Combustion Instabilities - Acoustics

In troubleshooting combustion instability related issues such as flashback, flameout (partial & full) or high acoustics events, it is important for control engineers to understand the basics of combustion theory. In certain gas turbine models focusing on DLE/DLN technology, a set of acoustics dynamic pulsation sensors or microphones are deployed to monitor & raise alarm, and in some cases automatically adjust fuel & trim flame temperature or even perform executive actions to prevent potential combustor damage.

In typical entry level combustion theory, we know that combustion instability is affected by other types of instabilities, including hydrodynamic and acoustic instabilities. ?Hydrodynamic instability is caused by flow perturbations i.e. fuel and/ or air, leading to hot gas wake which in turn, lead to unsteady heat release rate & subsequent pressure wave which feeds back/ impact the flow dynamics via acoustic wave (acoustic) pathway i.e. unsteady heat release acts as an acoustic monopole.

However note that the instability becomes sustained at specific operating regions (gain & phase) and not for the entire operating envelope of the machine temperature vs speed curve. Also, premixed type flame typically is more prone to acoustics as compared to diffused flame. The difference between the two lies in premixing location, i.e. fuel/air pre-mixed before/ at combustion zone. The relatively homogenous mixture with flame temperature controlled at lower temperature/ equivalence ratio contributes to the coupling dynamics between heat release and pressure wave. As mentioned the system gain & phase affect the dynamics (and tendency for positive feedback and sustained oscillation) and are different between the two flame types. It’s worth noting that over the entire operating envelope of a machine, typically both types of flame are present albeit in different amount.

.

It is suffice to say that the interaction and coupling between vortical wave, acoustic wave and entropic wave summarizes the stability dynamics. Attachments showed some interaction of the coupling whereby velocity is periodically perturbated with intermittent flashback to the inlet side. On the other hand, acoustics instability is showed by pressure fluctuations in static & rotating azimuthal mode of the combustor ring. It is worth mentioning here that the acoustic modes could be classified as longitudinal, transverse and/ or azimuthal. Longitudinal is typically related to lower frequency tone < 1200 Hz and could be classified as cold and hot tone, which amplitude directly/ inversely related to flame temperature. This mode is related to acoustics or sound wave propagation upstream & downstream of the forward/ axial hot gas flow, in which case the wave has to travel a longer distance before reflected off a boundary condition (thermal or physical) hence the slower frequency.

.

Transverse mode or screech typically occurs at a higher acoustic tone ~3.6 kHz and more dangerous and detrimental to the combustor hardware. This mode is related to sound wave propagation across the vertical cross section of the combustor can which is shorter than the length, hence the higher frequency (shorter travelling time). The 3rd mode is azimuthal, which is transverse direction propagation. This could be in the context of combustor can or annular combustor manifold/ ring itself as shown. In the example shown, the acoustic combustion instability is shown as a rotating (CW & CCW) pressure distribution on the manifold, the combustion coupling and feedback mechanism to the temperature profile via unsteady heat release with intermittent flashback. It is unlikely that the HPT exhaust distributed TCs could pickup the dynamics, given the instability frequency response vs typical thermocouple time constant. Also, the hot gas profile at HPT inlet stage and subsequent (till exhaust) will alter the distribution, unless the instability is sustained such that differential temperature or hot/cold spots are detected by the distributed TCs.

.

As there is no commercial sensors yet that could withstand the extreme condition of the combustor section, typically state estimation of combustor parameters ie flow, flame temperature etc are estimated by a combination of equations and tables which are extracted as part of experimental envelope under controlled condition, and accordingly adjusted. However, the modes can be monitored by acoustic sensor or microphone, which samples the tonal signal at particular spot in time domain and subsequently further data processing and FFT which transforms the signal to frequency domain for frequency analysis. Periodic rotating azimuthal mode could be observed from amplitude vs time trend.

Going back to the question of practicality in assisting with troubleshooting, for any given combustion instability event, one could specifically narrow down the investigation with fast trend (ms) of critical dynamics data, which include regulator mode, engine mode, AxCo discharge pressure, temperature, HPT speed and calculated air flow (if available), acoustics data (psig), exhaust data, fuel flow, fuel pressure, metering valve signals and as available, flame temperature(s), feed composition, energy and flow among other data. First-Outs typically marks the possible event causes, however some anomaly could be registered as alerts or alarms preceding the event.

Another good tool that can be used is drift or change overlaying method against known operating envelope. In this case typically one will have a set of correlations but not necessarily causation. Comparison of maps and adjustable & tunable of similar machines under similar operating condition might also highlight some critical differences or clues. Analogical comparison might work, especially if one is already equipped with a strong fundamental and experience in a particular machine and is trying to troubleshoot another machine from a different make/ model/ manufacturer. Surely not all know-how is transferrable and applicable. At times it is also worthwhile to dig deeper into manuals and application program and program libraries especially if there is any auto-intervention in active state and understand/ take into account the contribution into the analysis. Also remember to check the detail weightage of factors such as sensitivity analysis of flame temperature.

Last but not least, if nothing works go back to the first principle of combustion and build up the framework and reasoning from bottom up. Some might argue at this point why don’t one just call for an assistance from an expert but not everyone has the luxury of reasonable bosses with deep pocket to spare, or it could be the case that you’re the so called expert! Jokes aside, experts usually subscribe to certain technical fraternity or group of experts as well, so definitely ask & discuss. Afterall no one knows everything and learning is a lifelong process.