Fault Feature Extraction Method of Time-Frequency Image Based on Fractal Dimension

Author

Hao Zhihua ; Tian LiXin

Author_Institution

Dept. of Inf. Eng., Tangshan Coll., Tangshan, China

Volume

1

fYear

2009

fDate

11-12 April 2009

Firstpage

658

Lastpage

660

Abstract

A fault feature extraction method of time-frequency image based on fractal dimension is put forward in this paper. Theoretical calculation formulas of fractal dimension are introduced and the calculation method about box-counting dimension based on image processing is analyzed. Based on this, the vibration signals measured from diesel engine in the stage of deflagration are researched with smooth slice Wigner higher order moment spectrums. Then the dimensions of fractal images which were obtained from the signal waves and the Wigner higher order moment spectrums are calculated by the fractal dimension calculation program. The result indicates that different wearing stages of piston-cylinder can be discriminated depending on feature parameters of fractal dimension because the fractal dimensions of images are apparently different.

Keywords

Wigner distribution; diesel engines; engine cylinders; fault diagnosis; feature extraction; fractals; image processing; pistons; time-frequency analysis; vibrations; wear; Wigner higher order moment spectrums; diesel engine; engine cylinder; fault feature extraction method; fractal dimension; fractal images; image processing; pistons; signal waves; time-frequency image; vibration signals; wear; Automation; DH-HEMTs; Educational institutions; Feature extraction; Fractals; Geometry; Image processing; Mechatronics; Time frequency analysis; Vibration measurement; fault Feature Extraction; fractal dimension; time-frequency image; vibration signal;

fLanguage

English

Publisher

ieee

Conference_Titel

Measuring Technology and Mechatronics Automation, 2009. ICMTMA '09. International Conference on

Conference_Location

Zhangjiajie, Hunan

Print_ISBN

978-0-7695-3583-8

Type

conf

DOI

10.1109/ICMTMA.2009.449

Filename

5203058