How to Solve

Abnormal Combustion Noise Problems

Peter K. Baade, Consultant, Fayetteville, New York
This article discusses abnormal combustion noises in boil-
ers, burners and heating systems. An experimental method is
described to provide insight into the causes of such noises.
Several techniques are presented for the reduction or elimi-
nation of abnormal combustion noises.

During the development of a new boiler, burner or heating

system it is always possible that a prototype will emit an un-
acceptably loud noise that is clearly abnormal. Although the
occurrence might be infrequent, its very presence necessitates
elimination before the product can go to market. Since the noise
is so abnormal, it is tempting to think that a solution will be
found quickly by making some small design changes. If one is
very lucky, this cut-and-try approach may indeed work. Most
likely, however, it will be frustratingly slow and may not suc-
ceed at all.
Frustrations often escalate since the occurrence of the abnor-
mal combustion noise can be very elusive. In some cases it may
occur only for a few seconds during a cold start, while in oth-
ers it may occur only when the unit is shut down and restarted
while still hot. Others may occur only on days when the gas
utility is changing the composition of the supply for peak shav-
ing purposes. I even know of a case when, during the field tri-
als of a new furnace, three units out of 100 in widely scattered
locations had an abnormal noise problem, even though the pro-
totype tests had given no such indication. When one of the
noisy furnaces was brought back to the lab just a few miles
away, it was impossible to reproduce the abnormal noise de-
spite many efforts. Figure 1. Equipment used for the experiments.
While there is no “magic bullet” that will solve abnormal
noise problems in all cases, there are a number of useful tools.
The trick is to know what tools are most likely to work in a par-
ticular case, which requires a good insight into the causes of
abnormal noise and how the different parts of a unit are in-
volved.
Abnormal noise problems have received considerable study
in industrial gas turbines,1 but engineers in the heating indus-
try may find it difficult to apply this information to residen-
tial and small commercial equipment. This article will, there-
fore, use a series of bench-top experiments with the test rig
shown in Figure 1 to demonstrate the answers to many of the
questions that arise when dealing with abnormal combustion
Figure 2. Waveform of the sound pressure at the bottom end of the tube.
noise problems. The test rig consists of parts that are available
in any hardware store, requiring only a few simple modifica- the burner outside the tube, adjust the gas flow until the flame
tions. It is recommended that the reader construct such a de- is about 10 mm (0.4 in.) high, and let it heat up for at least two
vice for the learning advantages offered by performing the ex- minutes. Then lower the tube so that the burner top extends
periments rather than only reading about them. Construction about 50 mm (2 in.) into the tube. A loud noise will be heard.
details are available by e-mail from the author. The signal picked up by the small microphone near the bot-
The burner shown in Figure 1 does not lend itself to the tom of the tube will show the pressure to have the waveform
exploration of various burner port geometries nor to the explo- of a somewhat distorted sine wave with a period of about 3.09
ration of various lengths of the mixture supply system, both of milliseconds, which corresponds to a fundamental frequency
which are known to offer possible solutions for abnormal com- of about 324 Hz (Figure 2).
bustion noise problems in furnaces or boilers.2 Experiments for The corresponding spectrum in Figure 3 shows the expected
demonstrating these possibilities will be covered in a follow- spike at 324 Hz. There are also smaller spikes at 648 Hz and
ing article. Since it is often helpful to explore the effect of 972 Hz. Since the frequencies of these lower amplitude spikes
changing some parts of a system before actually making the are exact multiples of 324 Hz, these spikes are the second and
change, this separate article will also describe simple computer third harmonics of the fundamental frequency of 324 Hz. Such
simulation tools. harmonics are to be expected since the waveform in Figure 2
Question #1 – What Does Abnormal Combustion Noise is a somewhat distorted sine wave. The conclusion to note from
Sound Like? The equipment needed for this experiment is this experiment is that the abnormal combustion noise that
shown in Figure 1. Without the transparent window section at occurs when the flame is inserted into the confines of the tube
the bottom, adjust the sliding extension at the top of the tube has the nature of cyclical pressure oscillations with very little
so that the total length of the assembly is 610 mm (24 in.). Light variation from one cycle to the next.

