Sound solutions to combustion problems1 December 2002
New research is helping address the high amplitude acoustic perturbation problems that have plagued dry low NOx combustion systems.
Professor Chris Wilson, QinetiQ, Farnborough, UK
A key challenge facing power plant operators today is environmental legislation. By comparison with other prime movers, the industrial gas turbine has always been regarded as a relatively clean source of power. This is because it is an efficient technology with the high-tech appeal of its 'aero' ancestry, coupled with good quality fuel, producing an exhaust that cannot be seen or smelt.
But environmental concerns, led by the USA and Japan, have grown over the last 20 years, and industrial gas turbines now have to meet stringent legislative restrictions on invisible pollutants such as oxides of nitrogen (NOx) and carbon monoxide (CO). The expansion of the European Union has now added another strong voice to the debate.
A brief history of emission controls
In 1977 the US Environmental Protection Agency (EPA) issued proposed rules that limited the emissions of new, modified and reconstructed gas turbines to: l 75 vppm NOx at 15 per cent oxygen (dry basis); and l 150 vppm SOx at 15 per cent oxygen (dry basis), controlled by limiting fuel sulphur content to less than 0.8 per cent wt.
These standards applied to simple and regenerative cycle gas turbines, and to the gas turbine portion of combined cycle steam/electric generating systems. The 15 per cent oxygen level was specified to prevent the NOx ppm level being achieved by dilution of the exhaust with air.
At this time (1977), it was recognised that there were a number of ways to control oxides of nitrogen. These can be divided into two types:
l 'wet' controls - water/steam injection to reduce peak flame temperatures;
l 'dry' controls - advanced combustor design to control peak flame temperatures and catalytic exhaust cleanup.
'Wet' control became the preferred method since 'dry' controls were at a very early laboratory/combustor rig stage. There has been a gradual tightening of the NOx limits over the years, from 75 vppm down to 25 vppm, with single digit levels expected in the near future.
Advances in combustion technology now make it possible to control the levels of NOx production at source, removing the need for 'wet' controls. This of course opened up the market for the gas turbine to operate in areas with limited supplies of suitable quality water, eg deserts or marine platforms. Although water injection is still used, 'dry' control combustion technology has become the preferred method for the major players in the industrial power generation market. DLN (dry low NOx) was the first acronym to be coined, but with the requirement to control NOx while also controlling carbon monoxide and unburned hydrocarbons this has now become DLE (dry low emissions).
DLE and its problems
The majority of the NOx produced in the combustion chamber is called 'thermal NOx'. It is produced by a series of chemical reactions between the nitrogen (N2) and the oxygen (O2) in the air that occur at the elevated temperatures and pressures in gas turbine combustors.
The reaction rates are highly temperature dependent, and the NOx production rate becomes significant above flame temperatures of about 2100 K. Figure 1 shows schematically flame temperatures and therefore NOx production zones inside a 'conventional' combustor. This design deliberately burns all of the fuel in a series of zones going from fuel-rich to fuel-lean to provide good stability and combustion efficiency over the entire power range.
Presently, there are two avenues to pursue in order to reduce NOx emissions. One is to run very rich, passing quickly through the peak temperature. This is known as rich burn quench lean (RQL). Alternatively, there is the lean pre-mixed (LP) combustion approach, where the fuel and the air is mixed upstream of the combustion chamber and then burned in the combustion chamber.
This lean pre-mixed DLE approach, which has become the dominant technology, is to burn most (at least 75 per cent) of the fuel at cool, fuel-lean conditions to avoid any significant production of NOx.
Figure 2 shows a schematic layout of a typical lean pre-mixed DLE combustor. A small proportion of the fuel is always burnt richer to provide a stable 'piloting' zone.
If the combustor does not feature variable geometry, then it is necessary to turn on the fuel in stages as the engine power is increased. The number of stages will be determined by the expected operating range of the engine, but typically at least two or three are used.
Unfortunately, the lean pre-mixed DLE system produces several problems: l It results in a very narrow operating range in which to burn the fuel. For most hydrocarbon fuels this ends up being an air fuel ratio (AFR) of between 25 and 30 in order to reduce NOx emissions.
l In the pre-mixing process, it is necessary to mix thoroughly, but without taking too long, or auto-ignition will occur.
l Within the combustion chamber itself, one is operating very lean and any fluctuations in AFR can result in the flame itself being extinguished. This produces oscillation, as locally areas of the flame are alternating between being alight or extinguished.
These problems can result in sudden loss of power because a fault is sensed by the engine control system and the engine is shut down.
