Online Program

Parametric Bootstrap for Design of Experiment for Early-Life Reliability Screening of Electronics (304181)

Keywords: survival, bootstrap, reliability, design of experiment

We present a rigorous and practical methodology for design of experiment for early-life product reliability evaluations using the approach of parametric bootstrap Monte Carlo simulation. With intrinsically stochastic datasets such as product early-life reliability failure times used in determining optimal burn-in durations, the variation due to finite sampling can overwhelm other sources of variation such as wafer-to-wafer or production run-to-run differences. By inverting the cumulative distribution function of the lifetime distribution, failure times can be sampled parametrically and entire evaluations can be replicated to provide upfront understanding of confidence intervals relative to meaningful tolerances required for engineering decision making as well as providing data review teams with a deeper qualitative understanding of sample size effects on survival curves.