22 SOUND AND VIBRATION/JULY 2004

 –30
324 Hz
–40

Relative ampritude, dB
648 Hz
–50

–60 972 Hz

–70

–80

–90

–100
0 200 400 600 800 1000
Frequency, Hz
Figure 4. Map of the pressure field in the 610 mm long tube during
Figure 3. Sound pressure spectrum at the bottom end of the tube. oscillation. Note – The time scale represents the position in the tube
because the pressure was recorded with a moving probe tube micro-
With this experimental rig, the spikes in the spectrum are phone. There are about 1000 pressure cycles in the 3.3 sec record shown.
Thus, the individual cycles are so crowded that only the peaks can be
more distinct than they typically will be for an actual furnace
seen.
or boiler. Such units have other noise sources (e.g., fans) that
tend to fill in the spectrum between the spikes shown in Fig- overdriving the pressure transducer. Details of probe tube de-
ure 3. However, the existence of any discrete frequency spike signs can obtained from the author.
at a frequency that is unrelated to any of the other noise sources Question #3: Is It Possible to Calculate the Resonance Fre-
is “smoking gun” evidence of abnormal noise and the first step quency? While it is possible in principle to calculate a reso-
for solving such problems. nance frequency, it is often impractical for an actual unit. Such
To examine the spectrum requires an FFT analyzer, which calculations are likely to be very time consuming because the
also provides a means for recording waveshapes (like Figure walls of the combustion chamber are often not perfectly rigid.
2) and is an indispensable tool for solving abnormal combus- Thus, a rather elaborate mathematical model would be re-
tion noise problems. Any attempt to solve such problems with- quired.
out such a tool is like shooting blindfolded. Still, while an FFT It is instructive, however, to examine the calculation proce-
analyzer is essential, it will not answer all questions by itself. dure for the rigid walled experimental rig used here. The first
A good insight into the mechanism that causes the pressure os- step is to set up the equation for the wavelength of sound,
cillations is also required, which is exactly the purpose of the which is:
following experiments. c
l= (1)
Question #2 – What Determines the Frequency of Oscilla- f
tion? Listen to the change in the noise while adjusting the slid- where:
ing extension at the top of the tube to make the assembly longer l = wavelength in m
– the longer the tube, the lower the pitch. Also observe the c = speed of sound in m/s
changes in the spectrum. At a length of 610 mm, the frequency f = frequency in Hz.
of the fundamental component was 324 Hz, but at 810 mm (32 From the previous experiment, the frequency of oscillation
in.), it is 244 Hz. Since the frequency of oscillation is essen- is such that the wavelength of the pressure oscillations is equal
tially inversely proportional to the tube length, it appears that to about twice the length of the tube surrounding the flame.
the fundamental frequency may be the frequency of the low- Thus the resonance frequency could be found quite simply by
est half wavelength resonance of the gas space in the tube (rep- substituting 2(L+2 DL) for l in Equation 1 and solving for f. L
resenting the combustion chamber in these experiments). is the length of the tube and DL is the so called “end correc-
To check this hypothesis, insert from the top a probe tube tion,” 3 which takes into account the fact that the pressure am-
with a suitable pressure transducer until the tip extends to plitude does not go to zero at the open end, as shown in Fig-
about 50 mm (2 in.) below the bottom of the test rig, then with- ure 4.
draw it while recording the pressure-time history. If the probe There are several problems with this calculation. The speed
tube is withdrawn at a uniform rate the record will look like of sound increases in direct proportion with the square root of
that in Figure 4. The upper envelope should have the shape of the absolute temperature.4 The temperature in the test rig of
half a cycle of a sine wave, confirming that the fundamental Figure 1 is not constant over the length of the tube, but depends
frequency of oscillation is indeed that of the lowest half wave- on position. The bottom of the tube is at room temperature (20°
length resonance of the combustion chamber. C), a maximum exists just above the flame and then the tem-
The test of Figure 4 is difficult to perform due to the diffi- perature declines depending on heat transfer to the walls of the
culty of withdrawing the probe tube at a uniform rate. Also, it tube. Further complications arise because of temperature and
is necessary to pass the transducer signal through a bandpass velocity gradients in the area of the tube outlet.5
filter to remove the harmonics since only the map of the fun- So even for the simple geometry of the test rig of Figure 1, it
damental component is of interest. An easier, although some- would not be easy to calculate the resonance frequency with
what tedious, mapping procedure is to use the probe tube to any degree of accuracy.
obtain a series of spectra at fixed points along a traverse through Question #4: How Can the Resonance Frequency Be Mea-
the combustion chamber and then plotting the amplitude val- sured? Under cold conditions, measuring the resonance fre-
ues of the fundamental component vs. position. Since the op- quency is not difficult. With a homemade pulse generator (con-
erating conditions of the unit are likely to drift over the map- struction details available from the author), one can set off a
ping duration, the FFT analyzer should be set up to show the small explosive charge in the flame area with the cap of a toy
decibel difference between the signal from the probe tube and pistol. A microphone in the same area will pick up the pres-
that from a fixed microphone such as the one shown at the sure response. Figure 5 shows the response waveform for a 610
bottom of the tube in Figure 1. Such plots of amplitude vs. po- mm long tube at room temperature, taken with a sampling rate
sition are called “mode shape plots.” of 12,000 samples per second. The first 0.05 seconds show that
Mapping of the pressure field in the combustion chamber can many modes are excited, although all but the fundamental one
be a very useful tool for diagnosing some abnormal combus- die out quickly. In the time interval from 0.1 to 0.2 seconds,
tion noise problems in actual heating equipment. For either of Figure 5 shows about 28.5 cycles. Thus the lowest resonance
the test methods, the probe tube must be designed to prevent frequency of this system is about 285 Hz at room temperature.

