Feng Liu

Associate Professor, Mechanical & Aerospace Engineering
The Henry Samueli School of Engineering

PH.D., Princeton University

Phone: 949) 824-1750
Fax: (949) 824-8585
Email: fliu@uci.edu

University of California, Irvine
S3201, Engineering Gateway
Mail Code: 3975
Irvine, CA 92697
Research Interests
Fluid Mechanics, Computational Fluid Dynamics, Propulsion, Aerodynamics, Turbomachinery, Transonic and Hypersonic Flow
Research Abstract
Fluid/Structure Interaction

We are developing efficient numerical methods for calculating fluid/structure interactions. See The AGARD I-Wing 445.6

Multigrid method for solving the Reynolds-average Navier-Stokes equations with advanced turbulence models

We are interested in developing efficient numerical methods for simulating high Reynolds number flow with complex geometry. A staggered finite volume method with multigrid has been developed for the compressible Reynolds-averaged Navier-Stokes equations with algebraic and two-equations turbulence models. Multigrid method is applied to accelerate the convergence rate for both the solution of the Navier-Stokes and the turbulence model equations. Both upwind and central difference type algorithms are employed to discretize the differential equations in space. The resulting semi-discrete equations are marched in time by an explicit multi-stage method which are highly suitable for parallel processing.

Adaptive grid generation

Finite difference, finite volume and finite element methods all use a computational mesh to approximate a continuum problem. In areas of large gradient of velocity, presure, temperature or other flow parameters, such as in the neighborhood of a shock wave, very small grid sizes are required. In other areas, such as in the far field where the flow is rather uneventful, fewer grid points can be used without degrading the computational accuracy. For many practical problems the precise location of shock waves or other flow features are not known a prior. It is highly desirable to have a computational method that can automatically detect the areas of high and low flow gradients and refine and coarsen the computational mesh accordingly. A grid generation method has been proposed that can provide precise control of the mesh size distribution. This research will develop the method for arbitrary computational domains and couple it with a flow solver for steady and unsteady flow calculations.

Unsteady flow in turbomachinery cascades

Almost all design methods for turbomachinery cascades such as the compressor and turbine blades in a jet engine have been based on steady flow calculations and experimental correlation. Yet, turbomachines are essentially unsteady machines. Indeed, they depend on flow unsteadiness to produce work. Unsteady flows are extremely important to the performance, heat transfer, aeroelasticity and noise of jet engines. We have started to develop numerical methods that can simulate unsteady flow through moving cascade blade rows. Such calculation take a lot of computer time and memory. We are currently looking at a multigrid acceleration technique that may prove applicable and efficient for unsteady flows as well as steady flows. Particular applications may include possible rotor-stator interaction and the transport of hot streaks through turbine blade rows.

Aerodynamic Optimization of Turbine Blade Rows via CFD and Control Theory

Current aerodynamic design procedures involve long hours of manual iteration by a designer. The optimality of a design rests heavily on the designer's knowledge and skills and is not in any way guaranteed. This usually leads to non-optimal machines with long design cycle and large developmental costs. With the advance of computational fluid dynamics and nonlinear optimization theory, this research attempts to replace this manual iteration by a systematic computer optimization procedure based on rigorous mathematical and fluid dynamics principles, that is, by solving the fundamental fluid dynamic equations and using advanced nonlinear optimization methods. By formulating an aerodynamic shape optimization problem into a nonlinear control problem, the computational effort in achieving an improved design can be dramatically reduced compared to conventional sensitivity analysis. Gradient information needed for optimization can be obtained by solving a linear adjoint equation to the original nonlinear governing equations, the Euler or the Navier-Stokes equations. The optimization theory is combined with a fast and accurate Navier-Stokes analysis code for turbomachinery flows. The proposed research will explore various two- and three-dimensional shape optimization of compressor and turbine blades.
Available Technologies
Last updated