Date of Graduation

5-2016

Document Type

Thesis

Degree Name

Master of Science in Industrial Engineering (MSIE)

Degree Level

Graduate

Department

Industrial Engineering

Advisor/Mentor

Ed Pohl

Committee Member

Richard Cassady

Second Committee Member

Kelly Sullivan

Keywords

Applied sciences, Accelerated life testing, Reliability, Weibull

Abstract

In designing accelerated testing plans, cost is a factor that is missing in much of the literature. This paper explicits considers cost by developing an optimization model with the objective to minimize costs for a simple step stress accelerated test plan. Two methodologies are employed. One is an optimization approach in which an attempt is made to quantify the behavior of a series-parallel hardware system over all stages of testing using a response surface, and then an optimization model is used to determine the settings for stresses and failure mode modifications for all stages of testing prior to the start of testing. The second methodology or sequential stage approach is to generate a response surface using data from a completed test stage to determine the settings of stresses and failure mode modifications for the next stage. Then this process is repeated for all stages of testing.

When validating the results of the optimization model through simulation, the model overestimated costs. Assuming the simulated optimal settings are the true value of cost, the sequential approach produced suboptimal results. This is because each stage of testing results in narrowing the search parameters of a solution. However, it was found that the sequential stage approach had similar costs to that of the optimization model. Although the optimization model has a better solution, it requires much more data initially whereas the sequential stage approach does not require information about the system for all stages prior to testing. If information of the system's behavior is known for all stages prior to testing, then the optimization approach is more advantageous, yet most cases have limited information so the sequential stage approach should be utilized.

Share

COinS