Grinding wheel condition monitoring with boosted minimum distance classifiers

T. Warren Liao; Fengming Tang; J. Qu; P. J. Blau

doi:10.1016/j.ymssp.2007.06.005

Grinding wheel condition monitoring with boosted minimum distance classifiers

T. Warren Liao, Fengming Tang, J. Qu, P. J. Blau

Surface Engineering and Tribology

Research output: Contribution to journal › Article › peer-review

58 Scopus citations

Abstract

Grinding wheels get dull as more material is removed. This paper presents a methodology to detect a 'dull' wheel online based on acoustic emission (AE) signals. The methodology has three major steps: preprocessing, signal analysis and feature extraction, and constructing boosted classifiers using the minimum distance classifier (MDC) as the weak learner. Two booting algorithms, i.e., AdaBoost and A-Boost, were implemented. The methodology was tested with signals obtained in grinding of two ceramic materials with a diamond wheel under different grinding conditions. The results of cross-validation tests indicate that: (i) boosting greatly improves the effectiveness of the basic MDC; (ii) over all A-Boost does not outperform AdaBoost in terms of classification accuracy; and (iii) the performance of the boosted classifiers improves as the ensemble size increases.

Original language	English
Pages (from-to)	217-232
Number of pages	16
Journal	Mechanical Systems and Signal Processing
Volume	22
Issue number	1
DOIs	https://doi.org/10.1016/j.ymssp.2007.06.005
State	Published - Jan 2008

Funding

The authors acknowledge the support provided by two Programs: (i) the High Temperature Materials Laboratory User Program, Oak Ridge National Laboratory sponsored by the Assistant Secretary for Energy Efficiency and Renewable Energy, Office of Transportation Technologies, and (ii) the Louisiana Board's NSF ESPCoR Links Program with contract number NSF(2005)-LINK-06.

Keywords

AR model
Acoustic emission
Boosting
Condition monitoring
Grinding wheel
Minimum distance classifier
Wear

Access to Document

10.1016/j.ymssp.2007.06.005

Cite this

@article{8d55b9ae879249e0b1e10ff230a96c7f,

title = "Grinding wheel condition monitoring with boosted minimum distance classifiers",

abstract = "Grinding wheels get dull as more material is removed. This paper presents a methodology to detect a 'dull' wheel online based on acoustic emission (AE) signals. The methodology has three major steps: preprocessing, signal analysis and feature extraction, and constructing boosted classifiers using the minimum distance classifier (MDC) as the weak learner. Two booting algorithms, i.e., AdaBoost and A-Boost, were implemented. The methodology was tested with signals obtained in grinding of two ceramic materials with a diamond wheel under different grinding conditions. The results of cross-validation tests indicate that: (i) boosting greatly improves the effectiveness of the basic MDC; (ii) over all A-Boost does not outperform AdaBoost in terms of classification accuracy; and (iii) the performance of the boosted classifiers improves as the ensemble size increases.",

keywords = "AR model, Acoustic emission, Boosting, Condition monitoring, Grinding wheel, Minimum distance classifier, Wear",

author = "Liao, {T. Warren} and Fengming Tang and J. Qu and Blau, {P. J.}",

year = "2008",

month = jan,

doi = "10.1016/j.ymssp.2007.06.005",

language = "English",

volume = "22",

pages = "217--232",

journal = "Mechanical Systems and Signal Processing",

issn = "0888-3270",

publisher = "Elsevier",

number = "1",

}

TY - JOUR

T1 - Grinding wheel condition monitoring with boosted minimum distance classifiers

AU - Liao, T. Warren

AU - Tang, Fengming

AU - Qu, J.

AU - Blau, P. J.

PY - 2008/1

Y1 - 2008/1

N2 - Grinding wheels get dull as more material is removed. This paper presents a methodology to detect a 'dull' wheel online based on acoustic emission (AE) signals. The methodology has three major steps: preprocessing, signal analysis and feature extraction, and constructing boosted classifiers using the minimum distance classifier (MDC) as the weak learner. Two booting algorithms, i.e., AdaBoost and A-Boost, were implemented. The methodology was tested with signals obtained in grinding of two ceramic materials with a diamond wheel under different grinding conditions. The results of cross-validation tests indicate that: (i) boosting greatly improves the effectiveness of the basic MDC; (ii) over all A-Boost does not outperform AdaBoost in terms of classification accuracy; and (iii) the performance of the boosted classifiers improves as the ensemble size increases.

AB - Grinding wheels get dull as more material is removed. This paper presents a methodology to detect a 'dull' wheel online based on acoustic emission (AE) signals. The methodology has three major steps: preprocessing, signal analysis and feature extraction, and constructing boosted classifiers using the minimum distance classifier (MDC) as the weak learner. Two booting algorithms, i.e., AdaBoost and A-Boost, were implemented. The methodology was tested with signals obtained in grinding of two ceramic materials with a diamond wheel under different grinding conditions. The results of cross-validation tests indicate that: (i) boosting greatly improves the effectiveness of the basic MDC; (ii) over all A-Boost does not outperform AdaBoost in terms of classification accuracy; and (iii) the performance of the boosted classifiers improves as the ensemble size increases.

KW - AR model

KW - Acoustic emission

KW - Boosting

KW - Condition monitoring

KW - Grinding wheel

KW - Minimum distance classifier

KW - Wear

UR - http://www.scopus.com/inward/record.url?scp=35548978397&partnerID=8YFLogxK

U2 - 10.1016/j.ymssp.2007.06.005

DO - 10.1016/j.ymssp.2007.06.005

M3 - Article

AN - SCOPUS:35548978397

SN - 0888-3270

VL - 22

SP - 217

EP - 232

JO - Mechanical Systems and Signal Processing

JF - Mechanical Systems and Signal Processing

IS - 1

ER -

Grinding wheel condition monitoring with boosted minimum distance classifiers

Abstract

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this