A critical analysis of the six-nines reliability methodology

The current certification requirements of helicopter dynamic components demand that the probability of failure at the retirement life does not exceed one in a million, i.e. a reliability level of 0.999999 (six-nines reliability). It is therefore imperative that the component retirement time for fatigue-critical rotorcraft dynamic components be determined using probabilistic rather than deterministic methodologies. In the helicopter industry a method known as `Top of Scatter-mean minus three standard deviation (TOS/(µ - 3)) reliability method is commonly used for determining the component retirement time over which the probability of failure is 10-6. In this paper, the results of the TOS/(µ - 3) reliability method for an example of constant amplitude loading have been compared with those of an exact analytical solution, Monte Carlo simulation and Second-Moment methods. Examples using Monte Carlo simulation and a numerical approximation are evaluated for variable amplitude loading, where Miner's linear damage summation rule is assumed. Finally, this paper shows that the TOS/(µ - 3) reliability method is non-conservative for both constant and variable amplitude loading.