CN116432274B - Rod system structure optimization method and device, electronic equipment and storage medium - Google Patents

Rod system structure optimization method and device, electronic equipment and storage medium Download PDF

Info

Publication number
CN116432274B
CN116432274B CN202310208411.9A CN202310208411A CN116432274B CN 116432274 B CN116432274 B CN 116432274B CN 202310208411 A CN202310208411 A CN 202310208411A CN 116432274 B CN116432274 B CN 116432274B
Authority
CN
China
Prior art keywords
node
matrix
component
target
structural
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202310208411.9A
Other languages
Chinese (zh)
Other versions
CN116432274A (en
Inventor
王琛
张翀
陶慕轩
樊健生
宋凌寒
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tsinghua University
Original Assignee
Tsinghua University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Tsinghua University filed Critical Tsinghua University
Priority to CN202310208411.9A priority Critical patent/CN116432274B/en
Publication of CN116432274A publication Critical patent/CN116432274A/en
Application granted granted Critical
Publication of CN116432274B publication Critical patent/CN116432274B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/13—Architectural design, e.g. computer-aided architectural design [CAAD] related to design of buildings, bridges, landscapes, production plants or roads
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/10—Internal combustion engine [ICE] based vehicles
    • Y02T10/40—Engine management systems

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Geometry (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Hardware Design (AREA)
  • Theoretical Computer Science (AREA)
  • Civil Engineering (AREA)
  • Structural Engineering (AREA)
  • Computational Mathematics (AREA)
  • Architecture (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention relates to the technical field of buildings, in particular to a method and a device for optimizing a rod system structure, electronic equipment and a storage medium. Comprising the following steps: generating a node characteristic matrix and an adjacent matrix corresponding to the target building according to the component information and the topological connection relation of the rod system structure corresponding to the target building; inputting the node characteristic matrix and the adjacent matrix into a target structure response proxy model, and outputting a structural mechanical response corresponding to the target building; generating an objective function according to the construction cost of each component; and generating a constraint function according to the structural mechanical response, optimizing the objective function by taking the constraint function as a constraint condition, and outputting objective optimization variables corresponding to each component. The method can reduce the calculation complexity, improve the optimization convergence rate, shorten the optimization time, enlarge the application range, improve the efficiency and ensure the accuracy of the target optimization variables corresponding to each output component.

