Vibration Source Model Estimation and State Specificity Perception of a Rotor Structure
(0) Rotor structures in abnormal states will make its own vibration source non-stationary,these non-stationary features are often ignored by traditional signal processing algorithms,thereby the determination of rotor running states is affected.A new state specificity perception method based on kernel density function estimation,higher moments feature extraction and multi-class support vector machine is proposed.Firstly,for rotors with non-stationary vibration source,kernel density estimation algorithm is used to estimate the probability density model; this model fully considers band broadening features of non-stationary signals.Then higher order moments of non-stationary vibration source model are calculated,the values are assembled to state mark vectors.Finally,the multi-class support vector machine is used to classify the state mark vectors.Three types of abnormal states are set on cascade rotor system for experiments,acceleration data is collected and the state specificity perception process is carried out.All abnormal states are identified successfully.High accuracy proves the effectiveness of kernel density estimation-support vector machine (KDE-SVM) method.The proposed specificity extraction and perception method provides new ideas for running structure condition monitoring.
kernel density estimation SVM state specificity perception
Sai Ma ShunMing Li HaiLong Liu XiaoDong Miao Yong Wang TianWen Liu
College of Energy and Power Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 2 Beijing Institute of Aerospace Systems Engineering,Beijing 100076,China
国际会议
南京
英文
1-6
2013-08-20（万方平台首次上网日期，不代表论文的发表时间）