AN IMPROVED ENSEMBLE NEURAL NETWORKS MODEL FOR PREDICTION OF PUSHBOAT SHAFT POWER

This research intends to introduce new neural networks in the existing Ensemble Neural Networks (ENN) model and intends to build new ENN model for predicting pushboat shaft powers. As a continuation of the previous papers, research sets prerequisites for proper determination of shaft powers at a given navigating conditions. Authors made proposals that should contribute to improvement of the propulsive characteristics of the pushboats. Application of the ENN model may indirectly reduce fuel consumption and gas emissions. It is proposed that speed-power curves obtained by ENN model are compared with curves obtained by 3rd order polynomial. It is also expected that new ENN model finds new functional relationship between shaft powers and convoy speeds. Better generalization capability of the new ENN model is achieved through accuracy of additional component networks. It is shown that 4 additional component networks in the model can significantly improve the output results. All proposed conditions are satisfied while parameters like mean absolute error (MAE), root-mean square error (RMSE) and mean relative errors are notably reduced. Research finds out that new ENN model provides more accurate shaft powers than the previous ENN model. However, there are still incorrect parts of the speed-power curves which can be solved by further implementation of the component networks in the model and introduction of the new methodologies for the calculation of output results.