On the observation of the transient behavior of gas turbine combustion instability using the entropy analysis of dynamic pressure

Abstract

In the present study, two new assessment methods for combustion instability (CI) based on the concept of entropy are proposed and their performances are evaluated. These two methods are (1) the entropy of energy (EoE) method, which utilizes the probability distribution of the squared value of dynamic pressure (DP) and (2) the entropy of energy amplified by pressure magnitude (EoEPM) method, which considers both the averaged pressure magnitude and the energy distribution in the time domain to compensate for the limitation of the EoE method. In these methods, the entropy was defined as the energy distribution of the DP. Therefore, the assessment for CI can be performed by using the entropy characteristics: the unstable combustion status shows the relatively wider energy distribution than the stable one. The performances of these methods were validated using the DP data obtained from a gas turbine combustion experiment and were compared with those of the conventional methods, namely, the temporal kurtosis (TK) method, the root-mean-square method, and the continuous wavelet transform method. The EoE method showed a comparable performance with the TK method in terms of detection time and accuracy of CI detection in terms of time series plot trends. The EoEPM showed the best performance by precisely detecting the transition from the unstable to the stable combustion status, and its detection time was faster by 0.06–0.71 s than that of the other methods. This new assessment method is expected to contribute to the safe and reliable operation of combustion engines, such as gas turbines and rockets, by rapidly detecting the onset of CI and by preventing fatal accidents before a dynamic pressure fluctuation worsens.