Active combustion control using an evolution algorithm
Conference Paper
Overview
Research
Identity
Additional Document Info
View All
Overview
abstract
Self-tuning active instability control was applied to an atmospheric combustor with a swirl-stabilized premixed flame. A microphone was used as sensor while direct driven valves where used for actuation. Previous experiments have shown that good suppression of pressure amplitudes in the combustion chamber could be obtained using a gain-delay controller. The gain and the delay of this controller where adjusted manually by trial and error. Motivated by these previous results the need arose for a controller that tunes the controller parameters automatically, depending on operating condition. The control method needed to fulfill three design requirements: 1) the strategy should have no a-priori knowledge of the system to be controlled. 2) Because of the stochastic processes involved in the thermo-acoustic system, the control-strategy should be able to deal with noisy signals, having poor repeatability. 3) The control strategy should be capable to adapt itself to changes in the system (i.e. operating condition, ambient changes). The three design criteria could be fulfilled by using a relatively simple controller structure implemented on a fast DSP-board, while the free parameters in this controller where adjusted using an algorithm implemented on a PC. The algorithm is designed to minimize the power of a pressure signal by adjusting the controller parameters. An evolution strategy proved to be successful in converging to an optimal parameter setting, despite considerable amount of randomness in the system. Test of this control strategy on an atmospheric combustion facility showed pressure peak amplitude reduction of -26 dB.
