Environmental Engineering Reference
In-Depth Information
observed similar dominance of low-frequency oscillation and time-localized
intermittent events in acoustic data toward lean condition. Based on this observa-
tion, spectral, statistical, and thresholding signal processing schemes were devel-
oped for detecting blowout precursors with varying levels of time response,
sensitivity, and robustness.
Yi and Gutmark (
2007
) reported LBO prediction based on OH emissions in
partially premixed, liquid-fueled, multiswirl-stabilized gas turbine combustor.
A statistical analysis of the time series is used to justify two detection indexes,
namely the normalized chemiluminescence root mean square and the normalized
cumulative duration of LBO precursor events for blowout prediction.
Mukhopadhyay et al. (
2008
) used moving kurtosis and threshold-based tech-
niques applied on combustor pressure time series for early detection of
ame
extinction. More recently, Li et al. (
2007
) reported the use of a novel tunable diode
laser temperature sensor to acquire a temperature time series signal from partially
premixed, swirl-stabilized dump combustor. An increase in the power of low-
frequency temperature
fl
ed as a LBO precursor. The authors
further implement this LBO precursor as a control variable in active control system
and demonstrated prevention of LBO during power reduction and transient fuel
variation.
fl
fluctuations was identi
1.3 Objective of the Present Work
Several sensing strategies for predicting the proximity of combustor to LBO have
been reported by many researches in the past. In all
these studies,
the
fl
ame
dynamics near the blowout limit has been used for LBO prediction at a
xed degree
of premixing. However, the degree of premixing in
fluctuations
which are the primary cause of dynamic instability and should subsequently affect
the LBO limit (Cohen et al.
2001
). The optical signal-based techniques reported
earlier in literature are, however, reliant upon the
fl
uences heat release
fl
fl
fluctuation observed near LBO
limit. Williams (
1989
) reported that these
fluctuations are not persistent in diffusion/
partially premixed condition compared to observations in premixed
fl
fl
flame, which
motivates us to check the ef
cacy of those reported optical and acoustic signal-
based technique in the regime of variation of degree of premixing.
The statistical and the spectral approach reported earlier for sensing the prox-
imity of combustor to LBO needs a sample data window (N). A larger sample
window increases the accuracy of LBO predictor but decreases the ef
ciency of
sensing technique with respect to time response characteristics. The sensing tech-
niques use the quantitative blowout indicator calculated as an average of the sample
window and deviation from mean, which is again computationally inexpensive and
time-consuming and may lead to delay in real-time monitoring. Limitation of those
approaches motivates us to explore different approaches for sensing the proximity
of combustor to LBO, which are computationally inexpensive and economical. The
development of the proposed LBO detection techniques is based on the quanti
able
Search WWH ::
Custom Search