Large-Scale Finite Element Mesh Data Storage Index Method

Abstract

The large-scale finite element grid data storage index method includes: extracting and naming the geometric edges and geometric faces contained in the original geometry; extracting the geometric edges and their names, discretizing the geometric edges into discrete edges composed of line segments, and Name a line segment; extract the geometric surface surrounded by the geometric edge and its name, discretize the geometric surface into a discrete surface composed of triangular patches and obtain the normal direction of each triangular patch; save the basic description information of the discrete geometry to the element In the data file; the discrete geometry is meshed, and based on the number of mesh nodes that the mesh file can store, the mesh node data is split into multiple groups and stored in different mesh files in the entity data file. A row of data in the grid file is the coordinate data of a grid node; the grid file to which the grid data of each geometric surface of the discrete geometry belongs and the corresponding row number are stored in the metadata file, and the geometric surface and grid The association relationship is stored in the data index pool file.

Description

technical field [0001] The invention relates to the field of computer-aided engineering calculation software, in particular to a method for storing and indexing large-scale finite element grid data. Background technique [0002] Finite element analysis (FEA) software is a computer method that discretizes the physical characteristics of real-world scenes and performs analysis and prediction through numerical calculations. With the development of computer computing power and finite element analysis technology, finite element analysis has changed from the solution of single-disciplinary / physical field problems to the joint solution of multi-disciplinary / multi-physical fields; from simple or simplified 2 / 3-dimensional models to complex assembly or With the development of the calculation of complex geometric full models, the discrete mesh models generated by the finite element pre-processor are becoming more and more complex and the amount of data is increasing. [0003] In 2010...

AI Technical Summary

Problems solved by technology

[0007] 3) There is no corresponding relationship between the traditional finite element grid and its discretized geometry. This is because the early definition of the finite element grid was manually discretized according to the geometry. The two models have a corresponding relationship in processing logic, but in the data It is not required in terms of specifications, and geometric topology changes cannot be automatically reflected on discrete grids, so there is no need to establish internal relationships at the data level

[0008] 4) Traditional finite element grids are independent between different discretizations of the same geometry under different precision and different discipline requirements, and the corresponding loads and boundary conditions correspond to the nodes and units of the discrete grids (see figure 2 , E1, E2, E3, E4, and E5 represent the edges of the geometric model), that is, they do not have a corresponding relationship with each other after the discretization, which leads to the fact that when coupling between different disciplines, only the spatial position can be used for judgment. extremely inefficient

[0009] 5) Traditional finite element grids do not consider parallel mode in terms of data specifications because the early processors do not have parallel capability and the processing scale is small. Massively Parallel Processing Requirements

[0029] 1) The data format adopts a card / block data organization method for the convenience of manual reading, compilation, and error correction. The structure is simple, but the description of complex nodes and unit relationships is relatively weak

[0030] 2) During discrete grid processing and finite element calculation, the model needs to be read in at one time, which takes up a lot of memory and the data transmission efficiency is low during parallel calculation

[0031] 3) The correspondence between the original geometry and the grid is lost after being discretized, resulting in the discrete grid model being unable to trace back the associated information such as the original geometric features. When solving repeated iterations or optimization, it is necessary to re-mesh from the original geometry, and the reusability is extremely low

[0032] 4) Discrete geometry is arranged in the order of grid generation, and there is no inter-relationship within the grid, resulting in repeated traversal of grid information during program indexing and retrieval, which is inefficient

[0033] 5) Grids of different precisions and types generated by the same geometry have no other information except for spatial coordinate information for fast correspondence and retrieval between each other, and the memory overhead in subject coupling and other processing is large and the efficiency is extremely low

[0034] 6) Grid nodes and units are generated by sequential numbering, which cannot meet the needs of parallel grid generation, and additional processing of sequence numbers is required during parallel grid processing

[0035] 7) The single-file storage method limits the parallel reading and writing and on-demand access capabilities of large-scale data, forming a bottleneck for data input and output

However, with the large-scale increase in the amount of data, and now it is mainly suitable for specialized pre- and post-processing programs rather than manual data processing, the effect appears to do more harm than good

Images