Study On Extreme Short-term Motion Prediction Of Floaters

In complex sea conditions,the floaters on the sea will be produced a variety of surge,heave and roll motion,especially in the three degrees of freedom of heave,roll and pitch.Vigorous motion may bring a great impact to sea crew and work.Staff will feel uncomfortable,the efficiency of engineering operations will be reduced,some losses may even caused by the collision.Therefore,it is an important research topic to forecast the short-term motion of folaters in advance.Early control of the floaters attitude can avoid a lot of accidents and losses if based on the forecast results.In view of the above situation,this paper has carried out the research on the extrem short-term forecast method of sea surface buoy based on time series analysis.In this paper,three independent methods and two combined methods were used to analysis the motion data of different characteristics.First of all,according to the linear stationary motion data,the model is determined according to the correlation function method,the ARMA model is established,the order of the model is determined according to the information criterion of AIC and BIC respectively,and finally the rewults are calculated by least squares method.Then,the support vector machine regression(SVR)model is used to deal with the non-linear stationary motion data,and the selection of kernel function,kernel parameter and penalty factor parameter are discussed.Then,based on the Kalman filter theory,a Kalman predictor with wave hight input formula is derived,which is used to deal with the linear non-stationary motion data.Finally,the ARMA-EMD-SVR combination method based on empirical mode decomposition and the wavelet-SVR combination method based on wavelet decomposition are respectively used for nonlinear non-stationary motion data.And the boundary effect of empirical mode decomposition is discussed.On the basis of theoretical analysis,the heave,roll and pitch data of a drilling vessel and a semi-submersible platform in a basin model test are respectively predicted and analyzed by different methods,which verify that the above methods are accurate and effective.At the same time,it is found through the results that different methods of calculation and forecasting accuracy will be different for different characteristics of the data.For the data that the trend is not obvious and with strong non-stationary non-linearity,the two combined methods are more suitable than some single method.However,the computational efficiency of independent methods is higher than the combined methods,so in practice,the data features should be recognized to distinguish the method used.The research results and conclusions in this paper are of great significance to engineering application of extreme short-term motion prediction of floaters.