Description

Rod system structure optimization method and device, electronic equipment and storage medium
Technical Field
The invention relates to the technical field of buildings, in particular to a method and a device for optimizing a rod system structure, electronic equipment and a storage medium.
Background
The pole system structure has good material utilization efficiency, excellent bearing capacity and structural rigidity, and can be widely applied to projects such as frames, large-span space racks, power transmission towers, high-rise building support systems and the like. Therefore, the quick optimization of the system structure system of the rod system is realized, the engineering cost is reduced, and the method has considerable engineering value. In the engineering design optimization stage, engineers hope to reduce the material consumption of the structure as much as possible and reduce the construction cost on the premise of not reducing the stress performance of the structure, ensuring the safety of the structure and meeting the constraints of the appearance and the construction performance of the building.
The existing rod system structure optimization method generally calculates the mechanical response of the structure through finite element software, and adopts a heuristic random optimization algorithm to optimize the rod parameters.
However, the modeling process is low in automation degree, small in application range, long in optimization time and low in efficiency.
Disclosure of Invention
In view of the above, the embodiment of the invention provides a method, a device, an electronic device and a storage medium for optimizing a rod system structure body, which aim to solve the problems that the modeling process of the rod system structure body optimizing method in the prior art is low in automation degree, small in application range, long in optimizing time and low in efficiency.
According to a first aspect, an embodiment of the present invention provides a method for optimizing a rod system structure, including:
generating a node characteristic matrix and an adjacent matrix corresponding to the target building according to the component information and the topological connection relation of the rod system structure corresponding to the target building; the node characteristic matrix is used for representing the characteristics of each component in the target building and the characteristics of the corresponding crossing points of each component; the adjacency matrix is used for representing the connection relation between each component and each intersection point;
inputting the node characteristic matrix and the adjacent matrix into a target structure response proxy model, and outputting a structural mechanical response corresponding to the target building; the structural mechanical response is used for representing stress change conditions of each member, each intersection and the whole target building after a certain load is applied;
generating an objective function according to the construction cost of each component;
and generating a constraint function according to the structural mechanical response, optimizing the objective function by taking the constraint function as a constraint condition, and outputting objective optimization variables corresponding to each component.
According to the method for optimizing the rod system structure, the node characteristic matrix and the adjacent matrix corresponding to the target building are generated according to the component information and the topological connection relation of the rod system structure corresponding to the target building, and the accuracy of the node characteristic matrix and the adjacent matrix corresponding to the generated target building is guaranteed. Then, the node characteristic matrix and the adjacent matrix are input into a target structure response proxy model, and the structural mechanical response corresponding to the target building is output, so that the accuracy of the structural mechanical response corresponding to the output target building is ensured. According to the construction cost of each component, an objective function is generated, and the accuracy of the generated objective function is ensured. And then, generating a constraint function according to the structural mechanical response, optimizing the objective function by taking the constraint function as a constraint condition, outputting the objective optimization variables corresponding to each component, and ensuring the accuracy of the output objective optimization variables corresponding to each component. According to the rod system structure optimization method, the structural mechanical response of each node is calculated through the target structural response agent model instead of the finite element software, and the constraint function is generated according to the structural mechanical response, so that gradient information of the structural mechanical response along with the change of parameters of the rod piece and the cross point can be calculated, selection of an optimization algorithm is not limited, and various optimization algorithms can be selected. And then, generating a constraint function according to the structural mechanics response, optimizing the objective function by taking the constraint function as a constraint condition, and outputting objective optimization variables corresponding to each component. Therefore, the calculation complexity can be reduced, the optimization convergence rate is improved, the optimization time is shortened, the application range is enlarged, the efficiency is improved, and the accuracy of the target optimization variables corresponding to each output component is ensured.
With reference to the first aspect, in a first implementation manner of the first aspect, the component information includes a positional relationship and a structural feature of each component and each intersection, and the generating, according to the component information and the topological connection relationship of the rod system structure corresponding to the target building, a node feature matrix and an adjacent matrix corresponding to the target building includes:
generating a node characteristic matrix according to the position relation and the structural characteristics of each member and each cross point;
an adjacency matrix is generated based on the topological connection between each member and each intersection.
According to the method for optimizing the rod system structure, the component information comprises the position relation and the structural characteristics of each component and each cross point, and the node characteristic matrix is generated according to the position relation and the structural characteristics of each component and each cross point, so that the accuracy of the generated node characteristic matrix is ensured. According to the topological connection relation between each component and each cross point, an adjacent matrix is generated, and the accuracy of the generated adjacent matrix is ensured. And further, the accuracy of structural mechanical response corresponding to the output target building can be ensured by inputting the node characteristic matrix and the adjacent matrix into the target structural response proxy model.
With reference to the first aspect, in a second implementation manner of the first aspect, inputting the node feature matrix and the adjacency matrix into the target structure response proxy model, and outputting a structural mechanical response corresponding to the target building, where the method includes:
inputting the node characteristic matrix and the adjacency matrix into an encoder in a target structure response agent model;
the encoder performs feature recognition and extraction on the node feature matrix and the adjacent matrix, and performs transformation and integration on the extracted features to generate node embedded vectors or/and graph embedded vectors corresponding to the target building;
the node embedded vector or/and the graph embedded vector is input to a decoder in the target structure response proxy model, and the decoder decodes and calculates the node embedded vector or/and the graph embedded vector to output the structural mechanics response.
According to the rod system structure optimization method provided by the embodiment of the invention, the node characteristic matrix and the adjacent matrix are input to the encoder in the target structure response agent model. The encoder performs feature recognition and extraction on the node feature matrix and the adjacent matrix, performs transformation and integration on the extracted features, generates a node embedded vector or/and a graph embedded vector corresponding to the target building, and ensures the accuracy of the node embedded vector or/and the graph embedded vector corresponding to the generated target building. Then, the node embedded vector or/and the graph embedded vector is input to a decoder in the target structure response proxy model, the decoder decodes and calculates the node embedded vector or/and the graph embedded vector, the structural mechanical response is output, and the accuracy of the output structural mechanical response is ensured.
With reference to the second embodiment of the first aspect, in a third embodiment of the first aspect, the encoder includes k convolution layers and a pooling layer, the encoder performs feature recognition and extraction on the node feature matrix and the adjacent matrix, and performs transform integration on the extracted features to generate a node embedded vector or/and a graph embedded vector corresponding to the target building, where the method includes:
each convolution layer transforms and integrates the current node information and the neighbor node information of the previous layer to obtain the current node information of the current layer; the current node information is used for representing current information corresponding to any target node in each component and each cross point; the neighbor node information is used for representing the current information of the neighbor node of the target node;
after transformation integration processing of k convolution layers, obtaining node embedded vectors;
and the pooling layer aggregates the information in the node embedded vector to generate a graph embedded vector.
According to the rod system structure optimization method provided by the embodiment of the invention, the encoder comprises k convolution layers and one pooling layer, each convolution layer transforms and integrates the current node information and the neighbor node information of the previous layer to obtain the current node information of the current layer, and the accuracy of the current node information of the current layer obtained by transforming and integrating the current node information and the neighbor node information of the previous layer by each convolution layer is ensured. After transformation integration processing of k convolution layers, a node embedded vector is obtained, and accuracy of the obtained node embedded vector is guaranteed. The pooling layer aggregates the information in the node embedded vector to generate the graph embedded vector, so that the accuracy of the generated graph embedded vector is ensured, and the accuracy of the structural mechanical response output by the decoder can be further ensured.
With reference to the first aspect, in a fourth implementation manner of the first aspect, generating an objective function according to a construction cost of each component includes:
obtaining corresponding structural characteristics of each component;
calculating the construction cost corresponding to each component according to the corresponding structural characteristics of each component;
building costs corresponding to the components are added to generate an objective function.
According to the method for optimizing the rod system structure, provided by the embodiment of the invention, the structural characteristics corresponding to each component are obtained, the construction cost corresponding to each component is calculated according to the structural characteristics corresponding to each component, and the accuracy of the construction cost corresponding to each component obtained through calculation is ensured. And then, building costs corresponding to the components are added to generate an objective function, so that the accuracy of the objective function is ensured.
With reference to the first aspect, in a fifth implementation manner of the first aspect, optimizing the objective function with the constraint function as a constraint condition, and outputting a target optimization variable corresponding to each component includes:
combining the objective function and the constraint function into a Lagrangian function by utilizing Lagrangian multipliers;
updating the optimized variable sum and the Lagrange multiplier in the objective function by adopting a gradient descent method;
And outputting target optimization variables corresponding to the components until the Lagrangian function converges.
According to the method for optimizing the rod system structure, provided by the embodiment of the invention, the objective function and the constraint function are combined into the Lagrange function by utilizing the Lagrange multiplier, so that the accuracy of the generated Lagrange function is ensured. And then, updating the optimization variable sum and the Lagrange multiplier in the objective function by adopting a gradient descent method, thereby ensuring the accuracy and the rapidity of updating the optimization variable sum and the Lagrange multiplier in the objective function, reducing the calculation complexity, improving the optimization convergence rate and shortening the optimization time. And outputting the target optimization variables corresponding to the components until the Lagrangian function converges, so that the accuracy of the output target optimization variables corresponding to the components is ensured.
