Probabilistic fatigue damage prognosis using maximum entropy approach

Xuefei Guan, Ratneshwar Jha, Yongming Liu

Research output: Contribution to journalArticle

34 Citations (Scopus)

Abstract

A general framework for probabilistic fatigue damage prognosis using maximum entropy concept is proposed and developed in this paper. The fatigue damage is calculated using a physics-based crack growth model. Due to the stochastic nature of fatigue crack propagation process, uncertainties arising from the underlying physical model, parameters of the model and the response variable measurement noise are considered and integrated into this framework. Incorporating all those uncertainties, a maximum relative entropy (MRE) approach is proposed to update the statistical description of model parameters and narrow down the prognosis deviations. A Markov Chain Monte Carlo (MCMC) simulation is then employed to generate samples from updated posterior probability distributions and provide statistical information for the maximum relative entropy updating procedure. A numerical toy problem is given to demonstrate the proposed MRE prognosis methodology. Experimental data for aluminum alloys are used to validate model predictions under uncertainty. Following this, a detailed comparison between the proposed MRE approach and the classical Bayesian updating method is performed to illustrate advantages of the proposed prognosis framework.

Original languageEnglish (US)
Pages (from-to)163-171
Number of pages9
JournalJournal of Intelligent Manufacturing
Volume23
Issue number2
DOIs
StatePublished - Apr 2012
Externally publishedYes

Fingerprint

Fatigue damage
Entropy
Fatigue crack propagation
Markov processes
Probability distributions
Aluminum alloys
Crack propagation
Physics
Uncertainty

Keywords

  • Bayesian method
  • Fatigue crack growth
  • Markov Chain Monte Carlo
  • Maximum relative entropy
  • Model updating
  • Uncertainty

ASJC Scopus subject areas

  • Artificial Intelligence
  • Software
  • Industrial and Manufacturing Engineering

Cite this

Probabilistic fatigue damage prognosis using maximum entropy approach. / Guan, Xuefei; Jha, Ratneshwar; Liu, Yongming.

In: Journal of Intelligent Manufacturing, Vol. 23, No. 2, 04.2012, p. 163-171.

Research output: Contribution to journalArticle

@article{2703adfeb539404291c372f366973555,
title = "Probabilistic fatigue damage prognosis using maximum entropy approach",
abstract = "A general framework for probabilistic fatigue damage prognosis using maximum entropy concept is proposed and developed in this paper. The fatigue damage is calculated using a physics-based crack growth model. Due to the stochastic nature of fatigue crack propagation process, uncertainties arising from the underlying physical model, parameters of the model and the response variable measurement noise are considered and integrated into this framework. Incorporating all those uncertainties, a maximum relative entropy (MRE) approach is proposed to update the statistical description of model parameters and narrow down the prognosis deviations. A Markov Chain Monte Carlo (MCMC) simulation is then employed to generate samples from updated posterior probability distributions and provide statistical information for the maximum relative entropy updating procedure. A numerical toy problem is given to demonstrate the proposed MRE prognosis methodology. Experimental data for aluminum alloys are used to validate model predictions under uncertainty. Following this, a detailed comparison between the proposed MRE approach and the classical Bayesian updating method is performed to illustrate advantages of the proposed prognosis framework.",
keywords = "Bayesian method, Fatigue crack growth, Markov Chain Monte Carlo, Maximum relative entropy, Model updating, Uncertainty",
author = "Xuefei Guan and Ratneshwar Jha and Yongming Liu",
year = "2012",
month = "4",
doi = "10.1007/s10845-009-0341-3",
language = "English (US)",
volume = "23",
pages = "163--171",
journal = "Journal of Intelligent Manufacturing",
issn = "0956-5515",
publisher = "Springer Netherlands",
number = "2",

}

TY - JOUR

T1 - Probabilistic fatigue damage prognosis using maximum entropy approach

AU - Guan, Xuefei

AU - Jha, Ratneshwar

AU - Liu, Yongming

PY - 2012/4

Y1 - 2012/4

N2 - A general framework for probabilistic fatigue damage prognosis using maximum entropy concept is proposed and developed in this paper. The fatigue damage is calculated using a physics-based crack growth model. Due to the stochastic nature of fatigue crack propagation process, uncertainties arising from the underlying physical model, parameters of the model and the response variable measurement noise are considered and integrated into this framework. Incorporating all those uncertainties, a maximum relative entropy (MRE) approach is proposed to update the statistical description of model parameters and narrow down the prognosis deviations. A Markov Chain Monte Carlo (MCMC) simulation is then employed to generate samples from updated posterior probability distributions and provide statistical information for the maximum relative entropy updating procedure. A numerical toy problem is given to demonstrate the proposed MRE prognosis methodology. Experimental data for aluminum alloys are used to validate model predictions under uncertainty. Following this, a detailed comparison between the proposed MRE approach and the classical Bayesian updating method is performed to illustrate advantages of the proposed prognosis framework.

AB - A general framework for probabilistic fatigue damage prognosis using maximum entropy concept is proposed and developed in this paper. The fatigue damage is calculated using a physics-based crack growth model. Due to the stochastic nature of fatigue crack propagation process, uncertainties arising from the underlying physical model, parameters of the model and the response variable measurement noise are considered and integrated into this framework. Incorporating all those uncertainties, a maximum relative entropy (MRE) approach is proposed to update the statistical description of model parameters and narrow down the prognosis deviations. A Markov Chain Monte Carlo (MCMC) simulation is then employed to generate samples from updated posterior probability distributions and provide statistical information for the maximum relative entropy updating procedure. A numerical toy problem is given to demonstrate the proposed MRE prognosis methodology. Experimental data for aluminum alloys are used to validate model predictions under uncertainty. Following this, a detailed comparison between the proposed MRE approach and the classical Bayesian updating method is performed to illustrate advantages of the proposed prognosis framework.

KW - Bayesian method

KW - Fatigue crack growth

KW - Markov Chain Monte Carlo

KW - Maximum relative entropy

KW - Model updating

KW - Uncertainty

UR - http://www.scopus.com/inward/record.url?scp=84862213127&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84862213127&partnerID=8YFLogxK

U2 - 10.1007/s10845-009-0341-3

DO - 10.1007/s10845-009-0341-3

M3 - Article

VL - 23

SP - 163

EP - 171

JO - Journal of Intelligent Manufacturing

JF - Journal of Intelligent Manufacturing

SN - 0956-5515

IS - 2

ER -