A pipe truss optimization design method based on BIM technology

Abstract

The application belongs to the technical field of optimization design, and discloses a pipe truss optimization design method based on BIM technology, which comprises the following steps: identifying the connection relationship of pipe truss nodes and rod pieces in a BIM model to obtain a basic topology graph, giving the node geometry attribute and node type to the vertex, giving the rod piece section attribute and rod piece category to the edge to obtain an attribute topology graph; extracting the vertex local graph of each vertex; inputting the vertex local graph into a pre-trained node parameter adjustment amount prediction model to obtain an adjustment amount vector set with a confidence degree not lower than a preset threshold; optimizing the corresponding node according to the effective adjustment dimension determined by the adjustment amount vector set and the node type, and writing the optimization result into the BIM model. The application converts the traditional node optimization process which depends on the repeated iteration of finite element calculation and specification checking into prediction based on the pre-trained model, reduces the dependence of pipe truss node optimization on repeated finite element iteration calculation, and improves the design efficiency.

Description

Technical Field

[0001] This invention relates to the field of optimization design technology, and more specifically, to a method for optimizing the design of tubular trusses based on BIM technology. Background Technology

[0002] Tubular truss structures are widely used in engineering fields such as large-span industrial plants, stadiums, and bridges due to their high load-bearing efficiency and strong span capacity. The joints of a tubular truss are the core components for force transmission, and their load-bearing capacity directly affects the overall safety of the structure. In engineering design, the load-bearing capacity verification of joints needs to be checked against various failure modes one by one. When the verification fails to meet the specifications, the joint parameters must be adjusted and recalculated. Only after multiple iterations can a design scheme that meets the requirements be obtained.

[0003] In existing design processes, the iterative process heavily relies on alternating finite element analysis (FEM) calculations and manual verification according to specifications. When the parameters of a particular node are adjusted, it is typically necessary to re-execute the overall structural FEM to obtain the updated internal force distribution, and then substitute the new internal force results into the specification formulas for re-verification to determine the effectiveness of the adjustment. For complex tubular truss structures with numerous nodes, this iterative process is computationally expensive and time-consuming, severely hindering design efficiency.

[0004] In summary, how to reduce the reliance on repeated finite element iteration calculations in the optimization process of tubular truss nodes and achieve efficient and automated generation of node parameter optimization and adjustment schemes is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] To overcome the aforementioned problems in the prior art, this invention proposes a tube truss optimization design method based on BIM technology to solve the above problems.

[0006] This invention provides the following technical solution: A BIM-based method for optimizing the design of tubular trusses includes: Identify the connection relationships between tubular truss nodes and members in the BIM model to obtain a foundation topology diagram, wherein the vertices of the foundation topology diagram correspond to tubular truss nodes and the edges correspond to tubular truss members; Based on the truss node and member data, the vertices in the basic topology graph are assigned node geometric attributes and node type, and the edges are assigned member cross-sectional attributes and member category, thus obtaining the attribute topology graph. For each vertex in the attribute topology graph, extract the vertex local graph centered on that vertex; The vertex local graph corresponding to each vertex is input into a pre-trained node parameter adjustment prediction model to obtain a set of adjustment vectors with a confidence level not lower than a preset confidence threshold. The adjustment vectors include various dimensions corresponding to the outer diameter of the main pipe, the wall thickness of the main pipe, the outer diameter of the branch pipe, the wall thickness of the branch pipe, and the gap between the branch pipes. Based on the effective adjustment dimension determined by the set of adjustment vectors and node types, the corresponding truss nodes are optimized, and the optimization results are written into the BIM model to obtain the optimized truss model.

[0007] Preferably, the node geometric attributes include the main pipe outer diameter, main pipe wall thickness, outer diameter of each branch pipe, wall thickness of each branch pipe, angle between each branch pipe and the main pipe, and branch pipe gap; the node types include T-type nodes, Y-type nodes, K-type nodes, and KT-type nodes. The cross-sectional properties of the members include cross-sectional area, moment of inertia, and member length; the member categories include chord members, diagonal web members, and vertical web members.

[0008] Preferably, the extraction of the vertex local map centered on the vertex includes: Centered on the vertex, extract the vertices and their corresponding connecting edges in the attribute topology graph that are connected to the vertex with a number of edges not exceeding a preset number, and construct the vertex local graph corresponding to the vertex.

