Browse Prior Art Database

Reliability Testing Using History and Economics

IP.com Disclosure Number: IPCOM000034738D
Original Publication Date: 1989-Apr-01
Included in the Prior Art Database: 2005-Jan-27
Document File: 2 page(s) / 13K

Publishing Venue

IBM

Related People

Breyfogle, FW: AUTHOR [+3]

Abstract

A test process is described which employs a technique for obtaining a Bayesian prior utilizing history test data from other products and a technique for utilizing economic considerations when designing a reliability certification test. Tests are often performed to "certify" that a machine/card failure-rate criterion is not exceeded. Classical statistical sample size requirements are getting larger since failure-rate criteria are decreasing with new technologies. This new process reduces test sample size requirements. Bayesian statistical techniques can be used to reduce a test sample size if there is "prior information" about the product failure rate; however, often these tests are performed on new products that have no historical data.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 53% of the total text.

Page 1 of 2

Reliability Testing Using History and Economics

A test process is described which employs a technique for obtaining a Bayesian prior utilizing history test data from other products and a technique for utilizing economic considerations when designing a reliability certification test. Tests are often performed to "certify" that a machine/card failure-rate criterion is not exceeded. Classical statistical sample size requirements are getting larger since failure-rate criteria are decreasing with new technologies. This new process reduces test sample size requirements. Bayesian statistical techniques can be used to reduce a test sample size if there is "prior information" about the product failure rate; however, often these tests are performed on new products that have no historical data. Opinion data can be used to obtain this "prior information"; however, in reality this can yield much contention. In using opinion data for input to certify a failure criterion, one test group may design a difficult test to pass, while another group's test may be "easy" to pass. Economic considerations are not usually considered when designing a test since normal test objectives are to "verify the criterion." That is, the test objective is independent upon production volumes, test costs, and repair costs. The real test objectives should be to reduce the risk of having excessive field repair costs and customer dissatisfaction. The logic of this statement can be illustrated by an example. Assume that using classical statistical sampling techniques, a failure criterion test requires a sample size of 250. Given equal product cost considerations, it does not seem logical to use this same sample size to test the failure criterion of two products if the annual production volume projections were 1000 versus 400,000. Given all other factors being equal, the test sampling effort should address the fa...