CN109753726A - A kind of globe mill medium movement simulating method based on Bounding Box searching method and GPU - Google Patents
A kind of globe mill medium movement simulating method based on Bounding Box searching method and GPU Download PDFInfo
- Publication number
- CN109753726A CN109753726A CN201910006798.3A CN201910006798A CN109753726A CN 109753726 A CN109753726 A CN 109753726A CN 201910006798 A CN201910006798 A CN 201910006798A CN 109753726 A CN109753726 A CN 109753726A
- Authority
- CN
- China
- Prior art keywords
- particle
- gpu
- cellcontent
- calculating
- bounding box
- 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
Links
Abstract
The globe mill medium emulation mode based on boundary cassette method and GPU that the invention discloses a kind of.The present invention utilizes the many-core and multithreading advantage of GPU, the Large-scale parallel computing resource of GPU is called and distributed by the calculating kernel function based on discrete element method (DEM), realize the relevant calculation that a thread carries out a particle, judge including contact, the update of the states such as Force Calculation and position and speed, successively carry out time step iterative calculation, efficiency declines more problem when system emulation biggish to particle size differential for fundamental space decomposition method simultaneously, using a kind of more efficiently slightly colliding to searching method based on Bounding Box, to which the particle system simulation of fairly large particle is realized in realization on a common computer while guaranteeing computational efficiency when particle size distribution is uneven.The method efficiently solves the problems, such as that DEM computation model calculation amount is huge, and to the analog simulation for realizing extensive particle system in a short time and its simulation in real time is of great importance.
Description
Technical field
The present invention relates to belonging to particle system simulation technical field, more particularly to one kind based on Bounding Box searching method and
The globe mill medium movement simulating method of GPU.
Background technique
Ball mill is the key equipment that grinding is carried out to material, is the national warp such as mine, chemical industry, metallurgy, building materials and thermoelectricity
Most important grinding attachment in Ji basic activity, however ball mill huge energy consumption, and be a kind of disintegrating apparatus of inefficiency,
Only about 0.1%~1% energy is used for broken and grinding material, other energy in electric energy consumed by ball mill grinding process
Amount is all converted to noise, vibration and heat etc. and dissipates.Therefore, the efficiency for improving ball mill is the key that reduce total energy consumption, to drop
Low factory cost, important in inhibiting of increasing economic efficiency.
With the development of computer technology in recent years, carry out analog simulation using computer and become to study the solutions of many problems
Certainly method, due to the flexibility of computer simulation, many scholars by discrete element method (Discrete Element Method,
DEM) method is applied to the motion simulation of particle system, and still, the particle movement simulating method based on DEM is due to its own theory
Latent defect, computational efficiency is lower, so the numbers of particles that can be emulated is limited.As the sum of general-purpose computations GPU is based on GPU's
The development of parallel computational model, the many-core of GPU calculate advantage and powerful double precision computing capability gradually in scientific algorithm and people
Work smart field embodies, and solves the problems, such as the problem of many field computation-intensives calculate overlong time, and GPU is light due to it
The multi-threaded architecture of quantization makes the calculating of repeated intensity become simple and quick.However the media particle in practical ball mill
Size is unevenly distributed, and usually needs to study the influence that different medium matches Compare System, this will lead to the efficiency drop of emulation
It is low many, therefore the parallel computation mode of CUDA+GPU will be used herein, based on a kind of new spatial decomposition method and thick phase
It is imitative typically now to carry out size extensive globe mill medium movement unevenly distributed in fact for grain collision search method on computer
Very, substantially shorten and calculate the time, lay the foundation to study various phenomenons and the rule of real industrial particle system.
Summary of the invention
To solve the above problems, the present invention provides a kind of ball mill based on Bounding Box searching method and GPU
Medium motion emulation mode, and the search that slightly collides in the calculating of DEM model is improved using Bounding Box searching method, it can be realized
The high-speed simulation of extensive medium motion unevenly distributed to size, overcome media size gap it is larger when simulation velocity under
More defect drops, for this purpose, the present invention provide it is a kind of based on the globe mill medium of Bounding Box searching method and GPU move
Emulation mode includes the following steps, it is characterised in that:
S1, the simulation model based on discrete element method is established；
Material parameter and the coefficient of rolling friction of storeroom, the coefficient of sliding friction, restorer are set at the end CPU first
Number, constructs the geometric mould of large and small steel ball particle and roller；Then the generating mode for determining particle, if it is from file
Graininess information is read, then reads in file content, otherwise generates particle in the form of a grid, calculates particle position, imparting
Grain initial velocity etc. simultaneously stores；The size of space grid division is finally determined according to smallest particles diameter；
S2, the memory space that the information in S1 is imported to the end GPU；
In the relevant information of the end GPU distribution memory space storage particle and solid；
S3, thread distribution and relevant calculation are carried out using kernel function；
The kernel function for calculating correlation step is established, including grid position calculates and cryptographic Hash calculates, particle position sorts, thick
Collide detection, the detection that carefully collides, numerical density calculating and graininess update；
S4, it updates particle position and is visualized；
Using OPENGL by particle and cylinder visualize, and update its state by preset time step number；Determine program
Whether all default calculating and recording-related information are completed.
