A cyber-physical scheme for predicting tool wear based on a hybrid dynamic neural network

Haw Ching Yang, Yu Yung Li, Min Hsiung Hung, Fan Tien Cheng

研究成果: Article

2 引文 (Scopus)

摘要

It is costly to predict tool wear under various machining conditions. To address this challenge, a tool cyber-physical prediction (TCPP) scheme and a hybrid dynamic neural network (HDNN) model are proposed in this paper. This scheme enables users to build and use the models both in the cloud and at the factory by integrating the theoretical maximum tool life and the practical sensing features of the tool wear. Moreover, using features extracted from the sensors and controller, the HDNN model integrates the logistic regression and dynamic neural network to diagnose the tool break and predict tool wear simultaneously. In addition, the scheme presents a model-refreshing approach to tune the HDNN model to adapt to physical variation of the tool coating, the workpiece material, and the removal process in the similar cutting conditions. The experimental results demonstrate that the TCPP scheme with the HDNN model is promising for tool wear prediction while using only a few samples and the current features to adapt to various cutting conditions.

指紋

Wear of materials
Neural networks
Industrial plants
Logistics
Machining
Coatings
Controllers
Sensors

All Science Journal Classification (ASJC) codes

  • Engineering(all)

引用此文

@article{9ecea0b2466749629044631d3bb705da,
title = "A cyber-physical scheme for predicting tool wear based on a hybrid dynamic neural network",
abstract = "It is costly to predict tool wear under various machining conditions. To address this challenge, a tool cyber-physical prediction (TCPP) scheme and a hybrid dynamic neural network (HDNN) model are proposed in this paper. This scheme enables users to build and use the models both in the cloud and at the factory by integrating the theoretical maximum tool life and the practical sensing features of the tool wear. Moreover, using features extracted from the sensors and controller, the HDNN model integrates the logistic regression and dynamic neural network to diagnose the tool break and predict tool wear simultaneously. In addition, the scheme presents a model-refreshing approach to tune the HDNN model to adapt to physical variation of the tool coating, the workpiece material, and the removal process in the similar cutting conditions. The experimental results demonstrate that the TCPP scheme with the HDNN model is promising for tool wear prediction while using only a few samples and the current features to adapt to various cutting conditions.",
author = "Yang, {Haw Ching} and Li, {Yu Yung} and Hung, {Min Hsiung} and Cheng, {Fan Tien}",
year = "2017",
month = "10",
day = "3",
doi = "10.1080/02533839.2017.1372223",
language = "English",
volume = "40",
pages = "614--625",
journal = "Chung-kuo Kung Ch'eng Hsueh K'an/Journal of the Chinese Institute of Engineers",
issn = "0253-3839",
publisher = "Chinese Institute of Engineers",
number = "7",

}

TY - JOUR

T1 - A cyber-physical scheme for predicting tool wear based on a hybrid dynamic neural network

AU - Yang, Haw Ching

AU - Li, Yu Yung

AU - Hung, Min Hsiung

AU - Cheng, Fan Tien

PY - 2017/10/3

Y1 - 2017/10/3

N2 - It is costly to predict tool wear under various machining conditions. To address this challenge, a tool cyber-physical prediction (TCPP) scheme and a hybrid dynamic neural network (HDNN) model are proposed in this paper. This scheme enables users to build and use the models both in the cloud and at the factory by integrating the theoretical maximum tool life and the practical sensing features of the tool wear. Moreover, using features extracted from the sensors and controller, the HDNN model integrates the logistic regression and dynamic neural network to diagnose the tool break and predict tool wear simultaneously. In addition, the scheme presents a model-refreshing approach to tune the HDNN model to adapt to physical variation of the tool coating, the workpiece material, and the removal process in the similar cutting conditions. The experimental results demonstrate that the TCPP scheme with the HDNN model is promising for tool wear prediction while using only a few samples and the current features to adapt to various cutting conditions.

AB - It is costly to predict tool wear under various machining conditions. To address this challenge, a tool cyber-physical prediction (TCPP) scheme and a hybrid dynamic neural network (HDNN) model are proposed in this paper. This scheme enables users to build and use the models both in the cloud and at the factory by integrating the theoretical maximum tool life and the practical sensing features of the tool wear. Moreover, using features extracted from the sensors and controller, the HDNN model integrates the logistic regression and dynamic neural network to diagnose the tool break and predict tool wear simultaneously. In addition, the scheme presents a model-refreshing approach to tune the HDNN model to adapt to physical variation of the tool coating, the workpiece material, and the removal process in the similar cutting conditions. The experimental results demonstrate that the TCPP scheme with the HDNN model is promising for tool wear prediction while using only a few samples and the current features to adapt to various cutting conditions.

UR - http://www.scopus.com/inward/record.url?scp=85031290148&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85031290148&partnerID=8YFLogxK

U2 - 10.1080/02533839.2017.1372223

DO - 10.1080/02533839.2017.1372223

M3 - Article

AN - SCOPUS:85031290148

VL - 40

SP - 614

EP - 625

JO - Chung-kuo Kung Ch'eng Hsueh K'an/Journal of the Chinese Institute of Engineers

JF - Chung-kuo Kung Ch'eng Hsueh K'an/Journal of the Chinese Institute of Engineers

SN - 0253-3839

IS - 7

ER -