology diagram of the rebar nodes in the rebar project to be inspected; calculate the stress transfer weight between each rebar node based on the stress distribution map.

[0066] Specifically, to characterize the spatial connection relationships of various components in reinforced concrete engineering, a geometric relationship graph reflecting the topological characteristics of the nodes needs to be constructed. In practice, the reinforced concrete nodes are first represented as a vertex set V={vi} of the graph, where each vertex contains the node's spatial coordinates (x, y, z) and a local feature descriptor. Based on the connection relationships in the point cloud data, an edge set E={eij} is established to represent the physical connections between nodes, where eij represents the connection edge between nodes vi and vj. Subsequently, the stress transfer weight wij between nodes is calculated using a stress distribution map. The weight calculation formula is wij=σij·exp(-dij / d0), where σij is the average stress value between nodes, dij is the spatial distance between nodes, and d0 is a feature distance parameter (taken as 3 times the diameter of the reinforced concrete). This stress and distance-based weight calculation method can reasonably reflect the mechanical transfer characteristics between nodes.

[0067] Step 305: Superimpose the stress transfer weights onto the geometric topology diagram to obtain the stress-geometry coupling diagram.

[0068] Specifically, to achieve the organic integration of geometric features and mechanical properties, stress transfer weights need to be integrated into the geometric topology graph. In practice, a weighted adjacency matrix A is constructed, where the matrix element aij = wij represents the stress transfer intensity between nodes vi and vj. For non-directly connected node pairs, their indirect transfer weights are calculated using a shortest path algorithm. Based on the weighted adjacency matrix, a stress-geometry coupling graph G' = (V, E, W) is constructed, where W is the set of weight matrices. To improve computational efficiency, a sparse matrix storage method is adopted, retaining only connections with weights greater than a threshold wt (taken as 5% of the maximum weight). This coupling graph structure can simultaneously reflect the geometric connection relationships and stress transfer characteristics of reinforced concrete engineering.

[0069] Step 306: Based on the stress-geometry coupling diagram and morphological features, construct a finite element mesh model of the steel reinforcement project to be inspected; perform stress field interpolation on the finite element mesh model to generate a stress-geometry coupled digital twin model.

[0070] Specifically, to establish a digital model that reflects the complete mechanical behavior of reinforced concrete, a finite element model needs to be constructed based on coupling diagrams and morphological features, followed by stress field interpolation. First, an adaptive tetrahedral meshing algorithm is used to mesh the reinforced concrete members. The mesh size is dynamically adjusted according to the local stress gradient, with denser meshes used in areas with large stress gradients. For each mesh element, its material parameter E (elastic modulus) is corrected based on the local stress state σ: E = E0·f(σ), where E0 is the standard elastic modulus and f(σ) is the stress-related correction function. The deformation parameter is adjusted based on the morphological feature vector F = (κ, ψ) to ensure the model accurately reflects the actual deformation state of the reinforced concrete. Subsequently, the radial basis function (RBF) interpolation method is used to make the discrete stress data continuous. The interpolation function is φ(r) = exp(-r² / r0²), where r is the spatial distance and r0 is the influence radius parameter. Through interpolation, a continuous stress field distribution is obtained, ultimately generating a digital twin model containing complete geometric information, stress distribution, and deformation characteristics. This modeling method based on multi-source data fusion enables a comprehensive digital representation of the physical properties of reinforced concrete engineering, providing a reliable analysis platform for subsequent condition monitoring and performance evaluation.

[0071] Step 104: Dynamically map the digital twin model to the pre-calibrated BIM design model to determine the steel reinforcement deformation parameters in the stress anomaly area.

[0072] In this application, the reinforcement deformation parameters refer to a set of parameters used to quantitatively describe the degree to which the actual state of a reinforced concrete member deviates from its design state. Specifically, they include two core indicators: displacement deviation, which represents the spatial offset of the actual position of the reinforcement relative to its design position, calculated using the Euclidean distance between the measured coordinates and the design coordinates. This parameter reflects the spatial positional error of the reinforcement and is used to assess its installation accuracy. The angular deviation θ represents the angle between the actual direction of the reinforcement and its design direction, calculated using the angle between the measured axial vector and the design axial vector. This parameter reflects the directional error of the reinforcement and is used to assess its orientation accuracy.

