Treatment of Confidence levels When Allocating System Reliability Requirements 2005-01-1776
Reliability allocation of system objectives for Reliability validation purposes must account for Confidence levels. Misallocating Confidence levels can lead to unrealistic and unmanageable objectives, resulting in increased development times and associated costs. Therefore, it is necessary to correctly model both Reliability and Confidence levels.
Unfortunately, modeling for anything more complex than the simplest pass/fail test criteria can become quite complex in a multi-component system. The easiest case to model is time-censored testing with no failures. But time-censored testing with no failures is just a small subset of all viable validation strategies. Given that the validation strategy for each component can be different, trying to isolate a single one-size-fits-all model is extremely difficult. For these complex scenarios, computer simulation provides the best approach to calculating true system performance.
Rather than allocating a Reliability and confidence target for subsystem components, the authors recommend setting appropriate test objectives based on the system requirements. Perceptions about Confidence levels can be highly misleading. Instead of focusing on “making the numbers”, there should be a cohesive validation strategy that provides true insight into system Reliability, but within the context of practical “real-world” limitations.