Dmytro Iurashev

Humming is a dangerous, combustion-driven acoustic oscillation phenomenon which can take
place in gas-turbine burners. Any satisfactory description of humming should include both
acoustic- and convective-related events. In fact, the distance crossed by a fluid ”particle”
during one humming period is typically of the same order of the distance between the flame
and the inlet of the air-fuel mixture. Available models often postulate the Mach number to
vanish, which is a scarcely justifiable hypothesis. Furthermore, the combined effect of non-normality
and non-linearity might invalidate the familiar correspondence between humming
onset and the growth rate of the humming mode predicted by linear stability theory. The
prediction of humming amplitude, not available from linear theory, is required in order to
assess the impact of the phenomenon. Thus, a non-linear – albeit simplified – description is
required. We make use of a proprietary Ansaldo Energia model implemented in COMSOL
Multiphysics, a finite element commercial software. Nonlinear terms are introduced into the
heat release model and simulations are performed in the time domain. The initial condition in
the simulations is composed by a set of random frequencies. The employed model filters both
the fundamental frequency, the harmonics and the convective frequency from the initial signal;
the variables in the model, such as pressure, velocity and temperature perturbations, oscillate
in time without further application of external forces. The evolution scenarios of the pressure
perturbation depend on the flame position and the mean velocity distribution. Three possible
occurrences are observed: amplitude growth, decay and saturation.

Recently, because of environmental regulations, gas turbine manufacturers are restricted to produce machines that work in the lean combustion regime. Gas turbines operating in this regime are prone to combustion-driven acoustic oscillations referred as combustion instabilities. These oscillations could have such high amplitude that they can damage gas turbine
hardware. In this study, the three-step approach for combustion instabilities prediction is applied to an industrial test rig. As the first step, the flame transfer function (FTF) of the burner is obtained performing unsteady computational fluid dynamics (CFD) simulations. As the second step, the obtained FTF is
approximated with an analytical time-lag-distributed model. The third step is the time-domain simulations using a network model. The obtained results are compared against the experimental data. The obtained results show a good agreement with the experimental ones and the developed approach is able to predict thermoacoustic instabilities in gas turbines combustion
chambers.

Currently, gas turbine manufacturers frequently face the problem of strong acoustic combustion-driven oscillations inside combustion chambers. These combustion instabilities can cause extensive wear and sometimes even catastrophic damage of combustion hardware. This requires prevention of combustion instabilities, which, in turn, requires reliable and fast predictive tools. We have developed a two-step method to find a set of operating parameters under which gas turbines can be operated without going into self-excited pressure oscillations. As the first step, an unsteady Reynolds-averaged Navier–Stokes simulation with the flame speed closure model implemented in the OpenFOAM Õ environment is performed to obtain the flame transfer function of the combustion setup. As the second step time-domain simulations employing low-order network model implemented in Simulink Õ are executed. In this work, we apply the proposed method to the Beschaufelter RingSpalt test rig developed at the Technische Universität München. The sensitivity of thermoacoustic stability to the length of a combustion chamber, flame position, gain and phase of flame transfer function and outlet reflection coefficient are studied.