Reliability Modeling, Testing and Optimization of Systems with Inhomogeneous One-shot Devices

Abstract:

One-shot devices can only perform desired functionality once and typically experience long-term storage or standby status until needed for operational use. During the long storage period, one-shot devices are subjected to failure and/or degradation which inhibit devices’ operational performance with a variety of failure mechanisms. Therefore, it is critical to assess the one-shot devices’ storage reliability metrics as a function of time such as to guarantee their operational performance. In this presentation, we propose that performing repeated reliability tests randomly across one-shot devices’ entire life horizon is an effective approach to learn about their storage and operational reliability metrics. With the continuously produced one-shot devices and the repeatedly performed tests, the one-shot devices in storage have inhomogeneous and time-dependent characteristics.

Referring to the stored one-shot devices as a fault-tolerant system, we investigate the reliability modeling, testing and optimization of the system. Specifically, we consider different failure mechanisms and develop efficient and effective analytical models to assess the reliability metrics of the system. We then investigate the details of performing the reliability tests under different scenarios and environments, where optimal sequential testing plans are designed for a variety of types of reliability tests. Simulation studies validate that well-designed sequential testing plans provide accurate system reliability metrics estimation within short testing duration. At arbitrary time, a proportion of the one-shot devices may be retrieved from storage for operational use; we stochastically optimize the operational use of the stored one-shot devices by determining the retrieved devices’ characteristics and their launching procedure.