Hybrid Methods For The Computational Aeroacoustics Based On Acoustic Propagation Equations

Hybrid methods for the computational aeroacoustics based on acoustic propagation equations arehotspot and difficulty in computational aeroacoustics research. Typically, the methods forcomputational aeroacoustics can be divided into two subcategories, direct methods and hybridmethods. Direct methods such as direct numerical simulation and large eddy simulation are far fromsolving actual engineering problems for the huge need of computational resources, so hybrid methodsare developed. Using the hybrid methods, the computational region is divided into the source regionand the propagation region. As the propagation of acoustic is a linear process, the method used in thepropagation region can be more simplified than that used in the source region. Compared with thedirect methods, hybrid methods have the advantage of small calculation amount. There are two typesof methods for solving the propagation region, one is based on the traditional acoustic analogy theoryand the other is based on acoustic propagation equations. Hybrid methods based on the acousticpropagation equations are more accurate than that based on the traditional acoustic analogy theory,since the former can simulate the effect of non-uniform flow on the acoustic propagation, while thelatter assumes that the background flow field is uniform. The dissertation mainly focuses on theresearch on hybrid methods for the computational aeroacoustics based on acoustic propagationequations, and then a prediction program is implemented for solving broadband and tonal noiseproblems.The method for solving the propagation region is based on acoustic propagation equations, whichcan be coupled with different methods used in the source region to generate different hybrid methods.The dissertation mainly does research on two kinds of hybrid methods, one is based on coupling ofstochastic model and acoustic propagation equations, and the other is based on coupling of large eddysimulation and acoustic propagation equations.The acoustic propagation equations and its numerical methods are firstly discussed. Acousticpropagation equations contain the linearized Euler equations and acoustic perturbation equations. The7-point DRP scheme is used for the spatial discretization and optimized5-stage Rung-Kutta scheme isemployed for the temporal discretization. The far-field non-reflecting conditions are also adopted. Theacoustic propagation solver is developed based on the above methods, and the program is validated bysolving typical numerical examples.On this basis, the hybrid method based on the coupling of stochastic model and acousticpropagation equations is studied, and the specifics of the method are given as follows. Firstly the Reynolds-Averaged Navier-Stokes (RANS) equations are solved to obtain the time-averaged flowfield, which can be interpolated from the CFD grids to the CAA grids, and then the turbulent velocityfluctuations are synthesized using stochastic model, finally acoustic propagation equations withsource terms are solved. RANS equations are solved by the cell-centered finite volume method. Roe’sflux-difference splitting scheme is used for the spatial discretization and the4-stage Rung-Kuttascheme is employed for the time discretization, and the K turbulent model is also adopted. Theinterpolation techniques based on overset grids interpolation thought is proposed to make theinterpolation from the CFD grids to the CAA grids more efficient. A high-order finite differencemethod for multiblock structured grids is proposed based on generation of multiblock structured grids,which can be used for more complex shape aeroacoustic prediction.Against the problem that tonal noise can’t be simulated accurately by the hybrid method basedon the coupling of stochastic model and acoustic propagation equations, the hybrid method based onthe coupling of large eddy simulation and acoustic propagation equations is studied. Large eddysimulation equations are solved by finite difference method. The7-point stencil and6-order centraldifference scheme is used for the spatial discretization and optimized5-stage Rung-Kutta scheme isemployed for the temporal discretization. No-slip boundary conditions and non-reflecting boundaryconditions are used on the wall and the far field boundary separately. The coupling procedure betweenthe source region and propagation region adopts acoustic boundary conditions method. In theoverlapping region of the source region and the propagation region, the disturbance informationcalculated by the large eddy simulation is assigned directly to the propagation region. The temporalmixing layer is simulated using the hybrid method. Compared with the results calculated by the largeeddy simulation method, the hybrid method can be verified feasible. Finally the influences of gridsdensity, time step and other elements on the far field noise prediction result are discussed.