Investigation on Reciprocating Engine Pattern Recognition by Combining SVM and Hilbert Spectrum Entropy

Most of artificial intelligence methods used in pattern recognition for reciprocating engine is not very suitable for practical application because fault samples are very few. Support vector machine (SVM) is a new general machine-learning tool based on statistical learning method. It has good performance even when fault samples are few. In this paper, reciprocating engine pattern recognition based on SVM is discussed. To improve SVM recognition and reduce rejection area, several binary SVMs combined together for multi-class recognition are investigated. As for better description of engine vibration signal feature, Hilbert spectrum entropy (HSE) has been used as a tool for feature extraction. The effectiveness of the method is testified by the application to the pattern recognition for a reciprocating engine. It can be concluded that this method can contribute to reciprocating engine preventative maintenance development according to the result