Intelligent prediction algorithm of ship roll and pitch motion based on SSA- optimized BiLSTM network
Abstract
In maritime activities, real-time prediction of ship movements using artificial intelligence (AI) algorithms is crucial to ensure ship safety and improve operational efficiency. Consequently, the processing of bidirectional long short-term memory (BiLSTM) networks was further optimised using a sparrow search algorithm (SSA) to enhance the accuracy of ship motion prediction. Leveraging the superior performance of BiLSTM in time-series data analysis and the efficiency of SSA in global search optimisation, the accuracy of the prediction model can be significantly improved by tuning the network parameters. The results confirmed that the BiLSTM model optimised with the SSA predicted the transverse and longitudinal rocking motions of the ship more accurately than the classical long short-term memory (LSTM) and BiLSTM models. The root mean square error (RMSE) of transverse rocking was reduced to 5.6%, 3.4%, and 4.5% at peak spectral periods of 5 s, 10 s and 15 s, respectively, and the RMSE of longitudinal rocking was reduced to 9.2%, 6.0%, and 7.9% at the peak spectral period of 5 s, 10 s and 15 s. Moreover, the prediction accuracies of the SSA-BiLSTM model reduce with an increase in the peak spectral period. Hence, the adaptability of the SSA-BiLSTM model to a complex sea state and its effectiveness in ship motion prediction was verified, providing an important reference for improving the performance of wave compensation systems.