As a further improvement of that present invention, each calculating step in step S3 is to call GPU to complete by kernel function, is realized
The relevant calculation of one thread process, one unit slightly collides detection-phase in particle, and it is single for using with smallest particles diameter
The spatial decomposition method of position faster realizes thick mutually judgement, and concrete methods of realizing is:
S31, particle sequence is set as { x(n), (n=1,2 ..., N) }, wherein N is total number of particles, assigns it to GPU's
CUDA core is calculated, and is carried out due to calculating by basic unit of particle, so one-dimensional grid is divided into, wherein each
Thread block uses 64 threads, therefore thread number of blocks is N/64；
S32, according to the thread grid that sets, in conjunction with each kernel function for calculating step call the hardware resource of GPU into
Row calculates, wherein slightly collide detection method by taking two-dimensional space as an example, the diameter based on smallest particles when dividing space, according to
The coordinate and boundary position of grain calculate cellular unit shared by the particle, it is assumed that cellular unit shared by the particle that number is 1
For particlecell [1]={ 1,2,3,9,10,11 }, then the content that this particle is added to all shared cellular units is arranged
Table cellcontent [1]={ 1 }, cellcontent [2]={ 1 }, cellcontent [3]={ 1 }, cellcontent [9]
={ 1 }, cellcontent [10]={ 1 }, cellcontent [11]={ 1 }, other particles and so on are being carried out with particle
For unit slightly collide detect when, if there is other particles in cellular unit shared by the particle, be considered as potential collision pair, add
It is added in the neighbor list of the particle.
A kind of globe mill medium movement simulating method based on Bounding Box searching method and GPU of the present invention, including it is following excellent
Point:
1, the present invention can be realized carries out fairly large globe mill medium motion conditions mould under common hardware condition
It is quasi-.
2, present invention globe mill medium system non-uniform for media size is able to maintain higher computation efficiency,
Reduce the sensibility being distributed to media size.
Detailed description of the invention
Fig. 1 is flow chart of the invention；
Fig. 2 is the single calculation flow chart for calculating step of the invention；
Fig. 3 is the thick phase particles collision search routine figure the present invention is based on Bounding Box；
Fig. 4 is Bounding Box method schematic diagram of the present invention.
Specific embodiment
Present invention is further described in detail with specific embodiment with reference to the accompanying drawing:
The present invention provides a kind of globe mill medium movement simulating method based on Bounding Box searching method and GPU, and uses
Bounding Box searching method improves the search that slightly collides in the calculating of DEM model, can be realized big rule unevenly distributed to size
The high-speed simulation of mould medium motion, overcome media size gap it is larger when simulation velocity decline more defect.
1) simulation model based on discrete element method is established
Material parameter and the coefficient of rolling friction of storeroom, the coefficient of sliding friction, restorer are set at the end CPU first
Number, constructs the geometric mould of large and small steel ball particle and roller；Then the generating mode for determining particle, if it is from file
Graininess information is read, then reads in file content, otherwise generates particle in the form of a grid, calculates particle position, imparting
Grain initial velocity etc. simultaneously stores；The size of space grid division is finally determined according to smallest particles diameter.
2) information in S1 is imported to the memory space at the end GPU
In the relevant information of the end GPU distribution memory space storage particle and solid, thread grid is determined.If particle sequence
For { x(n), (n=1,2 ..., N) }, wherein N is total number of particles, and the CUDA core (stream multiprocessor) for assigning it to GPU carries out
It calculates.It is carried out due to calculating by basic unit of particle, so being divided into one-dimensional grid, wherein per thread block uses 64
A thread (needing according to actual hardware and system debug, the general smaller multiple computational efficiency for selecting 32 is higher), therefore thread
Number of blocks is N/64 (taking biggish closest to integer).
3) thread distribution and relevant calculation are carried out using kernel function
3.1) according to thread grid establish calculate correlation step kernel function, as shown in figure 3, include grid position calculate and
Cryptographic Hash calculates, and then carries out particle position sequence according to particle cryptographic Hash.
