Application of Wigner-Ville distribution and probability neural network for scooter engine fault diagnosis

Jian-Da Wu, Peng Hsin Chiang

Research output: Contribution to journalArticle

35 Citations (Scopus)

Abstract

An expert system for internal combustion engine fault diagnosis using Wigner-Ville distribution for feature extraction and probability neural network for fault classification is described in this paper. Most of the conventional techniques for fault signal analysis in a mechanical system are based chiefly on the difference of signal amplitude in the time and frequency domains. Unfortunately, in some conditions the performance is limited, such as when analysis signals are non-stationary. In the present study, the Wigner-Ville distribution is proposed for sound emission signal features classification, because it provides high resolution of instantaneous energy density both in time and frequency domains. Meanwhile, the instantaneous power spectrum is presented to obtain high-energy density when the engine fault condition occurs. These features of signals are classified using the probability neural network. To examine the efficiency of the probability neural network, both back-propagation and radial basis function neural networks are used in comparison with fault classification. The experimental results showed all three networks can achieve high recognition rate with feature extraction using Wigner-Ville distribution method. It also suggested the probability neural network can complete training in an extremely short time.

Original languageEnglish
Pages (from-to)2187-2199
Number of pages13
JournalExpert Systems with Applications
Volume36
Issue number2 PART 1
DOIs
Publication statusPublished - 2009 Jan 1

Fingerprint

Motor scooters
Wigner-Ville distribution
Failure analysis
Engines
Neural networks
Signal analysis
Feature extraction
Power spectrum
Internal combustion engines
Backpropagation
Expert systems
Acoustic waves

All Science Journal Classification (ASJC) codes

  • Engineering(all)
  • Computer Science Applications
  • Artificial Intelligence

Cite this

@article{3731f53494294e5fa64d3e3aec196574,
title = "Application of Wigner-Ville distribution and probability neural network for scooter engine fault diagnosis",
abstract = "An expert system for internal combustion engine fault diagnosis using Wigner-Ville distribution for feature extraction and probability neural network for fault classification is described in this paper. Most of the conventional techniques for fault signal analysis in a mechanical system are based chiefly on the difference of signal amplitude in the time and frequency domains. Unfortunately, in some conditions the performance is limited, such as when analysis signals are non-stationary. In the present study, the Wigner-Ville distribution is proposed for sound emission signal features classification, because it provides high resolution of instantaneous energy density both in time and frequency domains. Meanwhile, the instantaneous power spectrum is presented to obtain high-energy density when the engine fault condition occurs. These features of signals are classified using the probability neural network. To examine the efficiency of the probability neural network, both back-propagation and radial basis function neural networks are used in comparison with fault classification. The experimental results showed all three networks can achieve high recognition rate with feature extraction using Wigner-Ville distribution method. It also suggested the probability neural network can complete training in an extremely short time.",
author = "Jian-Da Wu and Chiang, {Peng Hsin}",
year = "2009",
month = "1",
day = "1",
doi = "10.1016/j.eswa.2007.12.012",
language = "English",
volume = "36",
pages = "2187--2199",
journal = "Expert Systems with Applications",
issn = "0957-4174",
publisher = "Elsevier Limited",
number = "2 PART 1",

}

Application of Wigner-Ville distribution and probability neural network for scooter engine fault diagnosis. / Wu, Jian-Da; Chiang, Peng Hsin.

In: Expert Systems with Applications, Vol. 36, No. 2 PART 1, 01.01.2009, p. 2187-2199.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Application of Wigner-Ville distribution and probability neural network for scooter engine fault diagnosis

AU - Wu, Jian-Da

AU - Chiang, Peng Hsin

PY - 2009/1/1

Y1 - 2009/1/1

N2 - An expert system for internal combustion engine fault diagnosis using Wigner-Ville distribution for feature extraction and probability neural network for fault classification is described in this paper. Most of the conventional techniques for fault signal analysis in a mechanical system are based chiefly on the difference of signal amplitude in the time and frequency domains. Unfortunately, in some conditions the performance is limited, such as when analysis signals are non-stationary. In the present study, the Wigner-Ville distribution is proposed for sound emission signal features classification, because it provides high resolution of instantaneous energy density both in time and frequency domains. Meanwhile, the instantaneous power spectrum is presented to obtain high-energy density when the engine fault condition occurs. These features of signals are classified using the probability neural network. To examine the efficiency of the probability neural network, both back-propagation and radial basis function neural networks are used in comparison with fault classification. The experimental results showed all three networks can achieve high recognition rate with feature extraction using Wigner-Ville distribution method. It also suggested the probability neural network can complete training in an extremely short time.

AB - An expert system for internal combustion engine fault diagnosis using Wigner-Ville distribution for feature extraction and probability neural network for fault classification is described in this paper. Most of the conventional techniques for fault signal analysis in a mechanical system are based chiefly on the difference of signal amplitude in the time and frequency domains. Unfortunately, in some conditions the performance is limited, such as when analysis signals are non-stationary. In the present study, the Wigner-Ville distribution is proposed for sound emission signal features classification, because it provides high resolution of instantaneous energy density both in time and frequency domains. Meanwhile, the instantaneous power spectrum is presented to obtain high-energy density when the engine fault condition occurs. These features of signals are classified using the probability neural network. To examine the efficiency of the probability neural network, both back-propagation and radial basis function neural networks are used in comparison with fault classification. The experimental results showed all three networks can achieve high recognition rate with feature extraction using Wigner-Ville distribution method. It also suggested the probability neural network can complete training in an extremely short time.

UR - http://www.scopus.com/inward/record.url?scp=56349099301&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=56349099301&partnerID=8YFLogxK

U2 - 10.1016/j.eswa.2007.12.012

DO - 10.1016/j.eswa.2007.12.012

M3 - Article

VL - 36

SP - 2187

EP - 2199

JO - Expert Systems with Applications

JF - Expert Systems with Applications

SN - 0957-4174

IS - 2 PART 1

ER -