Fujitsu Laboratories reports that it has completed trials of a new ship performance estimation technology system with Mitsui O.S.K. Lines (MOL) and Ube Shipping & Logistics, used to predict voyage parameters such as ship speed and fuel consumption to optimise routing and save on fuel.
The company says that the trials have demonstrated an estimated margin of error under 1.5 per cent, in a variety of weather and wave conditions.
The technology was developed by Fujitsu Laboratories and Tokyo University of Marine Science and Technology (TUMSAT), and will be included as an option in Fujitsu Mobility Solution’s SPATIOWL, a Cloud-based service that Fujitsu aims to introduce in 2018.
In May 2016, Fujitsu Laboratories worked with TUMSAT to jointly develop technology that could estimate a ship’s performance with high accuracy by incorporating Fujitsu’s artificial intelligence technologies.
The trials incorporated data generated during 2015 and 2016 on an MOL ocean going vessel and a domestic vessel operated by Ube, with dozens of different types of ship data collected.
A total of 90 per cent of the sample data was selected at random from the total amount of operational data used in the trial as training data. The remaining sample data was used as test data to calculate the mean absolute percentage error (MAPE). This error evaluation was repeated ten times, and the average of those results was taken as the estimated error rate compared to actual results.
When performing this error evaluation for ship speed, the model ‘learned’ the relationship between ship speed data and non-ship speed data, and applied this knowledge to predict ship speed. A similar error evaluation was performed for fuel consumption.
The results confirmed that the technology could predict ship performance with an estimated margin of error under 1.5 per cent, Fujitsu says.