[0009] Preferably, the training method for the node parameter adjustment prediction model includes: Generate a training sample set, wherein each training sample in the training sample set includes a vertex local graph and an adjustment vector of the corresponding node; Using the vertex local graph as input and the corresponding adjustment vector as the training objective, the mean square error between the predicted value and the label value on the effective dimension determined by the node type is used as the loss. The model parameters are updated by minimizing the loss, and a node parameter adjustment prediction model is trained.

[0010] Preferably, the method for generating the training sample set includes: Generate multiple training truss configurations, construct an attribute topology graph for each training truss configuration, and extract the vertex local graphs of each node; Finite element analysis was performed on each training truss configuration. The internal forces and geometric properties of the nodes were substituted into the standard bearing capacity formula corresponding to each failure mode to calculate the bearing capacity utilization rate of each failure mode. For node types with multiple branches or multiple branch gaps, the representative value of the corresponding dimension was determined according to preset rules. For failure modes where the load-bearing capacity utilization rate does not exceed the specification limit, the component of the corresponding dimension in the adjustment vector is set to zero; for failure modes where the load-bearing capacity utilization rate exceeds the specification limit, the target dimension is determined according to the preset priority, the target dimension value is calculated, and the difference between the target dimension value and the current dimension value is used as the initial value of the adjustment; when multiple failure modes are mapped to the same dimension, the maximum value of each initial value is taken as the final adjustment of that dimension. For node types without branch gaps, set the component of the branch gap dimension in the adjustment vector to zero. The training sample set is constructed using the vertex local graph and the adjustment vector.

[0011] Preferably, the generation of multiple training truss configurations includes: Sample within a topology parameter space consisting of a preset span range, height range, number of internodes range, and set of web member types to generate multiple topology parameter combinations; For each combination of topological parameters, within the range of cross-section models that satisfy the chord diameter-to-thickness ratio constraint and the web-to-chord diameter ratio constraint, the chord cross-section models and web cross-section models are sampled respectively. All nodes in the same truss configuration share the chord cross-section model, and web nodes of the same type share the web cross-section model. For node types with branch pipe gaps, each node independently samples the branch pipe gap value within the gap range that meets the specification requirements. By combining the aforementioned topological parameters with the corresponding chord section type, web section type, and branch gap value, multiple training truss configurations are obtained.

[0012] Preferably, the optimization of the corresponding truss node based on the effective adjustment dimension determined by the adjustment vector set and the node type includes: No optimization is performed when the set of adjustment vectors is empty; When the set of adjustment vectors is not empty, the set of adjustment vectors is converted into a set of candidate adjustment schemes according to the standard model specifications; When there is only one candidate adjustment scheme in the candidate adjustment scheme set, that scheme is directly used as the target adjustment scheme. When the number of candidate adjustment schemes in the candidate adjustment scheme set exceeds one, the adjustment cost of each candidate adjustment scheme is calculated, and the scheme with the minimum adjustment cost is selected as the target adjustment scheme. Update the node geometric attributes involved in the target adjustment scheme to the corresponding vertices of the attribute topology diagram, and write the updated node geometric attributes and associated member cross-sectional attributes back to the BIM model.

[0013] Preferably, converting the set of adjustment vectors into a set of candidate adjustment schemes based on standard model specifications includes: For each adjustment vector in the set of adjustment vectors, the effective adjustment dimension is determined based on the node type; For the non-zero components corresponding to the main pipe outer diameter, main pipe wall thickness, branch pipe outer diameter, and branch pipe wall thickness in the effective adjustment dimensions, find the smallest standard model in the standard model specifications that is not less than the sum of the current dimension value and the corresponding adjustment amount. For the non-zero component corresponding to the branch gap in the effective adjustment dimension, the sum of the current branch gap value and the corresponding adjustment amount is taken as the target gap value; The candidate standard models and target gap values ​​corresponding to each effective non-zero component are combined to generate a set of candidate adjustment schemes.

[0014] Preferably, the method for calculating the adjustment cost includes: For each candidate adjustment scheme, the adjustment cost is the sum of the product of the price difference per unit length between the adjusted model and the original model and the corresponding rod length. The adjustment cost for the branch pipe gap is not included in the adjustment cost. The option with the lowest adjustment cost is selected from the set of candidate adjustment options as the target adjustment option.

