A Stochastic Model of Reliability Growth.

Abstract

A stochastic reliability growth model is developed for one-shot devices which assumes reliability improvement occurs only after failures occur and the amount of improvement is proportional to the unreliability. The basic model is given by P(sub k + 1) = P sub k + C(1 - P sub k) where P sub k is the reliability at the kth failure and C is a learning constant. Given an initial reliability P sub 0 and the learning constant C, the expected reliability after n trials is calculated by Markov chain methods. Combining a discrete prior distribution of the initial reliability and learning constant with test data using Bayes Theorem, a posterior distribution is developed which then is used to calculate the expected reliability growth curve. The model was applied to test data from three missile development programs. The results compared favorably with the AMSAA model currently in use and showed more logical initial growth. The model does not appear to be unusually sensitive to priors or size of input matrices. (Author)

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Feb 18, 1980
Accession Number
ADA096508

Entities

People

  • Darrell D. Penrod

Organizations

  • Auburn University

Tags

Communities of Interest

  • Weapons Technologies

DTIC Thesaurus Topics

  • Bayes Theorem
  • Bayesian Networks
  • Classification
  • Computers
  • Digital Computers
  • Engineering
  • Equations
  • Failure Mode And Effect Analysis
  • Learning
  • Markov Chains
  • Mathematical Models
  • Mechanical Engineering
  • Models
  • Probabilistic Models
  • Probability
  • Probability Distributions
  • Reliability

Fields of Study

  • Engineering

Readers

  • Life Cycle Cost Analysis
  • Neural Network Machine Learning.
  • Statistical inference.