Environmental Engineering Reference
In-Depth Information
or events coincide with the occurrence of
flame loss (extinction) and reignition
events. The bursts (unsteady events) arise due to local extinction and strong reig-
nition of
fl
flame. These events can be called as precursor events to LBO, and they
grow in number near blowout. The corresponding PSD plot reveals some strong
dominating high-powered low-frequency events which are likely related to time-
scale of occurrence and duration of
fl
flame extinction and reignition events.
Similar tests were carried out by changing the fuel-injecting ports. As the fuel
entry point is raised, as discussed earlier, the length of the premixing section (L
fuel
)
available for mixing of fuel and air decreases. Thus, the
fl
flame obtained using a
lower mixing length (i.e., higher port) for fuel injection results in increasingly less
premixed
fl
ames.
Figure
8
shows the time series data for CH chemiluminescence for port 5 and the
corresponding PSD plots for different normalized equivalence ratio conditions.
Similar to port 1 and port 3 tests, here also the mean of the chemiluminescence
intensity decreases with the decrease in equivalence ratios due to reduced heat
release rate.
At higher equivalence ratio condition (i.e.,
fl
Φ
LBO
= 1.69), the CH oscillations
are strong in intensity and coherent in nature which is con
Φ
/
rmed from the corre-
sponding PSD where it exhibits a strong dominating frequency (
≈
20 Hz). At
Φ
Φ
LBO
= 1.35, the 20 Hz dominating combustion frequency diminishes from the
spectrum and a few lower frequencies are added. The spectrum shows a bunch of
frequencies in a broadband 1
/
50 Hz. As the combustor approaches blowout con-
-
dition (
Φ
LBO
= 1.06, 1.03), the CH chemiluminescence time traces are still
random and not coherent in nature. The corresponding PSD does not reveal any
dominating frequency and exhibits different frequencies in a band of 1
Φ
/
50 Hz. Near
-
blowout condition (i.e.,
Φ
/
Φ
LBO
= 1.0), the CH time trace does not exhibit sig-
ni
cant burst (unsteady events) from the mean. This corroborates our observation
that the
flame never lifts off from the dump plane and distinct extinction and
reignition events are absent.
From the above observation, it is evident that the near LBO behavior changes as
the degree of premixing varies. The previous LBO sensing techniques were tested
on premixed
fl
fl
ames. So,
first we should check the ef
cacy of those for different
degrees of premixing.
We tested statistical approach (Muruganandam
2006
; Yi and Gutmark
2007
),
spectral analysis (Yi and Gutmark
2007
), and direct signal analysis (Muruganan-
dam
2006
) for this purpose. Figure
9
shows variation of NRMS value as LBO
metric (as used by Yi and Gutmark
2007
) with normalized equivalence ratio at
different degrees of premixing. For port 1, the metric increases rapidly as the
equivalence ratio moves toward LBO (Fig.
9
a). This feature can be used as a
satisfactory metric for early detection of LBO. However, Fig.
9
b
e shows that this
rapidly rising feature of NRMS is absent for lesser degree of premixing (port 2 to
port 5). This shows that the metric is suitable for premixed
-
ames
with a larger degree of premixing. But, for partially premixed and non-premixed
fl
fl
flames or for the
fl
flame, the metric cannot provide satisfactory result. The other methods mentioned
Search WWH ::
Custom Search