[0015] This invention provides a method for optimizing the design of tubular trusses based on BIM technology, which has the following advantages: This invention transforms the connection relationships and geometric attributes of tubular trusses in a BIM model into an attribute topology graph. Using the vertex local graphs of each node as input and the node parameter adjustment amounts calculated by finite element analysis combined with the standard bearing capacity formula as training labels, a node parameter adjustment prediction model is pre-trained. During the inference phase, the vertex local graphs obtained from the BIM model analysis are directly input into the trained model to obtain the parameter adjustment scheme for each node. This eliminates the need to re-perform finite element calculations for the current design scheme, reducing the reliance on repeated finite element iterations for tubular truss node optimization and improving design efficiency.

[0016] This invention uses vertex local graphs, rather than single-node geometric parameters, as input to the prediction model for node parameter adjustment. The vertex local graph encompasses the geometric configuration and member cross-sectional information of adjacent nodes within the target node's neighborhood. This allows the model to learn the correlation between node stress and the surrounding structural topology and cross-sectional configuration from a large number of finite element calculations of truss configurations during training. This automatically incorporates the influence of the surrounding structure on the node's stress when predicting its adjustment, resulting in an adjustment scheme that is no longer limited to isolated judgments of single-node geometric parameters but reflects the node's stress state within the local truss structure, thus improving the accuracy of the adjustment scheme. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a BIM-based optimized design method for tubular trusses according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Example

[0019] Please see Figure 1 In this embodiment, a tube truss optimization design method based on BIM technology includes: S1. Identify the connection relationships between tubular truss nodes and members in the BIM model to obtain a foundation topology diagram, wherein the vertices of the foundation topology diagram correspond to tubular truss nodes and the edges correspond to tubular truss members. In this embodiment, it should be noted that the BIM model of the project is usually created by engineers based on the project's design scheme. It includes the geometric dimensions, cross-sectional types, spatial locations, and connection relationships between the components of the tubular truss. By analyzing the endpoint connection relationships of each component in the BIM model, each tubular truss node is mapped to a vertex in the topology graph, and the member connecting two nodes is mapped to an edge between the corresponding two vertices. This transforms the spatial connection relationship of the tubular truss into a graph structure representation, resulting in the basic topology graph.

[0020] S2. Based on the truss node and member data, assign node geometric attributes and node type to the vertices in the basic topology graph, and assign member cross-sectional attributes and member category to the edges to obtain the attribute topology graph. The node geometric attributes include the main pipe outer diameter, main pipe wall thickness, outer diameter of each branch pipe, wall thickness of each branch pipe, angle between each branch pipe and the main pipe, and branch pipe gap; the node types include T-type nodes, Y-type nodes, K-type nodes, and KT-type nodes. The cross-sectional properties of the members include cross-sectional area, moment of inertia, and member length; the member categories include chord members, diagonal web members, and vertical web members.

[0021] In this embodiment, node geometric attributes and member cross-sectional attributes can be directly read from the BIM model. The angle between each branch pipe and the main pipe is calculated from the angle between the axis directions of the branch pipe and the main pipe in the BIM model. The member length is calculated from the spatial coordinate difference between the two end nodes. Other attributes are directly extracted from the cross-sectional model parameters of the corresponding component in the BIM model. Node types are classified into T-type, Y-type, K-type, and KT-type according to engineering standards, determined by the number of branch pipes connected to the node and their geometric relationship. For example, a node with only one branch pipe intersecting the main pipe at an angle close to 90 degrees is determined as a T-type node; a node with only one branch pipe intersecting the main pipe at a non-90-degree angle is determined as a Y-type node, etc. Member categories are determined according to the member's position in the truss. Members along the upper and lower chord directions are chord members, web members diagonally connecting the upper and lower chords are diagonal web members, and web members vertically connecting the upper and lower chords are vertical web members.

[0022] S3. For each vertex in the attribute topology graph, extract the vertex local graph centered on that vertex; The extraction of the vertex local map centered on that vertex includes: Centered on the vertex, extract the vertices and their corresponding connecting edges in the attribute topology graph that are connected to the vertex with a number of edges not exceeding a preset number, and construct the vertex local graph corresponding to the vertex.

