Genetic programming control of thermoacoustic oscillations in a hydrogen-methane-fueled turbulent combustor

We experimentally investigate the application of genetic programming (GP) to suppress self-excited thermoacoustic oscillations in a turbulent premixed combustor operating on hydrogen-methane blends. The GP framework uses an evolutionary data-driven algorithm to breed successive generations of control laws via genetic operations. Each control law is evaluated using a cost function that considers both the thermoacoustic amplitude (state cost) and the actuation power (input cost). We implement GP in both closed-loop and open-loop forms, and benchmark it against conventional open-loop control. Our results show that GP closed-loop control can achieve significant reductions in thermoacoustic amplitude with minimal input cost, outperforming other control strategies with the lowest cost function value. The suppression mechanism is identified as synchronous quenching, which occurs without resonant amplification. The GP algorithm is found to be effective across various operating conditions, including different bulk reactant velocities, hydrogen power fractions, and combustor lengths. This versatility highlights the potential of GP as a model-free control strategy for mitigating thermoacoustic oscillations in combustion systems, including those operating on hydrogen-enriched fuels.