Article,
Prediction of Ship Main Particulars for Harbor Tugboats Using a Bayesian Network Model and Non-Linear Regression
Affiliations
- [1] Istanbul Technical University [NORA names: Turkey; Asia, Middle East; OECD];
- [2] Maritime Faculty, Istanbul Technical University, Istanbul 34940, Turkey;, omeremrekaracay@gmail.com, (Ö.E.K.);, karatug@itu.edu.tr, (Ç.K.);, uyanikt@itu.edu.tr, (T.U.);, arslanoglu@itu.edu.tr, (Y.A.)
- [3] Aalborg University [NORA names: AAU Aalborg University; University; Denmark; Europe, EU; Nordic; OECD]
Abstract
Determining the key characteristics of a ship during the concept and preliminary design phases is a critical and intricate process. In this study, we propose an alternative to traditional empirical methods by introducing a model to estimate the main particulars of diesel-powered Z-Drive harbor tugboats. This prediction is performed to determine the main particulars of tugboats: length, beam, draft, and power concerning the required service speed and bollard pull values, employing Bayesian network and non-linear regression methods. We utilized a dataset comprising 476 samples from 68 distinct diesel-powered Z-Drive harbor tugboat series to construct this model. The case study results demonstrate that the established model accurately predicts the main parameters of a tugboat with the obtained average of mean absolute percentage error values; 6.574% for the Bayesian network and 5.795%, 9.955% for non-linear regression methods. This model, therefore, proves to be a practical and valuable tool for ship designers in determining the main particulars of ships during the concept design stage by reducing revision return possibilities in further stages of ship design.