Study Of Premixed Flame Stability Based On OH—PLIF

In this paper, We combined planar laser induced fluorescence (PLIF) spectroscopy with flame radiation, extracting a lot characteristics of the combustion characteristics which can reflect changes, qualitative analysis of the impact of each characteristic variable combustion stability,constructed accelerating genetic algorithm based on real coding projection pursuit classification model for combustion stability diagnosis.In order to achieve a variety of combustion detection technology integration,we simulated gas turbine combustor monocular, designed and built the burning bench. The planar laser-induced fluorescence (PLIF) combined with flame radiation spectrometry,not only can diagnose the combustion flame extrinsic information from, but also to achieve a valid record transient flame structure.We are able to qualitatively analyze the distribution of products of combustion intermediate process and understanding the impact component of flame stability,explore aspects of combustion stability from the combustion mechanism.During the experiment, we focus on the different air-fuel ratio of methane-air combustion stability in the gas turbine combustor, established accelerating genetic algorithm based on real coding Projection Pursuit Model.Firstly, the paper extracted8features that affected the flame stability in the time and frequency domain, and the Real-coded Accelerating Genetic Algorithm Projection Pursuit Classification (RAGA-PPC) model was applied to analyze and discuss the flame features. We obtained the optimal projection direction that reflected the high-dimension data structure and gave the materiality level of the8features, which was that intensity mean had the lest contribution to the flame stability and the flicker frequency had the most contribution to the flame stability. The results coincided with the classical theory. Then an evaluation model of the flame stability was proposed based on the relationship of the projection values and the sample numbers. We gave the function of the evaluation model and analyzed the errors of the model. Finally, the testing sample was tested and the prediction result was given. The result showed that the evaluation model could meet the requirement of the flame combustion requirement.