A physics-based, uniform, meshless, high-performance point distribution method
Through Delaunay triangulation, quadratic point distribution method, filling algorithm and improved discrete unit method, combined with GPU and MPI technology, a uniform distribution of aptamers is achieved, and the uneven distribution and calculation speed in the existing technology are solved. The problem of insufficient improves computing efficiency and robustness.
Patent Information
- Application Number
- CN202411099760.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-12
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2044-08-12
AI Technical Summary
The existing gridless method is difficult to achieve uniformly distributed wall surfaces and internal fluid particle points of aptamer, and the calculation speed and robustness in complex geometry are insufficient, which cannot meet the needs of high-precision solution and large-scale computing.
Using the physical-based aptamer uniform gridless high-performance dot distribution method, Delaunay triangulation, quadratic dot distribution method, filling algorithm and improved discrete unit method, parallel accelerated calculations are performed in combination with GPU and MPI technology to ensure wall surface and internal fluids Uniform distribution and calculation speed of particles.
The isotropic distribution of aptamers is realized, the calculation speed and scale problems in the prior art are overcome, the robustness and computing efficiency are improved, and it is suitable for the pre-processing and subsequent solution of complex engineering geometric models.
Smart Images

Figure CN118627366B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of gridless method pre-processing, and in particular relates to a physics-based body-fitting uniform gridless high-performance point distribution method. Background Art
[0002] The meshless method requires the wall particles and internal fluid particles to be evenly distributed in a suitable body to obtain a high-precision solution. Among them, the suitable body means that the distribution of points is adapted to the physical characteristics of the geometric body, and the uniformity means that the distribution of the outer wall particles satisfies the isotropic distribution of the wall geometric model surface, and the distribution of the internal fluid particles satisfies the isotropic distribution of the fluid geometric model space. At the same time, due to the complexity of the three-dimensional engineering geometry, it is often necessary to arrange particles in the defects or sharp corners of the three-dimensional geometric model where the physical field changes drastically, or due to the large scale of the geometric file, the distribution scale often reaches tens of millions of scales, so the calculation speed, scale and robustness of the gridless method distribution are also crucial. At present, there is still a lack of a distribution method that can meet the isotropic requirements of the distribution of the uniform distribution of the body and the calculation speed, scale and robustness requirements of the distribution calculation. Therefore, it is crucial to propose a physics-based, uniform, meshless high-performance distribution method for the pre-processing field of the meshless method and the successful and accurate completion of the subsequent solution calculation work. Summary of the invention
[0003] In order to solve the problems in the prior art, the present invention proposes a physics-based, uniformly distributed gridless high-performance point distribution method. The present invention can obtain a robust point distribution of wall particles and internal fluid particles uniformly distributed with a specified spacing, achieving isotropy, and at the same time uses GPU and MPI technology for parallel accelerated calculation, expanding the scale of pre-processing point distribution calculation and improving the calculation speed.
[0004] The present invention first provides a physics-based, uniform, gridless, high-performance point distribution method for an adaptive body, comprising the following steps:
[0005] 1) Drawing a geometric model based on the physical geometric body to be drawn or cleaning and repairing an existing geometric model based on the physical geometric body, thereby obtaining a wall geometric model and a fluid geometric model of the physical model;
[0006] 2) Use the frontier advancing method to perform Delaunay triangulation on the surface of the wall geometry model to obtain triangular patches, and then use the quadratic point distribution method to perform secondary point distribution on the triangular patches to obtain outer wall particles that meet the isotropic distribution of the wall geometry model surface and adapt to the distribution of physical characteristics of the physical geometry;
[0007] 3) Use the filling algorithm and parallel accelerated computing technology to fill the fluid geometry model and obtain internal fluid particles;
[0008] 4) Use the improved discrete element method and parallel accelerated computing technology to adjust the distribution of internal fluid particles, thereby eliminating the jagged distribution of internal fluid particles and obtaining internal fluid particles that meet the isotropic distribution of the fluid geometric model space and adapt to the distribution of physical characteristics of the physical geometric body;
[0009] 5) Output the coordinates, unit normal vectors, and boundary values of the outer wall particles and the internal fluid particles to complete the body-fitting uniform gridless point distribution.
