Near blowout instabilities and identification based on multivariate analysis with high-speed multi-species optical diagnostic techniques

Abstract

Early prediction and depth analysis of lean blowouts are desirable for both aeroengine and stationary gas turbines. This study used a combination of high-speed multi-species optical diagnostic techniques and multivariate analysis methods to investigate near-blowout instabilities and identify combustion conditions. The flame root dynamic behavior and heat release pulsation characteristics were investigated using advanced modal decomposition methods. Results showed that a near-blowout flame exhibited a characteristic frequency (∼200 Hz) or “fingerprint” related to the temporal and spatial coupling characteristics of CH and OH radicals. Furthermore, flame image feature screening and multivariate fusion for combustion condition identification are reported for the first time. Significant differential features and effective fusion models were investigated and discussed. The identification accuracy of the multivariate fusion feature model compared with that of the geometric feature model improved by more than 24 %. Moreover, simplified fusion models of low-order moment and geometric features were constructed with relatively high identification accuracy (99 %) and low relative standard deviation value (14.1 %). These results demonstrate the significant potential of combining multi-species optical diagnostic techniques and multivariate analysis methods, thereby expanding their application to the field of combustion science.