Using machine learning to cut fuel costs

In addition to using machine learning to reduce fuel consumption and emissions, computer programs can also monitor hull performance, manage speed and optimise vessel routeing to lower fuel consumption by around 15%.

Ferry group DFDS is using GreenSteam’s Dynamic Trim Optimizer (DTO) to optimise vessel trim settings during roro voyages. DTO provides dynamic optimisation throughout the voyage to enable DFDS to improve fuel efficiency, says GreenSteam head of performance management Jonas Frederiksen.

DTO takes data from GreenSteam’s onboard sensors, along with data from vessel systems, weather and sea-state information. It uses this to calculate, in real-time, optimal vessel trim settings that will maximise operational efficiency. DTO provides a continuous update of trim advice to the crew, supporting actionable decision-making for the captain.

“When implemented, it produced annual fuel savings of up to 6%,” Mr Frederiksen tells Maritime Optimisation & Communications. He says other shipping companies can follow this example and make the most of trim optimising solutions.

“Shipowners should take the first step towards machine learning as soon as possible to optimise performance and gain financial benefits,” says Mr Frederiksen. “By reducing fuel wastage, shipowners can save costs, cut emissions and help contribute to reaching IMO’s CO2 reduction goals.”

GreenSteam offers machine learning tools including Discover, Fouling Analyser, Trim Planner and Speed Optimiser. DTO combines the historical baseline performance model generated by the Discover tool with real-time data from onboard sensors, vessel systems, weather and sea-state information.

“Through this data, the onboard console and shore-side operations monitor the trim setting, and calculate optimal trim for current sailing conditions,” says Mr Frederiksen. “This enables the crew to make operational adjustments to minimise potential fuel wastage.”

“To optimise these factors during a voyage, shipowners need to collect enough data on them to understand how they impact fuel consumption,” Mr Frederiksen explains. “Machine learning is the only approach capable of fully digesting the volume of data needed to ensure this level of understanding and build an accurate picture over time.”

Another option is large-scale sampling processes to convert big data into usable information. However, up to 90% of all data could be discarded in some cases, increasing the risk of making incorrect operational decisions based on poor data.

“Machine learning, on the other hand, allows us to use 100% of the data collected to provide the most accurate and effective insights possible, which in turn creates a solid basis for optimisation decisions to save money and cut emissions,” says Mr Frederiksen.

Further development

GreenSteam’s performance optimisation software uses machine learning to create performance benchmarks for individual vessels by identifying their optimal performance in ideal calm sea conditions.

“From this, the software allows shipowners to understand the precise impact of each different factor – including trim, speed and hull fouling – in real-life scenarios,” Mr Frederiksen explains, “and provide captains with solutions to mitigate where these are sub-optimal and manage their performance to improve fuel consumption.”

This transparency allows vessel operators to select the optimal trim, route and speed for individual vessels on specific journeys among others benefits, such as creating timely and cost-effective hull cleaning schedules.

GreenSteam is developing other machine learning technologies to enhance shipowners’ and operators’ ability to control and optimise vessel operations even further. “For example, by calculating the individual impact of each factor on different types of vessels in various sea conditions,” says Mr Frederiksen.

“In this way, shipowners and operators will be able to optimise their management of hull fouling on a preventative, rather than a reactive, basis.” They can use this for establishing more effective cleaning schedules, lowering both fuel and maintenance costs.

“This will enable vessel owners to avoid charter party claims on excessive consumption while also allowing charterers to control fouling and make more effective decisions that will help avoid fuel wastage,” says Mr Frederiksen.

Reducing flood risk through simulation

The European Commission is funding the Flooding Accident Response (Flare) project to improve passenger ship safety using software and simulation. This project is developing a risk-based methodology to enable live flooding risk assessment and control for passenger ships.

It will consider vessel draught and trim as factors related to ship stability and vulnerability to flooding.

This project published its first draft reports in April 2020, including Database of Operational Data and Statistical Analysis and Analysis of Permeabilities.

Partners in the project are assessing the vulnerability of an intact ship and its survivability in case of a flooding emergency. It is anticipated project findings will help improve monitoring, assessment and analysis of a live flood, and accelerate ship-shore response.

FLARE will improve vulnerability monitoring and survivability assessments across the passenger sector. This could also eventually lead to changes in SOLAS. Assessments during the next two years of this project will use Napa programmes including its emergency and loading computers.

An early result of FLARE studies demonstrated how passenger vessels generally operate within a much narrower draught range than presently assumed within SOLAS. It was found that the calculation draughts currently considered within SOLAS should be taken at intermediate locations within the vessel draught range, as they rarely operate at the extremities.

Napa is also involved in Mitsui OSK Lines (MOL)’s Focus project for enhancing the collection and application of ship operation data. This is a multi-year project developed to reduce the environmental impact from MOL’s ships and increase safety through digital applications. Other partners include Mitsui E&S Shipbuilding and Weathernews.

Napa will combine hydrodynamic modelling with big data analytics for voyage and vessel performance across a range of vessel types. Napa already analyses performance for nearly 100 of MOL’s time-chartered bulk carriers.

Another of the FLARE project partners, DNV GL, provides trim optimisation information to shipowners from its ECO Assistant on board and onshore software. This tool is built around a pre-calculated database of ship-specific resistance and power-demand data, providing answers in all operating conditions.

ECO Assistant 4.0 displays optimum trim and related fuel savings and it reports sailing conditions and shares trim performance between ship and shore. Its fuel consumption calculation identifies and benchmarks expected fuel consumption in various operating conditions.

Trelleborg Marine offers trim optimisation as part of a modular ship performance package, which can take up to 100 different inputs to enable all ship system configurations to be accommodated and to be upgraded. Interfaces are used to link with all applicable onboard systems. Trelleborg’s Trim Optimisation tool continuously calculates ship efficiency against optimum trim. It regularly updates crew on vessel efficiency and transmits data via a ship’s IT network for onshore monitoring.

Digitalisation and vessel optimisation will be discussed in depth during Riviera’s Vessel Optimisation Webinar Week 12-15 May