Chirplet OTA has much more resolution than WignerVille distribution

1
Diagnosis of Rotating Machinery using Order
Tracking Analysis with Chirplet and Particle
Filtering
S. M. Sohn, Prof. P. R. White
Introduction
Many diagnosis methods to be used when the
machine is running at steady state (constant
operation speed) have been developed. But the
dynamic characteristics of rotating machinery are
effected by speed the resonance frequency is
changed by speed. So, the signal characteristics
related with speed frequency is important to
understand the condition of machine. Mainly, we
use Bode plot for speed frequency using order
tracking analysis.
Frequency Analysis

• Order Tracking Analysis (OTA) Signal
  processing in the order domain relates the
  vibration and noise signals to the rotating speed
  of the shaft instead of an absolute frequency
  domain

Others Work
My Work

• Resampling
• Vold-Kalman filtering

• Chirplet OTA

• Fourier Transform
• Chirplet Transform

Much better resolution than Wigner-Ville
distribution
Frequency Analysis

• Sequential Importance Recursive Particle Filtering

• Automatically Search Rotating Speed Order
  Tracking
• System Model
• Measurement
• Model

• Importance

Order Analysis
Need for Precise Tacho Signal
No need for speed profile
Results
Resampling OTA
Speed
Frequency
Amplitude
Chirplet OTA
No. time bin
time
time
Speed profile comparison between real and
estimated
Resolution Comparison according to OTA method
Short time Fourier transform of real data
1X order Bode plot using SIR filter
Chirplet OTA
Amplitude
No. Component
Frequency
Resampling OTA
No. time bin
time
time
Resolution Comparison between Chirplet OTA and
W-V distribution
2X order Bode plot using SIR filter
No. component tracked using SIR filter
STFT of reconstructed data using SIR particle
filtering
Conclusions and Future Direction

• Chirplet OTA has much more resolution than
  Wigner-Ville distribution
• OTA using particle filtering does not need Tacho
  signal
• Dynamic model of rigid rotor with crack fault
  for non-stationary condition has been developed

• Simulation of rigid rotor according to various
  faults rubbing, oil whirl, looseness,
  misalignment
• Extract the pattern of fault signal
• Make a normalization of non-stationary signal
• Diagnosis using OTA model based