[0023] In this embodiment, the preset number of edges determines the neighborhood range covered by the vertex local graph. Its value needs to be balanced between local information coverage and computational complexity: if the value is too small, it will be difficult to capture the influence of a large range of structural topology on the stress of nodes; if the value is too large, the local graph will approach the global graph, increasing the computational complexity. For common span truss structures, the preset number of edges can be 2. In this case, for a node in the middle of the truss, the vertex local graph usually includes the node and about 7 to 9 adjacent nodes and corresponding connecting edges.

[0024] S4. Input the local graph of each vertex into the pre-trained node parameter adjustment prediction model to obtain a set of adjustment vectors with a confidence level not lower than the preset confidence threshold; wherein, the adjustment vectors include each dimension corresponding to the outer diameter of the main pipe, the wall thickness of the main pipe, the outer diameter of the branch pipe, the wall thickness of the branch pipe, and the gap between the branch pipes respectively. The training method for the node parameter adjustment prediction model includes: Generate a training sample set, wherein each training sample in the training sample set includes a vertex local graph and an adjustment vector of the corresponding node; Using the vertex local graph as input and the corresponding adjustment vector as the training objective, the mean square error between the predicted value and the label value on the effective dimension determined by the node type is used as the loss. The model parameters are updated by minimizing the loss, and a node parameter adjustment prediction model is trained.

[0025] The method for generating the training sample set includes: Generate multiple training truss configurations, construct an attribute topology graph for each training truss configuration, and extract the vertex local graphs of each node; Finite element analysis was performed on each training truss configuration. The internal forces and geometric properties of the nodes were substituted into the standard bearing capacity formula corresponding to each failure mode to calculate the bearing capacity utilization rate of each failure mode. For node types with multiple branches or multiple branch gaps, the representative value of the corresponding dimension was determined according to preset rules. For failure modes where the load-bearing capacity utilization rate does not exceed the specification limit, the component of the corresponding dimension in the adjustment vector is set to zero; for failure modes where the load-bearing capacity utilization rate exceeds the specification limit, the target dimension is determined according to the preset priority, the target dimension value is calculated, and the difference between the target dimension value and the current dimension value is used as the initial value of the adjustment; when multiple failure modes are mapped to the same dimension, the maximum value of each initial value is taken as the final adjustment of that dimension. For node types without branch gaps, set the component of the branch gap dimension in the adjustment vector to zero. The training sample set is constructed using the vertex local graph and the adjustment vector.

[0026] The generation of multiple training truss configurations includes: Sample within a topology parameter space consisting of a preset span range, height range, number of internodes range, and set of web member types to generate multiple topology parameter combinations; For each combination of topological parameters, within the range of cross-section models that satisfy the chord diameter-to-thickness ratio constraint and the web-to-chord diameter ratio constraint, the chord cross-section models and web cross-section models are sampled respectively. All nodes in the same truss configuration share the chord cross-section model, and web nodes of the same type share the web cross-section model. For node types with branch pipe gaps, each node independently samples the branch pipe gap value within the gap range that meets the specification requirements. By combining the aforementioned topological parameters with the corresponding chord section type, web section type, and branch gap value, multiple training truss configurations are obtained.

[0027] In this embodiment, the node parameter adjustment prediction model takes the vertex local graph as input and outputs a five-dimensional adjustment vector for the corresponding node. The five dimensions correspond to the adjustment amounts of the main pipe outer diameter, main pipe wall thickness, branch pipe outer diameter, branch pipe wall thickness, and branch pipe gap, respectively. Positive values ​​indicate that the corresponding geometric parameter needs to be increased, while negative values ​​indicate that the corresponding geometric parameter needs to be decreased. The node parameter adjustment prediction model can be implemented using a graph neural network. During training, the vertex local graph is used as input, and the adjustment vector of the corresponding node is used as the training target. Since T-type and Y-type nodes do not have branch gaps, the label of their branch gap dimension is always zero. Including this dimension in the loss calculation would introduce a large number of zero labels that would interfere with the prediction of other dimensions. Therefore, during training, binary weights are assigned to each dimension according to the node type of the target vertex. The weights of all five dimensions of K-type and KT-type nodes are 1, the weight of the branch gap dimension of T-type and Y-type nodes is 0, and the weights of the other four dimensions are 1. The loss function uses the average of the product of the mean square error of each dimension and the corresponding weight as the total loss. The weight of the branch gap dimension is 0, so that the gradient of the loss term of this dimension is zero during backpropagation, thereby automatically shielding the parameter update of this dimension.

