Machine fault detection by signal denoising—with application to industrial gas turbines

Zhang, Yu and Bingham, Chris and Yang, Zhijing and Ling, Wing-Kuen and Gallimore, Michael (2014) Machine fault detection by signal denoising—with application to industrial gas turbines. Measurement, 58 . pp. 230-240. ISSN 0263-2241

Documents
Machine fault detection by signal denoising—with application to industrial gas turbines

Request a copy
[img] PDF
1-s2.0-S026322411400339X-main.pdf - Whole Document
Restricted to Repository staff only

3MB
Item Type:Article
Item Status:Live Archive

Abstract

The paper proposes a new methodology of machine fault detection for industrial gas turbine (IGT) systems. The integrated use of empirical mode decomposition (EMD), principal component analysis (PCA) and Savitzky–Golay (S-G) adaptive filtering are applied to extract noise from underlying measurements. Through analysis of the resulting noise, along with the use of a developed power index, it is shown that transient measurements associated with system load or demand changes, for instance, can be effectively discriminated from those that are characteristic of emerging faults – the former being a primary contributor to generating ‘false alerts’ using more traditional techniques that are only robust when used with steady-state measurements. Comparative studies show the benefits of the hybrid technique compared to the more usual use of EMD and PCA alone. Three operational conditions are identifiable from the resulting noise analysis viz. normal behaviour, transient operation, and characteristics of emerging machine faults. The efficacy of the proposed approach is demonstrated using two experimental case studies (bearing wear and burner faults) on sub-15MW IGTs.

Keywords:Machine fault detection, empirical mode decomposition, principal component analysis, Savitzky–Golay adaptive filtering, spectral analysis, NotOAChecked
Subjects:G Mathematical and Computer Sciences > G700 Artificial Intelligence
G Mathematical and Computer Sciences > G760 Machine Learning
Divisions:College of Science > School of Engineering
ID Code:15048
Deposited On:26 Sep 2014 07:55

Repository Staff Only: item control page