CN112765866A - Method for rapidly analyzing sub-model based on finite element analysis - Google Patents

Method for rapidly analyzing sub-model based on finite element analysis Download PDF

Info

Publication number
CN112765866A
CN112765866A CN202110157958.1A CN202110157958A CN112765866A CN 112765866 A CN112765866 A CN 112765866A CN 202110157958 A CN202110157958 A CN 202110157958A CN 112765866 A CN112765866 A CN 112765866A
Authority
CN
China
Prior art keywords
unit
model
sub
node
area
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.)
Pending
Application number
CN202110157958.1A
Other languages
Chinese (zh)
Inventor
杜晟强
陈睿
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.)
Huzhou Zhongke Shenghe Information Technology Co ltd
Original Assignee
Huzhou Zhongke Shenghe Information Technology Co ltd
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 Huzhou Zhongke Shenghe Information Technology Co ltd filed Critical Huzhou Zhongke Shenghe Information Technology Co ltd
Priority to CN202110157958.1A priority Critical patent/CN112765866A/en
Publication of CN112765866A publication Critical patent/CN112765866A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Artificial Intelligence (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention belongs to the technical field of computer aided design, and particularly relates to a method for rapidly analyzing a sub-model based on finite element analysis. The method for rapidly analyzing the submodels introduces a fitting technology based on a neuron network, a surface node searching technology and an interpolation technology, and optimizes the analysis process of the submodels. The invention can simultaneously carry out the fitting and the pretreatment of the neuron network without extra time, and after the result file is obtained, a CAD model is not needed any more, and the setting of all sub models can be realized in the result file. Meanwhile, the user can randomly select any multiple concerned areas and simultaneously perform sub-model analysis. In addition, the method is generally used for various mainstream FEA software, and only needs to be properly adjusted in the aspect of data interface. Finally, the invention expands the application scene and the analysis efficiency of the sub-model technology.

