A multivariate stochastic model of sea-state parameters, such as wave height, wave period, wave direction, wind velocity and wind direction, is proposed and its simulation described. The stochastic model is described by a stochastic differential equation with some of its parameters based on a Markov process. In this model sea-state parameters that correlate to each other are transformed into independent components by eigenvalue decomposition. The model parameters are estimated from historical hindcast data. The validity of this model is tested by compareson of stochastic simulations and historical hindcast data.
A fundamental statistical model for long-term prediction of ship performances was proposed by Fukuda (1969) and Nordenstr ¨om (1973). Since then, more accurate models involving effects of ship speed reduction (1974, 1998), wave directionality (1995) and ship operation (2003) have appeared. Wave statistic tables, obtained from statistical analyses of observations and historical hindacst data, are used for sea-state information in these models. The confidence of prediction depends on the number of wave statistics available. For extrapolation beyond this confidence term, a mathematical probability density function (PDF) has been fitted to the wave statistic table. Several joint PDFs relating significant wave height and zero-upcrossing period were introduced by Mathiesen and Bitner-Gregersen (1990). The mathematical PDF of sea state enables evaluation of ship performance in a reasonably long term, e.g. ship life.
On the other hand, to evaluate seakeeping performances for several years more closely and in greater detail Dallinga et al. (2004) proposed and developed an oceangoing simulation method of ship responses in the time domain. This simulation requires long-term time series data of sea states having a period of several years with a time increment of several hours. These data are generally provided by the historical hindcast of sea states.
