Research on application of PCA and SVM to flame monitoring

Abstract

Seven characteristic values ,such as flame luminance, flame area, centroid offset and etc. are extracted in analysing the flame image. And then based on principal component analysis (PCA), a method for monitoring and diagnosing stability of flame is put forward. Two statistics of Hotelling T2 and Q are used to monitor time-to-time image data vectors, and check them whether they exceed their own controllable limit. As long as any one of them exceeds the limit, abnormity of combustion should be concluded. An experimental research shows that the method helps in on-line and real-time recognizing and judging the combustion status of the burning flame, and gives a visual result with figures of Q, of Hotelling T2 and PCA; at the same time, the characteristic vector and the original image data identified and sorted by using a method of support vector machine (SVM), the results show that two method, one is based on PCA and another is by support vector machine, are quite accordant.