Onset Of Combustion Instability Research Articles

Reducing Uncertainty in the Onset of Combustion Instabilities Using Dynamic Pressure Information and Bayesian Neural Networks

Abstract Modern low-emission combustion systems with improved fuel-air mixing are more prone to combustion instabilities and, therefore, use advanced control methods to balance minimum NOx emissions and the presence of thermoacoustic combustion instabilities. The exact operating conditions at which the system encounters an instability are uncertain because of sources of stochasticity, such as turbulent combustion, and the influence of hidden variables, such as unmeasured wall temperatures or differences in machine geometry within manufacturing tolerances. Practical systems tend to be more elaborate than laboratory systems and tend to have less instrumentation, meaning that they suffer more from uncertainty induced by hidden variables. In many commercial systems, the only direct measurement of the combustor comes from a dynamic pressure sensor. In this study, we train a Bayesain Neural Network to predict the probability of onset of thermoacoustic instability at various times in the future, using only dynamic pressure measurements and the current operating condition. We show that on a practical system, the error in the onset time predicted by the Bayesain Neural Networks is 45% lower than the error when using the operating condition alone and more informative than the warning provided by commonly used precursor detection methods. This is demonstrated on two systems: (i) a premixed hydrogen/methane annular combustor, where the hidden variables are wall temperatures that depend on the rate of change of operating condition, and (ii) full-scale prototype combustion system, where the hidden variables arise from differences between the systems.

Lean blow-out prediction in an industrial gas turbine combustor through a LES-based CFD analysis

The use of lean premixed flames to control combustion temperature and limit nitric oxides (NOx) formation is a common practice in all the applications of land-based gas turbines irrespective of their size and power production capacity. Despite the capability of lean premixed combustion to nominally reduce NOx emissions well below the most strict international regulations, such technology still requires improvements to overcome the impact on whole engine operability due to possible onset of combustion instabilities and flame lean blow-out (LBO). Both phenomena involve highly unsteady processes resulting in challenging numerical modeling. Particularly critical is the prediction of lean blow-out mechanisms where the role of finite rate chemistry and turbulence cannot be easily simplified. The current study aims at proving the capability of a CFD methodology based on Large Eddy Simulation (LES) and on an extended version of the Turbulent Flame Closure (TFC), to describe and properly catching the dynamics of flame extinction at the leanest operating conditions in typical turbulent lean premixed flames adopted in gas turbine combustors. The methodology is first validated on a laboratory test case where detailed experimental results are available. The extended TFC model well describes the LBO process observed in an atmospheric swirl stabilized burner investigated at the University of Cambridge. The proposed approach is then applied to the investigation of a real scale combustor operated with a swirl lean burner, representing the standard device adopted by Baker-Hughes on the heavy-duty 16 MW class engine Nova LT16. An accelerated numerical procedure is used to trigger the LBO, considering both constant firing temperature and constant pilot split conditions. An excellent agreement is observed with respect to the data obtained on a full-annular test rig, confirming the validity of the developed LBO modeling strategy which can be used in design phase to improve the overall engine operability.

Fast and Unbiased Determination of Dominant Frequencies and Amplitude of Thermoacoustic Oscillations

Reported is a three-step procedure (also referred to as the observer) for fast and unbiased determination of the dominant frequencies and amplitude of thermoacoustic oscillations. The first step is to determine the rang of the dominant frequency by computing the amplitude at frequency bins uniformly distributed from the lowest possible frequency to the Nyquist frequency (i.e., one-half of the sampling frequency) with a strategically specified, widely spaced frequency-bin separation. The other steps are to refine the estimation. The frequency estimation becomes unbiased after step 2, and the error (or uncertainty) is insensitive to the signal-to-noise ratio. The computational complexity can be orders-of-magnitude lower than the discrete Fourier transform and the fast Fourier transform, and the accuracy is no longer limited by , where and refer to the sampling frequency and the data length, respectively. This Paper recommends that the frequencies and amplitude be estimated every two pressure cycles. Cycle-to-cycle (or short-term) variations in the frequencies and amplitude are well resolved for thermoacoustic oscillations in three different real devices exhibiting quasi-steady oscillations before the onset of combustion instabilities and spontaneous instabilities (i.e., linearly unstable oscillations) with different levels of nonlinearities, respectively.