With reference to the first aspect, in a sixth implementation manner of the first aspect, the training process of the target structure response proxy model includes:
acquiring a training set, wherein the training set comprises a training node characteristic matrix, a training adjacent matrix and a real structural mechanics response corresponding to a training building;
inputting the training node characteristic matrix corresponding to the training building and the training adjacent matrix into an initial structure response proxy network, and outputting a virtual structure mechanical response corresponding to the training building;
And calculating a loss function based on the real structural mechanical response and the virtual structural mechanical response, and updating parameters of the initial structural response proxy network based on the calculation result to obtain a target structural response proxy model.
According to the method for optimizing the rod system structure, the training set is obtained, the training node characteristic matrix and the training adjacent matrix corresponding to the training building are input into the initial structure response proxy network, the virtual structure mechanical response corresponding to the training building is output, and the accuracy of the virtual structure mechanical response corresponding to the output training building is guaranteed. And then, calculating a loss function based on the real structural mechanical response and the virtual structural mechanical response, and updating parameters of the initial structural response proxy network based on the calculation result to obtain a target structural response proxy model, thereby ensuring the accuracy of the obtained target structural response proxy model. Therefore, the accuracy of the structural mechanical response corresponding to the target building can be ensured by inputting the node characteristic matrix and the adjacent matrix corresponding to the target building into the target structural response proxy model and outputting the node characteristic matrix and the adjacent matrix corresponding to the target building.
According to a second aspect, an embodiment of the present invention further provides a rod system structure optimization apparatus, including:
The first generation module is used for generating a node characteristic matrix and an adjacent matrix corresponding to the target building according to the component information and the topological connection relation of the rod system structure corresponding to the target building; the node characteristic matrix is used for representing the characteristics of each component in the target building and the characteristics of the corresponding crossing points of each component; the adjacency matrix is used for representing the connection relation between each component and each intersection point;
the output module is used for inputting the node characteristic matrix and the adjacent matrix into the target structure response proxy model and outputting the structural mechanical response corresponding to the target building; the structural mechanical response is used for representing stress change conditions of each member, each intersection and the whole target building after a certain load is applied;
the second generation module is used for generating an objective function according to the construction cost of each component;
and the optimization module is used for generating a constraint function according to the structural mechanics response, optimizing the objective function by taking the constraint function as a constraint condition, and outputting objective optimization variables corresponding to each component.
According to the device for optimizing the rod system structure, the node characteristic matrix and the adjacent matrix corresponding to the target building are generated according to the component information and the topological connection relation of the rod system structure corresponding to the target building, and the accuracy of the node characteristic matrix and the adjacent matrix corresponding to the generated target building is guaranteed. Then, the node characteristic matrix and the adjacent matrix are input into a target structure response proxy model, and the structural mechanical response corresponding to the target building is output, so that the accuracy of the structural mechanical response corresponding to the output target building is ensured. According to the construction cost of each component, an objective function is generated, and the accuracy of the generated objective function is ensured. And then, generating a constraint function according to the structural mechanical response, optimizing the objective function by taking the constraint function as a constraint condition, outputting the objective optimization variables corresponding to each component, and ensuring the accuracy of the output objective optimization variables corresponding to each component. According to the rod system structure optimization device, the structural mechanical response of each node is calculated through the target structural response agent model instead of the finite element software, and the constraint function is generated according to the structural mechanical response, so that gradient information of the structural mechanical response along with the change of parameters of the rod piece and the cross point can be calculated, the selection of the optimization algorithm is not limited, and various optimization algorithms can be selected. And then, generating a constraint function according to the structural mechanics response, optimizing the objective function by taking the constraint function as a constraint condition, and outputting objective optimization variables corresponding to each component. Therefore, the calculation complexity can be reduced, the optimization convergence rate is improved, the optimization time is shortened, the application range is enlarged, the efficiency is improved, and the accuracy of the target optimization variables corresponding to each output component is ensured.
According to a third aspect, an embodiment of the present invention provides an electronic device, including a memory and a processor, where the memory and the processor are communicatively connected to each other, and the memory stores computer instructions, and the processor executes the computer instructions, thereby executing the method for optimizing a rod system structure in the first aspect or any implementation manner of the first aspect.
According to a fourth aspect, embodiments of the present invention provide a computer-readable storage medium storing computer instructions for causing a computer to perform the method of optimizing a rod system structure of the first aspect or any one of the embodiments of the first aspect.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are needed in the description of the embodiments or the prior art will be briefly described, and it is obvious that the drawings in the description below are some embodiments of the present invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a method for optimizing a rod system structure provided by an embodiment of the present invention;
FIG. 2 is a flow chart of a method for optimizing a rod system structure provided by another embodiment of the present invention;
FIG. 3 is a schematic view of a frame-shear wall structure to which another embodiment of the present invention is applied;
FIG. 4 is a flow chart of a method for optimizing a rod system structure provided by another embodiment of the present invention;
FIG. 5 is a flow chart of a target structure response proxy model training process provided by another embodiment of the invention;
FIG. 6 is a flow chart of a target structure response proxy model training process provided by another embodiment of the invention;
FIG. 7 is a flow chart of a method of optimizing a rod system structure provided by another embodiment of the present invention;
FIG. 8 is a functional block diagram of a rod system structure optimizing apparatus to which an embodiment of the present invention is applied;
fig. 9 is a schematic diagram of a hardware structure of an electronic device according to an embodiment of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be noted that, the execution body of the method for optimizing a rod system structure body provided by the embodiment of the present application may be a device for optimizing a rod system structure body, and the device for optimizing a rod system structure body may be implemented as part or all of a computer device in a manner of software, hardware or a combination of software and hardware, where the computer device may be a server or a terminal, where the server in the embodiment of the present application may be a server or a server cluster formed by multiple servers, and the terminal in the embodiment of the present application may be other intelligent hardware devices such as a smart phone, a personal computer, a tablet computer, a wearable device, and an intelligent robot. In the following method embodiments, the execution subject is an electronic device.
In one embodiment of the present application, as shown in fig. 1, there is provided a method for optimizing a rod system structure, which is described as an example of application to an electronic device, including the steps of:
and S11, generating a node characteristic matrix and an adjacent matrix corresponding to the target building according to the component information and the topological connection relation of the rod system structure corresponding to the target building.
The node characteristic matrix is used for representing the characteristics of each component in the target building and the characteristics of the corresponding crossing points of each component; the adjacency matrix is used to characterize the connection between each component and each intersection point.
Specifically, the electronic device may receive the component information and the topology connection relationship of the rod system structure corresponding to the target building, which are input by the user, and may also receive the component information and the topology connection relationship of the rod system structure corresponding to the target building, which are sent by other devices, and the electronic device may identify the target building and determine the component information and the topology connection relationship of the rod system structure corresponding to the target building.
After the electronic device obtains the component information and the topological connection relation of the rod system structure corresponding to the target building, the node feature matrix and the adjacent matrix corresponding to the target building can be generated according to the component information and the topological connection relation of the rod system structure corresponding to the target building.
This step will be described in detail below.
S12, inputting the node characteristic matrix and the adjacent matrix into a target structure response proxy model, and outputting structural mechanical response corresponding to the target building.
The structural mechanical response is used for representing stress change conditions of all the components, all the crossing points and the whole target building after the whole target building is subjected to a certain load.
Specifically, after the electronic device generates the node feature matrix and the adjacent matrix corresponding to the target building, the node feature matrix and the adjacent matrix may be input to the target structural response proxy model, where the target structural response proxy model performs feature analysis and feature extraction on the input node feature matrix and adjacent matrix, and outputs structural mechanical response corresponding to the target building according to the extracted features.
This step will be described in detail below.
S13, generating an objective function according to the construction cost of each component.
Specifically, the electronic device may calculate the construction cost corresponding to each component according to the structural feature corresponding to each component. Then, an objective function is generated according to the construction costs corresponding to the respective components.
This step will be described in detail below.
S14, generating a constraint function according to the structural mechanics response, optimizing the objective function by taking the constraint function as a constraint condition, and outputting objective optimization variables corresponding to each component.
Specifically, after generating the objective function, the electronic device may generate the constraint function according to a structural mechanical response corresponding to the target building. Then, the electronic equipment can optimize the objective function by using the constraint function as a constraint condition and utilizing a preset optimization algorithm, and output objective optimization variables corresponding to all the components.
The preset optimization algorithm may be at least one of a gradient descent algorithm, an exponential weighted average algorithm, a momentum gradient descent algorithm, an RMSprop algorithm and an Adam optimization algorithm, and the preset optimization algorithm is not specifically limited in the embodiment of the present application.
According to the method for optimizing the rod system structure, the node characteristic matrix and the adjacent matrix corresponding to the target building are generated according to the component information and the topological connection relation of the rod system structure corresponding to the target building, and the accuracy of the node characteristic matrix and the adjacent matrix corresponding to the generated target building is guaranteed. Then, the node characteristic matrix and the adjacent matrix are input into a target structure response proxy model, and the structural mechanical response corresponding to the target building is output, so that the accuracy of the structural mechanical response corresponding to the output target building is ensured. According to the construction cost of each component, an objective function is generated, and the accuracy of the generated objective function is ensured. And then, generating a constraint function according to the structural mechanical response, optimizing the objective function by taking the constraint function as a constraint condition, outputting the objective optimization variables corresponding to each component, and ensuring the accuracy of the output objective optimization variables corresponding to each component. According to the rod system structure optimization method, the structural mechanical response of each node is calculated through the target structural response agent model instead of the finite element software, and the constraint function is generated according to the structural mechanical response, so that gradient information of the structural mechanical response along with the change of parameters of the rod piece and the cross point can be calculated, selection of an optimization algorithm is not limited, and various optimization algorithms can be selected. And then, generating a constraint function according to the structural mechanics response, optimizing the objective function by taking the constraint function as a constraint condition, and outputting objective optimization variables corresponding to each component. Therefore, the calculation complexity can be reduced, the optimization convergence rate is improved, the optimization time is shortened, the application range is enlarged, the efficiency is improved, and the accuracy of the target optimization variables corresponding to each output component is ensured.
In one embodiment of the present application, as shown in fig. 2, there is provided a method for optimizing a rod system structure, which is described as an example of application to an electronic device, including the steps of:
s21, generating a node characteristic matrix and an adjacent matrix corresponding to the target building according to the component information and the topological connection relation of the rod system structure corresponding to the target building.
The node characteristic matrix is used for representing the characteristics of each component in the target building and the characteristics of the corresponding crossing points of each component; the adjacency matrix is used to characterize the connection between each component and each intersection point.
In an optional embodiment of the present application, the component information includes a positional relationship and a structural feature of each component and each intersection, and the step S21 "generating a node feature matrix and an adjacent matrix corresponding to the target building according to the component information and the topological connection relationship of the rod system structure corresponding to the target building" may include the following steps:
s211, generating a node characteristic matrix according to the position relation and the structural characteristics of each member and each cross point.
Specifically, the electronic device may generate a bipartite graph structure according to the positional relationship and structural features of each member and each intersection, which is specifically as follows:
G＝(V t ,V s1 ,...,V sk ,E) (1)
Where Vt, vs1, …, vsk represent node sets, E represents edge sets, i.e., bar sets; wherein the Vt node set represents the crossing point of the rod in the rod system structure, and each Vt epsilon Vt takes at least the space coordinate as the crossing point characteristic (x, y, z), besides, the load and the boundary condition of the crossing point of the rod can be taken as the crossing point characteristic; vs1, …, vsk represent various classes of levers, each lever having its own unique component characteristics; edges exist only between the Vt node set and the Vs node set.
The electronic device may generate a cross point feature matrix and a skeleton feature matrix from the node set in the bipartite graph structure, and then generate a node feature matrix from the cross point feature matrix and the skeleton feature matrix.
S212, generating an adjacency matrix according to the topological connection relation between each component and each intersection point.
Specifically, after generating the node sets in the bipartite graph structure, the electronic device may assign a cross point number i to each cross point in each Vt set, and according to the topological connection relationship between each member and each cross point, arrange the nodes in all Vt cross point sets in a certain order, assign a node number j to each node, and generate the adjacency matrix. Wherein the adjacency matrix can be used to represent the edge set E in the bipartite graph structure.
Illustratively, assuming the adjacency matrix is a, the element aij=1 represents node v i And v j There are edges between, which means the rod v j Is located at node v i Where aij=0 represents node v i And v j There is no edge between them.
In order to better describe the node characteristic matrix and the adjacent matrix corresponding to the target building described in the embodiment of the application. A simple frame shear wall structure and its corresponding digital representation is shown in fig. 3, which has a total of 8 rod intersections and 3 types of rods: posts, walls and beams (walls can be seen as a special bar with 4 end points). The cross point of the rod piece is characterized by the space coordinates, each type of rod piece is also characterized by the structural characteristics of the column, namely (b, h, rx, ry), namely the reinforcement ratio in the x direction and the reinforcement ratio in the y direction, wherein the cross section is wide and the cross section is high; the structural characteristics of the wall are (t, rx, ry), namely wall thickness, reinforcement ratio in the horizontal direction and reinforcement ratio in the vertical direction; the structural characteristics of the beam are (b, h, rb, rt, P), namely the section width, the section height, the bottom reinforcement ratio, the top reinforcement ratio and the load linear density value.
Based on the positional relationship and structural features of each member and each intersection, a node feature matrix may be generated. An adjacency matrix is then generated based on the topological connection between each member and each intersection.
Wherein the adjacency matrix A is specifically as follows:
wherein, 9 columns in the matrix a represent 9 members corresponding to the frame shear wall structure in fig. 2, and 8 rows in the matrix a represent 8 intersections corresponding to the frame shear wall structure in fig. 2. From the first column next to the matrix a, it is known that the element 1, i.e. column 1, connects the intersection 1 and the intersection 5, i.e. node 1 and nodes 5, … …, the element 9, i.e. beam 9, connects the intersection 5 and the intersection 8, i.e. node 5 and node 8.
S22, inputting the node characteristic matrix and the adjacent matrix into a target structure response proxy model, and outputting structural mechanical response corresponding to the target building.
The structural mechanical response is used for representing stress change conditions of all the components, all the crossing points and the whole target building after the whole target building is subjected to a certain load.
In an alternative embodiment of the present application, the step S22 "inputting the node feature matrix and the adjacency matrix into the target structural response proxy model, and outputting the structural mechanical response corresponding to the target building" may include the following steps:
s221, inputting the node characteristic matrix and the adjacency matrix into an encoder in the target structure response agent model.
In particular, the electronic device may input the node feature matrix and the adjacency matrix to an encoder in the target structure response proxy model.
S222, the encoder performs feature recognition and extraction on the node feature matrix and the adjacent matrix, and performs transformation and integration on the extracted features to generate node embedded vectors or/and graph embedded vectors corresponding to the target building.
In an alternative embodiment of the present application, the encoder includes k convolution layers and a pooling layer, and the step S222 "the encoder performs feature recognition and extraction on the node feature matrix and the adjacent matrix, and performs transform integration on the extracted features to generate a node embedded vector or/and a graph embedded vector corresponding to the target building" may include the following steps:
(1) Each convolution layer transforms and integrates the current node information and the neighbor node information of the previous layer to obtain the current node information of the current layer.
The current node information is used for representing current information corresponding to any target node in each component and each cross point; the neighbor node information is used to characterize current information of neighbor nodes of the target node.
(2) After transformation integration processing of k convolution layers, a node embedded vector is obtained.
(3) And the pooling layer aggregates the information in the node embedded vector to generate a graph embedded vector.
Specifically, each convolution layer in the encoder can transform and integrate the current node information of the previous layer and the information of the neighbor nodes to obtain the current node information of the current layer, and the calculation mode is specifically as follows:
wherein,,an embedded vector representing a k-th layer v node; n (v) represents a set of contiguous nodes of v nodes; gamma ray θ Represents any differentiable function, such as a multi-layer perceptron MLP; psi phi type θ Representing any one of a number of differentiable and permutation independent functions,such as a summing function, an averaging function, or a maximum function. Taking a model of GNN-isomorphic neural network (Graph Isomorphism Network, GIN), its function is chosen as ψ θ ＝∑(·)，γ θ ＝MLP (k) (. Cndot.) then the kth convolutional layer can be written as:
wherein, MLP (k) Representing a multi-layer perceptron defining the number of hidden layers and neurons, MLP (k) The activation function in (a) adopts a ReLU function.
It should be noted that the input variables of the first convolutional layerThe node characteristic matrix of the current node is the node characteristic matrix. Because node characteristics of different node sets in the bipartite graph are different, all node characteristic dimensions are unified by first performing a transformation, such as MLP.
After passing through k convolution layers, a node embedded vector is obtained Sometimes, due to the problem, a pooling layer POOL is passed (·) The information in the node embedded vector is aggregated into a graph embedded vector or other intermediate level embedded vector. The pooling layer can select various operations such as summation, maximum value taking, average value taking and the like according to different problems.
It should be noted that, according to the requirements of the problem, the encoder may take different forms, the most common form is a multi-layer perceptron MLP, and the form of the encoder is not particularly limited in the embodiment of the present application.
S223, inputting the node embedded vector or/and the graph embedded vector into a decoder in the target structure response proxy model, and decoding and calculating the node embedded vector or/and the graph embedded vector by the decoder to output the structural mechanics response.
Specifically, a decoder in the target structural response proxy model may convert the node embedded vector or graph embedded vector obtained by the encoder into a final structural mechanical response.
The structural mechanical response may be, for example, the inter-layer displacement angle of the layers in the framework structure, or the displacement of the nodes in the space truss structure.
S23, generating an objective function according to the construction cost of each component.
For this step, please refer to the description of S13 in fig. 1, and a detailed description is omitted here.
S24, generating a constraint function according to the structural mechanics response, optimizing the objective function by taking the constraint function as a constraint condition, and outputting objective optimization variables corresponding to each component.
For this step, please refer to the description of S14 in fig. 1, and a detailed description is omitted here.
According to the method for optimizing the rod system structure, the component information comprises the position relation and the structural characteristics of each component and each cross point, and the node characteristic matrix is generated according to the position relation and the structural characteristics of each component and each cross point, so that the accuracy of the generated node characteristic matrix is ensured. According to the topological connection relation between each component and each cross point, an adjacent matrix is generated, and the accuracy of the generated adjacent matrix is ensured. And further, the accuracy of structural mechanical response corresponding to the output target building can be ensured by inputting the node characteristic matrix and the adjacent matrix into the target structural response proxy model.
The node feature matrix and the adjacency matrix are then input to an encoder in the target structure response proxy model. The encoder comprises k convolution layers and a pooling layer, each convolution layer transforms and integrates the current node information and the neighbor node information of the previous layer to obtain the current node information of the current layer, and the accuracy of the current node information of the current layer obtained after each convolution layer transforms and integrates the current node information and the neighbor node information of the previous layer is guaranteed. After transformation integration processing of k convolution layers, a node embedded vector is obtained, and accuracy of the obtained node embedded vector is guaranteed. The pooling layer aggregates the information in the node embedded vector to generate the graph embedded vector, so that the accuracy of the generated graph embedded vector is ensured, and the accuracy of the structural mechanical response output by the decoder can be further ensured. Then, the node embedded vector or/and the graph embedded vector is input to a decoder in the target structure response proxy model, the decoder decodes and calculates the node embedded vector or/and the graph embedded vector, the structural mechanical response is output, and the accuracy of the output structural mechanical response is ensured.
In one embodiment of the present application, as shown in fig. 4, there is provided a method for optimizing a rod system structure, which is described as an example of application to an electronic device, comprising the steps of:
s31, generating a node characteristic matrix and an adjacent matrix corresponding to the target building according to the component information and the topological connection relation of the rod system structure corresponding to the target building.
The node characteristic matrix is used for representing the characteristics of each component in the target building and the characteristics of the corresponding crossing points of each component; the adjacency matrix is used to characterize the connection between each component and each intersection point.
For this step, please refer to the description of S21 in fig. 2, and a detailed description is omitted here.
S32, inputting the node characteristic matrix and the adjacent matrix into a target structure response proxy model, and outputting structural mechanical response corresponding to the target building.
The structural mechanical response is used for representing stress change conditions of all the components, all the crossing points and the whole target building after the whole target building is subjected to a certain load.
For this step, please refer to fig. 2 for description of S22, and detailed description thereof is omitted herein.
S33, generating an objective function according to the construction cost of each component.
In an alternative embodiment of the present application, the step S33 "generating the objective function according to the construction cost of each component" may include the steps of:
S331, obtaining the corresponding structural characteristics of each component.
Specifically, the electronic device may receive the structural features corresponding to each component input by the user, or may also receive the structural features corresponding to each component sent by other devices, and the electronic device may further identify the target building and determine the structural features corresponding to each component in the target building.
The structural features of the column are (b, h, rx, ry), i.e. cross-section width, cross-section height, reinforcement ratio in x-direction, reinforcement ratio in y-direction; the structural characteristics of the wall are (t, rx, ry), namely wall thickness, reinforcement ratio in the horizontal direction and reinforcement ratio in the vertical direction; the structural characteristics of the beam are (b, h, rb, rt, P), namely the section width, the section height, the bottom reinforcement ratio, the top reinforcement ratio and the load linear density value.
And S332, calculating the construction cost corresponding to each component according to the structural characteristics corresponding to each component.
Specifically, the electronic device may calculate the construction cost corresponding to each component according to the structural feature corresponding to each component.
S333, building costs corresponding to the components are added to generate an objective function.
Specifically, after the electronic device calculates the building costs corresponding to each component, the building costs corresponding to each component may be added to generate the objective function.
By way of example, the objective function may be specifically as follows:
f(x)＝∑ v＝column b v h v L v [λ 1 +2λ 2 η(ρ x +ρ y )+∑ v＝beam b v h v L v [λ 1 +2λ 2 η(ρ b +ρ t )]
(5)
wherein v represents Liang Huozhe column, b v ，hv，L v The width, the height and the length of the section are respectively; for the column ρ x The reinforcement ratio in the x direction is ρ y For the beam, ρ is the reinforcement ratio in the y direction b Rate, ρ of reinforcement for bottom t Is a roofA portion reinforcement ratio; lambda (lambda) 1 Price of concrete per unit volume lambda 2 Is the price of the steel bar per unit weight, and eta is the density of the steel bar.
S34, generating a constraint function according to the structural mechanics response, optimizing the objective function by taking the constraint function as a constraint condition, and outputting objective optimization variables corresponding to each component.
In an optional embodiment of the present application, the step S34 "uses the constraint function as a constraint condition, optimizes the objective function, and outputs the objective optimization variables corresponding to each component" may include the following steps:
s341, combining the objective function and the constraint function into a Lagrangian function by utilizing Lagrangian multipliers.
S342, updating the optimized variable sum and the Lagrangian multiplier in the objective function by adopting a gradient descent method.
S343, outputting target optimization variables corresponding to the components until the Lagrangian function converges.
Specifically, after the electronic device further determines the structural mechanical response corresponding to the target building, a constraint function may be generated according to the structural mechanical response.
By way of example, the usual constraint function of a reinforced concrete frame structure is the inter-layer displacement angle constraint, the inter-layer displacement angle of each layer in rare earthquakes is not more than 0.02, and the constraint function is rewritten into the equality constraint, specifically as follows:
in θ i Is the interlayer displacement angle of the i-th layer.
The optimization problem can then be expressed as:
s.t.h(x)＝0 (7)
wherein f (x) is an objective function, and h (x) is a constraint condition.
The electronic device can solve the problem of optimization with constraint by adopting a Lagrange multiplier method, and combines an objective function and a constraint function into a Lagrange function by utilizing the Lagrange multiplier, and the method is specifically as follows:
L(x，λ)＝f(x)-λh(x) (8)
then, the optimization variable x and the lagrangian multiplier λ are updated using a gradient descent method:
λ′＝λ+λh(x) (9)
where β and γ are both predetermined learning rates.
The electronic device can continuously update the optimization variable x and the Lagrangian multiplier lambda until the Lagrangian function converges, at this time, the optimal objective function value meeting the constraint function is considered to be found, and the objective optimization variable corresponding to each component is output.
According to the method for optimizing the rod system structure, provided by the embodiment of the application, the structural characteristics corresponding to each component are obtained, the construction cost corresponding to each component is calculated according to the structural characteristics corresponding to each component, and the accuracy of the construction cost corresponding to each component obtained through calculation is ensured. And then, building costs corresponding to the components are added to generate an objective function, so that the accuracy of the objective function is ensured.
And then, generating a constraint function according to the structural mechanical response, and combining the objective function and the constraint function into a Lagrange function by utilizing a Lagrange multiplier, so that the accuracy of the generated Lagrange function is ensured. And then, updating the optimization variable sum and the Lagrange multiplier in the objective function by adopting a gradient descent method, thereby ensuring the accuracy and the rapidity of updating the optimization variable sum and the Lagrange multiplier in the objective function, reducing the calculation complexity, improving the optimization convergence rate and shortening the optimization time. And outputting the target optimization variables corresponding to the components until the Lagrangian function converges, so that the accuracy of the output target optimization variables corresponding to the components is ensured.
In an alternative embodiment of the present application, as shown in fig. 5, the training process of the target structure response proxy model includes:
S41, acquiring a training set.
The training set comprises a training node characteristic matrix, a training adjacent matrix and a real structural mechanical response corresponding to the training building. The real structural mechanical response corresponding to the training building can be obtained through finite element calculation or experimental, and the mode of obtaining the real structural mechanical response corresponding to the training building is not particularly limited in the embodiment of the application.
Specifically, the electronic device may receive the acquired training set input by the user, or may receive the acquired training set sent by other devices.
S42, inputting the training node characteristic matrix and the training adjacent matrix corresponding to the training building into the initial structure response proxy network, and outputting the virtual structure mechanical response corresponding to the training building.
Specifically, the electronic device may input a training node feature matrix and a training adjacency matrix corresponding to the training building to the initial structural response proxy network. Each convolution layer in the encoder in the initial structure response proxy network can transform and integrate the current node information of the previous layer and the information of the neighbor nodes to obtain the current node information of the current layer, and the calculation mode is as follows:
Wherein,,an embedded vector representing a k-th layer v node; n (v) represents a set of contiguous nodes of v nodes; gamma ray θ Represents any differentiable function, such as a multi-layer perceptron MLP; psi phi type θ Representing any one of a fair, permutation-independent function, such as a summation function, an averaging function, or a maximum function. By GA model of NN-isomorphic neural network (Graph Isomorphism Network, GIN) is exemplified, the function of which is selected as ψ θ ＝∑(·)，γ θ ＝MLP (k) (. Cndot.) then the kth convolutional layer can be written as:
wherein, MLP (k) Representing a multi-layer perceptron defining the number of hidden layers and neurons, MLP (k) The activation function in (a) adopts a ReLU function.
It should be noted that the input variables of the first convolutional layerThe node characteristic matrix of the current node is the node characteristic matrix. Because node characteristics of different node sets in the bipartite graph are different, all node characteristic dimensions are unified by first performing a transformation, such as MLP.
After passing through k convolution layers, a node embedded vector is obtainedFor the problem, the information in the node embedded vector is aggregated into a graph embedded vector or other intermediate level embedded vector through a pooling layer POOL (). The pooling layer can select various operations such as summation, maximum value taking, average value taking and the like according to different problems.
Then, the training node feature matrix and the node embedded vector and/or the embedded vector corresponding to the training adjacent matrix are input to a decoder. The decoder may convert the training node feature matrix and the node embedded vectors and embedded vectors corresponding to the training adjacency matrix into a final virtual structure mechanical response.
S43, calculating a loss function based on the real structural mechanical response and the virtual structural mechanical response, and updating parameters of the initial structural response proxy network based on the calculation result to obtain a target structural response proxy model.
Specifically, the electronic device may calculate a loss function based on the real structural mechanical response and the virtual structural mechanical response, and update parameters of the initial structural response proxy network based on the calculation result, so as to obtain the target structural response proxy model.
