New deep-learning-based intelligent generation method for yurt framework structure

Abstract

Disclosed in the present invention is a new deep-learning-based intelligent generation method for a yurt framework structure. The method comprises the following steps: acquiring data of a three-dimensional model of a yurt framework structure, performing topological transformation between the yurt framework structure and prefabricated components, constructing a framework structure generation model based on a graph neural network, generating a yurt framework structure model and matching and generating prefabricated components, testing the yurt framework structure model for stablilty, and outputting component drawings. The advantages lie in: realizing the automatic extraction of topological relationships among Khana components, Uni components and Toono components of the framework structure and among various components forming same; reflecting the connection relationships between the components; reducing the workload of identifying the connection relationships between the components; improving the working efficiency of architects; improving the intelligence and automation levels of the yurt framework structure and the accuracy of generating standardized prefabricated components; and reducing the difficulty of purchasing, producing and assembling of prefabricated components.

Description

New type of yurt skeleton structure intelligent generation method based on deep learning TECHNICAL FIELD

[0001] The application belongs to the technical field of building styles, and particularly relates to a new type of yurt skeleton structure intelligent generation method based on deep learning. BACKGROUND

[0002] Intelligent generation of building styles is one of important contents of intelligent construction, and is an important means to improve building design efficiency. Application of artificial intelligence technology can not only learn a large number of case samples, but also generate multiple design schemes for architects to compare and select in a short time, thereby accelerating the design process. At present, obtaining and organizing a large amount of building style data and performing effective learning and generation are important challenges for building style generation. The yurt is a special building, which is quite different from modern buildings and Chinese traditional wooden buildings in terms of component materials, component styles and skeleton structure systems. Traditional manual design methods consume a large amount of time and effort, and there is currently a lack of building style intelligent generation method for yurts, which cannot meet the needs of efficiently designing yurt buildings. TECHNICAL PROBLEM

[0003] In view of the deficiencies of the prior art, the application provides a new type of yurt skeleton structure intelligent generation method based on deep learning, which solves the problems of time-consuming and labor-consuming in traditional Mongolian building design, and provides a solution for intelligent generation of Mongolian building styles. TECHNICAL SOLUTION

[0004] The new type of yurt skeleton structure intelligent generation method based on deep learning comprises the following steps:

[0005] Step (1), collecting data of a three-dimensional model of a yurt building skeleton structure,

[0006] An object measurement of the skeleton of the yurt is performed using a laser scanning device, point cloud data of the yurt skeleton structure is collected, the point cloud data is classified through a point cloud classification algorithm to obtain point cloud data of a hana component, a nao component and a uni component, the point cloud data is input into a three-dimensional modeling software to generate a three-dimensional model of the yurt skeleton structure composed of the hana, the nao and the uni, and a three-dimensional model library of the yurt skeleton structure is constructed;

[0007] Step (2), topological conversion of the Mongolian building skeleton structure and the prefabricated component,

[0008] The three-dimensional shape attributes of the Hanan component, the Tou Nao component, and the Uni component in step (1) are calculated, including length, width, height, and cross-sectional shape, and a three-dimensional shape database of the components is constructed; the positions where each component is connected are identified and extracted as connection points, the three-dimensional data of the skeleton structure is converted into topological data according to the connection relationship between the components and the connection points, and the connection points are connected with the corresponding component data in the three-dimensional shape database of the components, to establish a sample topological data set, wherein the training data set accounts for 80%, and the test data set accounts for 20%;

[0009] Step (3), a skeleton structure generation model based on a graph neural network is constructed,

[0010] The training data set in step (2) is input into a deep learning workstation carrying a graph neural network deep learning model, the rules of the structural skeleton topological network are extracted and learned, and a Mongolian yurt skeleton structure generation model is constructed;

[0011] (4) Generation of a Mongolian architectural structure style and matching of prefabricated components,

[0012] According to the basic structural data of the Mongolian yurt, a Mongolian yurt basic model database is established, a Mongolian yurt skeleton structure generation system is constructed, the area and height data of the Mongolian yurt to be constructed are input into the system, the skeleton structure generation system can automatically match the basic model with the minimum error of the input area and height data, and input the basic model into the deep learning workstation in step (3) to automatically generate a skeleton structure model, and provide data information and required quantity information of the standardized prefabricated component three-dimensional model of the Hanan component, the Tou Nao component, and the Uni component;

[0013] Step (5), skeleton structure stability test,

[0014] The skeleton structure model generated in step (4) is input into an engineering simulation software, and a structural mechanics analysis is performed, and for the skeleton structure with structural displacement and deformation, the skeleton structure model is regenerated in step (4);

[0015] (6) Component drawing output,

[0016] The list and drawings of the required prefabricated components are output as excel and cad format files for the next step of procurement and construction.

[0017] Further, the point cloud classification algorithm in step (1) is a deep learning algorithm trained by a point cloud data set, which can automatically classify the unprocessed point cloud data.

[0018] Further, the identification and extraction of the positions of each component as connection points in step (2) means that the distance between the surfaces of the components is calculated to identify the connection points, and the distance of 0 is identified as the connection point.

