Support Vector Machines Applied to a Combustion Process

The following research aims to make the characterization of flames in the combustion process in an industrial boiler fossil fuel composed of one burner. The characterization of the flames is performed by analysis of electrical signals that are obtained through a flame detection sensor that measure the electromagnetic spectrum of the flame in the boiler as well as the acquisition of other variables involved in the combustion as are excess oxygen (%), flow, temperature and fuel density. After getting through a data acquisition system of the electromagnetic spectrum of the flame and the variables involved in combustion, it performs the signal processing of the spectrum by obtaining statistical moments and principal component analysis (PCA) to extract the most important characteristics. So get the patterns for training the support vector machine (SVM). After conducting the training of SVM was that the patterns obtained are suitable for proper classification of flames in a combustion process of a boiler, as the previously trained classifier has a high percentage of performance.