Phase Space Description of Unstable Dynamics of a Lean Premixed Gas Turbine Combustor

This work describes unstable phenomena of a lean premixed gas turbine combustor considering the geometrical properties in a phase space representation. Several experiments have been performed in which the air excess ratio λ and the percentage of pilot fuel PFP have been varied. The time series of experimentally measured pressure and heat release fluctuations inside the duct have been analysed using vectors collecting topological parameters of the attractors of experimental acoustical and optical sensors. Therefore, unsupervised Kohonen associative memories have been trained to create clusters of operating conditions characterised by similar dynamical behaviours. The results of the clustering underline the practical possibility of defining an algorithm for instability pattern recognition, which could be effective for either a diagnostic or a control system.