MATERIALS REFERENCE ISSUE 23

 Performing this test in the rig of Figure 1 shows that the peak
pressure developed by the explosion exceeds the dynamic
range of any ordinary microphone. Figure 5 indicates that the
microphone was overdriven for the first 0.04 seconds of the
record. Though undesirable from the standpoint of possible
damage to the microphone, it did not prevent extracting the
resonance frequency information from the test results. When
performing this test in an actual heating unit, overdriving is
not as likely to occur because the volume of the unit will be
much larger than that of the test rig, so that the explosion will
be much less confined and thus generate much lower pressures.
Remember that under actual operating conditions, the reso-
nance frequency will be considerably higher because the speed
of sound increases with temperature. Still, such measurements Figure 5. Pressure response of the 610 mm long tube in the test rig of
Figure 1 at room temperature when excited by the explosion of a toy
can be useful for diagnosing a problem of abnormal combus- pistol cap.
tion noise because they provide more insight into how the unit
behaves. By extrapolation, a rough estimate of the resonance is also considerable inherent damping due to static pressure
frequency of the hot system can be obtained, which is sufficient drop in various parts of the system. Thus, the desired increase
to determine whether the oscillations do indeed occur at the in damping can often be achieved by increasing the pressure
resonance frequency of the combustion chamber. drop. For some burner designs, this is best accomplished by
Question #5 – Is the Abnormal Combustion Noise a Reso- increasing the velocity through the burner. Increasing damp-
nance Problem? The previous experiments have shown that the ing by a sufficient amount will always work, although smaller
frequency of oscillation is determined by a resonance frequency increases will have little effect on the noise.
of the tube which, in these experiments, constitutes the com- To demonstrate this, slide the damping sleeve up very slowly
bustion chamber. It is therefore understandable that such ab- while listening to the noise. Until the critical position is
normal noises are often called “combustion resonances,” even reached where the oscillations are stopped, there is very little
though the term is misleading since it implies that the noise is reduction of the total noise. Beyond that position, the reduc-
a resonance problem. If that was the case, changing the reso- tion is quite sudden.
nance frequency should take the system out of resonance and Question #7 – What Is the Role of the Flame? When perform-
solve the problem. The experiment for Question #2 has, how- ing the above experiment, the flame should be observed closely
ever, shown that not to be the case, so the abnormal noise is while sliding the sleeve up and down repeatedly to stop and
clearly not a resonance problem. The most descriptive term for start the noise. When the noise is off, the light blue inner cone
the abnormal combustion noise explored in these experiments of the flame is sharply defined, but when the noise is on it be-
is “combustion-driven oscillations.” comes blurred. The appearance of the flame changes because
These experiments indicate that moderate changes of the the flame size grows and shrinks rapidly in a cyclical fashion
resonance frequency of a furnace or boiler will merely change during the oscillations. Since the flame oscillations occur at
the frequency of the noise without much effect on the ampli- the same frequency as the pressure oscillations, the eye can-
tude. There are well documented cases, however, in which not follow the rapid size changes, so the cone boundaries sim-
large changes did eliminate the combustion-driven oscillations. ply appear blurred.
