10 minute read
AI FOR SHIPPING
Artificial intelligence and machine learning are transforming how we use our planet’s oceans. When it comes to boats and shipping, the advances being made at Southampton feature bubbles, big data and behaviour influence.
Bring Out The Bubbles
Southampton is London-based air lubrication technology company Silverstream Technologies’ longest-standing University partner, having worked together since 2014. The company’s technology works by injecting a layer of tiny air bubbles underneath ships to reduce frictional resistance between the hull and the water.
Adam Sobey, Professor of Data-Centric Engineering, and his colleagues in Maritime Engineering have been developing artificial intelligence capabilities for Silverstream for the last three years. The work is focused on applying machine learning to the bubble technology, to improve Silverstream’s current efficiency savings of five to 10 per cent.
Adam said: “We have been working on using machine learning to interpret the bubbles’ behaviour and use that to optimise the control system.”
The work has been the subject of a two-year Knowledge Transfer Partnership (KTP), led by Southampton alumnus Dr Josef Camilleri, focusing on using data and machine learning techniques to characterise the behaviour of air lubrication systems in realistic conditions and to optimise their performance. Through the KTP, a system has also been developed to automate the controls to collect better data.
Adam added: “At the moment, Silverstream uses a single power for the bubbles, so we have been working on how to efficiently collect more data to optimise the performance according to where and how fast the ship is going.”
VOYAGE OPTIMISATION
Southampton colleagues and students work with Met-Ocean data company Theyr on voyage optimisation software.
“It’s like Garmin for ships,” explained Adam. “Ships can go pretty much anywhere, so if you are trying to calculate a route from Shanghai to Rotterdam over 30 days, for example, there are countless potential routes you could sail. We have created an algorithm that finds the most optimal route, taking into consideration weather forecasts, wind speed and direction, wave speed, current, the trim and draught of the vessel, and speed of the vessel.
“Our system out-performs other voyage optimisation software by five per cent in terms of fuel savings, seven per cent in terms of arrival time, and eight per cent improvement in Time Charter Equivalent, a proxy for profit.”
The system is installed on hundreds of ships.
“We’re currently working on improving it so it can tell you why it’s selected certain routes – providing an element of explanation and justification to ships’ captains,” added Adam.
Przemyslaw Grudniewski, a former Southampton PhD student who worked on this project, now works as lead AI Scientist Applied for Theyr.
Researchers at Southampton have developed an app that has been rolled out internationally to improve shipping efficiency by optimising the amount of fuel and power needed at any given moment. The app, called Just Add Water (JAWS), has been licensed to Kongsberg Maritime, which uses it on 42 liquefied natural gas (LNG) carriers around the world.
“JAWS uses machine learning to better understand ship behaviour, and to predict how to enable a vessel to perform most efficiently,” explained Adam. “Data is taken from ships, cleaned up, and neural networks are used to predict outcomes.”
Dr Amy Parkes developed the software during her PhD, supervised by Adam and Professor Dominic Hudson. Shell Shipping and Maritime supported the research, through the Centre for Maritime Futures.
OPTIMISING YACHT DESIGN
Artificial intelligence is also being employed in yacht design, via software developed by former Engineering student Thomas Savasta.
Adam said: “Yachts are very awkward spaces, so optimising the limited space available is very important. The algorithm developed at the University configures where the rooms should be, and places furniture in the most optimal positions to make use of the space.
“We use the same algorithm as in the voyage optimisation software, but rather than searching space in the ocean to work out routes, it searches potential layouts on a ship to determine how to best use the space.”
Yacht design company Olesinski, where Thomas now works as a Research and Development Engineer, continues to develop software which utilises this algorithm to inform its yacht layouts. This has taken the design times from two weeks to two days.
BLACKING OUT
There are millions of shipwrecks around the world, but the whereabouts of only about 10 per cent of them are known. Artificial intelligence developed at Southampton is helping to find the missing millions.
Locating shipwrecks is important for many reasons – historical, archaeological, environmental and ecological, to name a few.
Dr Alexandra Karamitrou, Research Fellow in Archaeology, has developed artificial intelligence techniques to locate so-called ‘black reefs’, as these give away the locations of hidden shipwrecks.
She outlined: “The fuel and corroded iron that can be released from shipwrecks creates discolouration in the reef, changing it to a dark brown and black colour, hence the term ‘black reef’. This is visible from satellite imagery such the ones used from Google Earth.”
Alexandra and her team explored all known locations of black reefs to develop an algorithm to identify other possible locations. Identifying a black reef identifies a shipwreck, as this is the cause.
UNESCO estimates there are three million shipwrecks around the world. Alexandra said: “It’s important to know where shipwrecks are located for historical and archaeological reasons, but also to be able to monitor them from an environmental point-of-view because they can be potential sources of pollution. It’s also important to monitor them from a
biological point of view to understand how these black reefs affect their surrounding areas. It’s important for ecology and marine ecosystems, for human life, local economies, and societies.”
