|
Activity Number:
|
707
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 4, 2016 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
|
Abstract #318573
|
View Presentation
|
|
Title:
|
Bayesian Signal-Response Data Nondestructive Inspection Test Planning
|
|
Author(s):
|
Yew-Meng Koh* and William Q. Meeker
|
|
Companies:
|
Hope College and Iowa State University
|
|
Keywords:
|
Linear regression ;
Nondestructive evaluation ;
Probability of detection ;
Optimum test plan ;
Compromise test plan
|
|
Abstract:
|
The most common question asked of a statistician is "How large should my sample be?" In Nondestructive Evaluation applications, the most common questions asked of a statistician are "How many specimens do I need and what should be the distribution of flaw sizes?" Although some useful general guidelines exist (e.g., in MIL-HDBK-1823) it is possible to use statistical tools to provide more definitive guidelines and to allow comparison among different proposed study plans. The Bayesian methods used in this paper allow for the specification of needed planning information into the design of a study. One can assess the performance of a proposed Probability of Detection (POD) study plan by obtaining computable expressions for estimation precision. This allows for a quick and easy assessment of tradeoffs and comparison of various alternative plans. We use a signal-response dataset obtained from MIL-HDBK-1823 to illustrate the methods.
|
Authors who are presenting talks have a * after their name.
