Stress–strength reliability under the generalized exponential model: General formula, censored-data inference, and engineering application

We study the stress–strength reliability 𝑅 = 𝑃 ⁡ ( 𝑋 2 < 𝑋 1 ) when strength 𝑋 1 and stress 𝑋 2 are independent members of O’Brien’s general exponential family, which contains the exponential, Rayleigh, gamma, Weibull, Maxwell, chi-square, Nakagami, Erlang, generalized-gamma and transformed-gamma laws. For a common shape parameter, 𝑅 is obtained in closed form as a Gauss hypergeometric function and, equivalently, as the regularized incomplete beta function 𝐼 𝛽 2 / ( 𝛽 1 + 𝛽 2 ) ⁡ ( 𝜈 2 , 𝜈 1 ) ; the proof is valid for every rate ratio. For unequal shapes we give a convergent series with its exact domain of convergence. We locate the series’ largest term and the number of digits lost to cancellation, which yields an explicit switching rule between the closed form, the series and a one-dimensional quadrature. We prove that the family converges in total variation to a power-function law as 𝑛 → ∞ . This gives closed-form lower bounds for the profile likelihood and a diagnostic that detects, before any optimization, data sets on which the unrestricted model has no interior maximum. Maximum likelihood estimation with Fisher-information asymptotics, logit-transformed intervals, Bayesian MCMC and a Lindley approximation are developed. A Monte Carlo study (1000 replications per cell) reports bias, MSE, interval length and coverage. In an application to aircraft windshields, with lifetime as strength and service time as demand, the diagnostic shows that the failure times alone admit no interior maximum, an artifact of ignoring the censored units. When censoring is taken into account, the unrestricted model is identified for the lifetimes and is the only candidate passing all goodness-of-fit tests for the service times. Maximum likelihood, Bayesian, bootstrap and Lindley estimates ( 0 . 6 9 9 - - 0 . 7 0 6 ) agree with a model-free Kaplan–Meier estimate ( 0 . 7 0 6 ), whereas ignoring censoring would understate 𝑅 by about two standard errors. The fitted model is then used for engineering decisions: it gives B-lives with bootstrap intervals, reveals a wear-out hazard that rises faster than the Weibull model predicts, yields optimal age-replacement intervals that cut long-run maintenance cost by 2 2 - - 5 4 % relative to run-to-failure, and quantifies the lifetime improvement ( 2 6 % and 8 4 % ) required to raise 𝑅 to 0 . 8 and 0 . 9 .

Authors

Institutions

Publication Details

Journal
Ain Shams Engineering Journal
Published
2026-11-17
DOI
https://doi.org/10.1016/j.asej.2026.104531
Primary Topic
Statistical Distribution Estimation and Applications
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Stress–strength reliability under the generalized exponential model: General formula, censored-data inference, and engineering application

M. E. Abd El-Monsef, Yakup Yıldırım, Ghareeb A. Marei, Hamdy M. Ahmed et al.
1 links
Ain Shams Engineering Journal
Statistical Distribution Estimation and Applications
article

Stress–strength reliability under the generalized exponential model: General formula, censored-data inference, and engineering application

M. E. Abd El-Monsef, Yakup Yıldırım, Ghareeb A. Marei, Hamdy M. Ahmed, M. Elsaid Ramadan, Taha Radwan, N.M. Kilany
article en

Abstract

We study the stress–strength reliability 𝑅 = 𝑃 ⁡ ( 𝑋 2 < 𝑋 1 ) when strength 𝑋 1 and stress 𝑋 2 are independent members of O’Brien’s general exponential family, which contains the exponential, Rayleigh, gamma, Weibull, Maxwell, chi-square, Nakagami, Erlang, generalized-gamma and transformed-gamma laws. For a common shape parameter, 𝑅 is obtained in closed form as a Gauss hypergeometric function and, equivalently, as the regularized incomplete beta function 𝐼 𝛽 2 / ( 𝛽 1 + 𝛽 2 ) ⁡ ( 𝜈 2 , 𝜈 1 ) ; the proof is valid for every rate ratio. For unequal shapes we give a convergent series with its exact domain of convergence. We locate the series’ largest term and the number of digits lost to cancellation, which yields an explicit switching rule between the closed form, the series and a one-dimensional quadrature. We prove that the family converges in total variation to a power-function law as 𝑛 → ∞ . This gives closed-form lower bounds for the profile likelihood and a diagnostic that detects, before any optimization, data sets on which the unrestricted model has no interior maximum. Maximum likelihood estimation with Fisher-information asymptotics, logit-transformed intervals, Bayesian MCMC and a Lindley approximation are developed. A Monte Carlo study (1000 replications per cell) reports bias, MSE, interval length and coverage. In an application to aircraft windshields, with lifetime as strength and service time as demand, the diagnostic shows that the failure times alone admit no interior maximum, an artifact of ignoring the censored units. When censoring is taken into account, the unrestricted model is identified for the lifetimes and is the only candidate passing all goodness-of-fit tests for the service times. Maximum likelihood, Bayesian, bootstrap and Lindley estimates ( 0 . 6 9 9 - - 0 . 7 0 6 ) agree with a model-free Kaplan–Meier estimate ( 0 . 7 0 6 ), whereas ignoring censoring would understate 𝑅 by about two standard errors. The fitted model is then used for engineering decisions: it gives B-lives with bootstrap intervals, reveals a wear-out hazard that rises faster than the Weibull model predicts, yields optimal age-replacement intervals that cut long-run maintenance cost by 2 2 - - 5 4 % relative to run-to-failure, and quantifies the lifetime improvement ( 2 6 % and 8 4 % ) required to raise 𝑅 to 0 . 8 and 0 . 9 .

Ain Shams Engineering JournalVol. 18(1)
Qassim University (SA), Tanta University (EG), El Shorouk Academy (EG), Biruni University (TR), Higher Institute of Engineering (EG), Islamic University of Madinah (SA), Egyptian Chinese University (EG), Menoufia University (EG), Near East University (CY)
Qassim University
Statistical Distribution Estimation and Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon