AbstractThis thesis discusses the development of an Artificial Neural Network (ANN) based Flame Monitoring and Control System (FMCS) to optimise the combustion of a pulverised coal flame.
A series of experiments were conducted on a 150 kW pulverised fuel burner rig based at Casella CRE Ltd. in the United Kingdom. These experiments systematically varied the burner swirl number and the secondary airflow rate over a significant range for two different coals so that both "satisfactory" and "poor" combustion conditions were obtained. The infrared emissions from the flame, the combustion noise and the acoustic emission generated in the burner body were measured with appropriate sensors, as were the fuel and airflow rates and pollutant emissions. The signals from the sensors were analysed using signal processing techniques to yield a number of features. These in turn were employed to train a neural network to predict the gaseous emissions, such as NOx and CO, from the rig.
In a separate set of experiments, the combustion process was placed in a poor condition, and the sensors together with the neural models were incorporated into an intelligent control system, which was able to alter the excess air level to improve the combustion process. In this fashion simultaneous lower NOx and CO levels were achieved with both coal types. The technique uses relatively low cost sensors and artificial intelligence techniques to control the combustion of the pulverised fuel burner. It is envisaged that this technique will be particularly attractive for multiple burner installations that are often fed from a common air supply manifold, so that the individual burner performance is often not known and cannot be optimised.
|Date of Award||Apr 2005|
|Supervisor||Steven Wilcox (Supervisor), John Ward (Supervisor) & CK Tan (Supervisor)|