[0028] It should be noted that training the node parameter adjustment prediction model requires a large number of training samples, typically no less than 10,000. To ensure a uniform distribution of training samples within the topology parameter space, methods such as Latin hypercube sampling can be used to generate topology parameter combinations. The topology parameter sampling range can be set according to the actual engineering situation. For each topology parameter combination, within the range of models that meet the constraints of the chord diameter-to-thickness ratio and the web-to-chord diameter ratio in the truss design code, the chord and web section models are sampled weighted according to the usage frequency of commonly used models in the project. This weighted sampling ensures that the cross-sectional distribution of the training data matches the commonly used configurations in actual projects. The chord diameter-to-thickness ratio constraint requires that the ratio of the main pipe's outer diameter to its wall thickness does not exceed the code limit; for example, for a circular hollow section main pipe, this limit is usually between 50 and 80. The web-to-chord diameter ratio constraint requires that the ratio of the branch pipe's outer diameter to the main pipe's outer diameter be within the range specified in the code, typically not less than 0.2 and not greater than 1.0. For K-type and KT-type nodes, the branch pipe gap value is independently sampled within the range that meets the construction requirements of the gap type node. The gap is usually required to be no less than the sum of the wall thicknesses of the two branch pipes to ensure the feasibility of welding construction.

[0029] For each training truss configuration, a corresponding finite element model was established using general-purpose finite element analysis software (such as ANSYS, Midas, or SAP2000), and design loads were applied to perform structural calculations. The verification internal forces of each node were extracted from the calculation results. These verification internal forces and node geometric properties were then substituted into the bearing capacity calculation formulas given in the CIDECT design guidelines for each node type and failure mode to calculate the bearing capacity utilization rate for each failure mode. The utilization rate is equal to the ratio of the actual internal force of the node to the standard bearing capacity. Each failure mode was calculated independently, and a utilization rate not exceeding 1.0 indicates that the bearing capacity under that mode meets the requirements. The failure modes and corresponding internal forces involved in different node types are different: K-type nodes consider plastic failure of the main pipe wall, punching shear failure of the main pipe, and failure of the effective width of the branch pipe, and the internal forces to be verified are the axial forces of the two branch pipes and the axial forces and shear forces of the main pipe section; T-type and Y-type nodes consider plastic failure of the main pipe wall and punching shear failure of the main pipe, and the internal forces to be verified are the axial forces of a single branch pipe and the axial forces and shear forces of the main pipe; KT-type nodes consider all three failure modes mentioned above, and the internal forces to be verified are the axial forces of the three branch pipes and the axial forces and shear forces of the main pipe.

[0030] For K-type and KT-type joints with multiple branches, when the cross-sectional parameters of the two branches are different, the branch parameter that maximizes the utilization rate of the bearing capacity for each failure mode is substituted into the calculation to ensure that the utilization rate calculation results for each failure mode are conservative values. For KT-type joints with gaps between two branches, the smaller gap value is taken as the representative value of the branch gap dimension. This is because, in the standard bearing capacity formula, the smaller the gap, the more concentrated the stress on the joint and the less favorable the bearing capacity.

[0031] The determination of the components of each dimension of the adjustment vector is as follows: For failure modes with a utilization rate not exceeding 1.0, the corresponding dimension component is set to zero; for failure modes with a utilization rate exceeding 1.0, the target geometric parameter values ​​required to reduce the utilization rate to exactly 1.0 need to be determined through back-calculation. The back-calculation process fixes all parameters in the standard bearing capacity formula except for the target dimension, and solves for the target dimension parameter values. For analytical formulas, direct algebraic solutions can be used; for implicit formulas, numerical solutions such as the bisection method can be used. The target dimension is determined according to a preset priority: wall thickness is selected first, followed by outer diameter, and finally branch pipe clearance. This is based on the fact that wall thickness adjustment significantly improves the bearing capacity of nodes while having a relatively small impact on the overall structure. The difference between the target dimension parameter value and the current value is used as the initial value of the adjustment for that dimension; a positive value indicates that the parameter needs to be increased, and a negative value indicates that the parameter needs to be decreased. When the back-calculation results of multiple failure modes are mapped to the same dimension, the initial value with the largest absolute value is taken as the final adjustment for that dimension to ensure that the requirements of all over-limit failure modes are met simultaneously after adjustment. For T-type and Y-type nodes, the branch gap dimension is a redundant space-occupying dimension, and the corresponding component is directly set to zero.

