Parameter Identification of a Model Scale Ship Drive Train

Simulation models of the ship propulsion system play an increasingly important role, for instance in controller design and condition monitoring. However, creation of such simulation models requires significant time and effort. In this paper, the application of deterministic identification techniques on a DC-electric ship drive train is explored as an alternative for data-driven identification techniques that require extensive measured data sets collected over long periods of ship operation. First, a nonlinear and a linear simulation model that represent the dynamic behavior of the propulsion plant are developed, and the main parameters to be identified are defined. Then, a set of experiments on a model scale boat in the bollard pull condition are conducted using an ad hoc experimental setup and data acquisition system. Subsequently, various types of identification techniques are applied, aiming to determine the unknown model parameters. Eventually, a comparison is made between experimental and simulated results, using the different sets of the estimated parameters. The value of the demonstrated approaches lies in the fast determination of unknown system parameters. These parameters can be used in simulation models, which in turn can be used for various purposes such as system controller development and tuning. Furthermore, periodic determination of system parameters can support condition monitoring to detect faults or degradation of the system. The latter point directly deals with the condition-based maintenance issue; in fact, monitoring the propulsion plant parameters over time could allow for better management (and timing) of maintenance. Although the developed ideas are far from ready to be used on the full-scale, the authors believe that the methodologies are promising enough to be developed further towards a full-scale application.

Automated summary

This paper explores the application of deterministic identification techniques in creating simulation models for ship propulsion systems, as an alternative to data-driven techniques requiring extensive time and effort. By conducting experiments and comparing experimental and simulated results, the value and potential of this approach are demonstrated.