Replacing the 610 mm (24 in.) long combustion tube with one A clear view of the cyclic variations of the flame size requires
that is only one third as long will eliminate the oscillations. the use of a mechanical stroboscope. A stroboscope is a slot-
The reason will be explored in a following article. ted disc driven at such a speed that the slot passing frequency
Question #6: How Does One Solve a Combustion-Driven Os- is about 0.5 Hz below the oscillation frequency. This will pro-
cillation Problem? Putnam has shown that combustion-driven duce a slow motion image of the changes in flame size that can
oscillations occur only for certain combinations of the proper- also be captured with a film or video camera. Figure 6 shows a
ties of the combustion chamber, flame and mixture supply.2 sequence of frames from a movie shot through a homemade
This experiment will explore a modification of the combustion stroboscope (construction details are available from the au-
chamber that will stop the oscillations quite spectacularly. thor). The significance of the flame size variations is that the
Some suitable modifications of the flame will be explored later. burning rate of the flame varies in direct proportion to the sur-
To investigate this modification of the combustion chamber, face area of the blue inner cone.
remove the sliding extension at the top of the tube and install Question #8 – Which Came First, the Flame Oscillations or
a short Plexiglas section at the bottom, about 75 mm long, so the Pressure Oscillations? The old “chicken-or-egg” scenario
that the flame can be observed. Then take an ordinary cloth – small oscillations of the flame size and burning rate cause
handkerchief, fold it twice to form a strip about 4 times longer pressure oscillations that cause larger oscillations of the burn-
than it is wide. Wrap it around the top part of the tube, leav- ing rate that cause larger pressure oscillations, and so on. It is
ing about 25 mm (1 in.) uncovered at the top. Now lower the a vicious circle (see Figure 7) that results in high amplitude
combustion chamber tube over the flame carefully so the combustion oscillations, limited only when the amplitude sur-
Plexiglas will not get damaged. Once the noise has started, slide passes the limits of linear behavior.
the cloth sleeve up so it extends about 25 mm (1 in.) beyond Question #9 – How Does the Vicious Circle Get Started?
the top of the combustion chamber, which will kill the noise Comparing the first few seconds of pressure oscillation under
instantly. various conditions can provide quite a bit of insight about self-
When doing this experiment for a live audience, I have of- excited combustion oscillations. The following three experi-
ten used the theatrics of removing the handkerchief, unfold- ments explore the effect of how far the tip of the burner extends
ing it, and asking the audience to convince themselves that into the combustion tube (adjusted to a total length of 610 mm).
there is nothing hidden inside it that might have produced the Before starting the burner, adjust the stand so that the burner
‘magic’ of killing the noise. tip extends exactly 50 mm into the tube. Rough adjustment is
The sleeve killed the noise by adding damping to the sys- made by loosening the band clamp and sliding the combustion
tem. I have solved several cases of combustion-driven oscilla- chamber tube up or down with the vertical adjustment screw
tion problems in actual heating systems by adding damping in in a position leaving about 13 mm between the two T-fittings
a very similar way. In our experimental rig without the damp- of the stand. Then use the vertical screw for fine adjustment,
ing sleeve, there is hardly any damping other than by acoustic making sure that coupling at the lower end of the screw remains
radiation from the inlet and outlet. In most heating units there in contact with the fixed (bottom) T-fitting and that the burner

