关键词: 联合概率密度, 燃煤种类, 神经网络, 特征值, 主成分分析

Abstract: A new approach for on-line identification of coal type is presented by combining the joint probability density arbiter and neural network techniques. In view of the fact that the combustion flames of different coals are of different oscillating features, three flame detectors are utilized to capture the flame oscillation signals from the IR, visible and UV spectral bands respectively. The flame features are extracted both in the time and frequency domains from each flame oscillation signal. The principal component analysis technique is utilized to transform each feature vector into an orthogonal and dimension-reduced feature vector. A joint probability density arbiter for each known fuel type and a neural network model of all the known fuel types are established by using the data of the orthogonal feature vectors. Then the joint probability density arbiters are used to determine whether the type of the fuel is new and the neural network model is used to identify the type of the fuel being burnt if it is not new. The data obtained from burning four different types of coal in a utility boiler demonstrate that combination of the joint probability density arbiter and neural network techniques is an effective approach for identifying known and unknown fuel types, and the average success rate is 96.7% in twenty trials.

Key words: Features, Fuel type, Joint probability density, Neural network, Principal component analysis