Signal processing techniques for machine condition monitoring

Steel, James (2016) Signal processing techniques for machine condition monitoring. [USQ Project]

[img]
Preview
Text (Main Project)
Steele_J_Leis.pdf

Download (11Mb) | Preview

Abstract

The purpose of this dissertation is to analyse and compare a wide range of wavelet de-noising parameters and determine which parameters are best suited to the de-noising of rolling element bearing vibration signals.

The condition of rolling element bearings is often monitored by recording the vibration of the bearing using accelerometers and analysing the signal for particular frequency content. When a bearing experiences a mechanical fault the vibration signal will contain frequencies relating to the failing component. Monitoring the bearing vibration can provide advanced warning that a bearing failure is imminent. However, when a mechanical fault is in the early stages of development the fault frequency can be very low in magnitude and difficult to detect. Improving the signal to noise ratio of these fault frequencies can provide earlier detection of the fault.

One method to improve the signal to noise ratio of a bearing fault frequency is to reduce the noise component in the vibration signal using wavelet theory. Wavelet de-noising has many parameters that can be varied which changes how the de-noising process modifies the vibration signal in the time domain. This dissertation makes comparison between the many de-noising parameters available and assesses which parameters provide the best increase in signal to noise ratio.

The wavelet de-noising process alone does not identify the frequencies relating to a bearing fault. The frequency content within the vibration time domain signal is required to be extracted and assessed to determine the effect of the wavelet de-noising process. Three frequency extraction methods were used to analyse the de-noised signals and indicate the magnitude of signal to noise ratio improvement achieved through de-noising. This dissertation shows that cepstrum analysis of the time domain signal responded best to the wavelet de-noising process and large improvements in signal to noise ratio were realised.


Statistics for USQ ePrint 31487
Statistics for this ePrint Item
Item Type: USQ Project
Item Status: Live Archive
Additional Information: Bachelor of Engineering (Honours) Major Electrical & Electronic Engineering project
Faculty / Department / School: Current - Faculty of Health, Engineering and Sciences - School of Mechanical and Electrical Engineering
Supervisors: Leis, John
Date Deposited: 24 Jul 2017 01:15
Last Modified: 24 Jul 2017 01:15
Uncontrolled Keywords: signal processing techniques; machine condition monitoring; wavelet de-noising parameters; vibration signals; accelerometers
Fields of Research : 09 Engineering > 0906 Electrical and Electronic Engineering > 090609 Signal Processing
URI: http://eprints.usq.edu.au/id/eprint/31487

Actions (login required)

View Item Archive Repository Staff Only