[0010] As a preferred embodiment of the present invention, when filling the fluid geometric model in step 3), it is necessary to first determine the geometric boundaries of the fluid particles to be filled; when the geometric boundaries of the fluid particles are on the outer wall, the outer wall particles obtained in step 2) are used as the boundary particles of the fluid region, thereby determining the geometric boundaries of the fluid particles; when the geometric boundaries of the fluid particles are a part of the area inside the outer wall, the geometric model of this part of the area is operated in step 2) to obtain the boundary particles of the fluid region to determine the geometric boundaries of the fluid particles; and then the filling algorithm and parallel accelerated computing technology are used to fill the internal fluid particles of the fluid region where the geometric boundaries of the fluid particles are determined.
[0011] As a preferred solution of the present invention, when filling the internal fluid particles of the fluid area that determines the geometric boundary of the fluid particles using the filling algorithm and the parallel accelerated computing technology, the parallel accelerated computing technology first allocates the thread index corresponding to the fluid particle according to the internal fluid particle serial number after the background grid index, and then calls the CUDA kernel function to perform the filling algorithm for each fluid particle, thereby realizing the parallelization of particle filling; wherein the overall boundary of the background grid will be determined according to the maximum or minimum of the three dimensions of the fluid area, the background grid size will be adaptively adjusted according to the pre-set filling spacing to take into account both the speed and stability of the calculation, and the serial number of the particle to be filled will be determined according to the index value generated by the background grid index where it is located.
[0012] As a preferred solution of the present invention, in the step 4), when the improved discrete element method and parallel accelerated computing technology are used to adjust the internal fluid particle distribution, the regions are divided according to the fluid boundaries, and the particles in different regions are assigned to different MPI processes. A GPU is assigned to each process for calculation, and the CUDA kernel function is called when each particle is calculated using the improved discrete element method, thereby eliminating the jagged distribution of the internal fluid particles.
[0013] The present invention also provides a system for implementing the above-mentioned aptamer uniform gridless high-performance point distribution method, comprising:
[0014] A model processing module is used to draw a geometric model based on the physical geometric body to be drawn or to clean and repair the existing geometric model to obtain a wall geometric model and a fluid geometric model;
[0015] The outer wall particle distribution module is used to perform Delaunay triangulation on the surface of the wall geometry model to obtain triangular patches, and use the quadratic distribution method to perform secondary distribution on the triangular patches to obtain outer wall particles that meet the isotropic distribution of the wall geometry model surface and adapt to the distribution of physical characteristics of the physical geometry;
[0016] Internal fluid particle distribution module, used to fill the fluid geometry model through filling algorithm and parallel acceleration computing technology to obtain internal fluid particles;
[0017] The body-fitting distribution adjustment module is used to adjust the internal fluid particle distribution through the improved discrete element method and parallel accelerated computing technology, eliminate the jagged distribution of the internal fluid particles near the wall, and obtain the internal fluid particles that meet the isotropic distribution of the fluid geometric model space and adapt to the physical characteristics of the physical geometric body;
[0018] The data output module is used to output the coordinates, unit normal vectors, and boundary values of the outer wall particles and the internal fluid particles to complete the body-fitting uniform gridless point distribution.
[0019] The present invention further provides a terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the body-fitting uniform gridless high-performance point distribution method when executing the computer program.