In summary, operating experience with these modern lean pre-mix DLE systems shows that, by their very nature, they are inherently prone to operating instability and structural fatigue, due to high amplitude acoustic perturbations.
The search for acoustic solutions to combustion problems has been underway for a long time. The conventional wisdom can be summarised as follows: "if there is noise, then drill a hole in it and if it is still making a noise then drill another hole in it." This was feasible with older style gas turbine combustors, which already have holes through the combustor liner. But it is not with today's machines: lean pre-mixed combustors place most of the air and fuel in the front of the combustor, which is surrounded by solid walls, and these then reflect the acoustic signal back into the combustion chamber.
This is the essence of the problem and the result is that today all the major manufacturers are having difficulties with acoustics.
There are various ways in which one can eliminate acoustic problems, and two measures are widely used:
l Passive control, where a tube with a resonator embedded in it is connected to the combustion system. The problem associated with this is that the hot gases pass through a small neck that must be cooled. Additionally, this approach only addresses the problem over a narrow frequency range when what is really needed is to be able to attenuate over a very large range.
l Another approach is installation of an active combustion control system that modulates the fuel flow so that it enters the combustor out of phase with the combustion instability.
To achieve active combustion control one needs to know how the combustion system is behaving and, in particular, how it behaves when the flow experiences perturbations. It is for this reason that one needs to measure the transfer function, ie the function that relates inlet changes to the combustion behaviour.
QinetiQ has embarked on a research programme to measure the transfer function in a combustion rig, one of a number of projects we have participated in with several partners over the years, with the aim of extending our knowledge of LP combustion.
In recent years, various researchers have used loudspeakers to excite the airflow. We, however, believed that the representative high pressure conditions could not be dealt with by even the most robust and loudest of loudspeakers. We therefore began to explore other ways of tackling the problem.
Our solution was an aerodynamic siren sited upstream of the rig to perturb some of the airflow going through the combustion chamber. The in-line siren, which is used in these experiments, is placed in a water-cooled housing and is required to survive high temperatures and pressures.
The siren was manufactured by using a 150 W brushless motor, which was surrounded by a water jacket. We were able to produce speeds approaching 10 000 rpm, giving us a maximum frequency of around 1.8 kHz. The airflow is modulated by a rotor attached to the motor. The motor assembly is fixed to a stator, which has another row of holes feeding the modulated airflow into the combustion section. The system was designed to have a pressure drop of approximately four per cent across the stator plate and to modulate up to around 20 per cent of the total flow.
The tests focused on conditions of between 7 and 15 bar inlet pressure and between 550 and 770 Kelvin inlet temperature. These conditions were selected to simulate conditions from engine idle to full power. The speed of the siren is changed thus varying the frequency of the airflow perturbations. Pressures upstream of the combustor under test and downstream in the combustor are measured over a range of frequencies. At the time of writing, these tests had been performed with the combustor running and we have begun to explore how the combustion and the air perturbation react with each other.
During the course of the tests, pressures were measured both upstream and downstream of the LP duct using a semi-infinite tube method. If the pressure transducers were to be placed on the combustor itself, high-temperature pressure transducers would be required. These are difficult to obtain and expensive. We therefore located the pressure transducers in a semi-infinite tube making certain that we calibrated and corrected for the location of these pressure transducers within the tube.
In addition heat release issues have been investigated by using a fibre optic boroscope which looked upstream, in order to detect CH emissions. This effectively gives a measure of the combustion intensity, which is related to areas of highest heat release.
One of the key problems we face with this type of investigation is that the combustion is a dynamic process, with the flame front position moving up and down the combustor. However, the optically derived heat release rate does correlate well with the pressure measurements taken from the transducers.
The flame transfer functions were calculated for a given combustor operating over a range of engine conditions. At each stage of the tests, the frequency of the siren was measured and measurements were taken of pressure and heat release rate.
The research programme has proved successful. The original objectives of the tests - to carry out flame transfer function measurements and to monitor the LP combustor at high pressure conditions - were achieved with no combustor damage. The instrumentation on the combustion system worked well and survived the harsh environments encountered.
The boroscope and pressure transducers provided us with the basics of the dynamic measurement. All these instruments worked well and agreed with each other. The siren proved capable of modulating the flow over a range of frequencies. Good coherence was found between the pressure and heat release signals. The flame transfer function tests showed a general decay in amplitude and a mainly constant time delay as the siren frequency was increased.
This work has provided critical transfer function information, which is now being incorporated into design tools used for the prediction of oscillation frequency and amplitude values.
The rig is now being used to help manufacturers characterise their particular combustor design frequency responses under full operating conditions, helping OEMs in their efforts to produce quiet, low emission, combustion systems.