UNBIASED ESTIMATES OF RELIABILITY WHEN TESTING AT ONLY ONE EXTREME STRESS LEVEL
Abstract
Based on the stress-strength concept of reliability for 'one-shot' items, it is assumed that an item cannot fail until the stress equals or exceeds the strength. From this premise and the following additional assumptions, methods are given for calculating unbiased estimates of nontime dependent reliability: (1) The relation between the stress and strength standard deviations are known approximately. (2) A single stress level is applied during testing at approximately three standard deviations from the average stress level. (3) The stress and strength distributions are normal. Calculations are included to show the effect of errors in the assumptions concerning the standard deviations, applied stress level, and rounding-off errors. This approach further reduces the sample size required to demonstrate high non-time dependent reliability in laboratory testing. It has the added advantage of obtaining unbiased estimates of reliability with the simplest of testing methods.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1963
- Accession Number
- AD0657676
Entities
People
- A. Bulfinch
Organizations
- Picatinny Arsenal