[0020] Compared with the prior art, the present invention has the following beneficial effects:
[0021] Because this technology uses the frontier advancement method and secondary point distribution technology for wall point distribution, and uses the filling algorithm and improved discrete unit method for internal point distribution, it overcomes the uneven wall point distribution and internal point distribution problems of the prior art, thereby achieving an isotropic point distribution effect with uniform distribution of the body;
[0022] The present invention uses GPU technology and MPI technology to perform high-performance parallel accelerated computing, thereby overcoming the problems of computing scale and computing speed in large-scale point distribution in the prior art, thereby achieving the effect of expanding the computing scale and improving the computing speed;
[0023] The present invention adopts processing methods such as adaptive background grid in program compilation, which overcomes the problem of insufficient robustness of point distribution at locations of complex physical geometric shapes such as defects and sharp corners in the prior art, thereby improving the robustness of the method. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] Figure 1 It is the overall flow chart of the present invention;
[0025] Figure 2The improved discrete element method calculation principle diagram used in the present invention;
[0026] Figure 3 This is a schematic diagram of the results before particle sorting of the present invention;
[0027] Figure 4 This is a schematic diagram of the results after the particles of the present invention are sorted;
[0028] Figure 5 A diagram showing the distribution of particles on the wall surface of a complex aircraft in an embodiment of the present invention;
[0029] Figure 6 A diagram showing the homogenization of the particle distribution on the wall of a complex aircraft in accordance with the particle number density in an embodiment of the present invention;
[0030] Figure 7 A diagram showing the distribution of wall particles obtained from the model in an embodiment of the present invention;
[0031] Figure 8 A diagram showing the homogenization of the wall particle distribution in terms of particle number density in an embodiment of the present invention;
[0032] Fig. 9 This is a diagram showing the overall distribution of wall particles and filled internal fluid particles with a filling interval of 70 mm obtained for a complex aircraft in an embodiment of the present invention;
[0033] Fig.10 This is a graph showing the internal fluid particle distribution result with a filling interval of 70 mm obtained for a complex aircraft in an embodiment of the present invention;
[0034] Fig.11 This is a general distribution diagram of wall particles and filled internal fluid particles with a filling interval of 30 mm obtained for a complex aircraft in an embodiment of the present invention;
[0035] Fig.12 This is a graph showing the distribution of internal fluid particles with a filling interval of 30 mm obtained for a complex aircraft in an embodiment of the present invention;
[0036] Fig.13 A filling result diagram obtained within a specified wavy free surface in a dam-break calculation example in an embodiment of the present invention;
[0037] Fig.14 A side view of a filling result obtained on a model in an embodiment of the present invention;
[0038] Fig.15 is a rear view cross-sectional diagram of a filling result obtained on a model in an embodiment of the present invention;
[0039] Fig.16A top view comparing the results before and after adjustment of the internal fluid fitness obtained by a 2D circular example in an embodiment of the present invention;
[0040] Fig.17 A top view comparing the results before and after the adjustment of the fitness of the internal fluid particles obtained by the model in an embodiment of the present invention;
[0041] Fig.18 This is a side view comparing the results before and after the adjustment of the fitness of the internal fluid particles obtained by the model in an embodiment of the present invention. DETAILED DESCRIPTION
[0042] The present invention is further described and illustrated below in conjunction with specific embodiments. The embodiments are merely exemplary of the present disclosure and do not define the scope of limitation. The technical features of each embodiment of the present invention may be combined accordingly without conflicting with each other.
[0043] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments.
[0044] The present invention provides a physics-based, uniform, gridless, high-performance point distribution method, which is mainly used for pre-processing point distribution of complex engineering geometric models, and can obtain uniform, isotropic particle distribution. The process is as follows: Figure 1 The specific description is as follows:
[0045] 1) Draw a geometric model based on the physical geometric body to be drawn or clean and repair the existing geometric model based on the physical geometric body to obtain a wall geometric model and a fluid geometric model; input the model into the present invention as an input file through an input interface. The interface of the present invention is universal, compatible with most commonly used file formats, and can be customized according to other needs. The formats supported by the present invention include STandard for the Exchange of Product data (step / stp) file format, Initial Graphics Exchange Specification (iges / igs) file format, Stereolithography (stl) file format, Boundary Representation (brep) file format, Standard ACIS Text (sat) file format and other formats.
[0046] 2) The wall geometry model is subjected to Delanuay triangulation with a specified spacing, and the obtained triangular facets are subjected to secondary distribution with a pre-set wall particle spacing. A fast and uniform wall particle distribution program is implemented using C++, which can achieve a distribution scale of tens of millions of wall particles.