24 SOUND AND VIBRATION/JULY 2004

Figure 6. Sequence of frames from a slow motion movie of the oscillat-
ing flame. Note – The frames show the flame size at instances 1/4 of a
cycle apart. The last frame shown is a repeat of the first.

Burning Rate Pressure

Oscillations Oscillations

Figure 8a. Onset of the oscillations during a hot start with the burner
inserted to a depth of 50 mm.
Figure 7. The vicious circle that gives rise to combustion oscillations.

is centered in the tube.

Lifting the top T-fitting and swinging the combustion tube
out of the way exposes the burner for lighting. Adjust the flame
to be 10 mm high, swing the top T-fitting to bring the tube into
position and lower it until the coupling in the adjustment screw
contacts the top of the bottom T-fitting. Note that it will take
about a minute for the oscillations to start and that the ampli-
tude will slowly grow for about another minute as the burner
continues to heat. This delay in the start of the oscillations is
of little practical concern, so long as the oscillations start once
the system has heated.
In order for these tests to provide valid comparisons, the
entire test rig must be at a constant temperature since any tem- Figure 8b. Expanded view of the oscillation build-up during a hot start
with the burner inserted to a depth of 50 mm.
perature drifts would introduce an additional variable. Let the
apparatus stabilize for another 2 minutes. During this time, ap-
ply the damping sleeve used in the experiment for Question
#6, then start recording the pressure signal as the sleeve is
quickly pulled off the top of the tube. A typical record is shown
in Figure 8a, which indicates that the oscillations do not start
instantly but take about half a second to build.
An expanded view of part of the oscillation build-up is
shown in Figure 8b, which shows that the amplitude of each
cycle is about 3% larger than that of the previous cycle. Math-
ematically, this is known as an exponential build-up, and the
growth per cycle is called the build-up rate. This building up
of the pressure amplitude shows that acoustic energy is gener-
ated by the oscillating flame faster than it is dissipated by
damping and by radiation from the ends of the combustion
tube. In the experiment for Question #6 we have seen that the
oscillations can be stopped by increasing damping to the point
where the acoustic energy is dissipated faster than it is gener-
Figure 9. Onset of the oscillations during a hot start with the burner
ated by the oscillating flame. inserted to a depth of 40 mm.
After obtaining this record, obtain a spectrum and note that
the fundamental frequency of the fully built-up oscillation is build-up rate of oscillations depends on the degree of confine-
324 Hz. Next, re-apply the damping sleeve and use the verti- ment of the flame. This may seem only of academic interest be-
cal adjustment screw to raise the combustion tube by 10 mm, cause in most furnaces and boilers flame confinement is im-
which will take 6.3 turns clockwise. Record the pressure sig- possible to change. I do, however, remember a furnace with an
nal again as the sleeve is quickly pulled off the top of the tube. oscillation problem that was solved simply by increasing the
A typical record is shown in Figure 9. Note that the oscillations clearance around the burners at the entrance to the heat ex-
build up much more slowly than those in Figure 8. Check the changer. Aside from that, the relation of oscillation build-up
spectrum of the fully built-up oscillation and note that the fun- to flame confinement is the key for gaining a deeper under-
damental frequency has shifted to 331 Hz. standing of what causes combustion oscillations.
Finally, re-apply the damping sleeve once more and use the The first step is to quantify the rather vague term “degree of
vertical adjustment screw to raise the combustion tube by an- confinement” used above. For this, we need to replace the dia-
other 10 mm. Again record the pressure signal as the sleeve is gram of Figure 7 with the one shown in Figure 10. In this sche-
quickly pulled off the top of the tube. With the insertion depth matic, the interactions of the flame and the pressure are ar-
of the burner reduced to 30 mm, oscillations do not build up ranged as a feedback loop that can be analyzed mathematically.
at all. The build-up rate has been reduced to zero. Under certain conditions, such loops become unstable, giving
Question #10 – How Can the Effect of Burner Insertion Depth rise to self excited oscillations, such as the generation of mu-
Be Explained? The previous experiments have shown that the sical sounds by flutes or organ pipes. These equations for the

