Scale effects in AR model real-time ship motion prediction

Abstract

Real-time prediction of ship motion is essential for decision making in shipborne maritime operations. Differences in ship hulls render different ship motion characteristics, which consequently affects the performance of real-time prediction models. In this study, the ship hull scale effects in real-time motion prediction are investigated using the AR model. The ship datasets are generated by applying the strip theory. These ship motions datasets with various spectral characteristics are used in real-time prediction simulations. This study explores how the spectrum bandwidth, peak frequency, and ship hull scale influence prediction performance, and conclusions are drawn based on numerical simulation results. Prediction accuracy shows a negative relation to spectrum bandwidth and peak frequency. The AR model performance is better for ships with larger principal dimensions where ship hulls are the same. A preliminary empirical formulation for evaluating the maximum predictable time duration is developed based on the above regularities.