Fatigue feature extraction analysis based on a K-means clustering approach

Research output: Contribution to journalArticle

10 Citations (Scopus)

Abstract

This paper focuses on clustering analysis using a K-means approach for fatigue feature dataset extraction. The aim of this study is to group the dataset as closely as possible (homogeneity) for the scattered dataset. Kurtosis, the wavelet-based energy coefficient and fatigue damage are calculated for all segments after the extraction process using wavelet transform. Kurtosis, the wavelet-based energy coefficient and fatigue damage are used as input data for the K-means clustering approach. K-means clustering calculates the average distance of each group from the centroid and gives the objective function values. Based on the results, maximum values of the objective function can be seen in the two centroid clusters, with a value of 11.58. The minimum objective function value is found at 8.06 for five centroid clusters. It can be seen that the objective function with the lowest value for the number of clusters is equal to five; which is therefore the best cluster for the dataset.

Original languageEnglish
Pages (from-to)1275-1282
Number of pages8
JournalJournal of Mechanical Engineering and Sciences
Volume8
DOIs
Publication statusPublished - 1 Jun 2015

Fingerprint

Feature extraction
Fatigue of materials
Fatigue damage
Wavelet transforms

Keywords

  • Clustering
  • Fatigue feature extraction
  • K-means
  • Objective function
  • Segments

ASJC Scopus subject areas

  • Energy Engineering and Power Technology
  • Fuel Technology
  • Computational Mechanics
  • Industrial and Manufacturing Engineering
  • Mechanical Engineering
  • Mechanics of Materials

Cite this

@article{3832a712a9d54fc2b11e9feeedf051b9,
title = "Fatigue feature extraction analysis based on a K-means clustering approach",
abstract = "This paper focuses on clustering analysis using a K-means approach for fatigue feature dataset extraction. The aim of this study is to group the dataset as closely as possible (homogeneity) for the scattered dataset. Kurtosis, the wavelet-based energy coefficient and fatigue damage are calculated for all segments after the extraction process using wavelet transform. Kurtosis, the wavelet-based energy coefficient and fatigue damage are used as input data for the K-means clustering approach. K-means clustering calculates the average distance of each group from the centroid and gives the objective function values. Based on the results, maximum values of the objective function can be seen in the two centroid clusters, with a value of 11.58. The minimum objective function value is found at 8.06 for five centroid clusters. It can be seen that the objective function with the lowest value for the number of clusters is equal to five; which is therefore the best cluster for the dataset.",
keywords = "Clustering, Fatigue feature extraction, K-means, Objective function, Segments",
author = "Yunoh, {M. F M} and Shahrum Abdullah and {Md Saad}, {Mohamad Hanif} and {Mohd Nopiah}, Zulkifli and Nuawi, {Mohd. Zaki}",
year = "2015",
month = "6",
day = "1",
doi = "10.15282/jmes.8.2015.2.0124",
language = "English",
volume = "8",
pages = "1275--1282",
journal = "Journal of Mechanical Engineering and Sciences",
issn = "2289-4659",
publisher = "Faculty of Mechanical Engineering, Universiti Malaysia Pahang",

}

TY - JOUR

T1 - Fatigue feature extraction analysis based on a K-means clustering approach

AU - Yunoh, M. F M

AU - Abdullah, Shahrum

AU - Md Saad, Mohamad Hanif

AU - Mohd Nopiah, Zulkifli

AU - Nuawi, Mohd. Zaki

PY - 2015/6/1

Y1 - 2015/6/1

N2 - This paper focuses on clustering analysis using a K-means approach for fatigue feature dataset extraction. The aim of this study is to group the dataset as closely as possible (homogeneity) for the scattered dataset. Kurtosis, the wavelet-based energy coefficient and fatigue damage are calculated for all segments after the extraction process using wavelet transform. Kurtosis, the wavelet-based energy coefficient and fatigue damage are used as input data for the K-means clustering approach. K-means clustering calculates the average distance of each group from the centroid and gives the objective function values. Based on the results, maximum values of the objective function can be seen in the two centroid clusters, with a value of 11.58. The minimum objective function value is found at 8.06 for five centroid clusters. It can be seen that the objective function with the lowest value for the number of clusters is equal to five; which is therefore the best cluster for the dataset.

AB - This paper focuses on clustering analysis using a K-means approach for fatigue feature dataset extraction. The aim of this study is to group the dataset as closely as possible (homogeneity) for the scattered dataset. Kurtosis, the wavelet-based energy coefficient and fatigue damage are calculated for all segments after the extraction process using wavelet transform. Kurtosis, the wavelet-based energy coefficient and fatigue damage are used as input data for the K-means clustering approach. K-means clustering calculates the average distance of each group from the centroid and gives the objective function values. Based on the results, maximum values of the objective function can be seen in the two centroid clusters, with a value of 11.58. The minimum objective function value is found at 8.06 for five centroid clusters. It can be seen that the objective function with the lowest value for the number of clusters is equal to five; which is therefore the best cluster for the dataset.

KW - Clustering

KW - Fatigue feature extraction

KW - K-means

KW - Objective function

KW - Segments

UR - http://www.scopus.com/inward/record.url?scp=84938562054&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84938562054&partnerID=8YFLogxK

U2 - 10.15282/jmes.8.2015.2.0124

DO - 10.15282/jmes.8.2015.2.0124

M3 - Article

AN - SCOPUS:84938562054

VL - 8

SP - 1275

EP - 1282

JO - Journal of Mechanical Engineering and Sciences

JF - Journal of Mechanical Engineering and Sciences

SN - 2289-4659

ER -