On-line identification of new fuel type using joint probability density arbiter

This paper presents a novel approach for on-line identification of new fuel type, which combines the principal component analysis technique and joint probability density arbiter. Because combustion flames of different coals have different oscillating features at the root area of the flame, a flame detector is utilized to capture the flame oscillation signals. Then the flame features are extracted in time and frequency domains from each flame oscillation signal, which form original feature vectors. The principal component analysis technique is utilized to transform each original feature vector into an orthogonal and dimension-reduced feature vector. Aiming at several known fuel types, a joint probability model is established for each fuel type using the data of the orthogonal feature vector. Then the joint probability density arbiters based on the models are used to determine whether the type of the fuel is new and identify the type of the fuel being burnt if it is one of the known fuel types.