Impact of water injection on thermoacoustic modes in a lean premixed combustor under atmospheric conditions

The goal of the study presented in this paper is to analyze how the stability margin of a premixed combustor changes when liquid water is injected. Within the scope of this work, experimental results are presented and serve as a validation basis for a numerical model of the combustor. Experiments comprise qualitative as well as quantitative analyses by means of damping rate measurements under varying water-to-fuel ratios. For this purpose, a natural gas fueled low-NO_X burner which is equipped with a coaxially placed water injection system is operated in a single burner test rig under atmospheric conditions. Waterto-fuel ratios investigated in this study range between 0 and 2 whereby either the adiabatic flame temperature or the equivalence ratio is held constant. With technical applications in mind, the former case is chosen to represent the conditions occurring in a gas turbine combustor when the thermal output is increased at a constant flame temperature level. A recently developed approach is applied to extract the net damping rates of the occurring modes from dynamic pressure measurements. More specifically, the autocorrelation function of the combustion noise signal is evaluated by an algorithm based on Bayesian statistics to ensure for robust extraction of the complex eigenfrequencies. Beyond that, the dynamic system response is measured for various operating conditions with and without water injection by means of Scattering Matrices (SM) using the Multi-Microphone Method (MMM). Measured SM's are deployed in a numerical model of the combustor and acoustic conservation equations are solved in frequency space using a Finite Element (FEM) approach. The results are, on the one hand, evaluated to analyze occurring thermoacoustic modes by comparing its spatial distribution with experiments. On the other hand, eigenvalue studies are carried out and experimentally identified eigenfrequencies and damping rates are used to benchmark the results of the numerical analyses.