MATERIALS REFERENCE ISSUE 25

onset of instability are derived in reference 6.
The blocks labeled Z, H and G in Figure 10 represent the Combustion Chamber
acoustic properties of the combustion chamber, the mixture
supply system, and the flame, respectively. These properties Z
can be formulated as equations that define the response of each qexternal qtotal
of these system components to an acoustic input. The response +

of any one of the components becomes the input to the next Oscillating p
Oscillating
Burning Rate Dq Pressure
one, and so on around the loop. The loop becomes unstable
when pressure oscillations are sustained even if there is no
external input, i.e. when qexternal = 0. This condition for insta- q1
G H
bility is that the magnitude of the product Z ¥ H ¥ G must be
greater than 1.0 at any frequency at which the sum of the phase Flame Mixture Supply
angles of Z, H and G is zero or 360°. 6
The block labeled Z in Figure 10 represents the confinement Figure 10. Schematic representation of the feedback loop (from Baade,
of the flame in the combustion chamber. The symbol Z is used 1978).
because this confinement amounts to an acoustic impedance. 3
For the test rig used in these experiments, the acoustic imped- 8 x 107
ance seen by the flame can be calculated, subject to some sim-

Magnitude of Impedance, Pa s/m3

7 x 107
plifying assumptions. The resulting plots of the magnitude and 50 mm Insertion
phase of Z are shown in Figure 11A and B for burner insertion 6 x 107
depths of 50 mm and 30 mm, and a total length of the combus- 5 x 107
tion tube of 610 mm.
Figure 11 shows that the magnitude of the acoustic imped- 4 x 107
30 mm Insertion
ance seen by the flame has a peak at any resonance frequency 3 x 107
of the combustion chamber and that the phase angle goes
2 x 107
though a 180° change at such a frequency. This jump in the
phase means that there is at least a 50/50 chance that, at these 1 x 107
frequencies, the above phase condition for self excited oscil- 0
lations will be satisfied. 320 325 330 335
While the phase condition must be satisfied for self excited Frequency, Hz
oscillations to build up, it alone is not sufficient for build-up
to occur. It is also necessary that the magnitude of the product Figure 11a. Magnitude of the acoustic impedance seen by the flame in
the test rig for insertion depths of 50 mm and 30 mm.
of Z ¥ H ¥ G must be larger than 1.0 at the frequency at which
the phase condition is satisfied. Putting it another way, it is nec-
100
essary that the magnitude of Z exceed the magnitude of 1/(H ¥
G).
Figure 11a shows that reducing the depth of insertion of the
50
Phase angle, degrees

