Data-driven identification of the critical transition to thermoacoustic instability in a full-scale solid rocket motor

Thermoacoustic instability is a persistent problem frequently observed in various types of combustors, resulting in damaging consequences. However, our understanding of the dynamics in industrial combustors undergoing thermoacoustic instability, particularly in solid rocket motors, still remains limited. Data-driven precursors for thermoacoustic instability in such systems are also unknown. In this study, we use recurrence network measures and spectral entropy to characterize the dynamics of pressure data obtained from a full-scale solid rocket motor transitioning to thermoacoustic instability and design data-driven precursors for thermoacoustic instability. We show the scale-free nature of combustion noise and that the dynamical transition from combustion noise to thermoacoustic instability can be detected using two complex network measures: the average path length and average betweenness centrality. We calculate the spectral entropy in the frequency domain and find it more sensitive to detecting the dynamical transition and computationally cheap, which is promising for flexible use as a new precursor in thermoacoustic instability prediction. Our work highlights the feasibility of employing complex network measures and spectral entropy for precursors in solid rocket motors, paving a new path for using data-driven measures to early warning of thermoacoustic instability in solid rocket motors.