To test the algorithm, the team used it to compare Google Earth images from 2005 and 2022 of a site in Japan that was used as a naval training base and is known to be home to many shipwrecks. “Putting the Google Earth shots into the algorithm resulted in two totally different images – one without any black reef, and the second picking up a black reef,” said Alexandra. “In those 17 years, a vessel that is visible in the first image has broken into parts and spread around the area and turned parts of the coral reef black.”
Detecting crannogs
Alexandra is also working with Fraser Sturt, Professor of Archaeology, and Dr Stephanie Blankshein, Research Fellow in Archaeology, to apply artificial intelligence to enable the auto-detection of crannogs in Scottish lochs.
Crannogs are islands made by humans in lakes in Scotland, Wales and Ireland. They were built from the Neolithic period (circa 3,700 BC) through to the early 18th century, with new sites still being found.
Alexandra said: “We would like to identify all the potential crannogs in Scotland. We searched all lochs and developed an algorithm that can identify all islets in lochs, resulting in a catalogue of all Scottish islets, and their sizes and shapes.
“Then, we trained a deep learning model to identify which islets are likely to be crannogs. Crannogs vary in size and shape, so our goal, as we get more information, is to feed the algorithm to be able to identify a crannog itself. There are around 500 known sites, but through this method, we have already found and proved new locations.”
LOOKING BACK TO THE FUTURE
Historic data is being used to inform the future in a ground-breaking simulator that could fast-track the shipping industry to net zero.
A team from the School of Electronics and Computer Science has built a powerful simulator – a ‘digital twin’ – to select optimal shipping routes for the future, and to show how and when to upgrade ships to achieve emission goals.
The digital twin is designed to enable the most effective ways for shipping to reach net zero.
The project, led by Professor Enrico Gerding, is funded by Shell, and is conducted through the University’s Centre for Maritime Futures.
Enrico explained: “Our simulator calculates which ships should take which routes for day-to-day shipping for the next 20 years, and figures out which ships should be replaced, when and how.
“The first role of the simulator is route optimisation. We have used five years’ worth of historic data to inform the future. You can also adjust the future, taking into account predictions for where trade might increase or decrease – you can play out different scenarios.
“The second element to the simulator is calculating which ships should be upgraded first, and how, to optimally reach emission goals.”
Jan Buermann, Post-doctoral Research Fellow on the project, added: “When comparing the business case of a shipping company using only its fleet’s past operation and emissions to that of our digital twin, we find that our digital twin can find a ship upgrade schedule with similar reductions at about two-thirds of the cost.”
The simulator is based on real data from historic routes and trades. A lot of historic trade data is confidential, so the team has collected ships’ GPS data, combined with data on how deep into the water the ships were sitting and at which times and locations. This has enabled them to infer what the trade data would be.
The project, which began in 2020, is continuing to advance. PhD student Hugo Webber is now developing a sister simulator to deduce which tax policies will provide the best incentives for shipping companies to adopt the simulator’s recommendations.
SYSTEMS FOR SUSTAINABLE RESEARCH
As the drive for green maritime accelerates, oceanographic methods are advancing to enable sustainable research of our oceans.
One key aspect to net zero oceanography will be replacing some of the capabilities of highemission research ships with autonomous marine vehicles – and designing automated sensor systems for these platforms.
Two Southampton research groups have recently developed sensor systems to go on autonomous submarines for deep-sea imaging and biogeochemical analyses, as part of a collaborative research programme called OCEANIDS.
The researchers tested their developments in 2022, taking them on two research expeditions into the Atlantic and launching them into the sea on the Autonomous Underwater Vehicle (AUV) famously named Boaty McBoatface.
A team led by Mark Moore, Professor of Biogeochemistry and Head of the School of Ocean and Earth Science, developed a new sensor for measuring primary production (the rate of photosynthesis in the ocean). This work was part of a project called STAFESAPP (Single Turnover Active Fluorescence of Enclosed Samples for Aquatic Primary Production).
Mark said: “The AUV was trialled down to depths of 600 metres with a suite of 11 new biogeochemical sensors on board, including ours and others developed by our colleagues within the National Oceanography Centre [NOC]. The validation testing we did was a great success, demonstrating the potential of these new systems. Our sensor is now being further developed by our commercial partner Chelsea Technologies within the EU-funded TechOceanS project, led by NOC.”
In a separate project, a team led by Blair Thornton, Professor of Marine Autonomy, developed an automated deep sea imaging system comprising cameras, lights and machine learning techniques. The imaging system, called BioCam, was sent down to a shallower depth at the edge of the continental shelf to capture images.
The two projects were part of the £16 million OCEANIDS programme, led by the National Oceanography Centre. OCEANIDS was set up in 2016 to support the drive for net zero oceanography. It was funded by UK Research and Innovation’s Industrial Strategy Challenge Fund, through the Natural Environment Research Council.