burner reduces the peak magnitude of the acoustic impedance

50 mm insertion
by about 50%. For this reason, oscillations built up with a
burner insertion depth of 50 mm in the experiments described 30 mm insertion
0
above, but did not build up with a burner insertion depth of
30 mm.
Figure 11 also shows that reducing the insertion depth of the
–50
burner moved both the calculated phase and the peak magni-
tude of Z to a somewhat higher frequency. In the above experi-
ment with a 40 mm insertion, an even larger shift of the oscil- –100
lation frequency was also observed. From the experiment in 315 320 325 330 335
Question #2, we know that such frequency shifts do not stop Frequency, Hz
the oscillation. Thus, reducing the height of the impedance
magnitude peak must be the reason there are no oscillations at Figure 11b. Phase angle of the acoustic impedance seen by the flame
in the test rig for insertion depths of 50 mm and 30 mm.
a burner insertion depth of 30 mm.
Question # 11: What Are the Practical Uses of the Feedback angles either since they are needed primarily for predicting the
Loop Model? Figure 11 has shown that the mathematical model already known frequency of oscillation.
of Figure 10 can be used for explaining the effect of the flame The magnitude of Z ¥ H ¥ G can be reduced by reducing the
confinement on the occurrence of self excited oscillations. magnitude of any one of these properties at the known fre-
Similar curves could also have been generated to show that the quency of oscillation. This was demonstrated in the experi-
effect of additional damping is simply to reduce the height of ments with the damping sleeve and insertion depth. Both of
the impedance peak at the resonance frequency. these experiments have shown that the oscillations can be
In principle, this model could be used to predict whether a avoided by reducing the magnitude of only the combustion
proposed boiler or furnace design will have an oscillation prob- chamber impedance. The changes made did not involve either
lem. That such predictions are possible has been demonstrated the mixture supply or the flame. Thus, the key to solving os-
for a gas turbine combustor7 as well as for typical heating prod- cillation problems in heating equipment is to determine which
ucts.8 The amount of effort required is rather substantial, how- part of the particular unit best lends itself to modifications that
ever, and is not likely to be justifiable since these problems do can reduce its response magnitude at the frequency of oscilla-
not occur very often in heating products and the stakes are not tion. Most often, this is more easly said than done. To help with
nearly as high as in the case of gas turbines. this task, one needs to understand the trends of the behavior
Thus, the primary use of the feedback loop model in typical of the various components of the system. These trends can be
heating products is to provide guidance for solving oscillation seen in the frequency response functions of the system com-
problems if and when they occur. For this purpose it is not nec- ponents in Figure 10.
essary to know the equations for Z, H and G with any degree Question #12: How Do the Pressure Oscillations Cause
of precision – approximating the trends involved is adequate. Flame Oscillations? So far, we have examined the frequency
In many cases, there is no need to be concerned with the phase response of the combustion chamber that governs the way in

26 SOUND AND VIBRATION/JULY 2004

 110 One very informative flame model for conical flames shows
575 Hz 587 Hz
not only the effect of the fuel/air ratio but also other factors
100 547 Hz 560 Hz 580 Hz involved in the driving potential of such flames. 11 These fac-
565 Hz 570 Hz
90
540 Hz tors will be addressed in a subsequent article.
552.5 Hz 572.5 Hz

527.5 Hz 557.5 Hz 557.5 Hz 552.5 Hz

80 560 Hz References
542.5 Hz
542.5 Hz
522 Hz 532 Hz 550 Hz 552.5 Hz 542.5 Hz 1. Polifke, W., 2004: “Combustion Instabilities,” VKI Lecture Series,
70
Q Air (l/min)

