Neural methods of interpretation of data obtained from optical sensor for flame monitoring

摘要

Burner systems and their control are getting more and more sophisticated and there is a growing need to obtain information about the course of combustion process in individual flames. Optical sensors offer the benefit of being selective, rapid and able to gather data from extremely hostile environments (e.g. the combustion zone of pulverised coal burners or gas turbines). Passive optical sensors offer the further advantage of simplicity, which make them attractive candidates. With the rapidly growing capability of these technologies for sensor hardware, there is an increased interest and need to develop data interpretation strategies that will allow optical flame emission data to be converted into meaningful combustor state information. The article describes various approaches to apply artificial neural network approaches to estimate parameters of combustion. One is acquiring information about emission of nitrogen oxides and carbon monoxide from fiberoptic systems for flame monitoring, developed in Department of Electronics of Lublin University of Technology and another is identification of flames in gas burners.