[0032] Taking a K-type node as an example: the current main pipe wall thickness is 8 mm, and the main pipe wall plastic failure utilization rate is 1.15. By inversely solving the CIDECT main pipe wall plastic failure bearing capacity formula, the required main pipe wall thickness to make the utilization rate equal to 1.0 is 10 mm, with an initial adjustment value of +2 mm. The main pipe punching shear failure utilization rate is 1.08, and the inverse solution yields a required main pipe wall thickness of 9 mm, with an initial adjustment value of +1 mm. Mapping both to the same dimension, +2 mm is taken as the final adjustment. At the same time, the branch pipe gap corresponds to a failure mode utilization rate of 1.08, and the inverse solution shows that the branch pipe gap needs to be increased from 50 mm to 65 mm, with an adjustment value of +15 mm. Finally, the adjustment vector for this K-type node is [0, +2, 0, 0, +15].

[0033] When using node parameter adjustment for prediction model inference, the vertex local graph of the node to be predicted is input into the trained model, and the model outputs an adjustment vector. The model's prediction reliability varies for different node configurations. For node configurations not fully covered in the training samples, the model's output adjustment vector has lower reliability and needs to be filtered using confidence scores. The confidence score is estimated using the Monte Carlo Dropout method: during the inference phase, the Dropout layer is kept active, and the same vertex local graph is forward-propagated multiple times (e.g., 20 times). The more consistent the prediction results are across multiple iterations, the higher the confidence score. Uncertainty is measured by the mean of the standard deviations of predictions in each effective dimension, and after normalization, a confidence score between 0 and 1 is obtained. The confidence threshold can be set between 0.8 and 0.9. Predictions below the threshold are not included in the adjustment vector set; the adjustment vectors of the remaining nodes are aggregated to form the adjustment vector set.

[0034] S5. Based on the effective adjustment dimension determined by the set of adjustment vectors and the node type, optimize the corresponding truss nodes, write the optimization results into the BIM model, and obtain the optimized truss model.

[0035] The optimization of the corresponding tubular truss nodes based on the effective adjustment dimension determined by the adjustment vector set and node type includes: No optimization is performed when the set of adjustment vectors is empty; When the set of adjustment vectors is not empty, the set of adjustment vectors is converted into a set of candidate adjustment schemes according to the standard model specifications; When there is only one candidate adjustment scheme in the candidate adjustment scheme set, that scheme is directly used as the target adjustment scheme. When the number of candidate adjustment schemes in the candidate adjustment scheme set exceeds one, the adjustment cost of each candidate adjustment scheme is calculated, and the scheme with the minimum adjustment cost is selected as the target adjustment scheme. Update the node geometric attributes involved in the target adjustment scheme to the corresponding vertices of the attribute topology diagram, and write the updated node geometric attributes and associated member cross-sectional attributes back to the BIM model.

[0036] The process of converting the set of adjustment vectors into a set of candidate adjustment schemes based on standard model specifications includes: For each adjustment vector in the set of adjustment vectors, the effective adjustment dimension is determined based on the node type; For the non-zero components corresponding to the main pipe outer diameter, main pipe wall thickness, branch pipe outer diameter, and branch pipe wall thickness in the effective adjustment dimensions, find the smallest standard model in the standard model specifications that is not less than the sum of the current dimension value and the corresponding adjustment amount. For the non-zero component corresponding to the branch gap in the effective adjustment dimension, the sum of the current branch gap value and the corresponding adjustment amount is taken as the target gap value; The candidate standard models and target gap values ​​corresponding to each effective non-zero component are combined to generate a set of candidate adjustment schemes.

[0037] The method for calculating the adjustment cost includes: For each candidate adjustment scheme, the adjustment cost is the sum of the product of the price difference per unit length between the adjusted model and the original model and the corresponding rod length. The adjustment cost for the branch pipe gap is not included in the adjustment cost. The option with the lowest adjustment cost is selected from the set of candidate adjustment options as the target adjustment option.

