Latent Variable Models for Aging Effects on Reliability of One-Shot Devices

• Activity Number: 146
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Defense and National Security
• Author(s): David Collins*
• Keywords: Accelerated aging ; Latent variables ; Binary regression ; One-shot device ; Reliability

Abstract

A one-shot device is a system or component designed to perform a specified function exactly once. After use, the device (such as an explosive component) cannot be reused, so reliability testing is destructive, and typically expensive. Devices may need to function after storage in a stockpile, so it is necessary to predict the effect of material aging on reliability. We propose a latent variable model for estimating the margin between nominal inputs to a one-shot device and the minimum levels required for correct functioning. In conjunction with designed experiments to model aging of materials making up the device, the model estimates changes in the operating margin of the device, which predict how reliability will decline with age. This methodology reduces testing cost, since reliability changes can be predicted without testing a large number of fully assembled devices. An application example involving explosive detonators is discussed.