CN-121598543-B - BIM-based automatic generation method and system for steel structure welding deepening model

CN121598543BCN 121598543 BCN121598543 BCN 121598543BCN-121598543-B

Abstract

The invention belongs to the technical field related to steel structure deepening design, and discloses a BIM-based automatic generation method and system for a steel structure welding deepening model, wherein the method comprises the steps of extracting part-level features and analyzing process semantics based on a BIM design model, and constructing a part-level feature set and an associated-level feature set; the method comprises the steps of obtaining a feature set, extracting a feature representation of the feature set by using a pre-training encoder, and respectively constructing a primary welding seam prediction function and a secondary welding seam prediction function by taking the feature representation as feature input, wherein the feature set comprises part geometric features, part attribute features and section process semantic features, the associated level feature set comprises connection types, part pair contact features and part pose features, and the primary welding seam specification parameters and the secondary welding seam specification parameters are predicted. The invention provides an intelligent welding deepening model generation method capable of meeting welding requirements, which is beneficial to improving the accuracy, consistency and automation level of welding design.

Inventors

YANG CHEN
ZHANG GAN
LU YIFAN
LIU QI
FANG WEILI

Assignees

华中科技大学

Dates

Publication Date: 20260508
Application Date: 20260130

Claims (9)

1. The automatic generation method of the steel structure welding deepening model based on BIM is characterized by comprising the following steps of: S1, extracting part-level features and analyzing process semantics based on a BIM design model, constructing a part-level feature set aiming at the part, and constructing an association-level feature set aiming at the part pair with connection; The part-level feature set comprises part geometric features, part attribute features and cross-section process semantic features, wherein the cross-section process semantic features comprise cross-section types, assembly sequences of all parts in a member to which the parts belong and standard welding joint forms of all welding seams in the member to which the parts belong; The association level feature set comprises a connection type, a part pair contact feature and a part alignment pose feature, wherein the part pair contact feature comprises a contact region overlapping rate, a connection angle and a contact part semantic mark; s2, extracting feature representations of a part level feature set and an associated level feature set by using a pre-trained encoder, respectively constructing a primary weld prediction function and a secondary weld prediction function by taking the feature representations as feature inputs, and realizing the prediction of primary weld specification parameters and secondary weld specification parameters by performing supervised learning on a sample set with labels so as to generate a welding deepening model; Constructing a welding seam semantic-space-topology fusion graph comprising nodes and connecting edges according to the part level feature set and the association level feature set; The node represents a part, and a connecting edge is arranged between the part pair which is judged to be connected; for any graph node, assigning an original feature vector for the node based on the part-level feature set corresponding to the node, and assigning an original feature vector for the connecting edge based on the corresponding association-level feature set for any connecting edge; Accordingly, extraction of the part-level feature representation from the encoder is performed based on the graph mask in S2.
2. The automatic generation method of the BIM-based steel structure welding deepening model according to claim 1, wherein the part geometrical characteristics comprise part position parameters, part posture parameters, part dimension parameters and space surrounding parameters, the part attribute characteristics comprise part types, material labels, part globally unique identifiers, component identification fields to which the parts belong and component section description character strings associated with components to which the parts belong, and the component section description character strings are used for representing construction types and section types of corresponding components; the part pairs with the connection comprise display part pairs, wherein the part pairs with the display connection are obtained through a BIM design model to serve as the display part pairs, and candidate part pairs are judged through the minimum space of the part space bounding boxes, and the corresponding part pairs are judged to be candidate part pairs when the minimum space of the space bounding boxes is smaller than a preset space threshold.
3. The automatic generation method of the BIM-based steel structure welding deepening model according to claim 1, wherein the construction of the semantic features of the cross-section process is specifically as follows: The method comprises the steps of judging that a component to which a part belongs is a welding combination type, traversing each part in the component according to the shape specification parameters of each function role in the preset component, and identifying the function role of each part in the section composition in the component as a part basic role; when the member to which the part belongs is judged to be a section steel type, the section type is the section steel type, and the assembly sequence and the standard welding joint form are both preset fixed quantities; When the part is judged to be an accessory type, the section type is obtained according to the BIM design model, and the assembly sequence and the standard welding joint form are both preset fixed quantities.
