Data-driven modeling of a thermoacoustic combustor system

Thermoacoustic instability is prevalent in thermal power machinery and devices. Accurate modal identification is essential for understanding instability mechanisms and optimizing combustor design and control. To address the modeling and identification of unstable thermoacoustic dynamics, this study proposes a physics-informed data-driven framework with a corresponding training strategy and verifies its feasibility using numerical simulations of a Rijke tube, a classical simplified thermoacoustic combustor model. By embedding the governing equations of the thermoacoustic combustor system within the PINN framework, the proposed data-driven PINN can accurately identify and separate the first and second acoustic modes under conditions of strong thermoacoustic coupling. Furthermore, under a well-designed training strategy, the framework can simultaneously accomplish mode identification and parameter discovery even when certain system parameters are unknown. Multiple training trials indicate that the method exhibits satisfactory robustness to external noise, suggesting potential for engineering applications in thermoacoustic combustor modeling and monitoring.