Reliability Prediction for Mechanical Systems.

Abstract

Two areas were investigated. A method for producing exact confidence bounds for the reliability of highly reliable series systems. The methods are based on finding suitable orderings of the sample outcome space, and are valid when the probabilities of component failure are small enough for the failure data to be distributed as Poisson random variables. Optimality criteria are invoked to insure that the orderings used produce confidence bounds which are as tight as possible. Tables to facilitate applications were produced. A new class of robust estimators of location were studied. These estimators (called P-estimators) are analogous to Pitman estimators in the same way that M-estimators are analogous to maximum likelihood estimators. Members of this class were evaluated by computer simulation and were found to perform with even higher efficiency than the bisquare M-estimator for both long and short tailed distributions. (Author)

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jul 11, 1979
Accession Number
ADA071835

Entities

People

  • M. Vernon Johns Jr

Organizations

  • Stanford University

Tags

Communities of Interest

  • Human Systems

DTIC Thesaurus Topics

  • Applied Mathematics
  • Computations
  • Computer Simulations
  • Confidence Limits
  • Data Science
  • Estimators
  • Information Science
  • Military Research
  • Normal Distribution
  • Probability
  • Random Variables
  • Reliability
  • Scientists
  • Security
  • Simulations
  • Statistical Estimation
  • Statistics

Readers

  • Statistical inference.
  • Systems Analysis and Design

Technology Areas

  • Space