A New Methodology for Early Detection of Thermoacoustic Combustion Oscillations Based on Permutation Entropy

We carry out a series of experimental investigations in a model combustor to detect a precursor of thermoacoustic combustion oscillations based on permutation entropy, which can amplify the subtle changes effected in the time sequence to identify the anomaly. By changing the flame's location or the fuel flow to a value, an abrupt switch from aperiodic small-amplitude oscillations to periodic large-amplitude oscillations would occur in pressure fluctuations. The characteristic frequency of combustion oscillation is obtained by spectral analysis, with which a modified algorithm of the permutation entropy is proposed. The impact evaluation on key parameters such as moving step sizes and window sizes reveals that the moving data permutation entropy has strong robustness, and can accurately detect the onset of thermoacoustic oscillations. Further nonlinear analysis exhibits peculiar dynamics of the combustion system, which result in specific patterns in the time series and provide a theoretical basis for anomaly detection. Our results suggest that the permutation entropy has a certain potential in early warning and detection of combustion oscillations.