Description

Method for rapidly analyzing sub-model based on finite element analysis
Technical Field
The invention belongs to the technical field of computer aided design, and particularly relates to a method for rapidly analyzing a sub-model based on finite element analysis.
Background
Finite Element Analysis (FEA) simulates real physical systems (geometry and loading conditions) by mathematical approximation, and by using simple and interacting elements (i.e. elements), an infinite unknown real system can be approximated by a Finite number of unknowns. In finite element analysis, the local model may be studied on the basis of the results of the global model analysis. Determining a maximum response area under the action of an excitation load through initial global model analysis and calculation, intercepting a local attention area model and refining a grid of the local attention area model, interpolating a global model node displacement result field to a model boundary node after the grid is refined, and performing finite element calculation again, thereby improving the analysis precision of the local area. The method comprises the steps of obtaining results around a local attention area by adopting a coarse grid model, obtaining displacement values of boundary nodes of the local area by utilizing an interpolation technology, driving a refined grid to calculate and obtaining a local analysis result.
The steps for performing sub-model analysis in the prior art abaqus are as follows: 1. carrying out grid division on the CAD model, and setting each calculation parameter; 2. submitting the global model to calculation and obtaining a calculation result of the global model; 3. subdividing the CAD model to obtain a CAD model of a concerned local area, and carrying out grid refinement on the CAD model; 4. setting boundary nodes needing interpolation in the local region model, and setting data required to be quoted by sub-model calculation as a global model calculation result obtained in the step 2; 5. submitting the local model to calculation, and automatically interpolating the required result in the step 2 to the boundary node set in the step 4 before calculating by a solver so as to drive the calculation; 6. and obtaining a local model calculation result for analysis. The above method has the following problems: 1. an original CAD model is required; 2. an original CAD model needs to be subdivided to obtain a concerned local area model, and when the shape of the area is complex or the number of the concerned areas in the model is too large, more time needs to be consumed in the subdivision process; 3. and the solver needs to check the data type during the calculation of the sub-model, so that the calculation time is long.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a method for rapidly analyzing a sub-model based on finite element analysis, aiming at solving the problems that the analysis of the sub-model in the prior art needs to divide an original CAD model to obtain a concerned local area model, and when the shape of the area is complicated or the number of the concerned areas in the model is too large, more time needs to be consumed on a dividing flow; and the solver needs to check the data type when the submodel is calculated, so that the calculation time is long.
The invention provides a method for rapidly analyzing a sub-model based on finite element analysis, which has the following specific technical scheme:
the method for rapidly analyzing the sub-model based on finite element analysis comprises the following steps;
s1, meshing, namely, meshing by using a first-order unit of one thousandth of the size of the model;
s2, exporting the grids in the step S1 as point cloud data or SDF data, and obtaining trained neuron networks through neuron network training;
s3, when training through the neural network, adopting the unit size meeting the calculation requirement to divide the grid again, and carrying out other pretreatment steps to submit the calculation to obtain a result file;
s4, analyzing the result in the step S3, selecting an area needing further analysis, obtaining all units in the area, and processing by utilizing surface node search to obtain the outer surface nodes of the units in the area;
s5, extracting the size and the position of the area appointed in the step S4, taking the obtained size and the position of the area as input conditions, predicting through the neuron network trained in the step S2, and obtaining an expression of point cloud distribution in the area, namely, under a Cartesian coordinate system, calculating a third coordinate value through any two coordinate values;
s6, the user specifies the discrete degree, the coordinate range of the area obtained in the step S4 is dispersed, a point cloud is generated through the expression of the step S5, and a grid is generated through the point cloud;
s7, processing by surface node search to obtain all the unit outer surface nodes generated in the step S6;
s8, interpolating the external surface node result obtained in the step S4 to the refined grid node obtained in the step S7 through interpolation processing;
s9, generating a calculation file, submitting calculation by the background and returning a result file to obtain the accurate calculation result of the region obtained in the step S4.
In some embodiments, in step S8, the interpolation process includes the steps of:
(1) obtaining the spatial coordinates of the single fine grid node in step S6;
(2) obtaining all grid centroid coordinates of the coarse grid in the region selected in step S4;
(3) obtaining 3 units with the shortest distance according to the distance between the centroid in the step (2) and the space coordinate obtained in the step (1);
(4) judging whether the node in the step (1) is positioned in 3 units in the step (3);
(5) and (3) if the unit where the node in the step (1) is located is found, carrying out an interpolation algorithm, and if the unit is not found, returning to the step (3) to search three next-nearest units.
Specifically, in step (4), the determination method is a PNPoly algorithm, and the method includes: and (2) introducing a ray from the node in the step (1), and calculating the intersection condition of the ray and each surface of the unit, wherein if the number of the intersection points is odd, the point is in the unit, and if the number of the intersection points is even, the point is outside the unit.
Specifically, in the step (5), the interpolation algorithm adopts a difference method based on a unit shape function.
The invention has the following beneficial effects: 1. the fitting and the pretreatment of the neuron network can be carried out simultaneously without extra time.
2. After the result file is obtained, the CAD model is not needed any more, and the setting of all the sub models can be realized in the result file.
3. The user can randomly select any multiple concerned areas and simultaneously perform sub-model analysis.
4. The method is generally used for various mainstream FEA software, and only needs to be properly adjusted in the aspect of data interface.
Drawings
FIG. 1 is a flow chart of a method for rapid analysis of a sub-model based on finite element analysis provided by the present invention;
FIG. 2 is the model and the grid map in step S1 in example 1;
fig. 3 is a visualization diagram of point cloud data derived from the model in step S2 in example 1;
FIG. 4 is a structural diagram of a neuron network in step S2 in embodiment 1;
FIG. 5 is a view showing a grid map repartitioning in step S3 in embodiment 1;
FIG. 6 is a view showing the calculation result submitted in step S3 in embodiment 1;
FIG. 7 is a region diagram for further analysis required in step S4 in example 1;
fig. 8 is a new point cloud location map in step S5 in embodiment 1.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings 1 in conjunction with specific embodiments.
Example 1
The method for rapidly analyzing the sub-model based on the finite element analysis provided by the embodiment has the following specific technical scheme:
the method for rapidly analyzing the sub-model based on finite element analysis comprises the following steps;