By way of example, the training process of the target structure response proxy model may be as shown in fig. 6. And inputting the training node characteristic matrix and the training adjacent matrix corresponding to the training building into an encoder in the initial structure response proxy network, and outputting the node embedded vector and/or the embedded vector corresponding to the training node characteristic matrix and the training adjacent matrix. The decoder may then convert the training node feature matrix and the node embedded vectors and embedded vectors corresponding to the training adjacency matrix into a final virtual structural mechanical response. The electronic equipment can calculate a loss function based on the real structural mechanical response and the virtual structural mechanical response, and perform parameter feedback on the initial structural response proxy network based on a calculation result to obtain a target structural response proxy model.
According to the method for optimizing the rod system structure, the training set is obtained, the training node characteristic matrix and the training adjacent matrix corresponding to the training building are input into the initial structure response proxy network, the virtual structure mechanical response corresponding to the training building is output, and the accuracy of the virtual structure mechanical response corresponding to the output training building is guaranteed. And then, calculating a loss function based on the real structural mechanical response and the virtual structural mechanical response, and updating parameters of the initial structural response proxy network based on the calculation result to obtain a target structural response proxy model, thereby ensuring the accuracy of the obtained target structural response proxy model. Therefore, the accuracy of the structural mechanical response corresponding to the target building can be ensured by inputting the node characteristic matrix and the adjacent matrix corresponding to the target building into the target structural response proxy model and outputting the node characteristic matrix and the adjacent matrix corresponding to the target building.
In order to better describe the method for optimizing the rod system structure provided by the embodiment of the application, as shown in fig. 7, a flowchart of the method for optimizing the rod system structure is provided. Specifically, the electronic device may acquire a rod optimization variable and a rod basic variable corresponding to the target building, and then generate a node feature matrix and an adjacency matrix corresponding to the target building according to the rod optimization variable and the rod basic variable. The electronic equipment inputs the node characteristic matrix and the adjacency matrix into a target structure response proxy model, and outputs structural mechanical responses corresponding to all nodes and rods in the target building. And then, calculating a constraint method according to structural mechanical responses corresponding to the nodes and the rods, and generating an objective function according to the optimization variables of the rods. And combining the objective function and the constraint function into a Lagrange function by utilizing the Lagrange multiplier, updating the optimization variables in the objective function and the Lagrange multiplier by adopting a gradient descent method until the Lagrange function converges, and outputting the objective optimization variables corresponding to each component.
It should be understood that, although the steps in the flowcharts of fig. 1, 2, and 4-5 are shown in order as indicated by the arrows, these steps are not necessarily performed in order as indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps of fig. 1, 2, and 4-5 may include multiple steps or stages that are not necessarily performed at the same time, but may be performed at different times, nor does the order in which the steps or stages are performed necessarily occur sequentially, but may be performed alternately or alternately with other steps or at least a portion of the steps or stages in other steps.
As shown in fig. 8, the present embodiment provides a rod system structure optimizing apparatus including:
a first generating module 51, configured to generate a node feature matrix and an adjacent matrix corresponding to the target building according to the component information and the topological connection relationship of the rod system structure corresponding to the target building; the node characteristic matrix is used for representing the characteristics of each component in the target building and the characteristics of the corresponding crossing points of each component; the adjacency matrix is used for representing the connection relation between each component and each intersection point;
The output module 52 is configured to input the node feature matrix and the adjacency matrix to a target structure response proxy model, and output a structural mechanical response corresponding to the target building; the structural mechanical response is used for representing stress change conditions of each member, each intersection and the whole target building after a certain load is applied;
a second generation module 53 for generating an objective function according to the construction costs of the respective components;
and the optimization module 54 is configured to generate a constraint function according to the structural mechanical response, optimize the objective function with the constraint function as a constraint condition, and output objective optimization variables corresponding to each component.
In one embodiment of the present application, the component information includes a positional relationship and a structural feature of each component and each intersection, and the first generating module 51 is specifically configured to generate a node feature matrix according to the positional relationship and the structural feature of each component and each intersection; an adjacency matrix is generated based on the topological connection between each member and each intersection.
In one embodiment of the present application, the output module 52 is specifically configured to input the node feature matrix and the adjacency matrix into the encoder in the target structure response proxy model; the encoder performs feature recognition and extraction on the node feature matrix and the adjacent matrix, and performs transformation and integration on the extracted features to generate node embedded vectors or/and graph embedded vectors corresponding to the target building; the node embedded vector or/and the graph embedded vector is input to a decoder in the target structure response proxy model, and the decoder decodes and calculates the node embedded vector or/and the graph embedded vector to output the structural mechanics response.
In one embodiment of the present application, the encoder includes k convolution layers and a pooling layer, and the output module 52 is specifically configured to transform and integrate the current node information and the neighbor node information of the previous layer by using each convolution layer to obtain the current node information of the current layer; the current node information is used for representing current information corresponding to any target node in each component and each cross point; the neighbor node information is used for representing the current information of the neighbor node of the target node; after transformation integration processing of k convolution layers, obtaining node embedded vectors; and the pooling layer aggregates the information in the node embedded vector to generate a graph embedded vector.
In one embodiment of the present application, the second generating module 53 is specifically configured to obtain structural features corresponding to each component; calculating the construction cost corresponding to each component according to the corresponding structural characteristics of each component; building costs corresponding to the components are added to generate an objective function.
In one embodiment of the present application, the optimization module 54 is specifically configured to combine the objective function and the constraint function into a lagrangian function by using lagrangian multipliers; updating the optimized variable sum and the Lagrange multiplier in the objective function by adopting a gradient descent method; and outputting target optimization variables corresponding to the components until the Lagrangian function converges.
In one embodiment of the present application, the output module 52 is specifically configured to obtain a training set, where the training set includes a training node feature matrix, a training adjacent matrix, and a real structural mechanical response corresponding to a training building; inputting the training node characteristic matrix corresponding to the training building and the training adjacent matrix into an initial structure response proxy network, and outputting a virtual structure mechanical response corresponding to the training building; and calculating a loss function based on the real structural mechanical response and the virtual structural mechanical response, and updating parameters of the initial structural response proxy network based on the calculation result to obtain a target structural response proxy model.
The specific limitations and advantages of the rod system structure optimization device can be found in the above limitations of the rod system structure optimization method, and are not described in detail herein. The various modules in the above described pole structure optimization device may be implemented in whole or in part by software, hardware, and combinations thereof. The above modules may be embedded in hardware or independent of a processor in the electronic device, or may be stored in software in a memory in the electronic device, so that the processor may call and execute operations corresponding to the above modules.
The embodiment of the invention also provides electronic equipment, which is provided with the rod system structure optimizing device shown in the figure 8.
Fig. 9 is a schematic structural diagram of an electronic device according to an alternative embodiment of the present invention, as shown in fig. 9, where the electronic device may include: at least one processor 61, such as a CPU (Central Processing Unit ), at least one communication interface 63, a memory 64, at least one communication bus 62. Wherein the communication bus 62 is used to enable connected communication between these components. The communication interface 63 may include a Display screen (Display) and a Keyboard (Keyboard), and the optional communication interface 63 may further include a standard wired interface and a wireless interface. The memory 64 may be a high-speed RAM memory (Random Access Memory, volatile random access memory) or a non-volatile memory (non-volatile memory), such as at least one disk memory. The memory 64 may also optionally be at least one storage device located remotely from the aforementioned processor 61. Where the processor 61 may be a device as described in connection with fig. 8, the memory 64 stores an application program, and the processor 61 invokes the program code stored in the memory 64 for performing any of the method steps described above.
The communication bus 62 may be a peripheral component interconnect standard (peripheral component interconnect, PCI) bus or an extended industry standard architecture (extended industry standard architecture, EISA) bus, among others. The communication bus 62 may be divided into an address bus, a data bus, a control bus, and the like. For ease of illustration, only one thick line is shown in fig. 9, but not only one bus or one type of bus.
Wherein the memory 64 may include volatile memory (English) such as random-access memory (RAM); the memory may also include a nonvolatile memory (english: non-volatile memory), such as a flash memory (english: flash memory), a hard disk (english: hard disk drive, abbreviated as HDD) or a solid state disk (english: solid-state drive, abbreviated as SSD); memory 64 may also include a combination of the types of memory described above.
The processor 61 may be a central processor (English: central processing unit, abbreviated: CPU), a network processor (English: network processor, abbreviated: NP) or a combination of CPU and NP.
The processor 61 may further include a hardware chip, among others. The hardware chip may be an application-specific integrated circuit (ASIC), a Programmable Logic Device (PLD), or a combination thereof (English: programmable logic device). The PLD may be a complex programmable logic device (English: complex programmable logic device, abbreviated: CPLD), a field programmable gate array (English: field-programmable gate array, abbreviated: FPGA), a general-purpose array logic (English: generic array logic, abbreviated: GAL), or any combination thereof.
Optionally, the memory 64 is also used to store program instructions. Processor 61 may invoke program instructions to implement the method of optimizing the rod system structure as shown in the embodiments of fig. 1, 2 and 4-5 of the present application.
The embodiment of the application also provides a non-transitory computer storage medium, which stores computer executable instructions, and the computer executable instructions can execute the rod system structure optimization method in any of the method embodiments. Wherein the storage medium may be a magnetic Disk, an optical Disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a Flash Memory (Flash Memory), a Hard Disk (HDD), or a Solid State Drive (SSD); the storage medium may also comprise a combination of memories of the kind described above.
Although embodiments of the present application have been described in connection with the accompanying drawings, various modifications and variations may be made by those skilled in the art without departing from the spirit and scope of the application, and such modifications and variations fall within the scope of the application as defined by the appended claims.