[0047] The secondary point distribution method is to distribute points along the side length and internally on the triangular face respectively. When distributing points along the side length, the three nodes of the triangle will be retained as wall particles first, and then the number of wall particles to be divided on the three sides will be calculated according to the side lengths of the three sides and the distance between the wall particles. The particles will be divided along the side length to obtain the wall particle distribution points on the side length. For internal point distribution, the unit normal vector of the triangle face will be calculated first, and the projection surface will be determined by the size of the unit normal vector in each direction. If the size of the unit normal vector on the Z axis is n z If the absolute value is greater than 0.57735, the triangle patch will be projected onto the XY coordinate plane. If the unit normal vector is n on the Z axis z If the absolute value is less than 0.57735, then the size n of the unit normal vector on the Y axis is determined. y Whether the absolute value is greater than 0.57735, if so, project it to the XZ plane, otherwise project it to the YZ plane. Assuming that it is projected to the XY plane, the calculation process is described below: first determine a rectangular area covered by the maximum and minimum coordinates of the triangle in the XY direction. In this rectangular area, start from the lower left node of the rectangular area with the wall particle spacing, and fill and distribute points along the positive direction of the X axis and the positive direction of the Y axis with the wall particle spacing; then use the same direction method to determine whether the filling point is located in the triangle patch, and determine whether the minimum distance from the filling point to the three sides of the triangle is greater than or equal to the wall particle spacing. The filling point that meets these two conditions at the same time is used as the wall particle distribution point, and its Z coordinate is calculated according to the unit normal vector of the triangle patch. Through the side length distribution and internal distribution, uniform wall particles and unit normal vectors of wall particles with specified spacing distribution will be obtained. In addition, the present invention can give boundary values to wall particles at different wall surfaces as needed and set boundary conditions; the same applies to projection to other planes.
[0048] Figure 5 and Figure 7 The results of the wall particle placement of two complex geometries are shown. Figure 6 and Figure 8 The dimensionless particle number density is used to reflect the uniformity of the distribution of wall particles.
[0049] In order to best characterize the normal vector of the wall particles, the area average method will be used to calculate the unit normal vector of the wall particles, and the normal vector points uniformly to the inside of the geometric body.
[0050] As required, boundary values can be given to the wall particles at different wall surfaces. For example, in a circular tube, the Type value of the wall particles at the inlet can be specified as -1, and the Type value of the wall particles at the outlet can be specified as 1. The solver will distinguish these particles based on the inlet and outlet boundary conditions.
[0051] The Material value of the given wall particles is 2. The Material value will be used to identify wall particles and fluid particles.
[0052] 3) Through ray functions and particle collision functions, it is determined whether the particles are located inside the complex geometry, and the background grid size is adaptively adjusted to achieve internal fluid particle distribution at different resolutions. C++ combined with CUDA is used to implement a single GPU accelerated program for internal fluid particle distribution in complex geometry.
[0053] When filling internal fluid particles, you will specify the filling spacing, the minimum distance from the boundary surface, and a filling seed point in the fluid area. The internal fluid particle filling will start from the seed point and extend in three dimensions with the size of the filling spacing. The minimum distance from the boundary surface, that is, the internal fluid particles whose distance from the wall is less than this distance, will not be retained.
[0054] When using the filling algorithm to fill the fluid geometric model, GPU technology is used to accelerate the calculation in parallel. The thread index corresponding to the particle is assigned according to the particle number after the background grid index, and then the CUDA kernel function is called to perform the filling algorithm for each fluid particle. This realizes the parallelization of particle calculation and fully calls the rich thread parallel computing resources of the GPU. The calculation time is greatly accelerated for the filling scale of tens of millions of fluid particles. The overall boundary of the background grid will be determined according to the maximum or minimum of the three dimensions of the fluid area, and the background grid size will be adaptively adjusted according to the filling spacing to take into account both the speed and stability of the calculation. The particle number to be filled will be determined according to the index value generated by the background grid index where it is located.
