Trend analysis of rotor-to-stator impact-rub based on smooth support vector regression. (1st January 2014)

• Record Type: Journal Article
• Authors: Chao, Zhang
• Is Part Of: International journal of computer applications technology. Volume 50:Number 3/4(2014)
• Page Start: 243
• Page End: 246
• Publication Date: 2014-01-01
• Subjects: SVM -- support vector machine -- regression -- smooth method -- impact-rub -- simulation -- trend analysis.; Technology -- Data processing -- Periodicals; 620.00285
• DOI: 10.1504/IJCAT.2014.066735
• Languages: English
• ISSNs: 0952-8091

Abstract

Traditional methods of mathematical modelling usually get complicated non-linear models for rub-impact fault. Besides, time-frequency domain analysis is very difficult to solve the problem because of its own limitation. Based on analysis need of degree and trend of rub-impact fault, support vector regression (SVR) arithmetic is imported, which adopts method of time series analysis. Smooth support vector regression (SSVR) using smoothing method in SVR suggests a best tradeoff between complexity of model and learning ability to establish linear model which is able to reflect the implicit mechanism of series data. Model of Jeffcott rotor was established to simulate actual rotor rub-impact fault, and SSVR was applied to analyse trend of time series data of rub-impact fault. The results indicate that SSVR is obviously superior to neural network, and is effective to analyse trend of rotor-to-stator impact-rub fault, which has guiding significance for using and maintenance of rotating machine.