Improving the generalization ability of an artificial neural network in predicting in-flight particle characteristics of an atmospheric plasma spray process

T. A. Choudhury, N. Hosseinzadeh, C. C. Berndt

Research output: Contribution to journalArticle

10 Citations (Scopus)

Abstract

This paper presents the application of the artificial neural network into an atmospheric plasma spray process for predicting the in-flight particle characteristics, which have significant influence on the in-service coating properties. One of the major problems for such function-approximating neural network is over-fitting, which reduces the generalization capability of a trained network and its ability to work with sufficient accuracy under a new environment. Two methods are used to analyze the improvement in the network's generalization ability: (i) cross-validation and early stopping, and (ii) Bayesian regularization. Simulations are performed both on the original and expanded database with different training conditions to obtain the variations in performance of the trained networks under various environments. The study further illustrates the design and optimization procedures and analyzes the predicted values, with respect to the experimental ones, to evaluate the performance and generalization ability of the network. The simulation results show that the performance of the trained networks with regularization is improved over that with cross-validation and early stopping and, furthermore, the generalization capability of the networks is improved; thus preventing any phenomenon associated with over-fitting.

Original languageEnglish
Pages (from-to)935-949
Number of pages15
JournalJournal of Thermal Spray Technology
Volume21
Issue number5
DOIs
Publication statusPublished - Sep 2012

Fingerprint

sprayers
flight
Neural networks
Plasmas
Coatings
stopping
education
simulation
coatings
optimization

Keywords

  • artificial neural network
  • atmospheric plasma spray
  • Bayesian regularization
  • cross-validation
  • early stopping
  • in-flight particle characteristics
  • Kernel regression

ASJC Scopus subject areas

  • Condensed Matter Physics
  • Materials Chemistry
  • Surfaces, Coatings and Films

Cite this

@article{bf85d59824ea4e3ab54fc0d6d7963027,
title = "Improving the generalization ability of an artificial neural network in predicting in-flight particle characteristics of an atmospheric plasma spray process",
abstract = "This paper presents the application of the artificial neural network into an atmospheric plasma spray process for predicting the in-flight particle characteristics, which have significant influence on the in-service coating properties. One of the major problems for such function-approximating neural network is over-fitting, which reduces the generalization capability of a trained network and its ability to work with sufficient accuracy under a new environment. Two methods are used to analyze the improvement in the network's generalization ability: (i) cross-validation and early stopping, and (ii) Bayesian regularization. Simulations are performed both on the original and expanded database with different training conditions to obtain the variations in performance of the trained networks under various environments. The study further illustrates the design and optimization procedures and analyzes the predicted values, with respect to the experimental ones, to evaluate the performance and generalization ability of the network. The simulation results show that the performance of the trained networks with regularization is improved over that with cross-validation and early stopping and, furthermore, the generalization capability of the networks is improved; thus preventing any phenomenon associated with over-fitting.",
keywords = "artificial neural network, atmospheric plasma spray, Bayesian regularization, cross-validation, early stopping, in-flight particle characteristics, Kernel regression",
author = "Choudhury, {T. A.} and N. Hosseinzadeh and Berndt, {C. C.}",
year = "2012",
month = "9",
doi = "10.1007/s11666-012-9775-9",
language = "English",
volume = "21",
pages = "935--949",
journal = "Journal of Thermal Spray Technology",
issn = "1059-9630",
publisher = "Springer New York",
number = "5",

}

TY - JOUR

T1 - Improving the generalization ability of an artificial neural network in predicting in-flight particle characteristics of an atmospheric plasma spray process

AU - Choudhury, T. A.

AU - Hosseinzadeh, N.

AU - Berndt, C. C.

PY - 2012/9

Y1 - 2012/9

N2 - This paper presents the application of the artificial neural network into an atmospheric plasma spray process for predicting the in-flight particle characteristics, which have significant influence on the in-service coating properties. One of the major problems for such function-approximating neural network is over-fitting, which reduces the generalization capability of a trained network and its ability to work with sufficient accuracy under a new environment. Two methods are used to analyze the improvement in the network's generalization ability: (i) cross-validation and early stopping, and (ii) Bayesian regularization. Simulations are performed both on the original and expanded database with different training conditions to obtain the variations in performance of the trained networks under various environments. The study further illustrates the design and optimization procedures and analyzes the predicted values, with respect to the experimental ones, to evaluate the performance and generalization ability of the network. The simulation results show that the performance of the trained networks with regularization is improved over that with cross-validation and early stopping and, furthermore, the generalization capability of the networks is improved; thus preventing any phenomenon associated with over-fitting.

AB - This paper presents the application of the artificial neural network into an atmospheric plasma spray process for predicting the in-flight particle characteristics, which have significant influence on the in-service coating properties. One of the major problems for such function-approximating neural network is over-fitting, which reduces the generalization capability of a trained network and its ability to work with sufficient accuracy under a new environment. Two methods are used to analyze the improvement in the network's generalization ability: (i) cross-validation and early stopping, and (ii) Bayesian regularization. Simulations are performed both on the original and expanded database with different training conditions to obtain the variations in performance of the trained networks under various environments. The study further illustrates the design and optimization procedures and analyzes the predicted values, with respect to the experimental ones, to evaluate the performance and generalization ability of the network. The simulation results show that the performance of the trained networks with regularization is improved over that with cross-validation and early stopping and, furthermore, the generalization capability of the networks is improved; thus preventing any phenomenon associated with over-fitting.

KW - artificial neural network

KW - atmospheric plasma spray

KW - Bayesian regularization

KW - cross-validation

KW - early stopping

KW - in-flight particle characteristics

KW - Kernel regression

UR - http://www.scopus.com/inward/record.url?scp=84865256576&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84865256576&partnerID=8YFLogxK

U2 - 10.1007/s11666-012-9775-9

DO - 10.1007/s11666-012-9775-9

M3 - Article

VL - 21

SP - 935

EP - 949

JO - Journal of Thermal Spray Technology

JF - Journal of Thermal Spray Technology

SN - 1059-9630

IS - 5

ER -