Analyzing ALT Data with Time-varying Stress Profile

Research output: Chapter in Book/Report/Conference proceedingConference contribution


We are interested in analyzing the accelerated life testing data obtained under time-varying stresses. This is a generalization of step-stress accelerated life test, where the stress levels are kept constant at each step. Our study shows that the time-dependent proportional hazard model, commonly appeared in the survival data analysis literature, is not applicable for ALTs, because this approach does not take accounts of the cumulative damage that the stress profile exerts on test units. Instead, we assume that products possess constant failure rates over small time intervals, and the change of failure rate by the stress variable still have the proportional hazard property. With these assumptions, it is possible to formulate the data according to a generalized linear model and statistical inferences on model parameters can be carried out. We demonstrate our models and inference procedures by using both synthetic and real datasets.

Original languageEnglish (US)
Title of host publication68th Annual Reliability and Maintainability Symposium, RAMS 2022
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781665424325
StatePublished - 2022
Externally publishedYes
Event68th Annual Reliability and Maintainability Symposium, RAMS 2022 - Tucson, United States
Duration: Jan 24 2022Jan 27 2022

Publication series

NameProceedings - Annual Reliability and Maintainability Symposium
ISSN (Print)0149-144X


Conference68th Annual Reliability and Maintainability Symposium, RAMS 2022
Country/TerritoryUnited States


  • accelerated life test
  • Poisson regression
  • proportional hazard
  • reliability prediction
  • stress profile

ASJC Scopus subject areas

  • Safety, Risk, Reliability and Quality
  • General Mathematics
  • Computer Science Applications


Dive into the research topics of 'Analyzing ALT Data with Time-varying Stress Profile'. Together they form a unique fingerprint.

Cite this