[0019] Further, the three-dimensional data of the framework structure in step (2) is converted into topological data, which means that the components in the framework structure are points in the topological network, and the connecting points are edges.

[0020] Further, the rule of the structural framework topological network in step (3) includes the connection relationship between the components, the number of connecting points, the connection relationship between the components and the components, the connection relationship between the components and the components, and the number of connecting points.

[0021] Further, the basic structure data of the yurt in step (4) includes the diameter of the bottom of the yurt, the height of the wall, the diameter of the brain, the height of the bottom edge of the brain, and the total height of the yurt.

[0022] Further, the basic model database of the yurt in step (4) is a database of the yurt framework structure model that stores the basic structure data of various yurts.

[0023] Further, the structural mechanics analysis in step (5) means analyzing the stress, strain and deformation of the structural framework under static load by engineering simulation software such as ANSYS.

[0024] Further, the precast component list and drawing in step (6) includes the type, number, diameter and length of the precast component. Advantages

[0025] By topologically converting the framework structure of the yurt, the topological relationship between the components, the components, the components and the components constituting the components is automatically extracted, the connection relationship between the components is reflected, the workload of identifying the connection relationship between the components is reduced, and the work efficiency of the architect is improved.

[0026] By using the graph neural network deep learning model, a large amount of topological data of the yurt framework structure can be automatically learned, the efficiency of topological data learning is improved, and the connection relationship of the yurt framework components can be more comprehensively understood compared with other machine learning algorithms, and the accuracy of topological network structure rule extraction and learning is improved.

[0027] The graph neural network-based yurt framework structure generation system generates the information of the framework structure model and the standardized prefabricated components, improves the intelligence and automation level of the yurt framework structure, improves the accuracy of the standardized prefabricated component generation, and reduces the difficulty of prefabricated component procurement, production and assembly. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 is a flowchart of the present application. Best mode for carrying out the invention

[0029] The method comprises the following steps:

[0030] Step (1), collecting data of a three-dimensional model of a Mongolian yurt building framework structure,

[0031] Object measurement is performed on the framework of the Mongolian yurt using a laser scanning device to collect point cloud data of the framework structure of the Mongolian yurt. The point cloud data is classified by a point cloud classification algorithm to obtain point cloud data of the Hanan component, the Taonao component and the Uni component. The point cloud data is input into a three-dimensional modeling software to generate a three-dimensional model of the framework structure of the Mongolian yurt composed of the Hanan, the Taonao and the Uni, and a three-dimensional model library of the framework structure of the Mongolian yurt is constructed.

[0032] Step (2), topology conversion of the framework structure of the Mongolian yurt and the prefabricated components,

[0033] The three-dimensional morphological properties of the Hanan component, the Taonao component and the Uni component in step (1) are calculated, including length, width, height and cross-sectional shape, to construct a three-dimensional morphological database of the components. The positions where the components are connected are identified and extracted as connection points. According to the connection relationship between the components and the connection points, the three-dimensional data of the framework structure is converted into topology data, and the connection points are connected with the corresponding component data in the three-dimensional morphological database of the components to establish a sample topology data set, wherein the training data set accounts for 80% and the test data set accounts for 20%.

[0034] Step (3), construction of a framework structure generation model based on a graph neural network,

[0035] The training data set in step (2) is input into a deep learning workstation equipped with a graph neural network deep learning model to extract and learn the rules of the structural framework topology network, and a Mongolian yurt framework structure generation model is constructed.

[0036] Step (4), generation of a Mongolian yurt building structure pattern and matching generation of prefabricated components,

[0037] A Mongolian yurt basic model database is established according to the basic structural data of the Mongolian yurt, and a Mongolian yurt framework structure generation system is constructed. The area and height data of the Mongolian yurt to be constructed are input into the system. The Mongolian yurt framework structure generation system can automatically match the basic model with the minimum error of the input area and height data, and input the basic model into the deep learning workstation in step (3) to automatically generate a framework structure model, and provide data information and required quantity information of standardized prefabricated component three-dimensional models constituting the Hanan component, the Taonao component and the Uni component.

[0038] Step (5), stability test of the framework structure,

[0039] The skeleton structure model generated in step (4) is input into engineering simulation software for structural mechanics analysis. For the skeleton structure that produces structural displacement and deformation, return to step (4) to regenerate the skeleton structure model;

[0040] (6) component drawing output,

[0041] The list and drawings of the required prefabricated components are output in excel and cad format files for the next step of procurement and construction.

[0042] Further, the point cloud classification algorithm in step (1) is a deep learning algorithm trained by point cloud data set, which can automatically classify the unprocessed point cloud data.

[0043] Further, the identification and extraction of the position of each component in step (2) is the position of the joint, which is identified by calculating the distance between the surfaces of the components. The distance of 0 is identified as the joint.

[0044] Further, the conversion of the three-dimensional data of the skeleton structure into topological data in step (2) refers to the components in the skeleton structure as points in the topological network, and the joint points as edges.