Predicting Onset of Combustion Instability in Gas Turbines Using Fiber Optic Sensors

A data driven method based on cross-entropy derived from symbolic time series analysis of the chemiluminescence and pressure variations has recently been found to be robust and computationally efficient among existing methods for predicting instability. In this article, we report the development and testing of compact and field-deployable optical fiber-based modules for sensing chemiluminescence and pressure variations in the combustor. The time-series data obtained from the above sensor modules are used to deduce the D-Markov cross-entropy for different protocols corresponding to typical combustor operation conditions. Such cross-entropy data is compared with similar data obtained from standard high speed camera and PZT-based pressure sensors. It is found that the fiber optic sensor modules compare favorably with the conventional sensors thereby providing an exciting new pathway for field-deployable solutions to monitor the onset of thermo-acoustic instability in combustors.

Multifractality in combustion noise: predicting an impending combustion instability

AbstractThe transition in dynamics from low-amplitude, aperiodic, combustion noise to high-amplitude, periodic, combustion instability in confined, combustion environments was studied experimentally in a laboratory-scale combustor with two different flameholding devices in a turbulent flow field. We show that the low-amplitude, irregular pressure fluctuations acquired during stable regimes, termed ‘combustion noise’, display scale invariance and have a multifractal signature that disappears at the onset of combustion instability. Traditional analysis often treats combustion noise and combustion instability as acoustic problems wherein the irregular fluctuations observed in experiments are often considered as a stochastic background to the dynamics. We demonstrate that the irregular fluctuations contain useful information of prognostic value by defining representative measures such as Hurst exponents that can act as early warning signals to impending instability in fielded combustors.

Impact of the Flame-Holder Heat-Transfer Characteristics on the Onset of Combustion Instability

In this article, we investigate the impact of heat transfer between the flame and the flame-holder on the dynamic stability characteristics of a 50-kW backward-facing step combustor. We conducted a series of tests where two backward step blocks were used, made of ceramic and stainless steel, whose thermal conductivities are 1.06 and 12 W/m/K, respectively. Stability characteristics of the two flame-holder materials were examined using measurements of the dynamic pressure and flame chemiluminescence over a range of operating conditions. Results show that with the ceramic flameholder, the onset of instability is significantly delayed in time and, for certain operating conditions, disappears altogether, whereas with the higher conductivity material, the combustor becomes increasingly unstable over a range of operating conditions. We explain these trends using the heat flux through the flame-holder and the change in the burning velocity near the step wall. Results suggest a potential approach using low-thermal-conductivity material near the flame-holder as passive dynamics suppression methods.

Flame Ionization Sensor Integrated Into a Gas Turbine Fuel Nozzle

Recent advances in lean premix gas turbine combustion have focused primarily on increasing thermodynamic efficiency, reducing emissions, and minimizing combustion dynamics. The practical limitation on increasing efficiency at lower emissions is the onset of combustion instability, which is known to occur near the lean flammability limit. In a laboratory environment there are many sensors available that provide the combustion engineer with adequate information about flame stability, but those sensors are generally too expensive or unreliable for widespread application in the field. As a consequence, engines must be commissioned in the field with adequate stability margin such that normally expected component wear, fuel quality, and environmental conditions will not cause the turbine to experience unstable combustion. Woodward Industrial Controls, in cooperation with the National Energy Technology Laboratory, is developing a novel combustion sensor that is integrated into the fuel nozzle such that low cost and long life are achieved. The sensor monitors flame ionization, which is indicative of air–fuel ratio and most importantly flame stability.

Investigation of self-excited combustion instabilities in two different combustion systems

The objective of this paper is to characterize dynamic pressure traces measured at self-excited combustion instabilities occurring in two combustion systems of different hardware. One system is a model lean premixed gas turbine combustor and the other a fullscale bipropellant liquid rocket thrust chamber. It is commonly observed in both systems that low frequency waves at around 300Hz are first excited at the onset of combustion instabilities and after a short duration, the instability mode becomes coupled to the resonant acoustic modes of the combustion chamber, the first longitudinal mode for the lean premixed combustor and the first tangential mode for the rocket thrust chamber. Low frequency waves seem to get excited at first since flame shows the higher heat release response on the lower frequency perturbations with the smaller phase differences between heat release and pressure fluctuations. Nonlinear time series analysis of pressure traces reveals that even stable combustion might have chaotic behavior with the positive maximum Lyapunov exponent. Also, pressure fluctuations under combustion instabilities reach a limit cycle or quasi-periodic oscillations at the very similar run conditions, which manifest that a self-excited high frequency instability has strong nonlinear characteristics.