Ship Motion Attitude Prediction Model Based on FMD-IBKA-BTGN

Area of Science:

• Marine Engineering
• Naval Architecture
• Oceanography

Background:

• Ship motion attitude prediction is crucial for maritime safety and operational efficiency.
• Marine environments present complex challenges like non-stationarity and stochasticity, hindering accurate predictions.

Purpose of the Study:

• To develop an advanced hybrid model for accurate ship motion attitude prediction.
• To overcome the limitations of existing methods in dynamic marine conditions.

Main Methods:

• Feature Mode Decomposition (FMD) for signal decomposition.
• Improved Black-winged Kite Algorithm (IBKA) for hyperparameter optimization.
• Bidirectional Temporal Convolutional Network with Gated Recurrent Unit (BTGN) for prediction.
• Ensemble learning combining predictions from all components.

Main Results:

• The proposed FMD-IBKA-BTGN model demonstrated superior performance compared to six other models.
• Significant reductions in prediction errors were achieved: 20.38% MAPE, 7.4% RMSE, 4.2% MAE, and 0.97% MSE versus LSTM.
• The model showed enhanced prediction accuracy and generalization capabilities in Sea State 4 conditions.

Conclusions:

• The hybrid FMD-IBKA-BTGN model offers a robust solution for ship motion attitude prediction.
• The integration of FMD, IBKA, and BTGN effectively addresses the complexities of marine environments.
• This approach enhances both the precision and reliability of ship motion predictions.