Developing The Immersed Boundary-Lattice Boltzmann Method To Simulate Flow Fields In Hydropower Engineering

The fluid-structure-interaction(FSI)is ubiquitous in scientific research,engineering application,and our daily life.There is a mutual dependence between the fluid and the structure parts,that is the flow patterns depend on the shape of the immersed structure and its motion,and the deformation of the structures is influenced by the hydrodynamic forces acted upon them by the surrounding fluid.Developing a set of precise mathematic formulas that can truly predict the FSI process and all its relevant details is in high demand in scientific research field and industry.Due to the intrinsic mechanic nonlinearity and the time-dependent nature of the problems,analytic solutions are very hard to derive.On the other hand,during the past decade,the numerical method has shown to be a powerful tool for simulating the FSI problems,and specially,as one of the hotspots in current research field,the immersed boundary-lattice Boltzmann(IB-LB)method is proven to be an accurate,stable and robust numerical method.It has been over a decade since the IB-LB method was first proposed,and many numerical issues exhibited by the original method have been properly and successfully solved,such as the non-slip boundary condition cannot be accurately satisfied.However,when it comes to applications,there are three drawbacks exhibited by the current IB-LB method which prevents it from successfully applying to engineering applications.They are:low computational efficiency,poor numerical stability when applied to simulate deformable structure with very high bending rigidity,and low numerical accuracy when simulating temperature fields.The aim of this dissertation is to developing current IB-LB method,and promote its application to hydropower engineering.The major works are summarized as follows:1.An approach to accelerate the simulations of FSI problems by implementing the IB-LB method on GPU platform is proposed.For the lattice Boltzmann(LB)method,the one-dimensional(1D)blocks and shared memory are used to solve the misalignment problems in global memory transactions.For the immersed boundary(IB)simulations,all sub-processes,including the velocity interpolation,boundary force computation and force spreading,are implemented by different kernels.Besides,to exploit the parallelism of Lagrangian points without violating the coalesced and aligned access pattern,the 1D arrays are used to store the information of all Lagrangian points.The performance of the proposed method is verified by a typical case,and the result shows that all the GPU kernels are memory bound and the achieved memory efficiencies for different kernels are all above 60%.2.To accurately simulate the heat transfer problems by the thermal IB-LB method,an iterative source correction procedure is introduced.The objective of this method is to remove the temperature jump across the immersed boundary when the conventional IB-LB method is applied to thermal flow simulations.The present method is based on Cheng's scheme which can easily incorporate the space-and time-dependent source term into the energy equation.Within every timestep,the next timestep's heat source term in Cheng's scheme is treated as unknown and is iteratively corrected to match its actual value.It is shown that the boundary temperature jump can be effectively removed for a certain range of LB relaxation time,while the first-order spatial convergence of the IB method is still maintained.3.Based on the former GPU acceleration,a ID shallow water coupled to the 3D single-phase free-surface lattice Boltzmann(SFLB)model is developed and applied to simulate the transient process of open channel flow for the first time.The proposed method has alleviated the shortages such as unable to include the transverse/vertical movement of bore wave and three-dimensional(3D)flow near the hydrostation intake when using traditional surge wave method or shallow water equation based method.The 3D topography elevation is extracted from the open-source GIS software,and through zooming and interpolation procedures,the digital elevation model can be readily used by the SFLB model,overcoming the defect of using bead roughness to represent the riverbed friction in traditional methods.After verfired by several simple cases,the proposed model is used to simulate a real engineering case-the transient process of Gezhuoba project when the Dajiang power units is in load rejection process.The time variation of water surface over the entire reservoir,the time series of water depths at several key monitoring points and the maximal/minimal bore wave are computed.Through Fourier analysis,the computed bore frequencies are in good agreement with those derived by physical laws,demonstrating that the SFLB model can accurately simulate the transient process under real engineering conditions.4.Serving as an application to hydropower engineering,the transient process of a Kaplan turbine at load rejection condition are simulated by the GPU accelerated IB-LB method.The guided vane and hydro turbine are treated as immersed boundary,while the spiral case,draft tube and pressure pipes are implemented by an efficient volumetric LB scheme.The histories of the turbine's angular velocity,moment and axial force are computed,and through it,the maximal angular velocity and the time in which the maximal angular velocity is reached are found.The results are found to qualitatively agree with those computed by CFD software,demonstrating that the proposed method has great prospect in numerical simulation of hydraulic machinery,and laying the foundation for real engineering simulations.Aiming at improving the stability,boundary accuracy and computational efficiency,this thesis develops current IB-LB method and applies it to the simulation of flow fields in hydraulic engineering for the first time.The future work may focus on developing high Reynolds turbulent model within the IB-LB framework,and by doing so,greatly promoting its usability in real engineering problems.