Extended and generalized fragility functions

Research output: Contribution to journalArticle

4 Citations (Scopus)

Abstract

Fragility functions indicate the probability of a system exceeding certain damage states given some appropriate measures that characterize recorded or simulated data series. Presented in two main parts, this paper develops fragility functions in their utmost generality, accounting for both (1) multivariate intensity measures with multiple damage states and (2) longitudinal damage state dependencies in time. Without adopting the limiting assumption of common variance to avoid improper function crossings, the first part presents what is here compactly termed as extended fragility functions. As shown, these are best supported by the softmax function for any arbitrary distribution of the exponential family to which the intensity measures of different states may belong, including the typically used normal distribution in the logarithmic scale of intensity measures. In the second part, generalized fragility functions are introduced for cases where multiple system state transitions need to be captured. To that end, dependent Markov and hidden Markov models are employed because they are able to portray longitudinal data dependencies and reveal intrinsic deterioration trends for multiple sequential events. Numerical results are presented, together with underlying implementation details, statistical properties, and practical suggestions.

Original languageEnglish (US)
Article number04018087
JournalJournal of Engineering Mechanics
Volume144
Issue number9
DOIs
StatePublished - Sep 1 2018

Fingerprint

Normal distribution
Hidden Markov models
Deterioration

All Science Journal Classification (ASJC) codes

  • Mechanics of Materials
  • Mechanical Engineering

Cite this

@article{beb5c332a98b426d89e4d4a77fe41dc7,
title = "Extended and generalized fragility functions",
abstract = "Fragility functions indicate the probability of a system exceeding certain damage states given some appropriate measures that characterize recorded or simulated data series. Presented in two main parts, this paper develops fragility functions in their utmost generality, accounting for both (1) multivariate intensity measures with multiple damage states and (2) longitudinal damage state dependencies in time. Without adopting the limiting assumption of common variance to avoid improper function crossings, the first part presents what is here compactly termed as extended fragility functions. As shown, these are best supported by the softmax function for any arbitrary distribution of the exponential family to which the intensity measures of different states may belong, including the typically used normal distribution in the logarithmic scale of intensity measures. In the second part, generalized fragility functions are introduced for cases where multiple system state transitions need to be captured. To that end, dependent Markov and hidden Markov models are employed because they are able to portray longitudinal data dependencies and reveal intrinsic deterioration trends for multiple sequential events. Numerical results are presented, together with underlying implementation details, statistical properties, and practical suggestions.",
author = "Andriotis, {C. P.} and Konstantinos Papakonstantinou",
year = "2018",
month = "9",
day = "1",
doi = "10.1061/(ASCE)EM.1943-7889.0001478",
language = "English (US)",
volume = "144",
journal = "Journal of Engineering Mechanics - ASCE",
issn = "0733-9399",
publisher = "American Society of Civil Engineers (ASCE)",
number = "9",

}

Extended and generalized fragility functions. / Andriotis, C. P.; Papakonstantinou, Konstantinos.

In: Journal of Engineering Mechanics, Vol. 144, No. 9, 04018087, 01.09.2018.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Extended and generalized fragility functions

AU - Andriotis, C. P.

AU - Papakonstantinou, Konstantinos

PY - 2018/9/1

Y1 - 2018/9/1

N2 - Fragility functions indicate the probability of a system exceeding certain damage states given some appropriate measures that characterize recorded or simulated data series. Presented in two main parts, this paper develops fragility functions in their utmost generality, accounting for both (1) multivariate intensity measures with multiple damage states and (2) longitudinal damage state dependencies in time. Without adopting the limiting assumption of common variance to avoid improper function crossings, the first part presents what is here compactly termed as extended fragility functions. As shown, these are best supported by the softmax function for any arbitrary distribution of the exponential family to which the intensity measures of different states may belong, including the typically used normal distribution in the logarithmic scale of intensity measures. In the second part, generalized fragility functions are introduced for cases where multiple system state transitions need to be captured. To that end, dependent Markov and hidden Markov models are employed because they are able to portray longitudinal data dependencies and reveal intrinsic deterioration trends for multiple sequential events. Numerical results are presented, together with underlying implementation details, statistical properties, and practical suggestions.

AB - Fragility functions indicate the probability of a system exceeding certain damage states given some appropriate measures that characterize recorded or simulated data series. Presented in two main parts, this paper develops fragility functions in their utmost generality, accounting for both (1) multivariate intensity measures with multiple damage states and (2) longitudinal damage state dependencies in time. Without adopting the limiting assumption of common variance to avoid improper function crossings, the first part presents what is here compactly termed as extended fragility functions. As shown, these are best supported by the softmax function for any arbitrary distribution of the exponential family to which the intensity measures of different states may belong, including the typically used normal distribution in the logarithmic scale of intensity measures. In the second part, generalized fragility functions are introduced for cases where multiple system state transitions need to be captured. To that end, dependent Markov and hidden Markov models are employed because they are able to portray longitudinal data dependencies and reveal intrinsic deterioration trends for multiple sequential events. Numerical results are presented, together with underlying implementation details, statistical properties, and practical suggestions.

UR - http://www.scopus.com/inward/record.url?scp=85047243466&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85047243466&partnerID=8YFLogxK

U2 - 10.1061/(ASCE)EM.1943-7889.0001478

DO - 10.1061/(ASCE)EM.1943-7889.0001478

M3 - Article

AN - SCOPUS:85047243466

VL - 144

JO - Journal of Engineering Mechanics - ASCE

JF - Journal of Engineering Mechanics - ASCE

SN - 0733-9399

IS - 9

M1 - 04018087

ER -