A Probabilistic Return-to-Flight Assessment of the Orbiter's LH 2 Feedline Flowliner

In May of 2002, three cracks were discovered in the flowliner of one of the orbiters near the interface with the low-pressure turbo pump. Cracking was identified as high-cycle fatigue due to flow-induced vibrations produced by the turbo pumps. Initial deterministic fracture mechanics analyses suggested that the fatigue cracks might lead to failure in a single flight. This result was believed to be conservative because of the inherent uncertainty in much of the input, which led to multiple worst-case assumptions in the deterministic analysis. The goal of the present work was to perform a probabilistic fracture mechanics analysis that explicitly accounted for the uncertainty in the input variables, rather than use the worst-case assumptions employed in prior deterministic analyses. Statistical models based on available data for key input variables to the analysis were developed to describe the uncertainty in dynamic loading and fatigue crack growth properties at cryogenic temperatures. New weight function solutions were also developed to more accurately describe the stress intensity factors driving the crack. Computational simulations using the Monte Carlo technique were used to establish the flowliner probability of failure (POF) per flight. A range of probability-of-detection (POD) curves was assumed for inspection of the flowliners since data were not available on this input variable at the time of the analysis. Results showed that relatively low POFs in the flowliner (0.001 to 0.0001) could be achieved provided cracks are detected using pre-flight inspections having adequate probability of detection (20 mil cracks with 50% POD, and 75 mil cracks with 99% POD). Thus, follow-up measurements to confirm the POD curves assumed in the analysis for pre-flight inspection are crucial. A flight rationale was recommended based on comparing the computed flowliner POFs with previously estimated POFs for engine failure and overall mission failure. A probabilistic sensitivity analysis showed that uncertainties in predicted flowliner fatigue failures are predominantly controlled by uncertainties in the values of the mean and standard deviation of the dynamic stress amplitudes. Based on these results recommendations are also provided on how best to increase the reliability of the orbiters LH2 feedline flowliner.