Reliability predictions — more than the sum of the parts

Abstract

Reliability predictions have been the subject of much discussion over the prior 20 years. Some articles have proclaimed them to be valueless while other articles suggest importance. Spending a great amount of time calculating numbers does not present value directly. Using the numbers as the basis for additional positive activities would seem to be one reason for predictions. Any reliability prediction should be considered as a single tool in a larger reliability improvement tool box that often feeds other more important activities. This role of predictions in a larger reliability world will be explored here. Examples of follow-on improvement activities include, lessons learned about components, identification of critical components, identification of critical design features, estimation of high-stress conditions, approaches for derating, design for reliability, design for manufacture, input to an FMEA, input to a verification test plan, and warranty and repair estimates. The prediction is not an end of the process, but rather the beginning of the larger reliability improvement and design review process. Here, the value of predictions will be tied to lessons learned and outcomes. Predictions have fundamentally changed over the last 20 years for several reasons. As Failure-in-Time (FIT) numbers have declined in most handbooks, the MTBF prediction didn't always match subsequent field data on an absolute scale. It is possible to be a factor of three different or more. Each successive issue of Telcordia or the Mil Handbook 217 (now 217Plus), appears rather similar to the prior ones. This simplicity masks some of the evolution in numerical content and models. There is much to be learned from a short review of the prediction process itself. Failure rate estimates from tables are not trustworthy for they depend upon experience, customer applications, models and other unknown items. At some point it is time to wrap up the prediction phase and move onto improvement and feed other reliability tools. The ldquoLessons Learnedrdquo based upon knowledge of the design, manufacture, customer environment or are valuable. Items in lessons learned might cover a variety of situations that can enhance or detract from estimated reliability. Other lessons learned are contained in design guidelines, derating standards. All of these should be addressed early in any project, once a Bill of Materials (BOM) has been generated. Each has an impact on the prediction estimate but are not overtly included in the process.