Evaluating Confidence Interval of Fatigue Damage from One Single Measured Non-stationary Time-History

This paper derives a confidence interval for the expected damage when only one single non-stationary time-history record is available. The proposed confidence interval is inspired on a solution existing in literature, which is restricted to stationary time-history. The proposed procedure divides the non-stationary time-history into stationary segments, and each segment is further partitioned into blocks. This technique allows the confidence interval for expected damage to be constructed. As a case study to check the proposed confidence interval, measured time-histories are obtained directly from a mountain-bike travelling on an off-road track with different speeds, seat and surface conditions over time. All measured time-histories are verified to be non-stationary by qualitative and quantitative approaches, as, for example, the ‘run test’ based on a sequence of damage values. Using the measured time-histories, a sort of ‘calibrator’ sample damage value is computed to estimate the expected damage, and then used to verify the validity of the proposed confidence interval. The obtained results confirm the correctness of the proposed confidence interval of fatigue damage from one single measured non-stationary time-history.