Fast Animation Of High-Resolution Fluids

With the development of computer technology, the physically-based fluid simulation method has made great achievements in relevant fields, such as film, game, virtual reality etc. Currently, the Eulerian grid-based technique has been able to simulate visually interesting results. However, this traditional method still has a lot of difficulties and one of bottlenecks is the low efficiency. So, in this thesis, a novel fluid simulation framework is proposed for fast animation of high-resolution fluids.First, we introduce the framework of Eulerian grid-based method, including advection, body force and projection. In addition, we analyze the limitations and bottlenecks of the conventional fluid simulation technique.Second, in view of the bottlenecks of the traditional technique, this paper presents a high-low resolution grid coupling simulation framework. In our framework, we firstly advect velocity field in high-resolution grid, and map this velocity field to coarse uniform grid by spline kernel function, then make the coarse grid velocity field divergence free by solving coarse grid Poisson equation, finally use MLI-upsampling(Modified Linear interpolation upsampling) method to map these velocities back to our fine simulation grid which still keeps the fine velocity field divergence free. The results show that our framework is effective and efficient because we accelerate simulation by solving one coarse grid Poisson equation and an upsampling operator instead of solving an expensive large Poisson equation. In addition, this paper also analyzes the limitations of MLI-upsampling operator.Finally, combined with Signal system theory, this paper proposes a FFT-upsampling(Fast Fourier Transform upsampling) method to substitute MLI-upsampling because FFT-upsampling method can provide near-perfect resampling for signals and it is also well suited for flow field reconstruction. Therefore, we present a high-low resolution grid coupling simulation framework based on FFT-upsampling operator. In this framework, we introduce FFT-advection(Fast Fourier Transform advection) technique, in addition, three different types of boundary conditions are dealt with by proper method. Experimental results show that, our method can simulate high-resolution fluid rapidly and produce the same level of details as the traditional high-resolution solver does. Moreover, compared with the results of conventional advection method, our FFT-advection technique has significant advantages. It can reduce the numerical dissipation and preserve the details in fluid simulation.