3.2) it slightly collides detection (establishing the neighbor lists of particle), when slightly colliding, using boundary cassette method, specifically
Step such as Fig. 3.By taking two-dimensional space as an example, the diameter based on smallest particles when dividing space, as shown in figure 4, according to the seat of particle
Mark and boundary position calculate cellular unit shared by the particle, it is assumed that the particle number in the upper left corner is 1 in Fig. 4, then shared by it
Cellular unit be particlecell [1]={ 1,2,3,9,10,11 }, this particle is then added to all shared cellular lists
Contents list cellcontent [1]={ 1 } of member, cellcontent [2]={ 1 }, cellcontent [3]={ 1 },
Cellcontent [9]={ 1 }, cellcontent [10]={ 1 }, cellcontent [11]={ 1 }, other particles are with such
It pushes away.Carry out using particle as unit slightly collide detect when, if in cellular unit shared by the particle contain other particles, depending on
It is potential collision pair, is added in the neighbor list of the particle, such as the particle 1 and particle 2 in figure, since they have jointly
Cellular unit, therefore they are potential collisions pair, and duplicate particle pair will be judged in program, will not repeat to record.
3.3) carefully collide detection (determining whether to collide and calculate contact distance), obtained according to last step
Grain neighbor lists, to each potential collision to calculating whether its particle centroid distance is less than the sum of particle radius and judges whether to occur
Contact, and store the particle pair being in contact.
3.4) numerical density calculates and graininess updates.According to resultant force suffered by DEM contact model calculating particle, and use Europe
Integral is drawn to update its state change in this time step.
S4, it updates particle position and is visualized
Using OPENGL by particle and cylinder visualize, and update its state by preset time step number；Determine program
Whether all default calculating and recording-related information are completed.
The above described is only a preferred embodiment of the present invention, being not the limit for making any other form to the present invention
System, and made any modification or equivalent variations according to the technical essence of the invention, still fall within present invention model claimed
It encloses.
Claims (2)
1. a kind of globe mill medium movement simulating method based on Bounding Box searching method and GPU, includes the following steps, feature
It is:
S1, the simulation model based on discrete element method is established；
Coefficient of rolling friction, the coefficient of sliding friction, the recovery coefficient of material parameter and storeroom, structure are set at the end CPU first
Build the geometric mould of large and small steel ball particle and roller；Then the generating mode for determining particle, if it is the reading from file
Grain status information, then read in file content, otherwise generates particle in the form of a grid, calculates particle position, and it is initial to assign particle
Speed etc. simultaneously stores；The size of space grid division is finally determined according to smallest particles diameter；
S2, the memory space that the information in S1 is imported to the end GPU；
In the relevant information of the end GPU distribution memory space storage particle and solid；
S3, thread distribution and relevant calculation are carried out using kernel function；
The kernel function for calculating correlation step is established, including grid position calculates and cryptographic Hash calculates, particle position sequence, slightly collides
Detection, the detection that carefully collides, numerical density calculating and graininess is hit to update；
S4, it updates particle position and is visualized；
Using OPENGL by particle and cylinder visualize, and update its state by preset time step number；Whether determine program
Complete all default calculating and recording-related information.
