Knowledge extraction from deep convolutional neural networks applied to cyclo-stationary time-series classification

Diego Cabrera; Fernando Sancho; Mariela Cerrada; René Vinicio Sánchez; Chuan Li

doi:10.1016/j.ins.2020.03.039

Knowledge extraction from deep convolutional neural networks applied to cyclo-stationary time-series classification

Diego Cabrera, Fernando Sancho, Mariela Cerrada, René Vinicio Sánchez, Chuan Li

Research and Development Group in Industrial Technologies (GIDTEC)

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

9 Scopus citations

Abstract

Modelling complex processes from raw time series increases the necessity to build Deep Learning (DL) architectures that can manage this type of data structure. However, as DL models become deeper, larger and more diverse datasets are necessary and knowledge extraction will become more difficult. In an attempt to sidestep these issues, in this paper a methodology based on two main steps is presented, the first being to increase size and diversity of time-series datasets for training, and the second to retrieve knowledge from the obtained model. This methodology is compared with other approaches reported in the literature and is tested under two configuration setups of Condition-Based Maintenance problems: fault diagnosis of bearing, and fault severity assessment of a helical gearbox, obtaining not only a performance improvement in comparison, but also in retrieving knowledge about how the signals are being classified.

Original language	English
Pages (from-to)	1-14
Number of pages	14
Journal	Information Sciences
Volume	524
DOIs	https://doi.org/10.1016/j.ins.2020.03.039
State	Published - 1 Jul 2020

Bibliographical note

Funding Information:
The work was sponsored in part by GIDTEC Project No. 003-002-2016-03-03 and the National Natural Science Foundation of China (Grant No. 51775112). The experimental work was developed at the GIDTEC Research Group Lab of the Universidad Politcnica Salesiana, Cuenca, Ecuador. This research was also partially supported by the projects TIN2017-82113-C2-1-R (VICTORY Project) and TIN2013-41086-P (LOCOCIDA Project), both from Ministerio de Economía, Industria y Competitividad of Spain, with FEDER funds from the European Union.

Funding Information:
The work was sponsored in part by GIDTEC Project No. 003-002-2016-03-03 and the National Natural Science Foundation of China (Grant No. 51775112 ). The experimental work was developed at the GIDTEC Research Group Lab of the Universidad Politcnica Salesiana, Cuenca, Ecuador. This research was also partially supported by the projects TIN2017-82113-C2-1-R (VICTORY Project) and TIN2013-41086-P (LOCOCIDA Project), both from Ministerio de Economía, Industria y Competitividad of Spain, with FEDER funds from the European Union.

Publisher Copyright:
© 2020 Elsevier Inc.

Keywords

Convolutional neural network
Cyclo-stationary time-series analysis
Deep learning
Fault diagnosis
Knowledge extraction

Access to Document

10.1016/j.ins.2020.03.039

Cite this

@article{a9a64209b4d0480e9f796b8cb137f364,

title = "Knowledge extraction from deep convolutional neural networks applied to cyclo-stationary time-series classification",

abstract = "Modelling complex processes from raw time series increases the necessity to build Deep Learning (DL) architectures that can manage this type of data structure. However, as DL models become deeper, larger and more diverse datasets are necessary and knowledge extraction will become more difficult. In an attempt to sidestep these issues, in this paper a methodology based on two main steps is presented, the first being to increase size and diversity of time-series datasets for training, and the second to retrieve knowledge from the obtained model. This methodology is compared with other approaches reported in the literature and is tested under two configuration setups of Condition-Based Maintenance problems: fault diagnosis of bearing, and fault severity assessment of a helical gearbox, obtaining not only a performance improvement in comparison, but also in retrieving knowledge about how the signals are being classified.",

keywords = "Convolutional neural network, Cyclo-stationary time-series analysis, Deep learning, Fault diagnosis, Knowledge extraction",

author = "Diego Cabrera and Fernando Sancho and Mariela Cerrada and S{\'a}nchez, {Ren{\'e} Vinicio} and Chuan Li",

note = "Publisher Copyright: {\textcopyright} 2020 Elsevier Inc.",

year = "2020",

month = jul,

day = "1",

doi = "10.1016/j.ins.2020.03.039",

language = "English",

volume = "524",

pages = "1--14",

journal = "Information Sciences",

issn = "0020-0255",

publisher = "Elsevier Inc.",

}

TY - JOUR

T1 - Knowledge extraction from deep convolutional neural networks applied to cyclo-stationary time-series classification

AU - Cabrera, Diego

AU - Sancho, Fernando

AU - Cerrada, Mariela

AU - Sánchez, René Vinicio

AU - Li, Chuan

PY - 2020/7/1

Y1 - 2020/7/1

N2 - Modelling complex processes from raw time series increases the necessity to build Deep Learning (DL) architectures that can manage this type of data structure. However, as DL models become deeper, larger and more diverse datasets are necessary and knowledge extraction will become more difficult. In an attempt to sidestep these issues, in this paper a methodology based on two main steps is presented, the first being to increase size and diversity of time-series datasets for training, and the second to retrieve knowledge from the obtained model. This methodology is compared with other approaches reported in the literature and is tested under two configuration setups of Condition-Based Maintenance problems: fault diagnosis of bearing, and fault severity assessment of a helical gearbox, obtaining not only a performance improvement in comparison, but also in retrieving knowledge about how the signals are being classified.

AB - Modelling complex processes from raw time series increases the necessity to build Deep Learning (DL) architectures that can manage this type of data structure. However, as DL models become deeper, larger and more diverse datasets are necessary and knowledge extraction will become more difficult. In an attempt to sidestep these issues, in this paper a methodology based on two main steps is presented, the first being to increase size and diversity of time-series datasets for training, and the second to retrieve knowledge from the obtained model. This methodology is compared with other approaches reported in the literature and is tested under two configuration setups of Condition-Based Maintenance problems: fault diagnosis of bearing, and fault severity assessment of a helical gearbox, obtaining not only a performance improvement in comparison, but also in retrieving knowledge about how the signals are being classified.

KW - Convolutional neural network

KW - Cyclo-stationary time-series analysis

KW - Deep learning

KW - Fault diagnosis

KW - Knowledge extraction

UR - http://www.scopus.com/inward/record.url?scp=85082022662&partnerID=8YFLogxK

UR - https://www.mendeley.com/catalogue/543ff335-c237-3539-a18e-183c510b36b1/

U2 - 10.1016/j.ins.2020.03.039

DO - 10.1016/j.ins.2020.03.039

M3 - Article

AN - SCOPUS:85082022662

SN - 0020-0255

VL - 524

SP - 1

EP - 14

JO - Information Sciences

JF - Information Sciences

ER -

Knowledge extraction from deep convolutional neural networks applied to cyclo-stationary time-series classification

Abstract

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this