and S1, meshing, namely meshing by using a tiny first-order unit. The specific operation is as follows: model processing is carried out in abaqus software, and the method comprises the following steps: geometry import, grid division, in which the grid can be refined as much as possible (the level of refinement affects the accuracy of subsequent interpolation). The cell size used in this example is 1mm, the cell type is a second order tetrahedral cell, and the model and mesh are shown in fig. 2.
And S2, exporting the grids in the step S1 as point cloud data or SDF data, and obtaining trained grid parameter files through neuron network training. The aim of fitting the geometry by utilizing the multi-layer neuron network is to store the characteristic information of the geometry into the neuron network, so that a large-scale geometric model can be stored in a very small storage space. In the finite element simulation process, after the mesh is generated, the spatial coordinate data of the unit and the unit-related nodes can be obtained. The unit and the node are rough expressions of smooth geometry, and can utilize spatial node data or convert the spatial node data into SDF expression, and then put the SDF expression into the existing multilayer neuron network for fitting. In this embodiment, the model is derived as point cloud data, the derived point cloud data is stored as a flat sequence, and the flat sequence is visualized through jupyter notebook, as shown in fig. 3. And submitting the point cloud data to a neural network for fitting, wherein the neural network is programmed by adopting a tensierflow based on python, the network structure is shown in figure 4, input is the Cartesian coordinates of the point cloud, and after training is completed, model configuration is automatically stored.
And S3, carrying out neural network training, simultaneously, adopting a proper unit size to divide the grids again, carrying out other pretreatment steps, submitting to calculation, and obtaining a result file. The embodiment is as follows: and (3) building a finite element calculation model while training the neural network. Firstly, adjusting the grid size, then, carrying out grid division by a user according to needs, wherein the grid division comprises redefinition of the global size and partial grid repartitioning, the global size is 2, the newly divided grid is shown in fig. 5, model building is continuously carried out, definition materials are respectively established according to the sequence, analysis steps are created, boundary conditions and loads are added, and calculation is submitted. The output is the misers stress, and the results are shown in fig. 6.
And S4, analyzing the result in the step S3, and selecting the area to be further analyzed by the user through a frame selection or one-by-one selection mode. In this embodiment, a user performs operations such as frame selection in software, selects a desired region as needed, and automatically filters out the outer surfaces of the cells in the selected region through the outer surface automatic selection function in the invention, where the selected region is a non-black region as shown in fig. 7. Substituting the obtained area coordinate range (namely the range of the area in the x, y and z axes) into the network obtained in the step S2 for prediction, and obtaining the geometric outline expression relationship of the area, namely obtaining the product geometric outline expression in the specific space range.
S5, extracting the size and position of the area in the step S4, obtaining the point cloud distribution of the area through the training of the neuron network, and the density degree of the point cloud can be specified by the user. In this embodiment, the degree of refinement is determined by the user, and the coordinates of the region in step S4 are substituted into the expression obtained in step S4, thereby obtaining the coordinates of the discrete points on the geometric outer contour in the region. The refinement degree is 1, i.e. it is consistent with step S1, and the resulting new point cloud location is shown as red area in fig. 8.
S6, a mesh is generated by the point cloud in step S5.
And S7, processing by surface node searching to obtain all the unit outer surface nodes. In the finite element model, nodes are the most basic components, the connecting lines of the nodes form edges, the edges mutually surround to form a surface, and the surfaces mutually surround to form a unit. The nodes are shared from cell to cell as shown. The outline of the region can be expressed in the form of a set of the outer surfaces of all the units in the region, and can also be expressed in the form of a set of the nodes of the outer surfaces. Based on this goal, two types of surface node search techniques can be formulated. Firstly, based on the normal included angle of the adjacent unit surface, by calculating the normal direction of each unit surface, on the basis of one unit surface, searching all the unit surfaces which are adjacent to the unit surface and have the normal included angle within a certain range, and then, on the basis of the obtained unit surface, continuing the search until the end. Secondly, the outer surface is searched under the condition that the unit surfaces are shared among the units, so that the outer surface can not be superposed with the surfaces of the other units, for example, the unit surfaces are represented by a node set, and s1 is { n1, n2 and n3}, after all the unit surfaces are expressed in the type, the completely nonrepeated surfaces can be selected, and the unit outer surfaces are obtained.
S8, the result of the boundary nodes from step S4 is interpolated to the nodes of the refined grid from step S7 by interpolation. The interpolation processing includes the steps of: (1) obtaining the space coordinates of a single fine grid node; (2) obtaining all grid centroid coordinates of the rough grid; (3) obtaining 3 units with the shortest distance according to the distance between the centroid in the step (2) and the space coordinate obtained in the step (1); (4) judging whether the node in the step (1) is positioned in 3 units in the step (3); (5) and (3) if the unit where the node in the step (1) is located is found, carrying out an interpolation algorithm, and if the unit is not found, returning to the step (3) to search three next-nearest units. Wherein, the judging method in the step (4) is PNPoly algorithm, and the method comprises the following steps: and (2) introducing a ray from the node in the step (1), and calculating the intersection condition of the ray and each surface of the unit, wherein if the number of the intersection points is odd, the point is in the unit, and if the number of the intersection points is even, the point is outside the unit. For the judging method in the step 4, the PNPoly algorithm is adopted for judging, and the method comprises the following steps: and (3) introducing a ray from the point, and calculating the intersection condition of the ray and each surface of the unit, wherein if the number of the intersection points is odd, the point is in the unit, and if the number of the intersection points is even, the point is outside the unit.
For the interpolation algorithm in step 5, a difference method based on a unit shape function is adopted, that is, a unit is divided by connecting lines of the point and each vertex of the unit, and the distribution relation of each vertex of the unit to the point can be determined by a volume ratio value formed by the division.
And S9, generating a calculation file, submitting calculation by a background and returning a result file to obtain an accurate calculation result of the region.
The above description is only for the purpose of illustrating preferred embodiments of the present invention and is not to be construed as limiting the invention, and the present invention is not limited to the above examples, and those skilled in the art should also be able to make various changes, modifications, additions or substitutions within the spirit and scope of the present invention.

