|
Activity Number:
|
70
|
|
Type:
|
Invited
|
|
Date/Time:
|
Monday, August 4, 2003 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Physical and Engineering Sciences
|
| Abstract - #301587 |
|
Title:
|
Accelerated Destructive Degradation Test Planning
|
|
Author(s):
|
William Q. Meeker*+
|
|
Companies:
|
Iowa State University
|
|
Address:
|
304C Snedecor Hall, Ames, IA, 50011-1210,
|
|
Keywords:
|
reliability ; Fisher matrix ; asymptotic variance ; simulation ; experimental design ; optimum design
|
|
Abstract:
|
Accelerated Destructive Degradation Tests (ADDT) are used to assess long term strength of materials. Samples of units are put into chambers at increased temperature. Units are removed from the chamber over time to be destructively evaluated. Then a nonlinear regression model for strength is fit to the data and used to estimate a time to first-crossing (failure) distribution. An ADDT plan requires specification of the number of units to put at each temperature and the time of evaluation for each unit. The problem is complicated because units at high levels of temperature and time can be right censored (because of serious physical deterioration) and units at low time/temperature combinations may be left censored (when the test fixture fails because its strength is less than the specimen). This talk describes methods ADDT planning. We show how to use large-sample approximations to obtain optimum test plans to provide insight for obtaining more useful compromise plans. Simulation allows fine-tuning of the plans and visualization of how the design affects statistical uncertainty. The methods are illustrated with an ADDT for an adhesive bond.
|
