DOSSIER

 

Toqua – making the best ship performance models


Toqua of Ghent, Belgium, aims to do one thing – make the best ship performance models. They are designed to be incorporated into software made by other companies, such as for vessel routing and optimisation.

Toqua of Ghent, Belgium, creates ship performance models, which capture the relationship between the ship’s speed and its fuel consumption. The company seeks to compete by making better ship performance models than anyone else.

A performance model is an essential component of any digital tool to work out the best vessel route and speed, the best time to clean a hull, or if you are on track for a certain CII score.

For example, a performance model is needed to calculate how much fuel could be saved by reducing speed by a certain amount, so whether the benefits from using less fuel are justified by the increased costs of the vessel taking longer for the voyage.

For your voyage management systems, the better you can estimate fuel consumption for a future voyage, the better your estimate of its profit or loss will be, and the better it can support your commercial decision making.

A hull performance monitoring system, which determines when fouling on the hull is sufficiently advanced to justify cleaning, also requires a ship performance model. A ship with a fouled hull will have a slower speed for the same fuel consumption.

Toqua was founded based on a belief that the shipping industry should have better models. “We believe it’s an area where a lot of improvement is still possible, it can lead to a lot of fuel savings in the short term,” said Casimir Morobé, Founder of Toqua. He was speaking at Digital Ship’s webinar on October 12, Integrated modular digital components for vessel performance.

The company’s technology and business grew out of a Master’s dissertation by Mr Morobé on how ship sensor data, combined with analytics, could be used to achieve fuel savings.

Uniquely among maritime software companies, Toqua does not build software to directly support decision making, such as in generating better routes or predicting CII scores. Instead, it provides its models as a ‘kernel’ for incorporation into software tools made by other companies.

“We don’ t want to build a software solution around the models, we want you to plug these models into the tools you already have.”

Mr Morobé believes that maritime software will move to a paradigm where companies are using several different ‘modular’ components which fit together, rather than products from a single software provider. “Over time it will overtake the traditional solutions,” he said.


Making better models

Most shipping companies only capture fuel consumption data in the noon day report, and so this is the only data they have available to make their performance models.

The noon day report is compiled manually, which means it can often contain inaccuracies, Mr Morobé said. Subsequently, any models built from noon day report data can often be inaccurate.

The other common source of data for ship performance models is from the sea trial, the tests the vessel is subjected to after it is first launched. But this model is only subsequently useful when the vessel is going at the same speed, draft, and weather conditions as in the sea trial, and other conditions such as hull fouling are the same. “The actual operating conditions of a vessel can be very different,” Mr Morobé said.

You can have a better model if you have sensors gathering data about vessel performance, such as a flowmeter showing how much the engine is consuming continuously, rather than using the noon day report. Toquâ s modelling is designed primarily for ships with sensor data.

But just having sensor data does not make it easy to build performance models. The ship’s performance is affected by many factors, including waves, wind, draft, trim, currents, and water salinity. A graph of fuel consumption vs speed taken directly from sensor data shows points of data all over the graph, not in a smooth curve. So very ‘noisy’ data, as data scientists would say.

The only way to resolve it is to take all these factors into consideration when building the model. It is a “high dimensional problem,” he said.

A second challenge is that the model changes over time. The hull gets steadily fouled and then cleaned; ‘energy saving devices’ (ESDs) may be added to try to improve performance. So, the model needs to be continuously updated.

Toqua applies what it calls ‘physics informed AI’ to understand the true performance of the vessel in terms of the speed which can be achieved with different levels of fuel consumption, considering whatever the conditions of the day are. Then this model is continually retrained as more data becomes available.

The model can be demonstrated as a standalone digital tool, which will tell you what fuel consumption you can achieve at given conditions, such as speed and waves. It can tell you about the revolutions per minute of the engine and its power output.

Toquâ’s models are built by its developers using software and programming tools such as Power BI and Python. The models themselves can be shared, for example if another company wants to deploy data scientists to try to find further insights into how it can improve performance.

The models by themselves will not save fuel, the value comes from how the models are applied. But if a company like Toqua can take on the task of making the models, staff will have more time for decision making.


Using the models

Tanker operator Euronav uses Toqua for its whole fleet. Euronav found that by using Toquâ s models in its routing system, it was able to double the benefits it was getting, Mr Morobé said.

Shipowners can typically reduce fuel costs by 2.5 or 3 per cent using routing systems, so doubling that means a very worthwhile return, he said.

Dutch shipping company Vroon, which operates a variety of vessels including tankers, offshore vessels, and livestock carriers, uses Toqua for ship performance modelling, using data provided by maritime shipboard electronics and sensor company Danelec.

Vroon uses the output of the models to monitor vessel performance and analyse the impact of a hull cleaning or maintenance during dry dock, he said. Vroon also uses it to analyse the impact of deploying specific energy saving devices, assess their decarbonisation decision making, and plan a decarbonisation ‘roadmap’ for their fleet, such as installing more ESDs.


Collaboration with charterers

Speed fuel models can assist with commercial decision making, and support collaboration between owners, charters and manages in their decision about how the vessel should be operated, Mr Morobé said.

The drive to decarbonise is encouraging more collaboration and data sharing between commercial partners. All parties are seeking to reduce fuel consumption. For them to be able to see the same data and models is very helpful in reaching agreement on how to achieve it.

And if the owner, charterer, and manager are each using separate software tools and data to work out what to do, it gets very hard to reach agreement.


Gathering data

Toquâ’s systems are created to take whatever data is available, such as from weather services, noon day reports, specific ‘events’ which improve operations such as maintenance, and shipboard sensors.

Many companies have been gathering sensor data for years and not using it until they start working with Toqua. There have been several times when Toqua has had to give a shipowner the bad news that their sensor data is useless due to a technical problem.

“We are often the first company who uses the sensor data or gets to see the data, so often the first to see how bad it can be,” he said.

It may need to say, ‘hey, you have been collecting this for 2 years, but given no-one has been looking at it, this sensor has been incorrect for 2 years. You cannot use the data.’

“You need someone guarding the data quality,” he said. Working with a company like Raa Labs (see previous article) which takes responsibility for the data can be very useful. Otherwise, “People start blaming each other.”
Understanding of technology

Over the past few years, Mr Morobé has seen a great deal of scepticism from shipping companies about software providers and their claims, much of which is justified, given that in the past many software companies have exaggerated what their solutions are capable of, and tried to make it sound easier than it is.
But now, the understanding between shipping companies and providers is improving, as buyers get a better understanding of what they are looking for, and providers are becoming more transparent about what they can and cannot do, he said.
A few years ago, Mr Morobé felt it was better not to mention AI in discussions with customers. But now it is almost expected, he said.

“You see a lot of legacy providers mention it in their marketing without understanding what it means.”

While AI is part of the modelling tool, it is not the defining factor, he said. Toqua uses a version it calls “physics informed machine learning,” where the model is constrained by what physics tells you must be true, in how it learns from the operational data.

“We use AI for a very specific purpose, and we try to break down how it works,” he said.


Ds

 

 

 

  LMB-BML 2007 Webmaster & designer: Cmdt. André Jehaes - email andre.jehaes@lmb-bml.be