A Bayesian tolerance interval estimation method for fatigue strength substantiation of rotorcraft dynamic components

• Benchmarking of tolerance interval estimators for very small sample sizes.
• Demonstration of Bayesian statistics for lognormal tolerance interval estimation.
• Bayesian analysis allows to take experience and prior design data into account.
• Realistic reduction factors with explicit and traceable analysis.
• Analysis justifies default 1/3 reduction factor for rotorcraft working curves.

Rotorcraft have many safety-critical components that are subject to repetitive loading and whose predicted fatigue life is substantiated by a statistically derived reduction factor for a working S-N curve. For cases where the statistical analysis must be performed based on very few samples, or even none at all, commonly used frequentist statistics are demonstrated to be inadequate. As an alternative, the value of Bayesian statistics is exemplified using real data and a simplified fatigue life substantiation model from industry. The case study exemplifies how well-established industry practise can be substantiated by traceable and explicit statistical analysis. It is thus exemplified how Bayesian statistics can reduce the number of critical engineering assumptions in fatigue life substantiation, or make them more explicit and traceable.