Proceedings of the Eighth International Conference on Probabilistic Safety Assessment & Management (PSAM)
38 Probabilistic Seal Leak Analysis for Space Shuttle External Tank (PSAM-0192)
Download citation file:
- Ris (Zotero)
- Reference Manager
The Shuttle's External Tank (ET) leak analysis described in this paper was performed by the authors as members of NASA's Shuttle Probabilistic Risk Assessment Team (SPRAT) . The method for predicting catastrophic consequences due to leakage of liquid hydrogen (LH2) or liquid oxygen (LO2) from the ET's flange seals and instrument and electrical feed-through penetrations requires estimates of leak frequency, leak size, and leak tolerance. In general, the combined effects of these variables can be modeled using a combination of probabilistic and deterministic physics-based simulations, empirical data analysis, or expert elicitation depending on the availability of time, data, and resources. Process engineers at Michoud Assembly Facility (MAF) provided deterministic estimates of key parameters that effect leak tolerance for candidate seal locations in the Shuttle's wake area. The MAF study was a screening analysis to identify sites having the potential for a fire/explosion due to leaks. Using phenomenological models, they determined whether a leak site under worst-case assumptions had the potential to reach threshold fluid concentrations in the wake area sufficient to cause a fire/explosion. Seven sites were identified. Five associated with LO2 and two with LH2.
Mission leak frequency at the system level was estimated from leak data contained in the National Aeronautics and Space Administration's (NASA's) Problem Reporting and Corrective Action (PRACA) Database from January 1993 through December 2002. These data were entered into a Fault Risk Analysis Spreadsheet (FRAS) , which applies discount factors for probability of detection and effectiveness of corrective action evaluated for each PRACA leak record. The system-level leak frequency was then allocated to individual leak sites.
The leak size distributions for the various sites were developed from historical telemetry leak data from the Orbiter Main Propulsion System (MPS) 17-inch disconnect between the ET and the Orbiter MPS. The medians of these datasets were scaled based on the ratio of the maximum leak rates for the seal and the 17-inch disconnect. The leak rate distribution was modeled as lognormal based on statistical goodness of fit tests. The dispersion of the leak-rate distribution was specified by equating an upper-tail percentile to the maximum leak rate distribution. The upper-tail percentile was determined by judgment through sensitivity analysis. A Monte Carlo simulation was created to model the uncertainty associated with several key parameters and to combine the mission leak frequency with the probability that a leak is critical to obtain a distribution for the overall probability of a critical leak. Within the Monte Carlo model, sampled data from three parameter distributions were used to generate leak size distributions and calculate the probability of a lognormal leak exceeding the critical leak rate. The probability of a critical leak at each site is the product of the mission leak frequency multiplied by the conditional probability that the leak is critical. The outcome from 20,000 trials of the simulation provided the data to generate empirical distributions.