Dynamic pressure method of detecting flame on/off in gas turbine combustion cans for engine protection

Patent metadata

Field	Value
Publication number	US-9599527-B2
Application number	US-201514691605-A
Country	US
Kind code	B2
Filing date	Apr 21, 2015
Priority date	Apr 21, 2015
Publication date	Mar 21, 2017
Grant date	Mar 21, 2017

Abstract

The flame status of a group of gas turbine combustors is acoustically monitored using dynamic pressure sensors within the combustor. Dynamic pressure sensor output signals are received from the sensors and processed to determine flame status. The signals are processed both by performing a correlation analysis within each combustor and by applying a wavelet-based flame detection algorithm to each output signal. A flame is determined to be present based on the correlation analysis and the wavelet-based flame detection algorithm. The wavelet-based flame detection algorithm is chosen based on whether the gas turbine combustors are in an ignition phase or a monitoring phase.

First claim

What is claimed is:

1. A method for monitoring flame status in a plurality of combustors of a gas turbine engine, comprising:
receiving dynamic pressure sensor output signals from dynamic pressure sensors arranged in the combustors, the dynamic pressure sensor output signals being indicative of acoustic oscillations within the combustors;
for each of the dynamic pressure sensor output signals, computing a preprocessed time-domain signal by digitizing and filtering the dynamic pressure sensor output signal;
for each of the dynamic pressure sensor output signals, computing a preprocessed wavelet signal using a discrete wavelet transform;
applying a wavelet-based flame detection algorithm to the preprocessed wavelet signals to determine a wavelet-based flame status for each of the dynamic pressure sensor output signals;
for each combustor, performing a correlation analysis on one or more of the preprocessed time-domain signals to determine a correlation-based flame status;
determining the flame status based on the correlation-based flame status and the wavelet-based flame statuses.

2. The method of claim 1 , further comprising:
selecting the wavelet-based flame detection algorithm based on a determination whether the combustors are in an ignition phase or a combustion phase.

3. The method of claim 2 , wherein selecting the wavelet-based flame detection algorithm further comprises:
selecting an algorithm for the ignition phase when a pilot oil valve is open for shorter than an oil ignition time or a pilot gas valve is open for shorter than a gas ignition time;
selecting an algorithm for the monitoring phase when a pilot oil valve is open for longer than the oil ignition time or a pilot gas valve is open for longer than a gas ignition time;
selecting no algorithm and terminating the monitoring if both the oil pilot valve and the gas pilot valve are closed.

4. The method of claim 1 , wherein the wavelet-based flame detection algorithm selected for ignition phase monitors a flame ignition signal (FI) defined as:

FI
=

1
π

⁢

tan

-
1

⁡

(

a
·
P

⁢

⁢
W
⁢

⁢
S

-

c
·

median
⁡

(

PWS

1
⁢

⁢
to
⁢

⁢
x

)

)

-
0.5

where PWS is the preprocessed wavelet signal, x is a number of combustors and a and c are empirical parameters to set a relative importance of signal spread to amplitude of the preprocessed wavelet signal values.

5. The method of claim 1 , further comprising:
removing narrow band distortions in the dynamic pressure sensor output signals by limiting an amplitude of each frequency component to 75% of a highest level of that frequency component among all combustors.

6. The method of claim 1 , further comprising:
removing combustion instabilities in the dynamic pressure sensor output signals by limiting a maximum impact of low frequencies to 1 mbar.

7. The method of claim 1 , wherein performing the correlation analysis further comprises:
computing a regularized flame correlation coefficient from auto correlation values of each of two sensors in a combustor and a cross-correlation value between the two sensors; and
comparing the preprocessed time domain signals with a threshold determined from the regularized flame correlation coefficient.

8. The method of claim 7 , wherein computing a regularized flame correlation coefficient further comprises computing

R
=

C
AB

max
⁡

(

[

C
AA

·

C
BB

,
1

]

)

where
C AB is a cross-correlation of a sensor A and a sensor B in the combustor;
C AA is a auto-correlation of sensor A with itself, and
C BB is a auto-correlation of sensor B with itself.

9. The method of claim 1 , wherein determining the flame status based on the correlation-based flame status and the wavelet-based flame statuses further comprises:
determining a flame-on status for a combustor only if the correlation-based flame status and each of the wavelet-based flame statuses indicate a flame-on condition.

10. The method of claim 1 , wherein determining the flame status based on the correlation-based flame status and the wavelet-based flame statuses further comprises:
determining a flame-on status for a combustor only if the correlation-based flame status and at least one of the wavelet-based flame statuses indicate a flame-on condition.

11. A system for monitoring a flame in a gas turbine engine combustion chamber, comprising:
a plurality of dynamic pressure sensors arranged for producing dynamic pressure sensor output signals indicative of acoustic oscillations within a plurality of combustors of the gas turbine engine comb

Assignees

• Siemens Energy Inc

Inventors

• Claussen Heiko
• Ulerich Nancy H
• Momin Zainul
• Rosca Justinian

Classifications

• G01L19/0663 Primary

  Flame protection; Flame barriers · CPC title

• G01M15/14 Primary

  Testing gas-turbine engines or jet-propulsion engines · CPC title