Claims (10)

1. A method of optimizing a rod system structure, comprising:
Generating a node characteristic matrix and an adjacent matrix corresponding to a target building according to component information and topological connection relation of a rod system structure corresponding to the target building; the node characteristic matrix is used for representing the characteristics of each component in the target building and the characteristics of the corresponding crossing point of each component; the adjacency matrix is used for representing the connection relation between each member and each intersection point;
inputting the node characteristic matrix and the adjacency matrix into a target structure response proxy model, and outputting a structural mechanical response corresponding to the target building; the structural mechanical response is used for representing stress change conditions of the components, the crossing points and the whole target building after a certain load is applied;
generating an objective function according to the construction cost of each component;
and generating a constraint function according to the structural mechanics response, optimizing the objective function by taking the constraint function as a constraint condition, and outputting objective optimization variables corresponding to the components.
2. The method according to claim 1, wherein the component information includes a positional relationship and a structural feature of each of the components and each of the intersections, and the generating the node feature matrix and the adjacent matrix corresponding to the target building according to the component information and the topological connection relationship of the rod system structure corresponding to the target building includes:
Generating the node characteristic matrix according to the position relation and the structural characteristics of each member and each intersection point;
the adjacency matrix is generated according to the topological connection relation between each component and each intersection point.
3. The method of claim 1, wherein inputting the node feature matrix and the adjacency matrix into a target structural response proxy model, outputting a structural mechanical response corresponding to the target building, comprises:
inputting the node feature matrix and the adjacency matrix to an encoder in the target structure response proxy model;
the encoder performs feature recognition and extraction on the node feature matrix and the adjacent matrix, and performs transformation integration on the extracted features to generate node embedded vectors or/and graph embedded vectors corresponding to the target building;
and inputting the node embedded vector or/and the graph embedded vector to a decoder in the target structure response proxy model, and decoding and calculating the node embedded vector or/and the graph embedded vector by the decoder to output the structural mechanics response.
4. A method according to claim 3, wherein the encoder comprises k convolution layers and a pooling layer, the encoder performs feature recognition and extraction on the node feature matrix and the adjacency matrix, performs transform integration on the extracted features, and generates a node embedding vector or/and a graph embedding vector corresponding to the target building, and the method comprises:
Each convolution layer transforms and integrates the current node information and the neighbor node information of the previous layer to obtain the current node information of the current layer; the current node information is used for representing current information corresponding to any target node in each component and each intersection; the neighbor node information is used for representing the current information of the neighbor node of the target node;
after transformation integration processing of k convolution layers, obtaining the node embedded vector;
and the pooling layer aggregates the information in the node embedded vector to generate the graph embedded vector.
5. The method of claim 1, wherein generating an objective function based on construction costs of each of the components comprises:
obtaining corresponding structural characteristics of each component;
calculating the construction cost corresponding to each component according to the structural characteristics corresponding to each component;
and adding the construction costs corresponding to the components to generate the objective function.
6. The method according to claim 1, wherein optimizing the objective function with the constraint function as a constraint condition, and outputting the objective optimization variable corresponding to each component includes:
Combining the objective function and the constraint function into a Lagrangian function by using Lagrangian multipliers;
updating the optimized variable sum and the Lagrangian multiplier in the objective function by adopting a gradient descent method;
and outputting target optimization variables corresponding to the components until the Lagrangian function converges.
7. The method of claim 1, wherein the target structure is responsive to a training process of a proxy model, comprising:
acquiring a training set, wherein the training set comprises a training node characteristic matrix, a training adjacent matrix and a real structural mechanics response corresponding to a training building;
inputting the training node characteristic matrix and the training adjacent matrix corresponding to the training building into an initial structure response proxy network, and outputting a virtual structure mechanical response corresponding to the training building;
and calculating a loss function based on the real structural mechanical response and the virtual structural mechanical response, and updating parameters of the initial structural response proxy network based on a calculation result to obtain the target structural response proxy model.
8. A bar system structure optimizing apparatus, comprising:
The first generation module is used for generating a node characteristic matrix and an adjacent matrix corresponding to a target building according to the component information and the topological connection relation of the rod system structure corresponding to the target building; the node characteristic matrix is used for representing the characteristics of each component in the target building and the characteristics of the corresponding crossing point of each component; the adjacency matrix is used for representing the connection relation between each member and each intersection point;
the output module is used for inputting the node characteristic matrix and the adjacency matrix into a target structure response proxy model and outputting a structural mechanics response corresponding to the target building; the structural mechanical response is used for representing stress change conditions of the components, the crossing points and the whole target building after a certain load is applied;
the second generation module is used for generating an objective function according to the construction cost of each component;
and the optimization module is used for generating a constraint function according to the structural mechanics response, optimizing the objective function by taking the constraint function as a constraint condition, and outputting objective optimization variables corresponding to the components.
9. An electronic device comprising a memory having stored therein computer instructions and a processor that, upon execution of the computer instructions, performs the method of rod system structure optimization of any one of claims 1-7.
10. A computer readable storage medium storing computer instructions for causing a computer to perform the rod system structure optimization method of any one of claims 1-7.
CN202310208411.9A 2023-02-27 2023-02-27 Rod system structure optimization method and device, electronic equipment and storage medium Active CN116432274B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310208411.9A CN116432274B (en) 2023-02-27 2023-02-27 Rod system structure optimization method and device, electronic equipment and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310208411.9A CN116432274B (en) 2023-02-27 2023-02-27 Rod system structure optimization method and device, electronic equipment and storage medium

