Comparing Reliability Predictions to Field Data for Plastic Parts in a Military, Airborne Environment

This paper examines two popular prediction methods and compares the results to field data collected on Plastic Encapsulated Microcircuits (PEMs) operating in a military, airborne environment. The comparison study focused on three digital Circuit Card Assemblies (CCAs) designed primarily with plastic, surface mount parts. Predictions were completed using MIL-HDBK217 models and PRISM, the latest software tool developed by the Reliability Analysis Center (RAC). The MIL-HDBK-217 predictions which correlated best to the field data were based on quality levels (πQ) of 2 and 3, rather than the typical πQ values of 5 or higher, traditionally assigned per the handbook's screening classifications for commercial, plastic parts. The initial findings from the PRISM tool revealed the predictions were optimistic in comparison to the observed field performance, meaning the predictions yielded higher Mean Time To Failure (MTTF) values than demonstrated. Further evaluation of the PRISM models showed how modifying default values could improve the prediction accuracy. The impact of the System Level Multiplier was also determined to be a major contributor to the difference between PRISM predictions and field data. Finally, experience data proved valuable in refining the prediction results. The findings from this study provide justification to modify specific modeling factors to improve the predictions for PEMs, and also serve as a baseline to evaluate future alternative prediction methods.