4. The automatic generation method of the BIM-based steel structure welding deepening model according to claim 3, wherein the part level feature set further comprises part role semantic identifiers, wherein the part role semantic identifiers are part basic roles for parts of which the parts belong to a welding combination type, and the part role semantic identifiers are preset fixed amounts for parts of which the parts belong to a section steel type and parts of which the parts belong to an accessory type; correspondingly, for the part of which the member is a section steel type, the equivalent functional role of the section steel geometric surface is identified according to the shape specification parameter of the equivalent functional role of the preset section steel geometric surface, and then the equivalent functional role is taken as a space reference standard, and the auxiliary functional roles of the other auxiliary parts are identified according to the shape specification parameter of the preset auxiliary functional role; And further, based on the part basic roles, the equivalent function roles and the accessory function roles, constructing a contact part semantic identifier in the association level feature set.
5. The automatic generation method of BIM-based steel structure welding deepening model according to claim 4, wherein the training of the graph mask self-encoder specifically comprises: Selecting a plurality of BIM design models to respectively construct a welding seam semantic-space-topology fusion diagram, and establishing a pre-training data set comprising a plurality of welding seam semantic-space-topology fusion diagrams; Respectively carrying out feature random mask processing on node original features and edge original features of samples in the pre-training data set, respectively carrying out feature extraction on the node mask features and the edge mask features after the random mask processing by using a graph mask self-encoder, constructing a node feature reconstruction mapping function, carrying out node feature reconstruction according to the extracted node features and obtaining node feature reconstruction loss, and constructing an edge feature reconstruction mapping function, carrying out edge feature reconstruction according to the extracted edge features and obtaining edge feature reconstruction loss; the comprehensive loss of graph mask self-encoder training includes node feature reconstruction loss and edge feature reconstruction loss.
6. The automatic generation method of a BIM-based steel structure welding deepening model according to claim 5, wherein the training of the graph mask self-encoder further includes: The node-level process role classification head is constructed to recognize the part function roles of the nodes according to the extracted node features, and the loss of role prediction tasks is constructed based on the part role semantic identifications in the part-level feature sets; the method comprises the steps that a side-level connection mode classification head is constructed, prediction of the connection type of the side and semantic identifications of contact parts is carried out according to the extracted side features, and loss of a side-level connection mode prediction task is constructed based on the connection type in the association-level feature set and the semantic identifications of the contact parts; The comprehensive penalty of graph mask self-encoder training also includes the penalty of role prediction tasks and the penalty of edge-level connected mode prediction tasks.
7. The automatic generation method of the BIM-based steel structure welding deepening model according to claim 1, wherein S2 specifically comprises: screening out parts of which the members are of a welding combination type according to the attribute characteristics of the parts, taking a part level characteristic set corresponding to the screened parts as input of an encoder obtained through pre-training, and inputting part characteristic representations output by the encoder into a primary weld prediction function to predict primary weld specification parameters; screening the part pairs with the connection, screening the part pairs with the secondary welding seams, taking the association level feature set of the screened part pairs as the input of the encoder obtained by pre-training, and inputting the association feature representation output by the encoder into a secondary welding seam prediction function to predict the specification parameters of the secondary welding seams.
8. The automatic generation method of the BIM-based steel structure welding deepening model according to claim 1, wherein the primary welding line specification parameters comprise primary welding line existence marks and primary welding line process parameters including primary welding line group intersection coordinates, primary welding line types and welding leg size parameters; A primary weld joint prediction function training stage, combining the primary weld joint existence identification with the marked primary weld joint specification parameter real label to construct primary weld joint existence prediction loss; constructing primary weld process parameter prediction loss according to the predicted value and the marked value of the primary weld process parameter, and carrying out weighted summation on the two losses according to a preset weight coefficient to obtain the comprehensive loss of primary weld prediction function training; the secondary weld specification parameters comprise secondary weld existence marks and secondary weld process parameters comprising secondary weld connection interface boundary tracks, secondary weld types and weld leg size parameters; And constructing the predicted loss of the secondary weld process parameters according to the predicted value and the marked value of the secondary weld process parameters, and carrying out weighted summation of the two losses according to a preset weight coefficient to obtain the comprehensive loss of the secondary weld predicted function training.
9. The automatic generation system of the BIM-based steel structure welding deepening model is characterized by comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the automatic generation method of the BIM-based steel structure welding deepening model according to any one of claims 1-8 when executing the computer program.

