Research On Fast Simulation Of Realistic Smoke Based On Vortex Particles

In the computer graphics community,fluid simulation is playing a more and more important role because of its wide application in many fields.The traditional process-based simulation method is gradually replaced by the more realistic physical-based simulation method.Physical-based fluid simulation method obtains various attributes of fluid at each time step by solving the governing equation of fluid motion,and then renders the animation effect frame by frame through rendering technology,which obviously shows better realistic and complex simulation results than the traditional simulation method based on parameter modeling.Although this method has made remarkable progress in recent years,there are still some shortcomings: for example,there is a certain numeri cal dissipation in solving the control equation,which leads to the loss of some details;the computational overhead is too big,and most of them can not be applied to areas with high response requirement.In the view of the above problems,the main research contents of this paper are as follows:(1)A simulation framework based on vortex particles is proposed.Solving the fluid control equation directly based on velocity field is easy to produce numerical dissipation,thus losing part of the fluid details,which reduces the reality of the simulation.Vortex particles are a tool for simulating vorticity,which represents the physical quantity of fluid rotation.Part of the fluid simulation work adds eddy particles to increase fluid details.Considering that the velocity field generated by the vortices is divergence-free,a simulation framework based on the vortices is proposed in this paper.By solving it,the details of the vortices can be better maintained and the numerical dissipation can be reduced.(2)Accelerate fluid simulation based on GPU architecture.In the process of solving fluid simulation based on vortex particles,it is necessary to convert the vorticity field into the velocity field,which includes the solution of Poisson equation.Solving Poisson equation takes up most of the time of fluid simulation.The preconditioned conjugate gradient method is used to solve Poisson equation.Through the study of the solving process,it is found that there are a lot of parallelizable computing,such as matrix multiplication and inner product operation.The solving process is transferred to GPU for parallel processing.paging locking,grid design,merge memory access,instruction optimization and flow processing used in optimization all accelerate the speed of fluid simulation to varying degrees.In this paper,the process of solving Poisson's equation is optimized to fit the parallel architecture of GPU by using the simulation method based on vortex particles.The experimental results show that,compared with the existing methods for solving the velocity field directly,the fluid simulation based on vortex particles exhibits more details of the vortices and is closer to the reality.The simulation speed can be accelerated by more than ten times as fast as possible,and the simulation speed can be improved significantly.