Neural ODE to model and prognose thermoacoustic instability.

Area of Science:

• Combustion science
• Fluid dynamics
• Nonlinear dynamics

Background:

• Thermoacoustic instability involves high-amplitude pressure fluctuations driven by heat release and acoustic field coupling.
• Turbulent flows exhibit transitions from chaos to periodic oscillations via intermittency, with heat release synchronizing to pressure fluctuations.
• Traditional models independently estimate heat source and acoustic subsystems, potentially missing nonlinear coupling effects.

Purpose of the Study:

• To develop a unified modeling framework for thermoacoustic systems.
• To capture the nonlinear coupling between unsteady heat release and acoustic pressure fluctuations.
• To establish an anomaly measure for early detection of thermoacoustic instability.

Main Methods:

• Introduction of a neural ordinary differential equation (neural ODE) framework.
• Simultaneous time-series measurements of heat release rate and pressure fluctuations.
• Utilizing neural ODE parameters to define an anomaly measure for system dynamics.

Main Results:

• The neural ODE framework models the thermoacoustic system as a whole, capturing coupled interactions.
• An anomaly measure derived from neural ODE parameters indicates proximity to limit cycle oscillations.
• The framework provides an early warning signal for the onset of thermoacoustic instability.

Conclusions:

• A neural ODE approach offers a more accurate representation of coupled thermoacoustic dynamics.
• The derived anomaly measure effectively predicts the onset of thermoacoustic instability.
• This method advances the understanding and prediction of combustion instabilities.