CN-122020821-A - Spatial relationship completion method and system based on BIM model

CN122020821ACN 122020821 ACN122020821 ACN 122020821ACN-122020821-A

Abstract

The invention discloses a spatial relationship completion method and system based on a BIM model, wherein the method comprises the steps of S1, extracting geometric data in a JSON file as original data, preprocessing the original data, S2, utilizing an octree data structure to perform collision detection and spatial relationship calculation of BIM components, S3, utilizing the existing collision detection result and spatial relationship data to predict undetected potential spatial relationships among BIM components, and completing the missing relationships to complete the whole spatial layout information. The invention cooperates with R-GCN through three-level geometric detection, and captures physical-level contact relation, azimuth relation and semantic-level association to greatly improve the recognition integrity of the spatial relation, thereby greatly shortening the data processing time, reducing the implementation cost and improving the design efficiency and construction quality.

Inventors

Deng Xudan
SUN SIYUAN
YAO FENGFENG
LIU ZHEN
ZHU QINGQING
WANG KAIJUN
LIU XINYU
WANG RAN
QI CHUNYU
YANG XUKUN
AN RAN
WU WEI
GENG CHONG
WANG ZICHAO

Assignees

中国铁路设计集团有限公司

Dates

Publication Date: 20260512
Application Date: 20260413

Claims (10)

1. A spatial relationship completion method based on a BIM model is characterized by comprising the following steps: S1, extracting geometric data in a JSON file as original data, loading a BIM model, extracting BIM component data, and performing node mapping between associated BIM components to complete data preprocessing; S2, performing collision detection and space relation calculation of the BIM component by utilizing the octree data structure; s3, predicting undetected potential spatial relationships among BIM components by utilizing the data obtained by collision detection and spatial relationship calculation, and complementing the missing relationships to complete the whole spatial layout information.
2. The method for spatial relationship completion based on BIM model as set forth in claim 1, wherein the specific method of S2 comprises: S21, constructing an octree for the data preprocessed in the S1; S22, recursively traversing the octree, and rapidly positioning to a subarea containing the set object; S23, checking whether a collision exists for BIM components in each node in the octree, discarding BIM components without collision detection, and calculating specific spatial relation between BIM component pairs for BIM component pairs with collision detection.
3. A method of spatial relationship completion based on a BIM model as recited in claim 2 wherein the method of constructing an octree includes defining a border of the octree, determining a root node border of the octree based on bounding boxes of all BIM members, inserting each BIM member into the octree, and assigning each BIM member to a corresponding child node based on its bounding box.
4. A method of spatial relationship completion based on BIM model as set forth in claim 3 wherein the octree partitioning is divided in two steps, starting from a bounding box containing all objects, into eight equal-sized sub-areas, each sub-area being further divided into eight smaller sub-areas forming a recursive tree structure, each object being assigned to its sub-area, a sub-area being further subdivided if the number of objects contained therein exceeds a predetermined threshold.
5. The method for spatial relationship completion based on BIM model as set forth in claim 4, wherein the specific method of S3 comprises: s31, acquiring existing spatial relationship data and extracting features related to the spatial relationship, wherein the spatial relationship data comprises an ID, a name, a collision type, a collision volume and a relative position of a collision BIM component; S32, based on the data in the S31, realizing spatial relationship prediction by an R-GCN spatial relationship prediction method; S33, complementing and optimizing the spatial relationship between BIM components.
6. The method for spatial relationship completion based on BIM model as set forth in claim 5, wherein the specific method of S32 comprises: S321, acquiring an initialization feature vector of each node in a map to be constructed, wherein the feature vector is derived from attribute information of the node or an embedded vector initialized randomly; s322, updating the feature vector of the current node by aggregating the features of the neighbor nodes by adopting the R-GCN; S323, continuously updating and spreading the feature vectors of the nodes, wherein the output features of each layer are used as input features of the next layer, and global position information of the nodes in the map and relation information of the nodes with other nodes are obtained through layer-by-layer feature extraction and updating; S324, for the node pairs needing to predict the spatial relationship, the R-GCN extracts the characteristics of the node pairs, predicts the spatial relationship type between the node pairs, and outputs the spatial relationship type with highest probability.
7. The method for spatial relationship completion based on BIM model of claim 6, wherein in R-GCN, the node update formula is: (1); Wherein, the Representing for a node Having relationship types in neighbors Is defined by a set of nodes of the set, Is a regularized constant, wherein The value of (2) is , Is a linear transformation function, and uses a parameter matrix for neighboring nodes of the similar type edge Transformation is carried out The activation function is represented as a function of the activation, Is shown in the first In the layer neural network, the current central node Is a certain neighbor node of (a) Is used for the feature vector of (a), Represent the first The linear transformation matrix of the nodes themselves in the layer, Represent the first Layer of the first layer The hidden vector of the individual nodes, R, represents the edge type matrix.
8. The method for spatial relationship completion based on BIM model as set forth in claim 7, wherein the relationship probability distribution pairs nodes Is mapped to a relationship type Probability of (2): (2); representing the type of relationship extracted from this distribution corresponding to the setting Is used to determine the probability value of (1), Representing a multi-layer perceptron whose inputs are two node feature vectors And Is provided for the splicing or the combination of (c), And Respectively represent the passes through After the R-GCN processing of the layer relation graph rolling network, nodes Sum node The final layer of the (a) output feature vector; The output of the last layer is then minimized as the cross entropy loss function: (3); Wherein, the Representing a labeled set of nodes, Q represents the total number of categories of relationship types, i.e., the number of all possible spatial relationship types that the model needs to predict, Represent the first The marked nodes are at the first The output of the individual positions is provided, Representing a real label of the tag, Index variables traversing all relationship categories are represented, with values ranging from 1 to Q.
9. A spatial relationship completion system based on a BIM model, to which the spatial relationship completion method based on a BIM model according to any one of claims 1 to 8 is applied, comprising: The data preprocessing and loading module is used for extracting geometric data from the JSON file, loading the BIM model and processing node name mapping between two associated BIM components; the collision detection and spatial relationship acquisition module is used for performing collision detection by utilizing octree, and analyzing the spatial relationship between two BIM components when collision occurs; and the spatial relationship prediction and supplement module predicts and supplements the spatial relationship among the BIM components according to the collision detection result and a small number of label labels, and perfects the overall spatial layout information.
10. The spatial relationship completion system based on BIM model of claim 9, wherein said spatial relationship between said completed BIM components comprises: The graph structure form according to the BIM model is defined as given an incomplete BIM model graph structure G, G= (V, E), wherein V represents nodes, E represents edges, partial edges are missing, and the aim is to predict missing edges and attributes thereof, so that the graph structure is completed.

