Multi-Frequency Signal Detection Based on Frequency Exchange and Re-Scaling Stochastic Resonance and Its Application to Weak Fault Diagnosis.

Jinjun Liu,Yonggang Leng,Zhihui Lai,Shengbo Fan

doi:10.3390/s18051325

Jinjun Liu, Yonggang Leng + Show 2 more

Open Access

https://doi.org/10.3390/s18051325

Copy DOI

Abstract

Mechanical fault diagnosis usually requires not only identification of the fault characteristic frequency, but also detection of its second and/or higher harmonics. However, it is difficult to detect a multi-frequency fault signal through the existing Stochastic Resonance (SR) methods, because the characteristic frequency of the fault signal as well as its second and higher harmonics frequencies tend to be large parameters. To solve the problem, this paper proposes a multi-frequency signal detection method based on Frequency Exchange and Re-scaling Stochastic Resonance (FERSR). In the method, frequency exchange is implemented using filtering technique and Single SideBand (SSB) modulation. This new method can overcome the limitation of "sampling ratio" which is the ratio of the sampling frequency to the frequency of target signal. It also ensures that the multi-frequency target signals can be processed to meet the small-parameter conditions. Simulation results demonstrate that the method shows good performance for detecting a multi-frequency signal with low sampling ratio. Two practical cases are employed to further validate the effectiveness and applicability of this method.

Highlights

Various types of rotating machinery, such as electric generators, water turbines, centrifugal compressors, pumps, etc., have played an important role in industry
Stochastic Resonance (SR) theory can provide a promising tool for weak signal detection and incipient fault diagnosis in engineering applications
Frequency exchange is carried out in time domain by a filter technique, which can prevent spectrum leakage caused by Fourier Transformation (FFT) in the frequency domain

Summary

Introduction

Various types of rotating machinery, such as electric generators, water turbines, centrifugal compressors, pumps, etc., have played an important role in industry. The detection of incipient faults has become the focus of a number of studies of fault diagnosis. Some methods, such as wavelet transform [4,5], empirical mode decomposition [6], and nonlinear dynamic theory [7,8], have been studied in depth and widely for weak fault feature extraction in many practical situations. While the weak fault signal submerged in heavy background noise is amplified by a linear amplifier, the background noise is amplified in the same proportion. The weak fault characteristic signal is still submerged in the background noise and cannot be identified

Methods

Discussion

Conclusion