[0073] Specifically, to achieve accurate comparative analysis between the actual state of the reinforcement engineering and the design requirements, it is necessary to dynamically map the constructed digital twin model with the standard BIM design model to identify stress anomaly areas and quantitatively assess the deformation state of the reinforcement. In practice, the Iterative Closest Point (ICP) registration algorithm is first used to spatially align the digital twin model and the BIM design model. During the registration process, reinforcement nodes are selected as feature points, and the rigid body transformation matrix T between the models is established by minimizing the Euclidean distance between feature point pairs. The registration error is evaluated using the root mean square error (RMSE), and initial registration is completed when the RMSE is less than a preset threshold ε (2 mm).

[0074] Subsequently, a dynamic mapping relationship between models is established based on the registration results. For each grid node pi in the digital twin model, its corresponding point qi is searched in the BIM design model, and the stress difference Δσi=|σi-σi'| is calculated, where σi is the actual stress value and σi' is the design stress value. By setting a stress difference threshold Δσt (taken as 20% of the design stress), stress anomaly regions R={pi|Δσi>Δσt} are identified. For the identified stress anomaly regions, the actual morphological features F=(κ, ψ) and design morphological features F'=(κ', ψ') of the reinforcing bars within the region are extracted, and the morphological feature deviation ΔF=(Δκ, Δψ) is calculated. Based on the morphological feature deviation, the set of deformation parameters D={δ, θ} of the reinforcing bars is calculated using a pre-established deformation parameter mapping function, where δ represents the displacement deviation and θ represents the angle deviation.

[0075] This dynamic mapping-based analysis method can not only accurately locate stress anomaly areas in steel reinforcement projects, but also quantitatively characterize the deformation state of steel reinforcement within the anomaly areas, providing reliable data support for quality assessment.

[0076] Based on the above embodiments, as an optional embodiment, step 104: dynamically mapping the digital twin model with the pre-calibrated BIM design model to determine the steel reinforcement deformation parameters in the stress anomaly area, may further include the following steps: Step 401: Construct the feature mapping matrix between the digital twin model and the BIM design model; based on the feature mapping matrix, determine the deviation between the measured coordinates and the design coordinates of the rebar nodes.

[0077] Specifically, to achieve a precise correspondence between the digital twin model and the BIM design model, a feature mapping matrix needs to be constructed to quantify the spatial transformation relationship between the two models. In practice, firstly, N feature-corresponding point pairs {(pi, qi)} are selected in each of the two models, where pi represents the node coordinates in the digital twin model and qi represents the corresponding node coordinates in the BIM design model. Based on these corresponding point pairs, a 4×4 homogeneous transformation matrix M is constructed, with matrix elements including a rotation matrix R and a translation vector t. The optimal transformation matrix is ​​solved by minimizing the objective function F = ∑||M·pi-qi||² using the SVD decomposition method. After obtaining the transformation matrix M, all node coordinates in the digital twin model are transformed, and the Euclidean distance d = ||M·pq|| between the transformed coordinates and the design coordinates is calculated as the deviation value.

[0078] Step 402: Filter out stress anomaly areas where the deviation value is greater than the deviation threshold; extract the elastic deformation and plastic deformation of the steel bars in the stress anomaly areas as steel bar deformation parameters.

[0079] Specifically, to accurately identify stress anomaly regions and quantitatively assess the deformation state of reinforcing bars, it is necessary to perform region screening and deformation parameter extraction based on deviation values. First, a deviation threshold dt = 5 mm is set, and the set of nodes with deviation values ​​greater than dt is selected as the stress anomaly region R. For the identified anomaly regions, the elastic deformation εe and plastic deformation εp are calculated separately. The elastic deformation is calculated using Hooke's Law: εe = σ / E, where σ is the current stress value and E is the material's elastic modulus. The plastic deformation is determined by the nonlinear part of the stress-strain curve: εp = εt - εe, where εt is the total strain. The calculated elastic and plastic deformations are combined to form the reinforcing bar deformation parameter set D = (εe, εp), which is used to characterize the actual deformation state of the reinforcing bars. This deformation parameter extraction method based on elastoplastic theory can comprehensively reflect the deformation characteristics of reinforcing bars in stress anomaly regions, with a deformation calculation accuracy better than 0.1%, providing a reliable quantitative basis for subsequent quality assessment.

[0080] Step 105: Generate a quality inspection report for the steel reinforcement project based on the steel reinforcement deformation parameters.

[0081] Specifically, to comprehensively assess the construction quality of the reinforcing steel structure under inspection and generate standardized inspection documents, a quality inspection report based on the steel deformation parameters needs to be generated. In practice, a quality assessment index system for reinforcing steel structures is first constructed, comprising three levels of assessment indicators: the overall deformation index Ig at the structural level, the local deformation index Il at the component level, and the stress anomaly index Is at the node level. The overall deformation index Ig is obtained by calculating the statistical characteristics of global steel deformation parameters: Ig = w1·μ(εe) + w2·μ(εp), where μ(·) represents the mean operation, and w1 = 0.6 and w2 = 0.4 are weighting coefficients. The local deformation index Il is calculated based on the deformation parameters within the stress anomaly region: Il = ∑(εe, i² + βεp, i²), where β = 1.5 is the penalty coefficient for plastic deformation. The stress anomaly index Is is determined by the spatial distribution characteristics of the anomaly region: Is = Na / Nt, where Na is the number of anomaly points and Nt is the total number of points.

[0082] Then, a graded evaluation rule was constructed based on the evaluation indicators. The quality of the reinforced concrete project was divided into four grades: Grade A (Excellent), Grade B (Qualified), Grade C (Needs Improvement), and Grade D (Unqualified). The evaluation rule adopted the fuzzy comprehensive evaluation method to establish the membership function between the evaluation indicators and the quality grades. For each evaluation indicator, its membership value to each quality grade was calculated, and the final quality rating result was obtained by weighted summation. The specific contents of the quality inspection report include: basic project information, description of the testing methods, stress anomaly area distribution map, reinforcement deformation parameter statistics table, calculation results of quality evaluation indicators, quality grade determination results, and improvement suggestions. Among them, the stress anomaly area distribution map uses a heat map to visually display the anomaly location, the deformation parameter statistics table lists the specific deformation data of key nodes, and the quality evaluation results are displayed in radar chart form to show the scores of each indicator.

[0083] This quality assessment method, based on a multi-level indicator system, can comprehensively reflect the construction quality status of reinforced concrete projects. The test report includes both quantitative assessment data and intuitive visualizations, facilitating understanding and use by engineering personnel. Practice shows that the accuracy rate of this assessment method exceeds 90%, and the assessment results show good consistency with manual inspection results, providing effective decision support for project quality management.

[0084] Reference Figure 2 This application provides a large-model-based steel reinforcement engineering quality inspection system, which includes: a data acquisition module, a digital twin model establishment module, a steel reinforcement deformation parameter determination module, and a quality inspection report generation module, wherein: The data acquisition module is used to acquire vibration response data and three-dimensional point cloud data of the steel reinforcement project to be inspected; The digital twin model building module is used to input vibration response data into a pre-trained vibration feature analysis model to obtain stress distribution maps and stress transmission paths. Based on the stress transmission paths, the three-dimensional point cloud data is segmented along the stress transmission direction to obtain segmented point cloud data. Based on the segmented point cloud data, the morphological features of each segment of steel bar in the steel bar project to be inspected are extracted, and a stress-geometry coupled digital twin model is established according to the stress distribution maps and morphological features. The reinforcement deformation parameter determination module is used to dynamically map the digital twin model and the pre-calibrated BIM design model to determine the reinforcement deformation parameters in the stress anomaly area. The quality inspection report generation module is used to generate a quality inspection report for the steel reinforcement project based on the steel reinforcement deformation parameters.