Claims (4)

1. The method for rapidly analyzing the sub-model based on finite element analysis is characterized by comprising the following steps;
s1, meshing, namely, meshing by using a first-order unit of one thousandth of the size of the model;
s2, exporting the grids in the step S1 as point cloud data or SDF data, and obtaining trained neuron networks through neuron network training;
s3, when training through the neural network, adopting the unit size meeting the calculation requirement to divide the grid again, and carrying out other pretreatment steps to submit the calculation to obtain a result file;
s4, analyzing the result in the step S3, selecting an area needing further analysis, obtaining all units in the area, and processing by utilizing surface node search to obtain the outer surface nodes of the units in the area;
s5, extracting the size and the position of the area appointed in the step S4, taking the obtained size and the position of the area as input conditions, predicting through the neuron network trained in the step S2, and obtaining an expression of point cloud distribution in the area, namely, under a Cartesian coordinate system, calculating a third coordinate value through any two coordinate values;
s6, the user specifies the discrete degree, the coordinate range of the area obtained in the step S4 is dispersed, a point cloud is generated through the expression of the step S5, and a grid is generated through the point cloud;
s7, processing by surface node search to obtain all the unit outer surface nodes generated in the step S6;
s8, interpolating the external surface node result obtained in the step S4 to the refined grid node obtained in the step S7 through interpolation processing;
s9, generating a calculation file, submitting calculation by the background and returning a result file to obtain the accurate calculation result of the region obtained in the step S4.
2. The method for sub-model fast analysis based on finite element analysis of claim 1, wherein in step S8, the interpolation process comprises the following steps:
(1) obtaining the spatial coordinates of the single fine grid node in step S6;
(2) obtaining all grid centroid coordinates of the coarse grid in the region selected in step S4;
(3) obtaining 3 units with the shortest distance according to the distance between the centroid in the step (2) and the space coordinate obtained in the step (1);
(4) judging whether the node in the step (1) is positioned in 3 units in the step (3);
(5) and (3) if the unit where the node in the step (1) is located is found, carrying out an interpolation algorithm, and if the unit is not found, returning to the step (3) to search three next-nearest units.
3. The method for sub-model rapid analysis based on finite element analysis of claim 2, wherein in step (4), the determination method is PNPoly algorithm, and the method comprises the following steps: and (2) introducing a ray from the node in the step (1), and calculating the intersection condition of the ray and each surface of the unit, wherein if the number of the intersection points is odd, the point is in the unit, and if the number of the intersection points is even, the point is outside the unit.
4. The method for sub-model fast analysis based on finite element analysis of claim 2, wherein in step (5), the interpolation algorithm adopts a difference method based on a unit shape function.
CN202110157958.1A 2021-02-04 2021-02-04 Method for rapidly analyzing sub-model based on finite element analysis Pending CN112765866A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110157958.1A CN112765866A (en) 2021-02-04 2021-02-04 Method for rapidly analyzing sub-model based on finite element analysis