[0038] In this embodiment, when the adjustment vector set is empty, it indicates that the model has no output with a confidence level higher than the preset value, usually meaning that the node does not need optimization. When the adjustment vector set is not empty, the effective adjustment dimensions are determined according to the node type: all five dimensions are effective for K-type and KT-type nodes; the branch pipe gap dimension is invalid for T-type and Y-type nodes. For the four effective adjustment dimensions of main pipe outer diameter, main pipe wall thickness, branch pipe outer diameter, and branch pipe wall thickness, the sum of the current geometric parameter value and the corresponding adjustment amount is taken as the target value, and the smallest standard model not less than the target value is found in the standard model specifications. Taking the above K-type node as an example, the target value of the main pipe wall thickness is 10 mm. If there is a standard model with 10 mm, it is selected; otherwise, the closest standard value not less than 10 mm, such as 12 mm, is selected. For the branch pipe gap dimension, the sum of the current branch pipe gap value and the adjustment amount, 65 mm, is directly taken as the target gap value.

[0039] When there are multiple adjustment vectors in the adjustment vector set, multiple candidate adjustment schemes are generated accordingly. The optimal scheme is selected based on the economic cost of model change: for each dimension where model change occurs, the difference in unit length price between the adjusted and unadjusted models is multiplied by the corresponding rod length. The total adjustment cost is obtained by summing all dimensions where model change occurs. The scheme with the minimum total adjustment cost is selected as the target adjustment scheme. The adjustment of branch pipe gap is not included in the cost.

[0040] Once the target adjustment scheme is determined, the updated node geometric attributes and the corresponding cross-sectional areas and moments of inertia of the members are synchronously written back to the BIM model. After all vertices have been processed, the optimized tubular truss model is obtained. Compared to the traditional node optimization process that relies on repeated finite element iterations, this method directly predicts the adjustment scheme of each node from the topological and geometric information of the BIM model through a pre-trained node parameter adjustment prediction model. This eliminates the need to re-execute finite element calculations during the inference phase, significantly reducing the reliance on repeated finite element iterations in the tubular truss node optimization process and achieving efficient and automated generation of node parameter optimization adjustment schemes.

[0041] In the several embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only one method, 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 interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0042] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

[0043] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for optimizing the design of tubular trusses based on BIM technology, characterized in that, include: Identify the connection relationships between tubular truss nodes and members in the BIM model to obtain a foundation topology diagram, wherein the vertices of the foundation topology diagram correspond to tubular truss nodes and the edges correspond to tubular truss members; Based on the truss node and member data, the vertices in the basic topology graph are assigned node geometric attributes and node type, and the edges are assigned member cross-sectional attributes and member category, thus obtaining the attribute topology graph. For each vertex in the attribute topology graph, extract the vertex local graph centered on that vertex; The vertex local graph corresponding to each vertex is input into a pre-trained node parameter adjustment prediction model to obtain a set of adjustment vectors with a confidence level not lower than a preset confidence threshold. The adjustment vectors include various dimensions corresponding to the outer diameter of the main pipe, the wall thickness of the main pipe, the outer diameter of the branch pipe, the wall thickness of the branch pipe, and the gap between the branch pipes. The training method for the node parameter adjustment prediction model includes: Generate a training sample set, wherein each training sample in the training sample set includes a vertex local graph and an adjustment vector of the corresponding node; Using the vertex local graph as input and the corresponding adjustment vector as the training objective, the mean square error between the predicted value and the label value on the effective dimension determined by the node type is used as the loss. The model parameters are updated by minimizing the loss, and a node parameter adjustment prediction model is trained. The method for generating the training sample set includes: Generate multiple training truss configurations, construct an attribute topology graph for each training truss configuration, and extract the vertex local graphs of each node; Finite element analysis was performed on each training truss configuration. The internal forces and geometric properties of the nodes were substituted into the standard bearing capacity formula corresponding to each failure mode to calculate the bearing capacity utilization rate of each failure mode. For node types with multiple branches or multiple branch gaps, the representative value of the corresponding dimension was determined according to preset rules. For failure modes where the load-bearing capacity utilization rate does not exceed the specification limit, the component of the corresponding dimension in the adjustment vector is set to zero; for failure modes where the load-bearing capacity utilization rate exceeds the specification limit, the target dimension is determined according to the preset priority, the target dimension value is calculated, and the difference between the target dimension value and the current dimension value is used as the initial value of the adjustment; when multiple failure modes are mapped to the same dimension, the maximum value of each initial value is taken as the final adjustment of that dimension. For node types without branch gaps, set the component of the branch gap dimension in the adjustment vector to zero. The training sample set is constructed using the vertex local graph and the adjustment vector. Based on the effective adjustment dimension determined by the set of adjustment vectors and node types, the corresponding truss nodes are optimized, and the optimization results are written into the BIM model to obtain the optimized truss model.