[0055] After obtaining the particle serial number, the particle will be assigned to a corresponding CUDA thread, and subsequent judgments will be performed using the CUDA kernel function. The particle serial number value will perform a binary AND operation with the position to be filled. If the result is true, it means that the position has not been filled. Therefore, it will be judged whether the particle is located in the fluid area. The method used is the ray method, that is, a ray is emitted from the particle to the outside. If the number of intersections between the ray and the fluid boundary is an odd number, it means that the particle is located in the fluid area. Then use the collision function to determine whether the distance between the particle and the boundary of the fluid area meets the minimum distance from the boundary surface. The collision algorithm will find a boundary surface closest to the particle and calculate the distance between the two. If the distance is greater than or equal to the minimum distance from the boundary surface, the particle will be judged as an internal fluid particle and filled, and the position will be marked as having been filled. If Figure 3 and Figure 4 As shown, they are schematic diagrams of the results before and after particle sorting in a specific embodiment of the present invention;
[0056] Using CUDA kernel functions for parallel computing on each particle greatly accelerates and expands the computing speed. Even when filling tens of millions of fluid particles, the computing time can be completed at a very fast speed.
[0057] like Fig. 9 and Fig.10 As shown in the figure, the internal fluid particle filling result diagram of a complex aircraft with a filling spacing of 70 mm is obtained, in which the number of internal fluid particles is 1389438, the wall particle spacing is 70 mm, and the number of wall particles is 48701; Fig.11 and Fig.12 As shown, the internal fluid particle filling result of a complex aircraft with a filling spacing of 30mm is obtained, where the number of internal fluid particles is 1773021, the wall particle spacing is 70mm, and the number of wall particles is 48701; it shows the internal fluid particle filling results of a complex aircraft at different filling spacings.
[0058] At the same time, the internal fluid particle placement program developed can specify free surfaces in complex geometric bodies and fill the specified area with internal fluid particles. For this operation, you only need to give the boundary particles on the free surface. The free surface boundary particles will form a closed area with the wall boundary particles, and the filling will be carried out in this area, such as Fig.13 As shown, in one embodiment of the present invention, a filling result diagram is obtained in a wavy free surface specified in a dam break example, wherein the blue particles are internal fluid particles, the filling spacing is 30 mm, the number of internal fluid particles is 792586, and the Material value of the fluid particles is set to 0.
[0059] At this time, the internal fluid particles obtained have a certain sawtooth distribution near the wall surface. In a specific embodiment of the present invention, Fig.14 and Fig.15 As shown in the cross-sectional view of the filling result obtained by the model, the white particles are wall particles, the blue particles are internal fluid particles, the wall particle spacing is 30 mm, the number of wall particles is 43946, the filling spacing is 30 mm, and the number of internal fluid particles is 138702; therefore, the internal fluid particles need to be further adjusted for fitness.
[0060] 4) The improved discrete element method is used to adjust the fitness of internal fluid particles and eliminate the jagged distribution near the wall: a domain particle search list is established to optimize GPU memory storage, particle partitioning is performed to complete dynamic load balancing during multi-GPU calculations, and a gridless method preprocessor for multi-GPU parallel accelerated calculations is implemented using C++ and combined with CUDA and MPI. The GPU and MPI technology are used to use the improved discrete element method to adjust the distribution of internal fluid particles in parallel to accelerate calculations, and the calculation domain is partitioned to complete dynamic load balancing during multi-GPU calculations. The calculation domain is divided into regions along one direction, and particles in different regions are assigned to different MPI processes; each process will be assigned a GPU for calculation, and then a domain particle search list is established to optimize GPU memory storage and logical branches. The steps of establishing the domain particle search list include: calculating the background grid index where the internal fluid particles are located; reordering the fluid particles according to the background grid index where the particles are located, and exchanging the particle coordinate information at the same time, so that the fluid particles in the same background grid are continuously sorted, so that they will be stored continuously in memory; finding the starting and ending positions of the particles in each background grid in memory.
[0061] The internal fluid particle adjustment program will first read the wall particles and fluid particles, as well as the set particle diameter. is the filling spacing, particle density , model related parameters , the time interval of the time step , the number of processes used, size and other calculation parameters are used, and the fluid particles are partitioned according to the set number of processes, and the particle input file of the internal fluid particle adjustment program is generated. Then, when the internal fluid particle adjustment program solves the calculation, in order to quickly determine the domain particle of each particle, the search algorithm will use the background grid covering the entire calculation domain, and the grid index is sorted in a one-dimensional array, including the following steps:
[0062] 1) Calculate the background grid index where the internal fluid particles are located;
[0063] 2) Reorder the fluid particles according to the background grid index where the particles are located, and exchange the particle coordinate information at the same time, so that the fluid particles in the same background grid are sorted continuously, so that they will be stored continuously in the memory;
[0064] 3) Find the starting and ending positions of the particles in each background grid in memory;
[0065] Particle reordering will produce two advantages. First, it will reduce the logic branches within the warp and avoid warp differentiation. Second, it will optimize memory access.
