Predicted vs Test MTBF's ... Why the Disparity?

Over the years there has been much concern and skepticism over the accuracy and usefulness of reliability predictions. This concern stems from the large variations often experienced between predicted and test observed Mean-Time-Between-Failure (MTBF's). At the Westinghouse Defense and Electronic Center a case study on an Electric Countermeasures (ECM) radar system was performed to evaluate the reasons for this difference. The problems experienced in test were analyzed in light of the prediction models of MIL-TIDBK-217, and it was concluded that certain assumptions inherent in these prediction models were largely responsible for this difference. These assumptions are related, but not limited to such factors as poor design and/or construction techniques, and poor quality control in the fabrication of the devices. As a result of this study a program has been instituted at the Westinghouse Defense and Electronic Center to identify and control these factors and it is identified in this paper.