Predicting Ship Diesel Engine Gas Emissions Using KAN Networks

Abstract

Maritime transport is the basis of trade by transporting goods and products in large volumes and at low cost around the world. The intense greenhouse gas emissions resulting from the use of low-quality diesel fuel by cargo ships have environmental impacts. The development of fuel use and energy efficiency increasing technologies for sustainable maritime transport is important to reduce greenhouse gas emissions. In this study, greenhouse gas emission estimation under different fault conditions and operating conditions is examined. In the study, in addition to the classical machine learning method (gradient boosting), deep learning (long short-term memory) and Kolmogorov-Arnold Networks were used. The results show that Kolmogorov-Arnold Networks are effective in time-series data analysis. The study also shows that greenhouse gas prediction is possible under different fault conditions and operating conditions. © 2025 Elsevier B.V., All rights reserved.