Ensemble Bayesian tensor factorization for debond thermal NDT

Peng Lu; Bin Gao; Qizhi Feng; Yang Yang; W. L. Woo; Jian Zhao; Xueshi Qiu; Liangyong Gu; Guiyun Tian

doi:10.1109/FENDT.2017.8584558

Ensemble Bayesian tensor factorization for debond thermal NDT

Peng Lu, Bin Gao, Qizhi Feng, Yang Yang, W. L. Woo, Jian Zhao, Xueshi Qiu, Liangyong Gu, Guiyun Tian

Computer Information Sciences Department

Research output: Chapter in Book/Report/Conference proceeding › Chapter › peer-review

Abstract

One of the common types of defects in the carbon fiber reinforced polymer (CFRP) is debond. The different feature extraction algorithms of optical stimulated infrared thermography are used to obtained the debond detection. However, the low detection accuracy as well as remain as challenges. In this paper, the ensemble variational Bayes tensor factorization (EVBTF) has been proposed to overcome the problems. The framework of the proposed algorithm is based on the Bayesian learning theory. It constructs spatial-transient multi-layer mining structure. Experimental tests have been proved that it can effectively improve the contrast ratio between the defective areas and the sound areas.

Original language	English
Title of host publication	Proceedings of 2017 IEEE Far East NDT New Technology and Application Forum, FENDT 2017
Publisher	IEEE
Pages	228-232
Number of pages	5
ISBN (Electronic)	9781538616154, 9781538616130
ISBN (Print)	9781538616161
DOIs	https://doi.org/10.1109/FENDT.2017.8584558
Publication status	Published - 20 Dec 2018
Event	2017 Far East NDT New Technology & Application Forum - Shaanxi Guesthouse, Xi’an, China Duration: 22 Jun 2017 → 24 Jun 2017 http://www.fendti.com/2017/

Conference

Conference	2017 Far East NDT New Technology & Application Forum
Abbreviated title	FENDT 2017
Country/Territory	China
City	Xi’an
Period	22/06/17 → 24/06/17
Internet address	http://www.fendti.com/2017/

Keywords

CFRP
Debond defects
Non-destructive testing
Tensor factorization

Access to Document

10.1109/FENDT.2017.8584558

Cite this

@inbook{f79773f5560f401b90332b2abc1f19fc,

title = "Ensemble Bayesian tensor factorization for debond thermal NDT",

abstract = "One of the common types of defects in the carbon fiber reinforced polymer (CFRP) is debond. The different feature extraction algorithms of optical stimulated infrared thermography are used to obtained the debond detection. However, the low detection accuracy as well as remain as challenges. In this paper, the ensemble variational Bayes tensor factorization (EVBTF) has been proposed to overcome the problems. The framework of the proposed algorithm is based on the Bayesian learning theory. It constructs spatial-transient multi-layer mining structure. Experimental tests have been proved that it can effectively improve the contrast ratio between the defective areas and the sound areas.",

keywords = "CFRP, Debond defects, Non-destructive testing, Tensor factorization",

author = "Peng Lu and Bin Gao and Qizhi Feng and Yang Yang and Woo, {W. L.} and Jian Zhao and Xueshi Qiu and Liangyong Gu and Guiyun Tian",

year = "2018",

month = dec,

day = "20",

doi = "10.1109/FENDT.2017.8584558",

language = "English",

isbn = "9781538616161",

pages = "228--232",

booktitle = "Proceedings of 2017 IEEE Far East NDT New Technology and Application Forum, FENDT 2017",

publisher = "IEEE",

address = "United States",

note = "2017 Far East NDT New Technology & Application Forum, FENDT 2017 ; Conference date: 22-06-2017 Through 24-06-2017",

url = "http://www.fendti.com/2017/",

}

Lu, P, Gao, B, Feng, Q, Yang, Y, Woo, WL, Zhao, J, Qiu, X, Gu, L & Tian, G 2018, Ensemble Bayesian tensor factorization for debond thermal NDT. in Proceedings of 2017 IEEE Far East NDT New Technology and Application Forum, FENDT 2017. IEEE, pp. 228-232, 2017 Far East NDT New Technology & Application Forum, Xi’an, China, 22/06/17. https://doi.org/10.1109/FENDT.2017.8584558

TY - CHAP

T1 - Ensemble Bayesian tensor factorization for debond thermal NDT

AU - Lu, Peng

AU - Gao, Bin

AU - Feng, Qizhi

AU - Yang, Yang

AU - Woo, W. L.

AU - Zhao, Jian

AU - Qiu, Xueshi

AU - Gu, Liangyong

AU - Tian, Guiyun

PY - 2018/12/20

Y1 - 2018/12/20

N2 - One of the common types of defects in the carbon fiber reinforced polymer (CFRP) is debond. The different feature extraction algorithms of optical stimulated infrared thermography are used to obtained the debond detection. However, the low detection accuracy as well as remain as challenges. In this paper, the ensemble variational Bayes tensor factorization (EVBTF) has been proposed to overcome the problems. The framework of the proposed algorithm is based on the Bayesian learning theory. It constructs spatial-transient multi-layer mining structure. Experimental tests have been proved that it can effectively improve the contrast ratio between the defective areas and the sound areas.

AB - One of the common types of defects in the carbon fiber reinforced polymer (CFRP) is debond. The different feature extraction algorithms of optical stimulated infrared thermography are used to obtained the debond detection. However, the low detection accuracy as well as remain as challenges. In this paper, the ensemble variational Bayes tensor factorization (EVBTF) has been proposed to overcome the problems. The framework of the proposed algorithm is based on the Bayesian learning theory. It constructs spatial-transient multi-layer mining structure. Experimental tests have been proved that it can effectively improve the contrast ratio between the defective areas and the sound areas.

KW - CFRP

KW - Debond defects

KW - Non-destructive testing

KW - Tensor factorization

U2 - 10.1109/FENDT.2017.8584558

DO - 10.1109/FENDT.2017.8584558

M3 - Chapter

SN - 9781538616161

SP - 228

EP - 232

BT - Proceedings of 2017 IEEE Far East NDT New Technology and Application Forum, FENDT 2017

PB - IEEE

T2 - 2017 Far East NDT New Technology & Application Forum

Y2 - 22 June 2017 through 24 June 2017

ER -

Ensemble Bayesian tensor factorization for debond thermal NDT

Abstract

Conference

Keywords

Access to Document

Fingerprint

Cite this