“A Long overdue paradigm shift for maritime inspection”

 

Vesselity Maritime Analytics, an emerging startup, aims to enhance ship efficiency by 5-10% through the innovative use of artificial intelligence, a sophisticated data platform, and underwater drone inspections. Their approach is set to cut the costs of traditional diving operations by half, marking a promising start in the industry.

 

 

You both started in September 2023 with Vesselity Maritime Analytics and hold 50% each. What is the story of your founding?  

Michael Stein: We already knew each other from our time at university. My first encounter with ROVs, or »underwater drones« was at the Europort trade fair 2016. The potential of this technology fascinated me so much that I started my own business in 2018 to continue working with drones and digitalization to advance maritime infrastructure. During the Lockdowns in 2020, David came forward and explained that he dealt intensively with machine learning and AI.

He asked if I could give him data to train a neural network. I gave him videos of my ROV operations and said: If you can automatically detect fouling in it, we could have a very interesting story. Three months passed before he came back to me with his first neural network. It became clear to me that the peace and quiet was now over and that we had think about a spin-off.

You provide underwater inspections with a drone over a cable connection where an AI detects fouling on the hull... 

Stein: We recognize overgrown sea chests, corroded anodes, cavitation on propeller blades, rust damage on the rudder and of course also chipped paint and marine fouling. Our AI calculates the number of fouling or paint damage down to the pixel and distinguishes whether it is a mussel, algae or barnacle. But more importantly: We can use our results with the master data of the ship and its AIS routes and thereby draw conclusions about the resistance in the water and thus determine the impact on fuel consumption.

 

 

The probably of marine fouling varies considerably depending on the region. Does the AI ​​recognize this too? 

David Kaiser: That depends on the data situation. So far, we inspect ships only when it is already too late and they are already overgrown. We will then look at the itinerary and can only speculate. It would be better if we could inspect the ships several times during the year and identify already the first stages of fouling.

Stein: Our goal is to reduce the excess fuel consumption caused by the hull’s roughness.  We do not necessarily identify the species of fouling, but rather look at the size and texture of marine growth. Through the AIS data for example, we can issue an early warning, if a ship lies in a warm region for too long and is exposed to a high risk of fouling.

Is “underwater” a special challenge for the AI technology?

Kaiser: Yes, compared to image recognition outside of the water it is significantly more challenging. You have to find ways to get the images prepared accordingly before they can be evaluated by the AI. We have developed a whole range of procedures for this to ensure that we can also perform inspections in low-visibility environments. Currently, our condition monitoring still requires cameras to be used. In the long term, we have other technologies in mind, for example Sonar. This means you could do inspections even in complete darkness or during a storm with a much higher density of data points to work with.

 

 

 

 

 

.

 

Where does such technology in shipping reach their limits? 

Kaiser: If the water is very murky or if it sediments have been agitated a lot. A passing ship can also make the inspection more difficult. Tidal range is also an issue because there is a lot of movement underneath the waterline. To effectively provide inspection services under these conditions, it's crucial to understand these factors well. That takes a lot of hands-on experience.

You do the inspection yourself and go to the ports. To attract business, you need more people to do that, right? 

Stein: We are in cooperation with initial partner companies and some inspection locations have already been set up, such as Spain, Ireland, Egypt, Singapore, Germany, Belgium, the Netherlands, Denmark and Poland. Canada, Italy and Greece will follow in 2024. If a ship does not pass our locations, we get on a plane and arrive within 6-24 hours at the ship. We are also considering a training center for ROV divers and are currently talking to relevant stakeholders.

 

 

You didn't build the drone yourself, you bought it in Norway … 

Kaiser: Exactly, not the drone is our product, it is the data platform in the background. We prioritize integrating various data sources, alongside AI-based analysis and the management of completed inspections. Furthermore, we are open to the possibility of future collaborations with providers of underwater cleaning services.

Many still rely on the classic diving approach... 

Stein: Bringing innovation to shipping has always been a fairy tale about donkeys and windmills. We do not intend to replace the classic diver! An 8kg ROV cannot do welding work and cannot tighten any bolts. What you can achieve, however, is an underwater inspection for 50% of the cost carried out during loading or unloading operations in port. The market will change, we will have a very sharp increase in demand for underwater inspections and at the same time decreased demand for divers.    

You recently attracted attention when being the first person to conduct a recognized intermediate class inspection of a ship carried out without any divers. For that you received the first ROV class certificate from Lloyd’s Register... 

Stein: Since 2021, I am the first and so far only person who has obtained several classes accepted “In-Water Surveys” (IWS) certificated solely using ROV technology. I have pushed forward a long overdue paradigm shift in underwater inspection and against all doubt realized in 2023 a transition of theory into shipping practice. It started with a pod-propelled ship but soon, the market will change and also ope up for other types of ships.

 

 

What else could the drone do in the future? Kaiser: Our stated goal is to gain more knowledge about the properties of the ship's hull into ship management and push forward “Predictive Hull Cleaning”. 

Stein: Until now, ships were usually inspected only every two and a half years as part of their intermediate class. You lose a lot of information if you are sailing blindly for 900 days. Also for charter contracts, for sales and purchases and before a call to a dry dock, the ship should be accompanied by a drone inspection. If a ship owner/operator can save several days of dry docking based on the knowledge of a 2-hour ROV inspection, because material and spare parts can be pre ordered, it will pay off big times.

How long does it take for your AI to evaluate the footage?

Kaiser: It takes minutes and runs fully automatically. Raw image or video data goes in, ready-evaluated data comes out. What we also offer is live streaming. Here, the technical manager can monitor what happens underwater without even leaving the office. We want our software be used as a management tool where inspection and voyage data is stored and remains for later comparison of years to come. This allows us to proactively alert if a ship spends too long in a port with high levels of fouling risk for immediate actions.

How would you generally rate the willingness of (German) ship owners to talk about something like that?  

Kaiser: The openness is already there, especially after the statement through the Scandlines class. But we reach our limits when it comes to the “data” argument. What we do is data-driven. There you often fight against resistance before partners are ready to provide us even with test data. The data quality is also a challenge because poor or incomplete operational performance data also dilutes the outcome of the best AI. 

Stein: Our strongest driver is the Carbon Intensity Index (CII), which will become 5% more difficult to achieve every year. It is no coincidence that we were founded in 2023. We have the ability to increase a ship's efficiency by 5-10%, by simply bringing together intelligent data. We do not even touched upon the topic of retrofitting a ship with intelligent machinery sensors for enhanced data collection, there lies an even much higher potential for fuel optimization in it.