[0066] After obtaining the neighboring particles of a particle, the normal elastic force on the particle is calculated using formulas (1) to (4), and its estimated position is calculated using formulas (5) and (6).
[0067] (1)
[0068] (2)
[0069] (3)
[0070] (4)
[0071] (5)
[0072] (6)
[0073] in, is the particle mass, is the particle density, is the circumference of a circle, is the particle diameter, is the elastic force stiffness, are model related parameters, is the time step interval, For particles With particles The normal contact length, For particles In the The current position vector at the time step, For particles In the The current position vector at the time step, It is a particle Due to the particle size within its diameter The normal elastic force on the contact is, For particles The resultant normal elastic force is For the The time step particle is The velocity generated by the resultant normal elastic force is For particles In the Time step position vector.
[0074] After many adjustments, the present invention proposes that the minimum distance from the boundary set in step 3) should be smaller than the filling spacing. For example, the minimum distance from the boundary is 0.8 filling spacing. In this way, the fluid particles near the wall will first be subjected to force due to their jagged distribution and their positions will be adjusted. Then, the change in the position of the particles near the wall will cause the forces on the particles further inside. After the calculation iteration converges, the overall fluid particles will be adaptively adjusted, and the adjusted spacing will be close to the required uniform distribution spacing.
[0075] In each time step calculation, particles will enter and exit several previously divided areas, so the background grid index will be updated, and the area division will also be updated to maintain a dynamic balance of the number of particles between areas. The communication between processes in different partitions is completed using MPI. When the calculation reaches the convergence condition, that is, the velocity of all internal fluid particles in the time step is 0, it means that the internal fluid particles have achieved an isotropic uniform distribution, and the fitness distribution of the internal fluid particles is completed. At this point, according to the input interface of the solver, the coordinates, unit normal vectors, boundary values, material identification values and other data of the corresponding wall particles and internal fluid particles can be output as the input file of the solver.
[0076] Fig.16 The results of the fitness distribution of internal fluid particles on a two-dimensional plane before and after adjustment are shown, with green representing boundary particles and purple representing fluid particles; Fig.17 and Fig.18 The distribution changes of internal fluid particles in a three-dimensional model before and after the fitness adjustment are shown in side view and top view respectively. The filling spacing is 30 mm and the number of internal fluid particles is 138702. It can be seen that after the adjustment, the jagged distribution of internal fluid particles near the wall is eliminated, satisfying the fitness distribution and achieving isotropy.
[0077] Based on the same inventive concept, a system for implementing the above-mentioned aptamer uniform gridless high-performance point distribution method is also provided in this embodiment, including:
[0078] A model processing module is used to draw a geometric model based on the physical geometric body to be drawn or to clean and repair the existing geometric model to obtain a wall geometric model and a fluid geometric model;
[0079] The outer wall particle distribution module is used to perform Delaunay triangulation on the surface of the wall geometry model to obtain triangular patches, and use the quadratic distribution method to perform secondary distribution on the triangular patches to obtain outer wall particles that meet the isotropic distribution of the wall geometry model surface and adapt to the distribution of physical characteristics of the physical geometry;
[0080] Internal fluid particle distribution module, used to fill the fluid geometry model through filling algorithm and parallel acceleration computing technology to obtain internal fluid particles;
[0081] The body-fitting distribution adjustment module is used to adjust the internal fluid particle distribution through the improved discrete element method and parallel accelerated computing technology, eliminating the jagged distribution of the internal fluid particles near the wall, so as to obtain the internal fluid particles that meet the isotropic distribution of the fluid geometric model space and adapt to the distribution of the physical characteristics of the physical geometric body;
[0082] The data output module is used to output the coordinates, unit normal vectors, and boundary values of the outer wall particles and the internal fluid particles to complete the body-fitting uniform gridless point distribution.
