Method for estimating parameters of practical ship manoeuvring models based on the combination of RANSE computations and System Identification

In this work a method for estimating parameters of practical ship manoeuvring models based on the combination of RANSE computations and System Identification procedure is investigated, considering as test case a rather slender twin screw and two rudders ship. The approach consists in the estimation of the hydrodynamic coefficients applying System Identification to a set of free running manoeuvres obtained from an in-house unsteady RANS equations solver, which substitute the usually adopted experimental tests at model or full scale. In this alternative procedure the numerical quasi-trials (in terms of kinematic parameters time histories and, if needed, forces time histories) are used as input for the System Identification procedure; the aim of this approach is to reduce external disturbances that, if not properly considered in the mathematical model, may compromise the identification results, or at least amplify the well-known “cancellation effects”. Furthermore, the CFD results provide information both in terms of flow field variables and hydrodynamic forces on the manoeuvring ship. These data may be adopted for a better understanding of the complex flow during manoeuvres, especially at stern, providing also additional information about the interaction between the various appendages (including rudders) and the hull. The identification procedure is based on an off-line genetic algorithm used for minimizing the discrepancy between the reference manoeuvres from CFD and those simulated with the system based modular model. The discrepancy was measured considering different metric functions and simplified formulations which consider only the main macroscopic parameters of the manoeuvre; the metrics have been analyzed in terms of their capability in reproducing the time histories and in limiting the cancellation effect of the hydrodynamic derivatives.