Proceedings of the Eighth International Conference on Probabilistic Safety Assessment & Management (PSAM)
Modeling Tools and Techniques: Bayesian Methods for Reliability Assessment I
Download citation file:
- Ris (Zotero)
- Reference Manager
From the inception of the National Aeronautics and Space Administration (NASA) Space Shuttle Main Engine (SSME) program, safety and reliability have been its highest priorities. Through the years, there has been an open question of how to quantitatively evaluate the risks and reliability of a rocket engine that is constantly being improved and also has a limited number of hot-fire tests. This paper presents an alternative approach for estimating rocket engine risks. It also identifies the major assumptions and describes the limitations of this approach.
With over a million seconds of hot-fire tests accumulated, the number of equivalent SSME missions is still considered relatively small from the standpoint of classical statistics. Over the years, a number of methods have been developed to overcome this weakness. These methods are aimed at using all available information applicable to the SSME safety/reliability estimates rather than relying exclusively on the limited number of hot-fire tests. One of the latest methods is the SSME failure mode level Probabilistic Risk Assessment (PRA) to support the 2002 Space Shuttle PRA effort. Recently, the Space Shuttle PRA team reexamined the SSME models and decided to develop an alternative approach for estimating the SSME catastrophic and benign shutdown risks.
The new approach uses a multi-step Bayesian-updating technique and incorporates failure discounting guidelines that were used successfully on jet engines. The failure discounting process starts with a review of failures, types of corrective actions taken, and operating time since implementation of each corrective action. This information is then used to estimate a range of the failure fraction that is applied to the appropriate failure for discounting purposes. The analysis is performed at the engine level, and the level of detail is consistent with the available engine-level test data. It is believed that the method presented in this paper offers more realistic estimates of the engine-level risks.
The engine-level analysis encompasses the complexity of a rocket engine and the numerous interactions among the engine components. This method avoids quantifying unknowns at the engine subcomponent level with no existing hard failure data. As a result, the subcomponent-level method may have a tendency to under-estimate the risk. However, the benefit of this approach is the possibility of a meaningful comparison of analytical results with the current engine level demonstrated reliability.