[0083] For the system embodiment, since it basically corresponds to the method embodiment, the relevant parts can refer to the partial description of the method embodiment, and the implementation methods of the remaining modules will not be repeated here. The system embodiment described above is only schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of the present invention. Ordinary technicians in this field can understand and implement it without paying creative work.
[0084] An embodiment of the present invention also provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the body-fitting uniform gridless high-performance point distribution method when executing the computer program.
[0085] The above-mentioned embodiments only express several implementation methods of the present invention, and the description is relatively specific and detailed, but it cannot be understood as limiting the scope of the present invention. For ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.
Claims
1. A physics-based, uniform, gridless, high-performance point distribution method, characterized in that: The following steps are involved: 1) Drawing a geometric model based on the physical geometric body to be drawn or cleaning and repairing an existing geometric model based on the physical geometric body, thereby obtaining a wall geometric model and a fluid geometric model of the physical geometric body; 2) Delaunay triangulation is performed on the surface of the wall geometric model to obtain triangular facets, and then the triangular facets are secondary distributed using the secondary distribution method to obtain outer wall particles that meet the isotropic distribution of the wall geometric model surface and adapt to the physical characteristics of the physical geometric body; the secondary distribution of the triangular facets using the secondary distribution method is specifically: side length distribution and internal distribution are performed on the triangular facets respectively, wherein the side length distribution is to divide the triangular facets along the direction of the side length according to the preset wall particle spacing, and the obtained nodes are the wall particles; the internal distribution is to calculate the three The unit normal vector of the triangle patch is selected, and the coordinate plane with the largest projection of the unit normal vector of the triangle patch is selected, and a rectangular area determined by the maximum coordinate and the minimum coordinate of the positive projection of the triangle patch is determined in the coordinate plane, starting from the lower left node of the rectangular area, along the positive direction of the two coordinate axes of the coordinate plane, filling points are arranged according to the set wall particle spacing, and the filling points located in the positive projection area of the triangle patch are retained, and the minimum distance from the filling point to the three sides of the triangle is controlled to be greater than or equal to the wall particle spacing, and the filling point is used as the wall particle, and the coordinate of the wall particle in another direction is obtained through the unit normal vector of the triangle patch; 3) Filling the fluid geometric model using a filling algorithm and parallel accelerated computing technology to obtain internal fluid particles; when filling the fluid geometric model, it is necessary to first determine the geometric boundaries of the filled fluid particles; when the geometric boundaries of the fluid particles are on the outer wall, the outer wall particles obtained in step 2) are used as the boundary particles of the fluid region, thereby determining the geometric boundaries of the fluid particles; when the geometric boundaries of the fluid particles are a part of the area inside the outer wall, the geometric model of this part of the area is operated in step 2) to obtain the boundary particles of the fluid region, thereby determining the geometric boundaries of the fluid particles; then, the filling algorithm and parallel accelerated computing technology are used to fill the fluid region with internal fluid particles for determining the geometric boundaries of the fluid particles; When filling the fluid region with the determined geometric boundary of the fluid particles with internal fluid particles using the filling algorithm and the parallel accelerated computing technology, the parallel accelerated computing technology first allocates the thread index corresponding to the fluid particle according to the internal fluid particle serial number after the background grid index, and then calls the CUDA kernel function to perform the filling algorithm for each fluid particle, thereby realizing the parallelization of particle filling; wherein the overall boundary of the background grid will be determined according to the maximum or minimum of the three dimensions of the fluid region, the background grid size will be adaptively adjusted according to the pre-set filling spacing, so as to take into account both the speed and stability of the calculation, and the serial number of the particle to be filled will be determined according to the index value generated by the background grid index where it is located; The filling algorithm comprises the following steps: 3.1) Define a seed point in the fluid region, which is also a filled fluid particle; 3.2) Determine whether the six adjacent points separated by a filling interval from the seed point meet the conditions of the ray function and the particle collision function at the same time. If they do, it means that the point meets the filling condition and will be filled as a fluid particle, and the point is marked as filled. If the filling condition is not met, the point will be marked as not needing to be filled to avoid repeated judgment; the six adjacent points are located on a three-dimensional coordinate axis with the seed point as the origin; 3.3) Repeat step 3.2) with the newly filled fluid particles as seed points until the entire fluid region is filled; 4) Use the improved discrete unit method and parallel accelerated computing technology to adjust the distribution of internal fluid particles, so as to eliminate the jagged distribution of internal fluid particles and obtain internal fluid particles that meet the isotropic distribution of the fluid geometric model space and adapt to the distribution of physical characteristics of the physical geometric body; the improved discrete unit method is specifically as follows: when calculating the force on the internal fluid particles, it is assumed that the internal fluid particles are all rigid particles with a particle diameter of and a particle density of, and the force on the internal fluid particles is the resultant normal elastic force obtained by other particles in contact with the internal fluid particles within the diameter range, and the gravity, tangential elastic force, normal damping force and tangential damping force on the internal fluid particles are shielded; the particles only move in parallel after being subjected to force, and the particle velocity is reset to 0 after the new position of the particles is calculated at each time step; when the uniform distribution condition is reached, the resultant normal elastic force on all fluid particles is 0, and no movement occurs, and this is the moment when the calculation converges; 5) Output the coordinates, unit normal vectors, and boundary values of the outer wall particles and the internal fluid particles to complete the body-fitting uniform gridless point distribution.
2. The method for high-performance uniform gridless point distribution of aptamers according to claim 1, characterized in that: The ray function is: send a ray from the point to be filled, determine the number of times the ray intersects with the boundary surface of the fluid area, if it is an odd number, it means that the point is located in the fluid area; the condition for satisfying the particle collision function is: find a boundary surface closest to the point to be filled, calculate the distance from the point to be filled to the boundary surface, if the distance between the two is greater than or equal to the set minimum boundary distance, it means that the point satisfies the particle collision function condition.
3. The method for high-performance uniform gridless point distribution of aptamers according to claim 1, characterized in that: When the improved discrete element method and parallel accelerated computing technology are used to adjust the internal fluid particle distribution in the step 4), the region is divided according to the fluid boundary, and the particles in different regions are assigned to different MPI processes. A GPU is assigned to each process for calculation. When each particle is calculated using the improved discrete element method, the CUDA kernel function is called to perform calculation, thereby eliminating the jagged distribution of the internal fluid particles.
4. A system for implementing the aptamer uniform gridless high-performance point distribution method according to claim 1, characterized in that: include: A model processing module is used to draw a geometric model based on the physical geometric body to be drawn or to clean and repair the existing geometric model to obtain a wall geometric model and a fluid geometric model; The outer wall particle distribution module is used to perform Delaunay triangulation on the surface of the wall geometry model to obtain triangular patches, and use the quadratic distribution method to perform secondary distribution on the triangular patches to obtain outer wall particles that meet the isotropic distribution of the wall geometry model surface and adapt to the distribution of physical characteristics of the physical geometry; Internal fluid particle distribution module, used to fill the fluid geometry model through filling algorithm and parallel acceleration computing technology to obtain internal fluid particles; The body-fitting distribution adjustment module is used to adjust the internal fluid particle distribution through the improved discrete element method and parallel accelerated computing technology, eliminate the jagged distribution of the internal fluid particles near the wall, and obtain the internal fluid particles that meet the isotropic distribution of the fluid geometric model space and adapt to the physical characteristics of the physical geometric body; The data output module is used to output the coordinates, unit normal vectors, and boundary values of the outer wall particles and the internal fluid particles to complete the body-fitting uniform gridless point distribution.
5. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the aptamer uniform gridless high-performance point distribution method as described in any one of claims 1 to 3 are implemented.
Citation Information
Patent Citations
High-precision geometric model particle discretization method, device and system based on meshless method
CN113761778A
A method and system for solving complex geometric flow domains with single-layer walls based on meshless method
CN113761812B
A coupling physics field rapid solution-oriented true and false dual particle model modeling method
CN114036815A
Geometric model boundary and internal particle uniform arrangement method and related device
CN116342837A
Automated geometric representation of discrete point sets
US8854366B1
