Abstract
Abstract
Damping is critical for accurately predicting wave-induced ship vibrations. Operational modal analysis (OMA) is useful for identifying damping estimates from measured outputs. In reality, periodic and non-stationary wave excitation violates fundamental OMA assumptions which, along with noisy finite output measurements, produce bias and random errors. To address uncertainty in empirical damping estimates of the two-node vertical bending (2nVB) modal damping of a polar vessel, OMA is performed on full-scale measurements obtained during dedicated open water test sequences. The effect of user-selected block size and model order for the data-driven stochastic subspace identification algorithm is investigated. Random errors in damping estimates generally decrease with increasing ship speed possibly due to higher signal-to-noise ratios in the higher speed cases. Further, higher encountered wave lengths, which occur at low ship speeds, may increase periodic wave excitation of the vessel and requires higher model orders to eliminate bias. In general, block sizes of 40 to 80, and model orders ≥ 80 yielded the most stable estimates. Uncertainty in damping estimates are estimated using a first-order sensitivity analysis and, even for the most stable estimates, relatively large random errors are obtained. A conservative estimate for the 2nVB damping is ≤ 2%. Damping estimates are observed to decrease with increasing ship speed possibly due to hydro-elastic effects.