Analysis of Machine Learning Techniques for Improvement Sea Margin Design Crew Boat

The application of the Sea Margin (SM) is inaccurate in some operational cases, which leads to avenues for improving ship design in such cases and the performance of the ship design can be improved. In this study, the speed and the power curves of crew boat engines have been calculated to predict the operational load conditions, which have been verified by sea trials. An engine power of 20% has been selected for the SM design based on prior references and experience in design results, considering operational conditions in towing tank tests. Ship performance has been evaluated by using operational data for one year. The boat has been equipped with various monitoring equipment, such as speed, engine, and flow meters. Data from these have been combined with environmental factors from a weather station. The data have been filtered via feature engineering, feature ranking, and selection techniques. Subsequently, they have been employed in the machine learning process. The prediction methods utilized in this study have encompassed Artificial Neural Network, Decision Tree, Random Forest, and K-Nearest Neighbor. The obtained Root Mean Square Error deviations have been 7.6%, 5.6%, 4.8%, and 5.3%. Thus, the evaluation of the SM design achieved during the displacement stage has been satisfactory, with a value of 20%. Pre-planing has achieved a value of 15%, whereas the planing phase has achieved a value of 5%. The change in SM during this phase is attributed to the waterjet's impact on rotation. When the engine RPM is low, the performance becomes less than optimal, and performance is enhanced for high RPM.