2. The method for optimizing the design of tubular trusses based on BIM technology according to claim 1, characterized in that, The node geometric attributes include the main pipe outer diameter, main pipe wall thickness, outer diameter of each branch pipe, wall thickness of each branch pipe, angle between each branch pipe and the main pipe, and branch pipe gap; the node types include T-type nodes, Y-type nodes, K-type nodes, and KT-type nodes; The cross-sectional properties of the members include cross-sectional area, moment of inertia, and member length; the member categories include chord members, diagonal web members, and vertical web members.

3. The method for optimizing the design of tubular trusses based on BIM technology according to claim 2, characterized in that, The extraction of the vertex local map centered on that vertex includes: Centered on the vertex, extract the vertices and their corresponding connecting edges in the attribute topology graph that are connected to the vertex with a number of edges not exceeding a preset number, and construct the vertex local graph corresponding to the vertex.

4. The method for optimizing the design of tubular trusses based on BIM technology according to claim 3, characterized in that, The generation of multiple training truss configurations includes: Sample within a topology parameter space consisting of a preset span range, height range, number of internodes range, and set of web member types to generate multiple topology parameter combinations; For each combination of topological parameters, within the range of cross-section models that satisfy the chord diameter-to-thickness ratio constraint and the web-to-chord diameter ratio constraint, the chord cross-section models and web cross-section models are sampled respectively. All nodes in the same truss configuration share the chord cross-section model, and web nodes of the same type share the web cross-section model. For node types with branch pipe gaps, each node independently samples the branch pipe gap value within the gap range that meets the specification requirements. By combining the aforementioned topological parameters with the corresponding chord section type, web section type, and branch gap value, multiple training truss configurations are obtained.

5. The method for optimizing the design of tubular trusses based on BIM technology according to claim 4, characterized in that, The optimization of the corresponding tubular truss nodes based on the effective adjustment dimension determined by the adjustment vector set and node type includes: No optimization is performed when the set of adjustment vectors is empty; When the set of adjustment vectors is not empty, the set of adjustment vectors is converted into a set of candidate adjustment schemes according to the standard model specifications; When there is only one candidate adjustment scheme in the candidate adjustment scheme set, that scheme is directly used as the target adjustment scheme. When the number of candidate adjustment schemes in the candidate adjustment scheme set exceeds one, the adjustment cost of each candidate adjustment scheme is calculated, and the scheme with the minimum adjustment cost is selected as the target adjustment scheme. Update the node geometric attributes involved in the target adjustment scheme to the corresponding vertices of the attribute topology diagram, and write the updated node geometric attributes and associated member cross-sectional attributes back to the BIM model.

6. The method for optimizing the design of tubular trusses based on BIM technology according to claim 5, characterized in that, The process of converting the set of adjustment vectors into a set of candidate adjustment schemes based on standard model specifications includes: For each adjustment vector in the set of adjustment vectors, the effective adjustment dimension is determined based on the node type; For the non-zero components corresponding to the main pipe outer diameter, main pipe wall thickness, branch pipe outer diameter, and branch pipe wall thickness in the effective adjustment dimensions, find the smallest standard model in the standard model specifications that is not less than the sum of the current dimension value and the corresponding adjustment amount. For the non-zero component corresponding to the branch gap in the effective adjustment dimension, the sum of the current branch gap value and the corresponding adjustment amount is taken as the target gap value; The candidate standard models and target gap values ​​corresponding to each effective non-zero component are combined to generate a set of candidate adjustment schemes.

7. The method for optimizing the design of tubular trusses based on BIM technology according to claim 6, characterized in that, The method for calculating the adjustment cost includes: For each candidate adjustment scheme, the adjustment cost is the sum of the product of the price difference per unit length between the adjusted model and the original model and the corresponding rod length. The adjustment cost for the branch pipe gap is not included in the adjustment cost. The option with the lowest adjustment cost is selected from the set of candidate adjustment options as the target adjustment option.

Images