Defect detection of helical gears based on time-frequency analysis and using multi-layer fusion network

被引:7
作者
Orimi, H. Ebrahimi [1 ]
Esmaeili, M. [2 ]
Oskouei, A. Refahi [3 ]
Mirhadizadehd, S. A. [4 ]
Tse, P. W. [5 ]
机构
[1] Concordia Univ, Dept Mech Engn, Montreal, PQ, Canada
[2] Islamic Azad Univ East Tehran, Mech Engn Grp, Tehran, Iran
[3] Shahid Rajaee Teacher Training Univ, Dept Mech Engn, Tehran, Iran
[4] Univ Northampton, Dept Sci & Technol, Northampton, England
[5] City Univ Hong Kong, Dept Syst Engn & Engn Management SEEM, Kowloon, Hong Kong, Peoples R China
关键词
Wavelet packet transforms (WPT); fusion network; MLP; helical gear; condition monitoring; classification; ARTIFICIAL NEURAL-NETWORK; ROLLING ELEMENT BEARINGS; DISCRETE WAVELET TRANSFORM; FAULT-DIAGNOSIS; MOTHER WAVELET; FEATURE-EXTRACTION; VIBRATION SIGNALS; MORLET WAVELET; CLASSIFICATION; PACKET;
D O I
10.1080/10589759.2016.1254211
中图分类号
TB3 [工程材料学];
学科分类号
0805 ; 080502 ;
摘要
Condition monitoring of rotary devices such as helical gears is an issue of great significance in industrial projects. This paper introduces a feature extraction method for gear fault diagnosis using wavelet packet due to its higher frequency resolution. During this investigation, the mother wavelet Daubechies 10 (Db-10) was applied to calculate the coefficient entropy of each frequency band of 5th level (32 frequency bands) as features. In this study, the peak value of the signal entropies was selected as applicable features in order to improve frequency band differentiation and reduce feature vectors' dimension. Feature extraction is followed by the fusion network where four different structured multi-layer perceptron networks are trained to classify the recorded signals (healthy/faulty). The robustness of fusion network outputs is greater compared to perceptron networks. The results provided by the fusion network indicate a classification of 98.88 and 97.95% for healthy and faulty classes, respectively.
引用
收藏
页码:363 / 380
页数:18
相关论文
共 44 条
  • [1] Properties determining choice of mother wavelet
    Ahuja, N
    Lertrattanapanich, S
    Bose, NK
    [J]. IEE PROCEEDINGS-VISION IMAGE AND SIGNAL PROCESSING, 2005, 152 (05): : 659 - 664
  • [2] Validation of a new method for the diagnosis of rotor bar failures via wavelet transform in industrial induction machines
    Antonino-Daviu, Jose A.
    Riera-Guasp, Martin
    Folch, Jose Roger
    Palomares, M. Pilar Molina
    [J]. IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, 2006, 42 (04) : 990 - 996
  • [3] Intelligent target recognition based on wavelet packet neural network
    Avci, E
    Turkoglu, I
    Poyraz, M
    [J]. EXPERT SYSTEMS WITH APPLICATIONS, 2005, 29 (01) : 175 - 182
  • [4] Speech recognition using a wavelet packet adaptive network based fuzzy inference system
    Avci, Engin
    Akpolat, Zuhtu Hakan
    [J]. EXPERT SYSTEMS WITH APPLICATIONS, 2006, 31 (03) : 495 - 503
  • [5] A new optimum feature extraction and classification method for speaker recognition: GWPNN
    Avci, Engin
    [J]. EXPERT SYSTEMS WITH APPLICATIONS, 2007, 32 (02) : 485 - 498
  • [6] A comparative study of acoustic and vibration signals in detection of gear failures using Wigner-Ville distribution
    Baydar, N
    Ball, A
    [J]. MECHANICAL SYSTEMS AND SIGNAL PROCESSING, 2001, 15 (06) : 1091 - 1107
  • [7] Amplitude and phase wavelet maps for the detection of cracks in geared systems
    Boulahbal, D
    Golnaraghi, MF
    Ismail, F
    [J]. MECHANICAL SYSTEMS AND SIGNAL PROCESSING, 1999, 13 (03) : 423 - 436
  • [8] Compression of biomedical signals with mother wavelet optimization and best-basis wavelet packet selection
    Brechet, Laurent
    Lucas, Marie-Francoise
    Doncarli, Christian
    Farina, Dario
    [J]. IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, 2007, 54 (12) : 2186 - 2192
  • [9] Coifman RR, 1991, WAVELETS THEIR APPL, P153
  • [10] Coifman RR, 1991, WAVELETS THEIR APPL, P453