Active control of combustion instabilities

A novel active control strategy for unstable combustion systems is presented. It consists of three main parts: a mode observer, an output predictor and a feedback controller. The mode observer is established using neural networks to reconstruct the measured output. To overcome the time delay of the system, which is often very large relative to the necessary sampling period, an output predictor is developed based on the mode observer. Unlike a classical Luenberger type state observer only a measure of the output is required. An output-feedback controller is introduced which uses the output of the predictor. The approach described is demonstrated through its application to the control of an experimental combustor having two dominant modes