Full-Scale Identification of Baseline and Extended Voith–Schneider Propulsion Models from Sparse Data

This paper describes the identification of low-order input–output models of a Voith–Schneider propulsion system based on a very limited set of full-scale data from a minehunter vessel. An analysis was conducted based on steady operating points at which data on the shaft speed, ship speed, hourly fuel consumption, and longitudinal blade setting angle were recorded. Using the measured values for the blade setting angle β and a monotonic mapping, an equivalent pitch ratio, h = (H/D)_eq, was determined and introduced as a second coordinate of the propulsion state. A baseline model that depended solely on shaft speed was compared with an extended model that also took into account the equivalent pitch ratio. Both models were zero-dimensional input–output models and described the relationship between the propulsion setting and the vessel’s operational response. The quality of the model was assessed using leave-one-out cross-validation, prediction error metrics, and a resampling-based uncertainty analysis. The results showed that the baseline model had predictive usefulness for both ship speed and fuel consumption, whereas the extended model yielded more significant improvements in the representation of fuel consumption than for ship speed. These findings indicate that the equivalent pitch ratio is a physically meaningful model variable, although its predictive value depends on how clearly the effects of shaft speed and blade setting are separated in the available data. The paper also provides practical guidance for planning future research.