[0085] Based on the above embodiments, the digital twin model building module is also used to construct a training sample library containing various types of reinforced concrete structures. Each training sample in the training sample library includes vibration response data, stress distribution ground truth data, and stress transmission path annotation data. The vibration response data in the training sample library is subjected to data augmentation processing to obtain an expanded training dataset. The expanded training dataset is input into a preset base model to obtain the predicted stress distribution spectrum and stress transmission path. The loss function between the predicted stress distribution spectrum and stress transmission path and the corresponding ground truth data is calculated. The model parameters of the preset base model are iteratively optimized based on the loss function until the preset base model converges. The converged preset base model is used as a pre-trained vibration feature analysis model.

[0086] Based on the above embodiments, the digital twin model building module is also used to obtain the direction vector and node coordinates of the stress transfer path; establish a local coordinate system based on the node coordinates and map the three-dimensional point cloud data to the local coordinate system; perform density clustering analysis on the mapped point cloud data along the direction vector to determine the segment boundaries of the point cloud data; and perform region growing processing on the point cloud data between the segment boundaries to obtain segmented point cloud data.

[0087] Based on the above embodiments, the digital twin model building module is also used to calculate the normal vector and curvature feature value of each point in the point cloud data; select the point with the smallest curvature feature value within each segment boundary as the seed point; calculate the search radius and the normal vector angle threshold based on the normal vector of the seed point, wherein the search radius is proportional to the point cloud density, and the normal vector angle threshold is inversely proportional to the curvature feature value of the seed point; traverse the neighboring points of the seed point, and sequentially add the neighborhood points within the search radius and whose normal vector angle is less than the normal vector angle threshold to the current region, and update the average normal vector of the current region. Using the updated average normal vector as the target search direction, the traversal process is repeated until no new neighborhood points are added to the current region, thus obtaining the segmented point cloud data.

[0088] Based on the above embodiments, the digital twin model building module is also used to perform main direction analysis on the segmented point cloud data to determine the axial direction of each segment of steel bar; for each steel bar, a sampling plane is established based on the axial direction, and the cross-sectional point cloud of the steel bar is obtained on the sampling plane at a preset interval; the cross-sectional point cloud is elliptical fitted to obtain the cross-sectional parameters of the steel bar; based on the cross-sectional parameters of the steel bar, the curvature and torsion angle of the steel bar are determined; the curvature and torsion angle are used as the morphological characteristics of the steel bar.

[0089] Based on the above embodiments, the digital twin model building module is also used to construct a geometric topology diagram of the steel reinforcement nodes in the steel reinforcement project to be inspected; calculate the stress transfer weight between each steel reinforcement node based on the stress distribution map; superimpose the stress transfer weight onto the geometric topology diagram to obtain a stress-geometric coupling diagram; construct a finite element mesh model of the steel reinforcement project to be inspected based on the stress-geometric coupling diagram and morphological features; and perform stress field interpolation on the finite element mesh model to generate a stress-geometric coupling digital twin model.

[0090] Based on the above embodiments, the rebar deformation parameter determination module is also used to construct the feature mapping matrix between the digital twin model and the BIM design model; based on the feature mapping matrix, determine the deviation value between the measured coordinates and the design coordinates of the rebar nodes; screen out stress anomaly areas where the deviation value is greater than the deviation threshold; and extract the elastic deformation and plastic deformation of the rebar within the stress anomaly area as rebar deformation parameters.

[0091] It should be noted that the above embodiments of the apparatus are only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0092] This application also discloses an electronic device. (See reference...) Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of this application. The electronic device 300 may include: at least one processor 301, at least one network interface 304, a user interface 303, a memory 305, and at least one communication bus 302.

[0093] The communication bus 302 is used to enable communication between these components.

[0094] The user interface 303 may include a display interface and a camera interface. Optionally, the user interface 303 may also include a standard wired interface and a wireless interface.

