Effects of noise intensity on early warning indicators of thermoacoustic instability: An experimental investigation on a lean-premixed combustion system

Abstract

In this work, we experimentally investigate the noise-induced dynamics of a lean premixed combustion system operating on natural gas–air mixtures that exhibit thermoacoustic instability via a subcritical Hopf bifurcation. The investigation is done before the bistable region with equivalence ratio ($\varphi$) as the control parameter. We analyze the acoustic pressure oscillations ($p^{\prime }$) in the combustor and fluctuations in the heat release rate ($q^{\prime }$) from the laminar quasi-flat flame at increasing levels of white noise. We show the effects of noise intensity on the reliability of various types of early warning indicators (EWIs) to predict the onset of the impending thermoacoustic oscillations. We investigate the indicators based on statistical measures (variance, skewness, and kurtosis), autocorrelation and spectral properties (coherence factor), system identification (growth/decay rates of $p^{\prime }$), multi-fractality (Hurst exponent), and time series complexity (permutation entropy and Jensen–Shannon complexity). The coherence factor, variance, and decay rates of $p^{\prime }$ always increases as the system approaches thermoacoustic instability, indicating their robustness as an EWI under most noise levels. An increase in kurtosis cannot be employed as an EWI. Implementing autocorrelation, skewness, Hurst exponent, permutation entropy and Jensen–Shannon complexity as effective EWIs has limitations: they can be estimated accurately only from pressure oscillations ($p^{\prime }$) data and work only above a particular threshold value of noise intensity. Our results have direct implication on early prediction and control of thermoacoustic instability in practical gas turbine combustors.

Novelty and significance statement

Developing effective early warning indicators (EWIs) to anticipate the onset of thermoacoustic instability is crucial for preventing potential damage and ensuring the reliable operation of lean premixed gas turbine combustion systems. In such systems, inherent noise dynamics undergo variations with changing operating conditions and combustor designs. Specifically, noise intensity increases as the system becomes more turbulent. In this study, we demonstrate that the inherent noise dynamics in a lean premixed combustion system play a crucial role in influencing the trends observed in early warning indicators of thermoacoustic instability. We address several key questions, including (a) whether comparative reliability assessments of different classes of EWIs exist, (b) the effect of variations in noise properties on the efficacy of EWIs, and (c) which EWIs are most reliable considering that noise characteristics depend on the system and may even change with operating conditions within a specific system. This information is critical for engine designers/users, facilitating the development of robust and effective monitoring systems for gas turbine combustion systems.