GB2149273A - Active control of flame noise - Google Patents

Abstract

Active noise control systems such as for regulating flame noise including a detector transducer responsive to optical radiation from a turbulent flame source, means for phase changing and amplifying signals generated by said transducer, and means for expressing the phase changed amplified signal as an acoustic emission from an output transducer, eg a loudspeaker, located in the region of the flame source.

Description

SPECIFICATION Active control of flame noise This invention relates to the control of noise generated as a result of the combustion gases.

Noise generated during combustion may be of two types viz (i) Flame Noisy the noise generated in a turbulent flame due to local random fluctuations in the rate of reaction between the combusting substances and hence fluctuations in the local velocity and pressure changes within the combustion gas, and (ii) Combustion Oscillations-standing waves of a single frequency set up in the burner, combustion chamber, flue, etc. The interaction of the wave with the flame causes pulsations in the combustion process and this feed back effect causes energy to be supplied continuously to the wave. In addition to creating flame instability the acoustic effect can damage the fabric of the combustion chamber. In aerospace technology this phenomenon is encountered in aeroengine reheat systems and is known as 'reheat buzz'.

Perfectly laminar flames in a homogenous gas are practically silent, the noise of a flame being dependent upon its turbulence. Paradoxically, however, the most efficient flame, in terms of heat out put per volume is the turbulent flame. Noise Control techniques have of necessity been directed to the use of sound absorbent materials and surfaces to attenuate the noise. However, the sound output from flames is of low frequency and, since the efficiency of sound absorbing materials decreases as frequency decreases, effective attenuation tends to be both costly and bulky.

Over the past thirty years interest in the acoustics of flames was first directed towards understanding combustion oscillations and a number of solutions have been proposed for remedying the ocurance of the single frequency standing wave, including: (i) Improving flame stability and/or decreasing flame intensity.

(ii) Dimensioning equipment to ensure no coincidence of the acoustic frequency with any natural frequencies or modifying equipment to prevent standing wave formation or maintenance, eg drilling holes at the antinodes or use of quarter wave tubes or Helmholtz resonators.

For the last decade research has been directed to active control techniques (as opposed to passive techniques such as attenuation by absorption). One promising technique is the application of 'antisound' in which the noise is replicated through a phase shifted feedback loop. The phase shifted sound cancels out the noise sound thus reducing the overall noise level in a given volume of space.

Some disadvantages have been noted in connection with this sytem of control. A principle disadvantage lays in the delay between omission of the original noise sound and play back of the antisound. Thus although the sound is phase shifted it may not be the antiphase replica of the current sound and the canceliation effect may not be appreciable.

The reasons for the delay are inherent and due to the relatively slow speed of sound and to time elapse for signal processing and playback.

It has been suggested that there is a correlation between the noise emitted from a flame and the light emitted by it. We have confirmed that there is a corelation and that the relationship approaches linearity.

We have thus determined that it is possible to use the light emitted from the flame to trigger the phase changed signal loop and playback antisound. Since the speed of light is infinite, for practical purposes, the inherent delays can be considerably reduced.

Based on these findings we have found that low frequency noise associated with combustion flames can be significantly attenuated with a single degree of freedom active control system.

Thus the present invention provides a control system for regulating noise associated with combustion flames said systems including: -a detector transducer responsive to optical radiation from a turbulent flame source, means for phase changing and amplifying signals generated by said transducer, and means for expressing the phase changed amplified signal as an acoustic emission from an output transducer located in the region of the flame source.

The detector transducer may be a suitable photomultiplier associated with a focusing system and is arranged so as to receive substantially all the light radiated from the flame. It is essential that substantially all the light emission impinges on the photodiode assembly of the photomultiplier since significant sound producing regions of the flame are at its base and if these areas are shielded there is a much reduced coherence between the signal associated with the light emission and that of the acoustic emission.

The signal output from the photomultiplier is in analogue form and it may be processed as such. However recent developments in the digital filtering of low frequency sound indicate that there are a significant number of advantages in processing a digital signal.

Thus in accordance with an embodiment of the invention it is preferred to employ a digital active control loop for regenerating the flame noise.