[0095] The network interface 304 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0096] The processor 301 may include one or more processing cores. The processor 301 connects to various parts of the server using various interfaces and lines, and performs various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in the memory 305, and by calling data stored in the memory 305. Optionally, the processor 301 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 301 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface graphics, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 301 and may be implemented as a separate chip.

[0097] The memory 305 may include random access memory (RAM) or read-only memory. Optionally, the memory 305 may include a non-transitory computer-readable storage medium. The memory 305 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 305 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 305 may also be at least one storage device located remotely from the aforementioned processor 301. (Refer to...) Figure 3 The memory 305, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program for a method of inspecting the quality of reinforced concrete engineering based on a large model.

[0098] exist Figure 3In the illustrated electronic device 300, the user interface 303 is mainly used to provide an input interface for the user and acquire user input data; while the processor 301 can be used to call an application program stored in the memory 305 for a steel reinforcement engineering quality inspection method based on a large model. When executed by one or more processors 301, the electronic device 300 performs one or more methods as described in the above embodiments. It should be noted that, for the foregoing method embodiments, for the sake of simplicity, they are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, because according to this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0099] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0100] In the various embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some service interface; the indirect coupling or communication connection between apparatuses or units may be electrical or other forms.

[0101] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0102] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0103] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0104] The above are merely exemplary embodiments of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Other embodiments of this disclosure will be readily apparent to those skilled in the art upon consideration of the specification and practical disclosure.

[0105] This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure. The specification and embodiments are to be considered exemplary only.

Claims

1. A method for inspecting the quality of reinforced concrete engineering based on a large model, characterized in that, include: Acquire vibration response data and 3D point cloud data of the steel reinforcement project to be inspected; The vibration response data is input into a pre-trained vibration feature analysis model to obtain a stress distribution map and stress transmission path. Based on the stress transmission path, the three-dimensional point cloud data is segmented along the stress transmission direction to obtain segmented point cloud data. Based on the segmented point cloud data, the morphological features of each segment of steel bar in the steel bar project to be inspected are extracted, and a stress-geometry coupled digital twin model is established according to the stress distribution map and morphological features. The digital twin model is dynamically mapped to the pre-calibrated BIM design model to determine the steel reinforcement deformation parameters in the stress anomaly area; Based on the steel bar deformation parameters, a quality inspection report for the steel bar project to be inspected is generated.

2. The method for inspecting the quality of reinforced concrete engineering based on a large model according to claim 1, characterized in that, The training process of the pre-trained vibration feature analysis model includes: A training sample library containing various types of reinforced concrete structures is constructed. Each training sample in the training sample library includes vibration response data, stress distribution true value data, and stress transfer path annotation data. Data augmentation processing is performed on the vibration response data in the training sample library to obtain an expanded training dataset; The expanded training dataset is input into the preset base model to obtain the predicted stress distribution map and stress transmission path; Calculate the loss function between the predicted stress distribution map and stress transfer path and the corresponding true data; The model parameters of the preset base model are iteratively optimized based on the loss function until the preset base model converges. The converged preset base model is then used as a pre-trained vibration feature analysis model.

3. The method for inspecting the quality of reinforced concrete engineering based on a large model according to claim 1, characterized in that, Based on the stress transmission path, the three-dimensional point cloud data is segmented along the stress transmission direction to obtain segmented point cloud data, including: Obtain the direction vector and node coordinates of the stress transfer path; A local coordinate system is established based on the node coordinates, and the 3D point cloud data is mapped to the local coordinate system; Density clustering analysis is performed on the mapped point cloud data along the direction vector to determine the segment boundaries of the point cloud data. The point cloud data between the segment boundaries is subjected to region growing processing to obtain segmented point cloud data.