Description

BIM-based automatic generation method and system for steel structure welding deepening model Technical Field The invention belongs to the technical field related to steel structure deepening design, and particularly relates to a BIM-based automatic generation method and system for a steel structure welding deepening model. Background In the deepening design stage of steel structure engineering, a design model needs to be converted into an executable machining manufacturing and field installation model, wherein the deepening of welding is a key link. The type, position, size, sequence and other information of the primary welding line (assembly welding among the sub-boards in the cross section of the component) and the secondary welding line (connection welding among the components or parts) directly influence the processing precision of the component, the stress reliability of the structure, the site construction efficiency and the feasibility of the subsequent welding robot operation. Common complex nodes in the steel structure, such as beam column nodes, bracket nodes, column foot nodes and the like, relate to various component forms and complex spatial relationships, and have complicated weld joint layout. The welding relation of the components contained in the nodes is usually manually judged by relying on experience of engineering personnel, manual modeling is time-consuming, key welding seams are easy to miss, and accordingly design quality is inconsistent, and construction efficiency and safety are affected. Currently, automatic welding is deepened by adopting a mode based on manual rules, assembly library weld joint through-drawings or fixed logic, for example, default weld joint configuration is set according to the type of a component or the size of a section. The method is applicable to simple and standardized nodes, but lacks the capability of describing the contact relation between a complex section structure and actual geometry, and is difficult to distinguish the specific welding requirements of a primary welding line and a secondary welding line, and the adaptability and generalization capability are insufficient, so that the method is difficult to cover different node types, engineering conditions and construction process requirements. With the increase of engineering scale and node types, the maintenance and updating of rule bases are more complicated, and the automatic generation of the track and the technological parameters of the welding robot are difficult to directly support, so that the overall automation efficiency is low. In summary, the prior art has technical bottlenecks of high artificial dependency, low intelligent level, insufficient processing capacity of complex nodes and the like in the steel structure welding deepening process. Disclosure of Invention Aiming at the defects or improvement demands of the prior art, the invention provides a BIM-based automatic generation method and system for a steel structure welding deepening model, which are used for solving the technical bottlenecks of high artificial dependency, low intelligent level and insufficient complex node processing capacity in the steel structure welding deepening process in the prior art. In order to achieve the above object, according to one aspect of the present invention, there is provided a method for automatically generating a deep model of welding a steel structure based on BIM, comprising: S1, extracting part-level features and analyzing process semantics based on a BIM design model, constructing a part-level feature set aiming at the part, and constructing an association-level feature set aiming at the part pair with connection; The part-level feature set comprises part geometric features, part attribute features and cross-section process semantic features, wherein the cross-section process semantic features comprise cross-section types, assembly sequences of all parts in a member to which the parts belong and standard welding joint forms of all welding seams in the member to which the parts belong; The association level feature set comprises a connection type, a part pair contact feature and a part alignment pose feature, wherein the part pair contact feature comprises a contact region overlapping rate, a connection angle and a contact part semantic mark; s2, extracting feature representations of a part level feature set and an associated level feature set by using the encoder obtained through pre-training, respectively constructing a primary weld prediction function and a secondary weld prediction function by taking the feature representations as feature inputs, and predicting the primary weld specification parameters and the secondary weld specification parameters by executing supervised learning on the sample set with the labels so as to generate a welding deepening model. According to the BIM-based automatic generation method for the steel structure welding deepening model, the part geometric featur