Prior to analogue-to-digital conversion, in order to avoid ambiguities of aliasing during conversion, the signal frequency is limited by low pass fi terin. The purpose of the antialiasing filter is to reduce the high frequency content of the signal such that when analogue-to-digital conversion takes place high frequency components are not mistaken for low frequency ones. Such filters are necessary because a digital filter system can only distinguish data in a limited frequency band (up to the Nyquist frequency). Thus without antialiasing frequencies higher than the Nyquist frequency would be misinterpreted as principle low frequency signals. Conversely after phase shifting and any necessary amplification, the digital output signal is subjected first to digital-to-analogue conversion followed by antialiasing filtering to produce the phase shifted replica signal.

Typical antialiasing filters include those of the Butterworth type and have a phase response which is approximately linear with frequency, at least up to frequencies of 1 kHz.

Although antialiasing is necessary for digital control systems there is some time delay through the filter. Thus the phase shift through the filter means there will be equivalent delay in the introduction of the antisound, in addition to the calculation time in the digital filter. However, these factors for Butterworth filters can be computed in a known manner.

Any errors in replicating the antisound are less significant in controlling combustion oscillations since the originating sound is of a single frequency. However with noise associated with combustion frequency changes are continuous and random. Thus it is desirable that the response time should be as small as possible. Digital control systrems provide some amelioration of the problem compared with analogue versions, however in order to be as quick as possible only a few coefficients can be analysed.

The digital filter aims to model the flame in antiphase so that the correct antisound signal appears at its output. Ideally, therefore, some prior knowledge of the control system characteristics is required if the output is to be predicted with reasonabie accuracy from known input data. Input/output methods are preferred for active control but they rely on introducing known disturbances at the input and measuring the output, and since the basic combustion process is of itself not subject to disturbance it is not possible to make direct observations, measurements and consequential predictions. However, since flame emissions are stationary stochastic processes, measurements of the flame are emenable to analysis by spectral analysis, thereby to allow estimation of the frequency response, required by any active noise control systems and system identification.System identification allows a model of the antisound system to be made which can be established in a digital filter.

Another advantage of using a digital filter as the controller is that a large number of system identification algorithms are known which produce solutions which are readily implimented on such digital filters. The identification techniques have their basis in modelling the system as an autoregressive/moving average (ARMA) process utilizing least squares estimation to build a model from time domain information. The model can be readily implimented on a recursive or infinite impulse response digital filter. The advantage of ARMA modelling is that once the necessary coefficient have been found the speed of impiementation is high.

Once the model for the antisound has been built on the filter, the digitalised output signal can be emitted by a reverse procedure to the inputing procedure. Thus the digital output is processed through digital-to-analogue conversion and the resultant analogue signal subjected to antialiasing filtering prior to exitation of the output transducer. The antisound is generated by acoustic emissions from a loudspeaker activated by the output from the digital filter.

Since no real time advantage can be gained in controlling the noise from turbulent flames whilst generating antisound, the key to effective one-dimensional active control is to use the inherent time delay to process the control signal. It has been found, in practice, that the light emission, at the C2 free radical wavelength, used as the control signal to a hardwired non recursive or finite impulse resonse (FIR) filter, produced antisound which resulted in a reduction in noise of 10 dB in the band 200-800 Hz.

In controlling combustion oscillations we have found that the present invention will achieve 30 di3 reduction of the fundamental mode of resonance in a flame driven Rijke tube. Reductions in the harmonic resounces were also observed even though the antisound loudspeaker is activated just at the fundamental.

Claims (9)

1. A control system for regulating noise associated with combustion flames including a detector transducer responsive to optical radiation from a turbulent flame source, means for phase changing and amplifying signs generated by said detector transducer, and means for expressing the phase changed amplified signal as an acoustic emission from an output transducer located in the region of the flame source.

2. A system as claimed in Claim 1 including means for converting an analog signal from the detector transducer to a digital output.

3. A system as claimed in Claim 2 including an antialiasing filter prior to A/D conversion of the detector transducer signal.

4. A system as claimed in any of Claims 1 to 3 including means for D/A conversion and antialiasing of a digital output from the phase changing and amplifying means.

5. A system as claimed in either of Claims 3 or 4 in which the antialiasing filter is a Butterworth type.

6. A system as claimed in any one of the preceding claims in which the phase changing-amplifying means is a recursive or infinite impulse response digital filter, adapted to model and output a phase-shifted amplified replication of the received input signal.

7. A system as claimed in Claim 6 wherein the digital filter is adapted to model on an auto regressive-moving average basis utilizing time domain information.

8. A system according to Claim 1 and substantially as hereinbefore described.

9. A combustion system including means for combusting gases and an associated control system as defined in any one of claims 1 to 8.