Research About The Parallel Algorithm Of Multiphase Lattice Boltzmann Method And Simulation Of Droplet Movement

Computer simulation which has the advantage of high precision,high efficiency,and low cost,is a bridge between theoretical analysis and experiment research and also an effective means to investigate complex nonlinear processes deeply.High-Performance Computing(HPC)has advanced by leaps and bounds in recent years,and it also prompts computer simulation to become a new paradigm of scientific research.Especially multi-core parallel technology typified by graphics Processing Unit(GPU)has dramatically promoted the size and efficiency of scientific computation.In the field of multiphase flow,the lattice Boltzmann method(LBM)not only satisfies thermodynamic consistency and Galilean invariance,but also has the characteristics that the phase change can be achieved by automatic evolution,without resorting to any techniques to track interface,and becomes a reliable simulation technology to study the phase transformations and flow of multiphase flow.The combination of GPU and LBM will develop into a very powerful tool for describing the details of the complex movement of multiphase flow and accurately capturing the process of nonlinear evolution.The droplet movement widely found in nature and industrial applications is a typical multiphase flow system.It is relevant not only to the interaction of gas-liquid-solid three-phase but also to the chemical composition and the physical structure in interface area.The process of droplet movement is nonlinear.To understand the peculiar way of moving droplets in nature and precisely control droplet movement in various application areas,this research proposes a parallel algorithm for multiphase LBM and uses this algorithm to study the complex rules of droplet dynamics.The main contribution and innovations of this paper are listed as follows:(1)The parallel algorithms of the 2D and 3D chemical-potential multiphase LBM solver are presented based on GPU.Firstly,in the parallel algorithms of 2D multiphase LBM,a data load algorithm in which the periodic and wetting boundary conditions are considered is designed with overlapped tiling,and data reuse is realized with shared memory in gradient computation.Then,in parallel algorithms of 3D multiphase LBM,the autotuning of thread block dimensioning is obtained by particle swarm optimization.Referring to the block pattern of standard Stencil calculations,we compare the effects of the parallel algorithm of multiphase LBM in two blocking schemas of 3D-Blocking and 2.5D-Blocking.And an effective kernel fusion is proposed by analyzing data dependence between the kernel function.The results show that the performance of the parallel algorithm of 2D multiphase LBM is double the GPU algorithm without optimization of the gradient computation,and is more the 400 times of the CPU algorithm.The performance of the parallel algorithms of 3D multiphase LBM is more than 60 times of the CPU+Open MP algorithm.(2)Based on mesoscopic LBM and the theory of surface wetting,we propose the algorithms to quantitatively analyze the macroscopical force acting on the drop and the microcosmic force acting at the three-phase contact line as the droplet is moving on heterogeneous surfaces.We construct the model of droplet sliding on an inclined heterogeneous surface using the parallel optimization algorithm of 2D,and three types of depinning behavior are reproduced.It is observed that at the macroscopic scale the dynamic equilibrium of the droplet before rolling reflects the competition between gravity and the capillary force in the whole drop,and the unilateral depinning is accompanied by a sudden fluctuation in the capillary force.At the microscopic scale,the local force balance in the contact line region is maintained by the unbalanced Young’s force and the resistance of the substrate in the slow-moving stage,while the abrupt change of unbalanced Young’s force provides the driving force for the motion of the contact line in the fast-moving stage.These algorithms are appropriate for analyzing the force in different phenomena of droplet motion taking place on heterogeneous surfaces,such as,drop rebound,drop splitting,and drop sliding on a curved surface.(3)Based on the parallel algorithm of 3D multiphase LBM,we construct the numerical model of passive droplet-splitting on two different heterogeneous surfaces and study the mechanisms of droplet-splitting.For a hydrophilic surface decorated with a superhydrophobic strip,two splitting cases with single and double-liquid-bridge are contrasted and analyzed.It is found that the splitting time did not linearly shorten with the increase in strip width.The unbalance surface tension at the hydrophobic/hydrophilic surface during the droplet splitting process is shown quantitatively by the above-mentioned tool of force analysis.For the Y-shaped wettability-gradient track,the role of the track structure is discussed and the wedge-shaped branch is introduced to shorten the splitting time of the droplet.(4)We present an EOS(equation of state)-based multiphase lattice Boltzmann model with large density ratios and the corresponding algorithm.The theoretical analyses and numerical results showed that the present model satisfies thermodynamics,Galilean invariance,and the Young-Laplace equation.We use the model to study the rupture of liquid plugs and the formation of aerosol microdroplets in the pulmonary airway during breathing motions.And we present the algorithms to quantitatively analyze the wall pressure,wall shear stresses,and momentum distributions.The results showed that more force that acts on the airway wall appears after the rupture.The results of momentum distributions reveal that the aerosol takes away most of the momentum,following retraction of the remainder of the broken liquid plug,and a considerable impact force to the airway wall.