Gas turbines generate electricity and power our aeroplanes. They will be a long term presence in the overall energy mix, complementing renewable but inconsistent sources of power, such as wind turbines. It is clearly important to make them as clean, quiet and efficient as possible. Unfortunately, the very conditions under which gas turbines produce ultra low NOx emissions (NOx cause air pollution and breathing problems) also make the combustor extremely prone to thermoacoustic instability. The associated high amplitude pressure and flame oscillations lead to damaging vibration, and make operation impractical. Thermoacoustic instability is caused by acoustic waves and unsteady heat release in the combustor mutually affecting one another, leading to positive feedback. There is an urgent need to be able to predict it as part of the gas turbine design process, so that it can be "designed out". This is currently not possible.

Computationally simulating thermoacoustic instability requires many length scales to be resolved - from the tiny chemical reaction scales, to flame-front wrinkling, to the very long acoustic waves. Simulating compressible reacting flow over this range of scales is prohibitively expensive. Low order network models provide a computationally fast alternative. They approximate the combustor geometry as a network of simple connected modules, and exploit the fact that the acoustic wave behaviour is linear (even for large oscillations, nonlinearity comes only from the flame). Furthermore, at low frequencies the acoustic waves behave at most two-dimensionally. This means that simple analytical models for the acoustic waves can be coupled with more complex models, from computational flow simulations or experiments, for the flame.

Recent work, in which the investigators have played a key role, has shown that (i) the flame nonlinearity can typically be captured via a weakly nonlinear modeling approach and (ii) that these weakly nonlinear flame models can be obtained by "incompressible" large eddy simulations (LES), which capture large turbulent flow features (although "incompressible", the density can change with flow temperature). The use of incompressible simulations saves roughly an order of magnitude in time compared to compressible simulations. For single turbulent flame combustors, major progress was recently achieved by coupling a low order thermoacoustic network model with incompressible LES of the flame, and accurately predicting not only the frequency of thermoacoustic instability, but also its limit cycle amplitude. It is now timely to attempt fast and accurate predictions for increasingly realistic combustor setups.

The proposed work is two-fold. It will firstly seek to develop new methods for low order thermoacoustic network modeling and for incompressible LES of flames. These are needed to facilitate fast and accurate predictions for multi-burner annular combustors, more representative of real gas turbine set-ups. We will use data from a world-class experimental combustion facility at our collaborators, NTNU in Norway, for validation. We will consider flames which burn independently, and more complicated cases when they are close enough to interact with one another. We will investigate the flow physics at play - the combined effect of circumferential acoustic waves, flame nonlinearity and flame-flame interactions.

At the same time, we will develop a UK-based configurable optical combustion rig to measure forced flame response using advanced laser imaging. This will have the option for multiple burners, and for the first time, multi-phase fuel, interchangeable combustor wall materials and variable exit conditions. It will provide insights into instability in practical combustors, and high-fidelity flame and flow data for modellers. We aim to predict thermoacoustic instability in the presence of these phenomena, moving predictive capability towards increasing industrially relevant combustors.

Potential Impact:
Gas turbines generate electricity and power our aeroplanes. They will be a long term ingredient in the overall energy mix, complementing renewable but inconsistent sources of power, such as wind turbines. It is clearly important to make them as clean, quiet and efficient as possible. Unfortunately, the very conditions under which gas turbines produce ultra low NOx emissions (NOx cause air pollution and breathing problems) also make the combustor extremely prone to thermoacoustic instability. The associated high amplitude pressure and flame oscillations lead to damaging levels of vibration, and make operation impractical. The proposed research aims to develop predictive tools and scientific understanding that will contribute to allowing thermoacoustic instability to be "designed out" of the next generation of gas turbine combustors. This will contribute to the overarching aim of achieving ultra-low NOx emissions from gas turbines. The environmental benefits will be improved air quality near airports and power stations.

The economic benefits will be improved competitiveness of gas turbines manufacturers, from which the UK stands to benefit through its significant activity in this field. The proposed work is ambitious, high-risk research with the potential to feed into gas turbine combustor designs in the medium term. Our collaboration UK-based gas turbine company, Siemens (who already support both investigators with funded PhD students, in ASM's (Imperial) case as official partners in the Centre for Doctoral Training in Fluid Dynamics across Scales) will be mutually beneficial, ensuring in particular a direct industrial uptake route of tools and insights developed in the project, to the benefit of the UK gas turbine industry.

The organisation of a workshop at the end of the project will enhance dialogue between UK academics, European academics and gas turbine manufacturers more widely.