0
Research Papers

A Selective Multiclass Support Vector Machine Ensemble Classifier for Engineering Surface Classification Using High Definition Metrology

[+] Author and Article Information
Shichang Du

State Key Lab of Mechanical
System and Vibration,
School of Mechanical Engineering,
Shanghai Jiao Tong University,
Shanghai 200240, China
e-mail: lovbin@sjtu.edu.cn

Changping Liu

School of Mechanical Engineering,
Shanghai Jiao Tong University,
Shanghai 200240, China
e-mail: lcpsky89@gmail.com

Lifeng Xi

State Key Lab of Mechanical
System and Vibration,
School of Mechanical Engineering,
Shanghai Jiao Tong University,
Shanghai 200240, China
e-mail: lfxi@sjtu.edu.cn

1Corresponding author.

Contributed by the Manufacturing Engineering Division of ASME for publication in the JOURNAL OF MANUFACTURING SCIENCE AND ENGINEERING. Manuscript received June 3, 2013; final manuscript received July 30, 2014; published online November 26, 2014. Assoc. Editor: Robert Gao.

J. Manuf. Sci. Eng 137(1), 011003 (Feb 01, 2015) (15 pages) Paper No: MANU-13-1247; doi: 10.1115/1.4028165 History: Received June 03, 2013; Revised July 30, 2014; Online November 26, 2014

The surface appearance is sensitive to change in the manufacturing process and is one of the most important product quality characteristics. The classification of workpiece surface patterns is critical for quality control, because it can provide feedback on the manufacturing process. In this study, a novel classification approach for engineering surfaces is proposed by combining dual-tree complex wavelet transform (DT-CWT) and selective ensemble classifiers called modified matching pursuit optimization with multiclass support vector machines ensemble (MPO-SVME), which adopts support vector machine (SVM) as basic classifiers. The dual-tree wavelet transform is used to decompose three-dimensional (3D) workpiece surfaces, and the features of workpiece surface are extracted from wavelet sub-bands of each level. Then MPO-SVME is developed to classify different workpiece surfaces based on the extracted features and the performance of the proposed approach is evaluated by computing its classification accuracy. The performance of MPO-SVME is validated in case study, and the results demonstrate that MPO-SVME can increase the classification accuracy with only a handful of selected classifiers.

Copyright © 2015 by ASME
Your Session has timed out. Please sign back in to continue.

References

Figures

Grahic Jump Location
Fig. 1

The architecture of the proposed feature extraction and classification approach

Grahic Jump Location
Fig. 2

Framework of the DT-CWT

Grahic Jump Location
Fig. 3

The two set of wavelet filters (a) 2D discrete wavelet filters and (b) 2D dual-tree complex wavelet filters

Grahic Jump Location
Fig. 4

A geometric interpretation of binary classification of SVM

Grahic Jump Location
Fig. 5

A component multiclass SVM classifier combined by binary SVMs

Grahic Jump Location
Fig. 6

The framework of the proposed selective ensemble classifiers

Grahic Jump Location
Fig. 7

The random subspace algorithm

Grahic Jump Location
Fig. 8

The MPO-SVME algorithm

Grahic Jump Location
Fig. 9

Engine cylinder blocks processed by a major domestic car manufacturer

Grahic Jump Location
Fig. 10

Two samples of the color-coded measurement results from data sets A and B

Grahic Jump Location
Fig. 11

Selection of small surface samples through a grid chart

Grahic Jump Location
Fig. 12

Six surfaces selected from set A and set B

Grahic Jump Location
Fig. 13

The generation of training samples and testing samples

Grahic Jump Location
Fig. 14

The CCPs using DT-CWT and DWT with different levels

Grahic Jump Location
Fig. 15

The CCPs of single multiclass SVMs using surface samples at 20 different locations of each surface

Grahic Jump Location
Fig. 16

Classification results of MPO-SVME at each time (30 times totally)

Grahic Jump Location
Fig. 17

The NSC at each time

Grahic Jump Location
Fig. 18

The process of selective classifiers by MPO-SVME (T = 30, d = 30)

Grahic Jump Location
Fig. 19

Comparison of CCPs with different numbers of random features (T = 20)

Grahic Jump Location
Fig. 20

Comparison of CCPs for different multiclass SVM classifier T (n = 40)

Grahic Jump Location
Fig. 21

CCPs for different sizes of training set

Grahic Jump Location
Fig. 22

Boxplots of the classification results using different strategies

Tables

Errata

Discussions

Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging and repositioning the boxes below.

Related Journal Articles
Related eBook Content
Topic Collections

Sorry! You do not have access to this content. For assistance or to subscribe, please contact us:

  • TELEPHONE: 1-800-843-2763 (Toll-free in the USA)
  • EMAIL: asmedigitalcollection@asme.org
Sign In