Chaos suppression via genetic programming control in a self-excited thermoacoustic system

We demonstrate how genetic programming (GP) control can be used to suppress low-dimensional chaotic oscillations in a self-excited thermoacoustic system comprising a ducted laminar premixed flame. We start by initializing a generation of candidate model-free control laws, which are evaluated based on a predefined cost function that balances the reduction in thermoacoustic amplitude with the actuator power consumed. To evolve the control laws from one generation to the next, we use a tournament algorithm involving genetic operations such as replication, elitism, crossover, and mutation. We compare the performance of both closed-loop and open-loop forms of GP control with that of classic open-loop control based on time-periodic excitation. Our results show that GP closed-loop control outperforms both GP and classic open-loop control, achieving the highest reduction in thermoacoustic amplitude with the lowest actuator power. We also analyze the optimal GP-controlled state using various tools from complex systems theory, which enables us to identify the key chaos suppression mechanisms and gain new insights into how the flame--acoustic coupling in a chaotic thermoacoustic system can be disrupted.