Publications (2)

Publication Number Publication Date
CN116432274A CN116432274A (en) 2023-07-14
CN116432274B true CN116432274B (en) 2023-09-26

Family

ID=87080514

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310208411.9A Active CN116432274B (en) 2023-02-27 2023-02-27 Rod system structure optimization method and device, electronic equipment and storage medium

Country Status (1)

Country Link
CN (1) CN116432274B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117612145B (en) * 2023-12-15 2024-06-21 上海青翼工业软件有限公司 Automatic part machining method and device, computer equipment and storage medium
CN118036839A (en) * 2024-01-15 2024-05-14 北京国联视讯信息技术股份有限公司 Big data processing method and system applied to intelligent logistics scheduling

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110688795A (en) * 2019-09-25 2020-01-14 国网湖南省电力有限公司 Transformer box damping vibration attenuation method, system and medium based on topology optimization
CN110941887A (en) * 2019-12-13 2020-03-31 杭州昕华信息科技有限公司 Base station layout method, device, medium and equipment
CN112686971A (en) * 2020-12-29 2021-04-20 博锐尚格科技股份有限公司 Method and system for undirected graph orientation of building system relation topology
CN113032889A (en) * 2021-05-31 2021-06-25 北京盈建科软件股份有限公司 Method and device for splicing foundation structure and superstructure into combined building model
CN115659758A (en) * 2022-11-09 2023-01-31 中建三局第一建设工程有限责任公司 Shield tunnel rock-soil parameter inversion and tunneling parameter optimization method based on approximate model

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP4036822A1 (en) * 2021-01-28 2022-08-03 Tata Consultancy Services Limited Method and system for graph signal processing based energy modelling and forecasting

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110688795A (en) * 2019-09-25 2020-01-14 国网湖南省电力有限公司 Transformer box damping vibration attenuation method, system and medium based on topology optimization
CN110941887A (en) * 2019-12-13 2020-03-31 杭州昕华信息科技有限公司 Base station layout method, device, medium and equipment
CN112686971A (en) * 2020-12-29 2021-04-20 博锐尚格科技股份有限公司 Method and system for undirected graph orientation of building system relation topology
CN113032889A (en) * 2021-05-31 2021-06-25 北京盈建科软件股份有限公司 Method and device for splicing foundation structure and superstructure into combined building model
CN115659758A (en) * 2022-11-09 2023-01-31 中建三局第一建设工程有限责任公司 Shield tunnel rock-soil parameter inversion and tunneling parameter optimization method based on approximate model

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
新型盒式模块化建筑钢结构 体系开发、力学性能及设计研究;冯超;中国优秀硕士学位论文全文库;摘要 *

Also Published As

Publication number Publication date
CN116432274A (en) 2023-07-14

Similar Documents

Publication Publication Date Title
CN116432274B (en) Rod system structure optimization method and device, electronic equipment and storage medium
Ergen et al. Online training of LSTM networks in distributed systems for variable length data sequences
CN111027772B (en) Multi-factor short-term load prediction method based on PCA-DBILSTM
CN113627529B (en) Air quality prediction method and device, electronic equipment and storage medium
CN111428854A (en) Structure searching method and structure searching device
WO2015193894A1 (en) Method and system for determining a configuration of a model having a collection of entities and satisfying a set of constraints
CN114912578A (en) Training method and device of structure response prediction model and computer equipment
Li et al. Stg-mamba: Spatial-temporal graph learning via selective state space model
CN115206457A (en) Three-dimensional molecular structure generation method, device, equipment and storage medium
CN115222046A (en) Neural network structure searching method and device, electronic equipment and storage medium
KR20220042315A (en) Method and apparatus for predicting traffic data and electronic device
CN117952272A (en) Digital service network-oriented flow prediction model training method, device and equipment
US20230342626A1 (en) Model processing method and related apparatus
Sorjamaa et al. Long-term prediction of time series using NNE-based projection and OP-ELM
CN117389780A (en) Converter station fault analysis method, device, computer equipment and storage medium
CN116992607A (en) Structural topology optimization method, system and device
US20210365828A1 (en) Multi-pass system for emulating sampling of a plurality of qubits and methods for use therewith
WO2020054402A1 (en) Neural network processing device, computer program, neural network manufacturing method, neural network data manufacturing method, neural network use device, and neural network downscaling method
CN116305461B (en) Structure response calculation method, device, electronic equipment and storage medium
CN116305995B (en) Nonlinear analysis method, nonlinear analysis device, nonlinear analysis equipment and nonlinear analysis medium of structural system
CN115859815B (en) Short-term adjustable load prediction method and system based on SA-TCN model
US20240028783A1 (en) Automated design of architectural structures for fabrication with standard components
CN114201822A (en) Portal steel frame modeling method and device, computer equipment and readable storage medium
CN118504151A (en) Structural mode calculation method based on embedded physical information graph neural network
CN118036544A (en) Global layout method, system, equipment and medium based on RSMT topology

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant