An early detection indicator of combustion instability for an industrial gas turbine combustor

Abstract

Detection of combustion instability is crucial for the safety and reliability of gas turbines. In this paper, the recurrence quantification analysis (RQA) and multi-fractal analysis (MFA) methods are applied to investigate the transition process from combustion noise to combustion instability in an industrial-scale combustor. Based on the dynamic pressure (DP) obtained from high pressure and high temperature tests, a novel method is proposed to construct early detection indicators (EDI) of combustion instability. The method is mainly based on the three-dimensional map of the recurrence rate, Hurst exponent, and root mean square ratio. A regression method and SVM are applied to define the classification boundary. For three test cases, the results showed that the proposed EDI can effectively detect the onset of combustion instability. Compared to the conventional method based on the root mean square levels of dynamic pressure, the EDI has capability to forecast the onset of combustion instability approximately a few hundred milliseconds in advance.