Transfer learning of deep neural networks for predicting thermoacoustic instabilities in combustion systems

The intermittent nature of operation and unpredictable availability of renewable sources of energy(e.g.,wind and solar)would require the combustors in fossil-fuel power plants,sharing the same grid,to operate with large turn-down ratios.This brings in new challenges of suppressing high-amplitude pressure oscillations(e.g.,ther-moacoustic instabilities(TAI))in combustors.These pressure oscillations are usually self-sustained,as they occur within a feedback loop,and may induce severe thermomechanical stresses in structural components of combus-tors,which often lead to performance degradation and even system failures.Thus,prediction of thermoacoustic instabilities is a critical issue for both design and operation of combustion systems.From this perspective,it is important to identify operating conditions which can potentially lead to thermoacoustic instabilities.In this regard,data-driven approaches have shown considerable success in predicting the instability map as a function of operating conditions.However,often the available data are limited to learn such a relationship efficiently in a data-driven approach for a practical combustion system.In this work,a proof-of-concept demonstration of transfer learning is provided,whereby a deep neural network trained on relatively inexpensive experiments in an electrically heated Rijke tube has been adapted to predict the unstable operating conditions for a swirl-stabilized lean-premixed laboratory scaled combustor,for which data are expensive to obtain.The operating spaces and underlying flow physics of these two combustion systems are different,and hence this work presents a strong case of using transfer learning as a potential data-driven solution for transferring knowledge across domains.The results show that the knowledge transfer from the electrically heated Rijke tube apparatus helps in formulating an accurate data-driven surrogate model for predicting the unstable operating conditions in the swirl-stabilized combustor,even though the available data are significantly less for the latter.

Influence of the Acoustic Liner in Large Eddy Simulation of Longitudinal Thermoacoustic Instability in a Model Annular Combustor

Although extensive efforts have been made to dampen the thermoacoustic instability,successfully controlling the pressure oscillations in modern gas turbines or aeroengines remains challenging.The influence of the acoustic liner on the longitudinal thermoacoustic mode in a model annular combustor is investigated by Large Eddy Simulation(LES) in this work.The result of the self-excited longitudinal thermoacoustic instability without the liner agrees well with the frequency and acoustic analysis of the pressure mode based on experimental data.Three different bias flow velocities of the liner located downstream of the combustor are then simulated.The results reveal that the existence of the liner influences not only the acoustic field but also the flow field.When the bias velocity is large,it leads to intense turbulence-induced fluctuations,and the pressure oscillation is modulated intermittently.It shows that the weak coupling between flow and pressure oscillations plays a significant role in the onset of the intermittency of a thermoacoustic system.Based on the dynamic analysis of the thermoacoustic system with the acoustic liner,this intermittency is caused by the influence of the flow field on the flame-acoustic coupling.Finally,a low-order modeling method based on Van der Pol(VdP) oscillator with additive stochastic forcing is conducted to reproduce the evolving dynamics of the thermoacoustic system.Although the numerical cases demonstrated in this work are relatively simpler than those in a practical combustion system,the results are helpful for us to understand the effect of the acoustic liner and show the attractive potential to apply this device to suppress thermoacoustic instability.

Effect of CO_(2) Opposing Multiple Jets on Thermoacoustic Instability and NO_(x) Emissions in a Lean-Premixed Model Combustor

This paper experimentally studied the effect of CO_(2) opposing multiple jets on the thermoacoustic instability and NO_(x) emissions in a lean-premixed model combustor.The feasibility was verified from three variables:the CO_(2) jet flow rate,hole numbers,and hole diameters of the nozzles.Results indicate that the control effect of thermoacoustic instability and NO_x emissions show a reverse trend with the increase of open area ratio on the whole,and the optimal jet flow rate range is 1-4 L/min with CO_(2) opposing multiple jets.In this flow rate range,the amplitude and frequency of the dynamic pressure and heat release signals CH~* basically decrease as the CO_(2) flow rate increases,which avoids high-frequency and high-amplitude thermoacoustic instability.The amplitude-damped ratio of dynamic pressure and CH*can reach as high as 98.75% and 93.64% with an optimal open area ratio of 3.72%.NO_(x) emissions also decrease as the jet flow rate increases,and the maximum suppression ratio can reach 68.14%.Besides,the flame shape changes from a steep inverted "V" to a more flat "M",and the flame length will become shorter with CO_(2) opposing multiple jets.This research achieved the synchronous control of thermoacoustic instability and NO_(x) emissions,which could be a design reference for constructing a safer and cleaner combustor.

Effects of swirler position on flame response and combustion instabilities

This study is concerned with the experimental and theoretical investigation of the combustion instabilities in a premixed swirl combustor.It is focused on the effects of the swirl mixing distance on the intrinsic thermoacoustic mode.The swirler as an origin of the swirling flow is also the source of the flow disturbance,which has effects on the flame response.The location of the swirler is varied in the experiment to study the effect on combustion instabilities and flame transfer functions.A low order model is built to analyze the thermoacoustic instabilities of the combustion system.The experimental results show that the ITA switches from an unstable state to a stable state as the swirl mixing distance changes with an increment of 15 mm;while the instability of the quarter-wave mode is not varied.The measured Flame Transfer Functions(FTFs)show that the gain curves of the frequency-dependent FTFs seem to be stretched or compressed with the modulation of the swirler position,which has effects on frequencies and instabilities of thermoacoustic modes.With the low order model,the effects of flame response on combustion instabilities are analyzed and the flame dominant nature of the ITA mode is confirmed.

Adaptive Neural Network Control of Thermoacoustic Instability in Rijke Tube: A Fully Actuated System Approach

Thermoacoustic instability phenomena often encounter in gas turbine combustors,especially for the premixed combustor design,with many possible detrimental results.As a classical experiment,the Rijke tube is the simplest and the most effective illustration to study the thermoacoustic instability.This paper investigates the active control approach of the thermoacoustic instability in a horizontal Rijke tube.What’s more,the radial basis function(RBF)neural network is adopted to estimate the complex unknown continuous nonlinear heat release rate in the Rijke tube.Then,based on the proposed second-order fully actuated system model,the authors present an adaptive neural network controller to guarantee the flow velocity fluctuation and pressure fluctuation to converge to a small region of the origin.Finally,simulation results demonstrate the feasibility of the design method.