4. The method for inspecting the quality of reinforced concrete engineering based on a large model according to claim 3, characterized in that, The process of performing region growing on the point cloud data between the segment boundaries to obtain segmented point cloud data includes: Calculate the normal vector and curvature eigenvalue of each point in the point cloud data; Within each segment boundary, select the point with the smallest curvature eigenvalue as the seed point; Based on the normal vector of the seed point, calculate the search radius and the normal vector angle threshold, wherein the search radius is proportional to the point cloud density and the normal vector angle threshold is inversely proportional to the curvature feature value of the seed point; Traverse the neighboring points of the seed point, and sequentially add the neighboring points within the search radius whose normal vector angle is less than the normal vector angle threshold to the current region, and update the average normal vector of the current region. Using the updated average normal vector as the target search direction, the traversal process is repeated until no new neighborhood points are added to the current region, thus obtaining segmented point cloud data.

5. The method for inspecting the quality of reinforced concrete engineering based on a large model according to claim 1, characterized in that, The step of extracting the morphological features of each segment of steel bar in the steel bar project to be inspected based on the segmented point cloud data includes: Perform main direction analysis on the segmented point cloud data to determine the axial direction of each segment of steel reinforcement; For each of the aforementioned reinforcing bars, a sampling plane is established based on the axial direction, and cross-sectional point clouds of the reinforcing bars are obtained on the sampling plane at preset intervals; Ellipse fitting is performed on the point cloud of the cross section to obtain the cross-sectional parameters of the reinforcing bar; Based on the cross-sectional parameters of the reinforcing bar, determine the bending degree and torsion angle of the reinforcing bar; The curvature and the torsion angle are used as morphological characteristics of the reinforcing bar.

6. The method for inspecting the quality of reinforced concrete engineering based on a large model according to claim 1, characterized in that, The step of establishing a stress-geometry coupled digital twin model based on the stress distribution spectrum and morphological characteristics includes: Construct a geometric topological relationship diagram of the rebar nodes in the rebar project to be inspected; Based on the stress distribution map, the stress transfer weight between each steel reinforcement node is calculated. The stress transfer weights are superimposed on the geometric topology diagram to obtain a stress-geometry coupling diagram; Based on the stress-geometry coupling diagram and the morphological features, a finite element mesh model of the steel reinforcement project to be inspected is constructed. Stress field interpolation is performed on the finite element mesh model to generate a stress-geometry coupled digital twin model.

7. The method for inspecting the quality of reinforced concrete engineering based on a large model according to claim 1, characterized in that, The dynamic mapping between the digital twin model and the pre-calibrated BIM design model to determine the reinforcement deformation parameters in the stress anomaly area includes: Construct the feature mapping matrix between the digital twin model and the BIM design model; Based on the feature mapping matrix, the deviation between the measured coordinates and the design coordinates of the rebar nodes is determined; Filter out stress anomaly regions where the deviation value is greater than the deviation threshold; The elastic and plastic deformation of the reinforcing bars within the stress anomaly region are extracted as reinforcing bar deformation parameters.

8. A steel reinforcement engineering quality inspection system based on a large model, characterized in that, The system includes: The data acquisition module is used to acquire vibration response data and three-dimensional point cloud data of the steel reinforcement project to be inspected; The digital twin model building module is used to input the vibration response data into a pre-trained vibration feature analysis model to obtain a stress distribution map and stress transmission path. Based on the stress transmission path, the three-dimensional point cloud data is segmented along the stress transmission direction to obtain segmented point cloud data. Based on the segmented point cloud data, the morphological features of each segment of steel bar in the steel bar project to be inspected are extracted, and a stress-geometry coupled digital twin model is established according to the stress distribution map and morphological features. The steel reinforcement deformation parameter determination module is used to dynamically map the digital twin model with the pre-calibrated BIM design model to determine the steel reinforcement deformation parameters in the stress anomaly area. The quality inspection report generation module is used to generate a quality inspection report for the steel reinforcement project to be inspected based on the steel reinforcement deformation parameters.

9. An electronic device, characterized in that, The device includes a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to enable the electronic device to perform the large-model-based steel reinforcement engineering quality inspection method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed, perform the large-model-based steel reinforcement engineering quality inspection method as described in any one of claims 1-7.

Images