A new rotating machine diagnosis method based on Complex Signal Bilateral Spectrum and SVM

Aimed at solving the shortage of fault sample and enhancing the reliability of diagnosis results, this paper proposes a new diagnosis method based on the combination of Complex Signal Bilateral Spectrum (CSBS) and Support Vector Machine (SVM). SVM is a classical and effective method for fault classification of small sample, while it has difficulties in recognizing fault types, and its classification process is quite complicated. However, through analyzing the amplitude spectrums of CSBS can confirm the fault types more reliable and easy. Therefore, CSBS is merged into SVM for fault diagnosis. Firstly, CSBS utilize a couple of channels to synthesize the complex signals to acquire the parameters of positive and backward precession. The amplitude spectrums of each frequency are able to be composed through making full use of positive and backward precession parameters. Then, amplitude characteristic vectors can be employed as the input of SVM for fault status classification, and the amplitude spectrums of CSBS can confirm the types. There are two obvious advantages of this new method. On the one hand, CSBS can extract features more comprehensive that make the classification accuracy higher. On the other hand, through analyzing the amplitude spectrums of CSBS can get a more reliable diagnosis result. Finally, through a case study, this new method proved to be more reliable and efficient for fault diagnosis.