Description

Spatial relationship completion method and system based on BIM model Technical Field The invention relates to the technical field of graph data processing, in particular to a spatial relationship completion method and system based on a BIM model. Background Foundation pit excavation, side slope support, underground structure and earthwork are basic links of building engineering construction, and scientificity and accuracy of design directly determine overall safety, stability and economical efficiency of engineering. Along with the continuous expansion of the existing engineering scale, the construction environment is more and more complex, the engineering often faces the challenges of dense surrounding buildings, complicated pipelines, changeable geological conditions and the like, and the related regulations are more and more strict in technical standards for foundation treatment, foundation pit support and underground structure construction, so that extremely high requirements are put on the accuracy, safety and economy of engineering design. At present, in the scheme design stage of foundation pit excavation, side slope support, underground structure and earthwork, the three-dimensional modeling technology has been widely used, compared with the traditional two-dimensional drawing design mode, the three-dimensional modeling technology has obvious advantages in visual presentation, scheme comparison, cooperative communication and other aspects, can preliminarily realize three-dimensional visual expression of engineering structures (including pile foundations, underground continuous walls, anchor rods, steel support and other support members, underground pipelines, underground main structures and the like), assists designers in combing design ideas, arranging and finding out obvious design conflicts, and improves design efficiency and scheme rationality to a certain extent. In the related art, a designer builds a three-dimensional model through mainstream software such as Revit, tekla, midas GTS NX and the like, integrates core design elements such as geological survey data, supporting structure parameters, pipeline information and the like, tries to realize digital modeling of all engineering elements, however, the existing three-dimensional modeling technology still has a plurality of defects in practical application, can not fundamentally solve the core problems of complex spatial relationship and undefined mutual position among various components, pipelines, supporting structures and earth excavation surfaces, and has great difference in construction precision requirements. Secondly, the pipeline is used as a surrounding environment element which needs important consideration in engineering design, the pipeline is various (including a water supply and drainage pipeline, a power cable, a communication line and the like), parameters such as burial depth, pipe diameter and trend are different, the existing three-dimensional modeling is mainly used for simply marking the pipeline position, the spatial range and depth of foundation pit excavation cannot be fully combined, the spatial interference relation between the pipeline and an excavation face and between the pipeline and a supporting structure in a foundation is accurately analyzed, hidden dangers such as collision between the supporting structure and the underground pipeline and touching of the pipeline in the excavation process are easily caused, and the pipeline damage can cause safety accidents such as water cut, power failure and communication interruption, pipeline migration and change in later period are also needed, and engineering cost and construction period burden are increased. Thirdly, the space collaborative design of underground structural components (such as pile foundations, bearing platforms, underground continuous walls and the like) and earth excavation surfaces and peripheral pipelines is provided with a short plate, the arrangement of various components in the existing three-dimensional model is based on theoretical design parameters, the influence of soil stress strain change on the component position in the earth excavation process is not fully considered, the safety distance between the components and the excavation surfaces and between the components and the pipeline is not accurately calculated, the arrangement of partial components is unreasonable, the material waste is caused, the construction convenience is influenced, the connection requirement of subsequent construction procedures cannot be met even, the construction reworking probability is increased, and the engineering construction efficiency is reduced. Thus, there is a need for an intelligent complementation method that can automatically learn and infer BIM models and potential relationships between all pairs of components therein using a small number of known (calculated or labeled) spatial relationships. The method aims at solving the core problem of how to effi