Uncertainty analysis of a neural network used for fatigue lifetime prediction

S.G. Pierce, K. Worden, A. Bezazzi

Research output: Contribution to journalArticle

39 Citations (Scopus)

Abstract

The application of interval set techniques to the quantification of uncertainty in a neural network regression model of fatigue lifetime is considered. Bayesian evidence training was implemented to train a series of multi-layer perceptron networks on experimental fatigue life measurements in glass fibre composite sandwich materials. A set of independent measurements conducted 2 months after the training session, and at intermediate fatigue loading levels, was used to provide a rigorous test of the generalisation capacity of the networks. The robustness of the networks to uncertainty in the input data was investigated using an interval-based technique. It is demonstrated that the interval approach allowed for an alternative to probabilistic-based confidence bounds of prediction accuracy. In addition, the technique provided an alternative network selection tool, and also allowed for an alternative to estimating the lifetime prediction error that was found to be a significant improvement over the Bayesian-derived estimate of confidence bound.
Original languageEnglish
Pages (from-to)1395-1411
Number of pages16
JournalMechanical Systems and Signal Processing
Volume22
Issue number6
DOIs
Publication statusPublished - Aug 2008

Fingerprint

Uncertainty analysis
Fatigue of materials
Neural networks
Multilayer neural networks
Glass fibers
Composite materials
Uncertainty

Keywords

  • lifetime prediction
  • fatigue
  • neural networks
  • uncertainty
  • interval analysis

Cite this

@article{779031ebe7f74c8397129dd80f7d0e3e,
title = "Uncertainty analysis of a neural network used for fatigue lifetime prediction",
abstract = "The application of interval set techniques to the quantification of uncertainty in a neural network regression model of fatigue lifetime is considered. Bayesian evidence training was implemented to train a series of multi-layer perceptron networks on experimental fatigue life measurements in glass fibre composite sandwich materials. A set of independent measurements conducted 2 months after the training session, and at intermediate fatigue loading levels, was used to provide a rigorous test of the generalisation capacity of the networks. The robustness of the networks to uncertainty in the input data was investigated using an interval-based technique. It is demonstrated that the interval approach allowed for an alternative to probabilistic-based confidence bounds of prediction accuracy. In addition, the technique provided an alternative network selection tool, and also allowed for an alternative to estimating the lifetime prediction error that was found to be a significant improvement over the Bayesian-derived estimate of confidence bound.",
keywords = "lifetime prediction, fatigue, neural networks, uncertainty, interval analysis",
author = "S.G. Pierce and K. Worden and A. Bezazzi",
year = "2008",
month = "8",
doi = "10.1016/j.ymssp.2007.12.004",
language = "English",
volume = "22",
pages = "1395--1411",
journal = "Mechanical Systems and Signal Processing",
issn = "0888-3270",
publisher = "Elsevier",
number = "6",

}

Uncertainty analysis of a neural network used for fatigue lifetime prediction. / Pierce, S.G.; Worden, K.; Bezazzi, A.

In: Mechanical Systems and Signal Processing, Vol. 22, No. 6, 08.2008, p. 1395-1411.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Uncertainty analysis of a neural network used for fatigue lifetime prediction

AU - Pierce, S.G.

AU - Worden, K.

AU - Bezazzi, A.

PY - 2008/8

Y1 - 2008/8

N2 - The application of interval set techniques to the quantification of uncertainty in a neural network regression model of fatigue lifetime is considered. Bayesian evidence training was implemented to train a series of multi-layer perceptron networks on experimental fatigue life measurements in glass fibre composite sandwich materials. A set of independent measurements conducted 2 months after the training session, and at intermediate fatigue loading levels, was used to provide a rigorous test of the generalisation capacity of the networks. The robustness of the networks to uncertainty in the input data was investigated using an interval-based technique. It is demonstrated that the interval approach allowed for an alternative to probabilistic-based confidence bounds of prediction accuracy. In addition, the technique provided an alternative network selection tool, and also allowed for an alternative to estimating the lifetime prediction error that was found to be a significant improvement over the Bayesian-derived estimate of confidence bound.

AB - The application of interval set techniques to the quantification of uncertainty in a neural network regression model of fatigue lifetime is considered. Bayesian evidence training was implemented to train a series of multi-layer perceptron networks on experimental fatigue life measurements in glass fibre composite sandwich materials. A set of independent measurements conducted 2 months after the training session, and at intermediate fatigue loading levels, was used to provide a rigorous test of the generalisation capacity of the networks. The robustness of the networks to uncertainty in the input data was investigated using an interval-based technique. It is demonstrated that the interval approach allowed for an alternative to probabilistic-based confidence bounds of prediction accuracy. In addition, the technique provided an alternative network selection tool, and also allowed for an alternative to estimating the lifetime prediction error that was found to be a significant improvement over the Bayesian-derived estimate of confidence bound.

KW - lifetime prediction

KW - fatigue

KW - neural networks

KW - uncertainty

KW - interval analysis

UR - http://dx.doi.org/10.1016/j.ymssp.2007.12.004

U2 - 10.1016/j.ymssp.2007.12.004

DO - 10.1016/j.ymssp.2007.12.004

M3 - Article

VL - 22

SP - 1395

EP - 1411

JO - Mechanical Systems and Signal Processing

JF - Mechanical Systems and Signal Processing

SN - 0888-3270

IS - 6

ER -