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110157958.1A CN112765866A (en) 2021-02-04 2021-02-04 Method for rapidly analyzing sub-model based on finite element analysis

Publications (1)

Publication Number Publication Date
CN112765866A true CN112765866A (en) 2021-05-07

Family

ID=75705042

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110157958.1A Pending CN112765866A (en) 2021-02-04 2021-02-04 Method for rapidly analyzing sub-model based on finite element analysis

Country Status (1)

Country Link
CN (1) CN112765866A (en)

Similar Documents

Publication Publication Date Title
CN112434448B (en) Proxy model constraint optimization method and device based on multipoint adding
Gálvez et al. A new iterative mutually coupled hybrid GA–PSO approach for curve fitting in manufacturing
Coupez et al. Parallel meshing and remeshing
CN100468418C (en) Method and program for generating volume data from boundary representation data
US5373457A (en) Method for deriving a piecewise linear model
CN109684714B (en) Building design method based on machine learning and BIM technology
CN116227043B (en) Aircraft numerical simulation method, system, equipment and computer storage medium
CN113128095A (en) Method and device for constructing three-dimensional grid model
US20230062600A1 (en) Adaptive design and optimization using physics-informed neural networks
US20020177985A1 (en) Computer system and method for radial cooled bucket optimization
Sieger et al. A comprehensive comparison of shape deformation methods in evolutionary design optimization
CN104361185A (en) Wire distribution space automatic generating method for cable virtual design
JP2012194910A (en) Mesh number prediction method, analyzer, and program
CN115688276A (en) Aircraft appearance automatic optimization method, system, equipment and medium based on discrete companion method
CN115758938A (en) Boundary layer grid generation method for viscous boundary flow field numerical simulation
CN114676522A (en) Pneumatic shape optimization design method, system and equipment integrating GAN and transfer learning
CN112464040B (en) Graph structure recognition, visual display and display operation method and device
CN112560212A (en) Spraying system based on BIM model, spraying head and pipeline generation type design method
CN112765866A (en) Method for rapidly analyzing sub-model based on finite element analysis
CN104240301A (en) Geological surface reconstruction method and device
JP2002056038A (en) Accelerator for obtaining numerical solution of linear simultaneous equation resulting from mesh/grid generation and accelerating method
Piotrowski et al. Enhanced interactive parallel coordinates using machine learning and uncertainty propagation for engineering design
CN115835931A (en) Method for providing a parameter value of a heat source for producing a weld seam between two plates, corresponding computer program and corresponding device
CN111210877A (en) Method and device for deducing physical property parameters
JPH11203330A (en) Shape deformation mode generation system, shape optimization analyzing system and record medium in which program used for the same is recorded

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