No abstract

A probabilistic assessment for an integrally bladed disk design is a system reliability problem where a failure in at least one blade constitutes a system failure. Turbine engine fan and compressor blade life is dominated by High Cycle Fatigue (HCF) which is initiated either by pure HCF or Foreign Object Damage (FOD). The system model includes frequency prediction, model stress variation, mistuning amplification, FOD effect, and random material capability. Although the analysis time for these submodels would have ruled out any sampling method as recently as two years ago, novel reduced order models have made this process tractable. A large system problem has more random variables than some probability integration methods can handle; this system has nearly 700 variables, 29 for each of the 24 blades. Sampling methods, such as the Latin Hypercube Sampling or Monte Carlo used in this analysis, can handle such large numbers of variables easily. Additionally, both methods construct a confidence interval around the most likely failure rate estimate. The results converge to the same statistic with equal confidence intervals for all analysis methods.

A new initiative was started to address Department of Defense safety and affordable readiness for legacy turbine engines. By using the GOTChA/ApPRoVal methods, a baseline research plan was established that is based on prognostics and health management. This paper outlines the decision process, investment strategy, projected return on investment, validation, and transition strategies. .

Probabilistic failure assessments for integrally bladed disks are system reliability problems where a failure in at least one blade constitutes a rotor system failure. Turbine engine fan and compressor blade life is dominated by High Cycle Fatigue (HCF) initiated either by pure HCF or Foreign Object Damage (FOD). To date performing an HCF life assessment for the entire rotor system has been too costly in analysis time to be practical. Although the substantial run-time has previously precluded a full-rotor probabilistic analysis, reduced order models make this process tractable as demonstrated in this work. The system model includes frequency prediction, modal stress variation, mistuning amplification, FOD effect, and random material capability. The model has many random variables which are most easily handled through simple random sampling.

No abstract