[0045] Further, the extraction and learning of the topology of the structural skeleton in step (3) includes the connection relationship between the components and the number of joint points, the connection relationship between the components and the number of joint points, the connection relationship between the components and the number of joint points, the connection relationship between the components and the number of joint points, and the connection relationship between the components and the number of joint points.

[0046] Further, the basic structure data of the yurt in step (4) includes the diameter of the bottom of the yurt, the height of the wall, the diameter of the brain, the height of the bottom edge of the brain, and the total height of the yurt.

[0047] Further, the basic model database of the yurt in step (4) is a database of the basic structure model of the yurt, which stores the basic structure data of a variety of yurts.

[0048] Further, the structural mechanics analysis in step (5) refers to the analysis of stress, strain and deformation of the structural skeleton under static load by engineering simulation software such as ANSYS.

[0049] Further, the prefabricated component list and drawings in step (6) include the type, quantity, diameter and length of the prefabricated components.

Claims

1. A novel intelligent generation method of a yurt skeleton structure based on deep learning, characterized in that, The method comprises the following steps: Step (1), data acquisition of the three-dimensional model of the yurt skeleton structure, Use a laser scanning device to measure the skeleton of the yurt, collect point cloud data of the yurt skeleton structure, classify the point cloud data by point cloud classification algorithm, obtain point cloud data of the hana component, the set brain component and the uni component, input the point cloud data into a three-dimensional modeling software to generate a three-dimensional model of the yurt skeleton structure composed of the hana, the set brain and the uni, and construct a three-dimensional model library of the yurt skeleton structure; Step (2), topology conversion of the yurt skeleton structure and the prefabricated component, Calculate the three-dimensional shape attributes of the hana component, the set brain component and the uni component in step (1), including length, width, height and cross-sectional shape, and construct a three-dimensional shape database of the components; identify and extract the positions of the interfaces of each component as the interface points, convert the three-dimensional data of the skeleton structure into topology data according to the interface relationship between the components and the interface points, connect the interface points with the corresponding component data in the three-dimensional shape database of the components, and establish a sample topology data set, wherein the training data set accounts for 80% and the test data set accounts for 20%; Step (3), construction of the skeleton structure generation model based on graph neural network, Input the training data set in step (2) into a deep learning workstation equipped with a graph neural network deep learning model, extract and learn the rules of the structure skeleton topology network, and construct a yurt skeleton structure generation model; Step (4), generation of the yurt skeleton structure model and matching of the prefabricated components, According to the basic structure data of the yurt, establish a yurt basic model database, construct a yurt skeleton structure generation system, input the area and height data of the yurt to be constructed into the system, the yurt skeleton structure generation system can automatically match the basic model with the minimum error of the input area and height data, input the basic model into the deep learning workstation in step (3), automatically generate the skeleton structure model, and provide data information and required quantity information of the standardized prefabricated component three-dimensional model of the hana component, the set brain component and the uni component; Step (5), stability test of the yurt skeleton structure model, Input the skeleton structure model generated in step (4) into an engineering simulation software, perform structural mechanics analysis, and return to step (4) to regenerate the skeleton structure model for the skeleton structure with structural displacement and deformation; Step (6), component drawing output, Output the required prefabricated component list and drawing in excel and cad format files for the next step of procurement and construction.

2. The deep learning-based intelligent generation method of a new yurt framework structure according to claim 1, characterized in that, The point cloud classification algorithm in step (1) is a deep learning algorithm trained by a point cloud data set, which can automatically classify untreated point cloud data. 3.The method of claim 1, wherein, In step (2), the positions of the interfaces of each component are identified and extracted as interface points, which means that the distance between the component surfaces is calculated to identify the interface points.

4. The deep learning-based intelligent generation method of a new yurt framework structure according to claim 1, characterized in that, In step (2), the three-dimensional data of the skeleton structure is converted into topology data, which means that the components in the skeleton structure are taken as points in the topology network, and the interface points are taken as edges.

5. The deep learning-based intelligent generation method of a new yurt framework structure according to claim 1, characterized in that, The extraction and learning of the structure skeleton topology network in step (3) includes the connection relationship between the Hanan components, the number of connection points, the connection relationship between the Hanan components and the Uni components, the connection relationship between the Uni components and the Tou Nao components, the number of connection points, the connection relationship between the Tou Nao components, and the number of connection points.

6. The deep learning-based intelligent generation method of a new yurt framework structure according to claim 1, characterized in that, The basic structure data of the yurt in step (4) includes the diameter of the bottom of the yurt, the height of the Hanan wall, the diameter of the Tou Nao, the height of the bottom edge of the Tou Nao, and the total height of the yurt.

7. The deep learning-based intelligent generation method of a new yurt framework structure according to claim 1, characterized in that, The basic model database of the yurt in step (4) is a database of yurt skeleton structure models that stores the basic structure data of various yurts.

8. The method of claim 1, wherein the method is characterized by, The structural mechanics analysis in step (5) refers to analyzing the stress, strain, and deformation of the structure skeleton under static load through engineering simulation software. 9.The method of claim 1, wherein, The precast component list and drawings in step (6) include the type, quantity, diameter, and length of the precast components.

Images