2. a kind of globe mill medium movement simulating method based on Bounding Box searching method and GPU according to claim 1,
It is characterized by: each calculating step in step S3 is to call GPU to complete by kernel function, one list of a thread process is realized
The relevant calculation of member slightly collides detection-phase in particle, uses the spatial decomposition method as unit of smallest particles diameter, more
Fast to realize thick mutually judgement, concrete methods of realizing is:
S31, particle sequence is set as { x(n), (n=1,2 ..., N) }, wherein N is total number of particles, assigns it to the CUDA core of GPU
The heart is calculated, and is carried out due to calculating by basic unit of particle, so be divided into one-dimensional grid, wherein per thread block
Using 64 threads, therefore thread number of blocks is N/64；
The thread grid that S32, basis are set calls the hardware resource of GPU to be counted in conjunction with each kernel function for calculating step
It calculates, wherein detection method is slightly collided by taking two-dimensional space as an example, the diameter based on smallest particles when dividing space, according to particle
Coordinate and boundary position calculate cellular unit shared by the particle, it is assumed that cellular unit shared by the particle that number is 1 is
Particlece ll [1]={ 1,2,3,9,10,11 } then arranges the content that this particle is added to all shared cellular units
Table cellcontent [1]={ 1 }, cellcontent [2]={ 1 }, cellcontent [3]={ 1 }, cellcontent [9]
={ 1 }, cellcontent [10]={ 1 }, cellcontent [11]={ 1 }, other particles and so on are being carried out with particle
For unit slightly collide detect when, if there is other particles in cellular unit shared by the particle, be considered as potential collision pair, add
It is added in the neighbor list of the particle.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910006798.3A CN109753726A (en) | 2019-01-04 | 2019-01-04 | A kind of globe mill medium movement simulating method based on Bounding Box searching method and GPU |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910006798.3A CN109753726A (en) | 2019-01-04 | 2019-01-04 | A kind of globe mill medium movement simulating method based on Bounding Box searching method and GPU |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109753726A true CN109753726A (en) | 2019-05-14 |
Family
ID=66405334
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910006798.3A Pending CN109753726A (en) | 2019-01-04 | 2019-01-04 | A kind of globe mill medium movement simulating method based on Bounding Box searching method and GPU |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109753726A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111539103A (en) * | 2020-04-20 | 2020-08-14 | 东南大学 | Method for quantifying segregation degree of particles of ball mill based on lacey method |
| CN113641470A (en) * | 2021-07-23 | 2021-11-12 | 曙光云计算集团有限公司 | GPU-based thread arrangement method, device, equipment and storage medium |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103324780A (en) * | 2012-12-20 | 2013-09-25 | 中国科学院近代物理研究所 | Particle flow simulation system and method |
-
2019
- 2019-01-04 CN CN201910006798.3A patent/CN109753726A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103324780A (en) * | 2012-12-20 | 2013-09-25 | 中国科学院近代物理研究所 | Particle flow simulation system and method |
| WO2014094410A1 (en) * | 2012-12-20 | 2014-06-26 | 中国科学院近代物理研究所 | Particle flow simulation system and method |
Non-Patent Citations (2)
| Title |
|---|
| 常新正: "基于GPU的颗粒离散元计算方法研究", 《万方数据库学位论文库》 * |
| 李鸿程: "基于离散单元法的球磨机工作参数优化", 《中国优秀硕士学位论文全文数据库（电子期刊）工程科技Ⅰ辑》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111539103A (en) * | 2020-04-20 | 2020-08-14 | 东南大学 | Method for quantifying segregation degree of particles of ball mill based on lacey method |
| CN113641470A (en) * | 2021-07-23 | 2021-11-12 | 曙光云计算集团有限公司 | GPU-based thread arrangement method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | Real-time collision culling of a million bodies on graphics processing units | |
| EP2521113A1 (en) | Simplified smoothed particle hydrodynamics | |
| US8554527B2 (en) | Particle simulator and method of simulating particles | |
| US10354099B2 (en) | Particle simulation device, particle simulation method, and particle simulation program | |
| Pazouki et al. | Parallel collision detection of ellipsoids with applications in large scale multibody dynamics | |
| Tracy et al. | Efficient large-scale sweep and prune methods with AABB insertion and removal | |
| CN109753726A (en) | A kind of globe mill medium movement simulating method based on Bounding Box searching method and GPU | |
| Hermann et al. | Ray-traced collision detection for deformable bodies | |
| JP2010238030A (en) | Particle simulation device and particle simulation method | |
| US8717382B2 (en) | Method of detecting collisions in graphics processing unit | |
| Fan et al. | A hierarchical grid based framework for fast collision detection | |
| Cintra et al. | A hybrid parallel DEM approach with workload balancing based on HSFC | |
| Lettich et al. | Processing streams of spatial k-NN queries and position updates on manycore GPUs | |
| Ye et al. | Modeling and rendering of real-time large-scale granular flow scene on GPU | |
| CN111104728B (en) | Simulation prediction method and device for structural surface wear | |
| Mainzer et al. | Collision detection based on fuzzy scene subdivision | |
| Kim et al. | Efficient collision detection among moving spheres with unknown trajectories | |
| Hesam et al. | Gpu acceleration of 3d agent-based biological simulations | |
| CN107610231A (en) | A kind of dynamic crash detection method | |
| EP2509009A1 (en) | Particle simulator and method of simulating particles | |
| Zhao et al. | A data allocation strategy for geocomputation based on shape complexity in a cloud environment using parallel overlay analysis of polygons as an example | |
| Geleri et al. | Parallelizing broad phase collision detection algorithms for sampling based path planners | |
| Venetillo et al. | GPU-based particle simulation with inter-collisions | |
| Yalçın et al. | GPU algorithms for diamond-based multiresolution terrain processing | |
| Baryshev et al. | Optimization of particle search algorithm for CFD-DEM simulations |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190514 |