Research Of A Heave Compensation Predictive Model Algorithm For An Offshore Crane

Influenced by the various bad condition such as varying wind and waves whichinevitably causes significant translational displacement, swing, and heave of the ship crane,which may reduce the work efficiency and even threaten the security of the production system,the equipment and the staff. So far, many marine work ship have been equipped with dynamicpositioning system which make the ship sway, surge under control in certain extent. Howerer,the heave motion in the vertical direction is still very difficult to control. In China, althoughextensive research in the field of heave compensation system has been carried out, the relatedresearch starts late and is not significant. To compensate the system dead-time, crane heavemovement is predicted in advance to eliminate the delay phenomenon in this paper. Based onhistorical data of the crane heave motion, forecasting the future motion in a very short time sothat the crane heave motion can be compensated in advance can not only solve the delayphenomenon during heave compensation process, but also improve crane safety andefficiency in maritime operations, thus ensuring crane safety at sea in real-time to completethe task. Therefore, the study of the movement of short-term forecasting of the crane heavehas great theoretical and practical value.The study focuses on crane heave motion prediction using time series analysis, Elmanneural network and SVR prediction algorithm, in which particle swarm is used to prevent theblindness of SVR parameter optimizaion. With the algorithms mentioned above, anexperiment of ship crane heave motion was carried out. And analysing the result, we canconclude as: AR prediction model of time series have a better accuracy for the prediction oflinear heave motion, but hysteresis phenomenon exists with the increase in t he number of thesteps, and it has poor prediction for non-linear heave motion; The Elman neural network andPSO-SVR prediction algorithm can help to acheive satisfactory prediction of the crane heavemovement. Because the heave crane movement is he non-linear and non-stationary caused bythe random waves, in this paper wavelet multi-resolution analysis is introduced to optimizethe the Elman and PSO-SVR algorithms, and simulations are included to illustrate the higherprediction accuracy and stability of the improved prediction algorithm, comparing with the PSO-SVR and Elman neural network prediction model, which provide theory foundation forthe critical issues in the follwing heave motion compensation.