SAE AIR5006/2

Description / Abstract: The authors of this guidebook and the solid propulsion industry at large have long been troubled by an inability to absolutely "demonstrate" SRB reliability because, under classic reliability theory, large numbers of tests are required to establish "statistically significant" reliability values. 240 tests with no failures, for instance, are required to demonstrate 0.99 reliability to a 90% confidence level if a binomial model is used. Since the testing cost for hundreds of large motors like the Space Shuttle SRBs would be prohibitive, the solids industry has always been confronted with a vexing dilemma: (1) what is the fewest number of SRBs that can be tested to establish reliability with high confidence level for commitment to flight and (2) what should the reliability that was established with fewer-than-classical numbers of tests be called (it obviously is not "demonstrated reliability" in the strictest sense of the definition)?

Allocation, Prediction, Verification, and Demonstration of Reliability:

To help resolve this dilemma, the authors of this guidebook suggest a new definition of the reliability demonstration process for SRBs which is divided into four steps as follows:

RELIABILITY ALLOCATION: This is the distribution or "trickling down" of the overall system reliability goals to the lowest component part of the system and the identification of a target reliability for each of these component parts.

RELIABILITY PREDICTION: This is the calculation of expected component and motor reliabilities based on specific design characteristics. This step may use data from a variety of sources including prior experience on similar systems, engineering estimates, new technologies, etc.

RELIABILITY VERIFICATION: This is the calculation of an interim reliability based on probabilistic assessment methods and a limited-but-statistically-valid combination of component, subscale, and full-scale motor tests. The scope of verification testing on the development program is influenced by any differences between the allocated and predicted reliabilities. Hardware testing usually continues until the component and system reliabilities are shown to exceed the predicted and allocated values.

RELIABILITY DEMONSTRATION: This is the final calculation of system and component reliability that is based on all uses of the system during its lifetime. SRB demonstration is always based on the actual flight or static test results. Data collected for the reliability demonstration are attribute data, namely "success" or "failure". Using these attribute data, statistical tools such as binomial distribution may be used to estimate the SRB reliability at some confidence level. Data used for SRB reliability demonstration must come from homogeneous sample data. Using this method to demonstrate SRB reliability, at high confidence level, requires large sample size attribute data which are often not available on large SRB programs. In this case, the confidence level of the calculated reliability will necessarily be lower.