Researches On Accuracy Of The Complex Flows Numerical Simulation And Parallel Computations Based On Navier-Stokes Equations

With the development of the computer and computing techniques, computational fluid dynamics (CFD) has made a great progress in solving various engineering problems. It plays a very important role in aerospace, vehicle, oil and energy industries. To date, CFD has been becoming a compulsory tool for engineers when designing various fighters. Although CFD has made a great progress during the past decades, it has some challenge problems. The accuracy and efficiency of CFD methods are still the difficult issues when using CFD to solve the problems of complex flows. This research focuses on improving the accuracy and efficiency of the CFD simulations. A CFD application is also conducted for the flow fields around a complex configuration of air-to-air missiles (AAM). The major works and achievements are described as the following:1. A general approach generating high quality multi-block structured patched grids around wing-body configurations is developed using the hypercube concept. As an example, three sets of multi-block structured grids around the DLR-F4 Wing-Body configuration are generated by solving elliptic using this approach. The three sets of grids have different grid densities. The computations are conducted by solving Reynolds-averaged Navier-Stokes (RANS) equations coupled with the Spalart-Allmaras and Baldwin-Lomax turbulence models. The computed results using generated meshes agree well with the experiment and are better than those using other software and other meshes when compared with experimental data. Our results show that the proposed grid generation approach using hypercube concept is feasible. The generated mesh has high quality. This approach is general and can be applied to generate meshes around any other similar wing-body configuration.2. The accuracy of the computed drag around both the DLR-F6 wing-body and wing-body-nacelle-pylon configurations, and the effects of grid and turbulence models on aerodynamic characteristics are investigated by solving Reynolds-averaged Navier-Stokes (RANS) equations coupled with three kinds of turbulence model:the Spalart-Allmaras (SA) one equation model, Wilcox's k-?(Wilcox) model and Menter's k-?SST (SST) model. A set of multi-block structured patched grid with different density (coarse, medium and fine) provided by the second international AIAA Drag Prediction Workshop (DPW?) is used in this study. The Roe scheme is used to capture the discontinuities. In order to improve the accuracy and avoid numerical oscillations near discontinuities, MUSCL (Monotone Upstream centered Scheme for Conservation Laws) interpolation method and A third-order upwind-biased flux limiter are employed. The sensitivity of CFD to grid size and the behavior of three different turbulence models are examined in the series of test cases designed. The computed results agree well with the experiments. For WB, the predicted drag using S A model is better than those using other codes when compared with experimental data. This study shows that the grid density has obvious effects on pressure drag and total drag, but It has small effects on the pressure distribution on the wing and nacelle using the SA turbulence model. Turbulence models have certain effects on the positions of the shock wave on the wing surface, obvious effects on drag, especially friction drag. The different turbulence models also have effects on the wing-root separation bubble size, SA predicted the largest bubble, Wilcox the smallest. In the general, the computed results using SST model have the highest accuracy compared with the experiment. It consistently predicted lower drag than SA and Wilcox. The predicted lift using SA model agrees well with the experiment. The predicted lift and drag using Wilcox model are over-estimated. The present computations show that performing the CFD calculation at the same angle of attack as experiment resulted in good comparisons with wing surface pressures, but matching lift (CL=0.5) did not.3. The performance of a low diffusion E-CUSP (LDE) upwind scheme for compressible flow is investigated systemically by simulating the flow fields around several standard models. The widely used Roe scheme is employed to do the comparison. The governing equations are the Euler equations or the Reynolds averaged Navier-Stokes equations coupled with SA one equation turbulence model. The fully implicit unfactored Gauss-Seidel Line relaxation method is used for time marching. The computed results show that LDE scheme can capture crisp shock waves accurately. The results using LDE scheme agree well with the experimental data. Without the matrix operation, LDE scheme has a less computation and faster speed of convergence Compared with Roe scheme. The LDE scheme coupled with SA turbulence model is applied to simulate the supersonic flow field around a slender body at high-angle-of-attack. The leeward flow separations and vortices developing process are accurately simulated. The computation captures the subtle secondary vortices and cross separation with a promise resolution. The predicted aerodynamic forces and moments agree well with experimental data. The study shows that LDE scheme coupled with SA one equation turbulence model is accurate, efficient and robust, and can be used to simulate complex separated flows. 4. Numerical simulation of the flow fields around a complex configuration of air-to-air missiles (AAM) is performed by solving the Reynolds averaged Navier-Stokes equations (RANS). The S-A one equation turbulence model is coupled to closure the governing equations. The finite volume multistage Rung-Kutta time-stepping scheme is used in the computation. A patched multi-block structured grid is generated around the complex configuration by solving elliptic grid generation together with an algebraic method. Aerodynamic forces and moments of the AAM with or without angle of roll are calculated using the formulae derived in this paper. The present results agree well with the experiment and are better than those using other codes.5. Parallel machine scheduling problem limited splitting manner for job to minimize makespan is proposed. Different from the researches in the literature, the splitting manner for job is limited. This problem is NP-hard. An approximate algorithm for solving the problem is developed and applied to solve the problems of the load balance of parallel computations in a heterogeneous parallel computer system. Simulation results show that compared with traditional method, this algorithm is reliable and practical. A load balance algorithm for multi-zone CFD parallel computation using heterogeneous parallel computer systems is developed by applying the research achievements of the scheduling problems. The data splitting and task allocating problems of CFD parallel computations are solved. By designing the communication manner between the processors in an optimal way, the numbers of the communication and the waiting time between the processes are decreased, and then the rate of computation/communication is increased. The parallel computation is implemented in a cluster of workstations. The scalability of the parallel computation is estimated using the standards of the equal efficiency and equal velocity respectively. The results show that the parallel computation has a good scalability. The load balance algorithm developed performs well with high parallel computation efficiencies. It can be applied to solve the load balance problems of CFD parallel computations on MIMD architecture.