527Hz 547.5 Hz 552 Hz 547.5 Hz “Advances in Acoustics and Applications,” March 15th-19th, 2004,
530 Hz 540 Hz 532 Hz 530 Hz Brussels, Belgium
60 507 Hz 517.5 Hz 537 Hz 535 Hz 2. Putnam, A. 1971: Combustion-Driven Oscillations in Industry,
517.5 Hz
50
512.5 Hz 522.5 Hz 517 Hz Elsevier.
497 Hz 525 Hz 520 Hz
3. Kinsler, L. E., A. Frey, A. B. Coppens, and J. V. Sanders, 1982: Fun-
507 Hz 497.5 Hz500 Hz
40
485 Hz damentals of Acoustics, John Wiley & Sons, p. 202.
505 Hz
497 Hz 4. Kapur, A., A. Cummings, and P. Mungur, 1972: “Sound Propagation
30 in a Combustion Can with Axial Temperature and Density Gradi-
ents,” J. Sound and Vibration, Vol 25 (1), pp. 129-138.
20 5. Cummings, A., 1977: “High Temperature effects on the Radiation
Impedance of an Unflanged Duct Exit,” J. Sound and Vibration, Vol
10 22 (2), pp 299-304.
6. Baade, P. K., 1978: “Design Criteria and Models for Preventing Com-
0 bustion Oscillations,” ASHRAE Transactions, Vol. 84, Part I, pp.
0 4 8 12
449-465.
Q Gas (l/min)
7. Sattinger, S. S., et al, 2000: “Sub-Scale Demonstration of the Ac-
Figure 12. Instability plot for the experimental system used by Elsari tive Feedback Control of Gas-Turbine Combustion Instabilities,”
and Cummings (2003). Trans. ASME, J. of Engineering for Gas Turbines and Power, Vol. 122,
pp. 262-268
which pressure oscillations are caused by flame oscillations. 8. Ogrzewalla, J., K. Schmillen, W. Bomscheuer, and F. Altendorf,
2003: “Analysis of Self-Excited Combustion Oscillation,” Proc. 3rd
We have seen that, in some cases, this can be exploited to solve
European Conf. on Small Burner Technology and Heating Equip-
a problem. In the majority of cases, however, the solution will ment, pp. 159-166.
require changing the way in which flame oscillations are 9. Polifke, W., M. Bockholts, P. Flohr, and C. O. Paschereit: “Fluid Dy-
caused by pressure oscillations, which depend on the type of namic Instabilities in a swirl Stabilized Burner and their Effect on
Heat Release Fluctuations,” Proc. of Flow Induced Vibrations,
burner. Flames of some burners are prone to vortex shedding.2
Luzern, CH, 2000.
Swirl type burners are also subject to similar fluid dynamic 10. Elsari, M. and A. Cummings, 2003: “Combustion Oscillations in Gas
instabilities.9 Fired Appliances: Eigen-Frequencies and Stability Regimes,” Ap-
For the majority of burners used in residential and small plied Acoustics, Vol 64, No. 6, pp. 565-580.
11. Schuller, T., S. Ducruix, D. Durox and S. Candel, 2002: “Modeling
commercial heating units, however, such flow instabilities are
Tools for the Prediction of Premixed Flame Transfer Function,” Pro-
not likely to be important. With these types of burners the os- ceedings of the Combustion Institute, 29:107-113.
cillating pressure acting on the burner ports will modulate the
flow of the air/fuel mixture into the flame. This modulated The author may be contacted at: peterbaade@aol.com.
input results in modulating the burning rate of the flame. This
process involves both the properties of the mixture supply and
those of the flame, as depicted in the bottom part of the feed-
back loop in Figure 10. To reduce these modulations, either the
mixture supply system or the flame have to be modified.
In many heating units, the mixture supply system may be the
easiest to modify. Suitable modifications are to lengthen or
shorten the mixture supply conduit. Unfortunately, such modi-
fications cannot be explored with the burner shown in Figure
1. Experiments with other burners will be documented in a sub-
sequent article. The mixture supply is also the only part of a
heating unit that lends itself well to modeling. Thus, the sub-
sequent article will include methods for predicting the effect
of proposed changes to avoid costly trial and error efforts.
Meanwhile, the reader can find some relevant information
in Chapter 3 of Putnam.2
Question #13: How Can the Flame Be Modified to Solve an
Oscillation Problem? The air/fuel ratio is a significant variable
for the the “driving potential” of a flame.2 In Figure 10, the
driving potential is represented by the amplifier G. Figure 1210
shows that, in systems operating on the lean side, the air has
to be increased to stop the oscillations; while for systems with
Bunsen-type flames, the primary air has to be reduced. Years
ago, many Bunsen-type systems had a spoiler in the burner
mixing tube for reducing the primary air.
The torch used in these experiments can be easily modified
for exploring the effect of changing the air/fuel ratio further
away from the stochiometric ratio. All that is needed is to add
a sliding sleeve that will partly block the air inlet holes (con-
struction details are available from the author). By doing so,
the abnormal noise can be stopped quite easily.
While any substantial changes in air/fuel ratio may not be
acceptable as a permanent solution, it is useful to explore this
in a real unit if permitted by the burner design. Doing so will
make it quite clear that the flame drives the oscillations even
though practical solutions can often be found without chang-
ing the flame.

MATERIALS REFERENCE ISSUE 27