Publications

The isentropic exponent is one of the most important properties affecting gas dynamics. Nonetheless, its effect on turbine performance is not well known. This paper discusses a series of experimental and computational studies to determine the effect of isentropic exponent on the flow field within a turbine vane. Experiments are performed using a newly modified transient wind tunnel that enables annular cascade testing with a wide range of working fluids and operating conditions. For the present study, tests are undertaken using air, CO2, R134a, and argon, giving a range of isentropic exponent from 1.08 to 1.67. Measurements include detailed wall static pressures that are compared with computational simulations. Our results show that over the range of isentropic exponents tested here, the loss can vary between 20% and 35%, depending on vane exit Mach number.

The mechanisms of blade row interaction affecting rotor film cooling are identified to make recommendations for the design of film cooling in the real, unsteady turbine environment. Present design practice makes the simplifying assumption of steady boundary conditions despite intrinsic unsteadiness due to blade row interaction; we argue that if film cooling responds nonlinearly to unsteadiness, the time-averaged performance will then be in error. Nonlinear behavior is confirmed using experimental measurements of flat-plate cylindrical film cooling holes. Unsteady computations are used to identify the blade row interaction mechanisms in a high-pressure turbine rotor, and a quasi-steady model is used to predict unsteady excursions in momentum flux ratio. It is recommended that the designer should choose a cooling configuration that behaves linearly over the expected excursions in momentum flux ratio as predicted by a quasi-steady hole model.

In this paper, we study the effect of rotor-stator axial gap on midspan compressor loss using high-fidelity scale-resolving simulations. For this purpose, we mimic the multi-stage environment using a new numerical method that recycles wake unsteadiness from a single blade passage back into the inlet of the computational domain. As a result, a type of repeating-passage simulation is obtained such as observed by an embedded blade-row. We find that freestream turbulence levels rise significantly as the size of the rotor-stator axial gap is reduced. This is because of the effect of axial gap on turbulence production, which becomes amplified at smaller axial gaps and drives increases in dissipation and loss. This effect is found to raise loss by between 5.5% and 9.5% over the range of conditions tested here. This effect significantly outweighs the beneficial effects of wake recovery on loss.

We quantify the sensitivity of turbine acoustic impedance to aerodynamic design parameters. Impedance boundary conditions are an influential yet uncertain parameter in predicting the thermoacoustic stability of gas turbine combustors. We extend the semi-actuator disk model to cambered blades, using non-linear time-domain computations of turbine vane and stage cascades with acoustic forcing for validation data. Discretising cambered aerofoils into multiple disks improves reflection coefficient predictions, reducing error by up to an order of magnitude compared to a flat plate assumption. A parametric study of turbine stage designs using the analytical model shows acoustic impedance is a weak function of degree of reaction and polytropic efficiency. The design parameter with the strongest influence is flow coefficient, followed by axial velocity ratio and Mach number. We provide the combustion engineer with improved tools to predict impedance boundary conditions.

Further improvements in aero-engine efficiencies require accurate prediction of flow physics and incurred loss. Currently, the computational requirements for capturing these are not known leading to inconsistent loss predictions even for scale-resolving simulations depending on the chosen convergence criteria. This work investigates two aspects of loss generation using high-fidelity simulation. In the first case study, we look at the effect of resolution on capturing entropy generation rate by simulating a Taylor-Green vortex canonical flow. The second case study focuses on the effect of resolution on flow physics and loss generation and uses a compressor cascade subjected to freestream turbulence. The results show that both resolving local entropy generation rate and capturing the inception and growth of instabilities are critical to accuracy of loss prediction. In particular, the interaction of free-stream turbulence at the leading-edge and development of instabilities in the laminar

Every supersonic fan or compressor blade row has a streamtube, the “sonic streamtube,” which operates with a blade relative inlet Mach number of one. A key parameter in the design of the “sonic streamtube” is the area ratio between the blade throat area and the upstream passage area, Athroat/Ainlet. In this article, it is shown that one unique value exists for this area ratio. If the area ratio differs, even slightly, from this unique value, then the blade either chokes or has its suction surface boundary layer separated due to a strong shock. Therefore, it is surprising that in practice designers have relatively little problem designing blade sections with an inlet relative Mach number close to unity. This article shows that this occurs due to a physical mechanism known as “transonic relief.”

Direct numerical simulations (DNS) are performed of a cantilevered stator blade to identify the unsteady and turbulent flow structure within compressor tip flows. The simulations were performed with clearances of 1.6% and 3.2% of chord. The results show that the flow both within the gap and at the exit on the suction side highly unsteady phenomena controlled by fine-scale turbulent structures. The signature of the classical tip-leakage vortex is a consequence of time-averaging and does not exist in the true unsteady flow. Despite the complexity, we are able to replicate the flow within the tip gap using a validated quasi-three-dimensional (Q3D) model. This enables a wide range of Q3D DNS simulations to study the effects of blade tip corner radius and Reynolds number. Tip corner radius is found to radically alter the unsteady flow in the tip; it affects both separation bubble size and shape, as well as transition mechanisms in the tip flow. These effects can lead to variations in tip ma

This paper describes a simple and efficient physics-based method for designing optimal transonic multistage compressor rotors. The key to this novel method is that the spanwise variation of the parameter which controls the three-dimensional shock structure, the area ratio between the throat and the inlet, ‘Athroat /Ainlet’, is extracted directly from the 3D CFD. The spanwise distribution of the area ratio is then adjusted iteratively to balance the shock structure across the blade span. Because of this, the blade design will be called ‘aerodynamically balanced’. The new designmethod converges in a few iterations and is physically intuitive because it accounts for the real changes in the 3D area ratio that directly controls the shock structure. Specifically, changes in both the spanwise 3D flow and in the rotor’s operating condition; thus aiding designer understanding.

The prevailing view is that high fidelity simulation, particularly DNS (direct numerical simulation), is not something for the practical turbomachinery aerodynamicist — requiring too much computational and personal effort to make it worth it. The aim of the ‘Desktop-DNS’ toolkit described in this paper is to change this by greatly lowering the barrier to entry for running DNS. The paper shows how, using an efficient high-order Navier-Stokes computer code, it is becoming increasingly possible to solve testcases of industry relevance with high fidelity LES and DNS, making use of the latest advances in single compute node performance. This is achievable using both efficient algorithms and GPU acceleration. The paper will use a compressor blade testcase to illustrate how, in some cases, high-fidelity simulations can be performed at relatively low costs on a small number of computer nodes. This raises the possibility of a much more widespread use of DNS to inform early design choices, enhan