Skip to main content

DE-RISKING COMPLIANCE WITH MACHINE LEARNING

Tuesday, April 28, 2020 

The Industrial revolution 4.0 is in full swing and the maritime sector stands to make significant gains if it can work holistically to harness its full benefits. This rings particularly true as we hurtle towards a zero emissions future.

The challenge is no longer availability of advanced technological options, but rather figuring out what is and isn’t useful. People getting excited by the latest technology is all well and good, but if it fails to simplify, is hard to implement or isn’t practical, then little has been achieved.

Disruption

The legacy of the 2008 recession is still with us today. World events, public and social concerns combined with the tendency to regulate, present a myriad of challenges for owners and operators.

It was also in 2008 that the International Maritime Organization (IMO) began to formalise discussions around the significant reduction of sulphur emissions from shipping. The 2020 sulphur cap represents the first in a series of broad emissions focused steps towards the IMO’s 2050 goal to slash greenhouse gas (GHG) emissions to 50% of 2008 levels.

This is not the first time the industry has had to adopt new higher standards of safety and efficiency.

 

The next decade will prove critical if the industry is to keep pace with the growing demand for greener shipping solutions, coming from industry bodies, government organisations and society. Environmental concerns have climbed to the top of the agenda and the scientific evidence to back climate claims speaks for itself. However, with a large installed base of tonnage, change takes time and the shipping industry is cautious and sceptical when it comes to making choices of significance.

Adversity to change is sometimes unavoidable. Back in 2012, the introduction of mandated Electronic Chart Display and Information Systems, highlighted the rigidity of some mindsets when it came to upgrading paper charts to a digital version, despite the evidenced safety and efficiency benefits.

During these times of change and uncertainty those with a flexible but challenging mindset have the opportunity to gain most as “early adopters”. Targeting GHG reduction, delivering operational efficiencies by using data and machine learning is essential for the existing fleet and newbuild vessels.

State of the nation

In the wake of the sulphur cap implementation, it would appear that market analysis of fuel pricing is lining up with previous predictions. According to Maritime Strategies International (MSI), the movement in bunker prices in combination with MSI forecasts for price spread levels indicates that scrubber adopters look set to reap financial benefits from high sulphur fuel oil price.

For low sulphur fuels, the outlook is less certain, with the first six-months of 2020 expected to be turbulent for compliant fuel pricing. Those who could afford to will have already secured their compliant fuel of choice for a fixed fee last year and for owners looking to buy on the spot markets, it is likely they will be paying a premium. However, these are still early days so we should be cautious.

In most cases, and certainly in the case of meeting sulphur emission regulations, compliance costs money. Sitting in late January 2020, the decision to invest in scrubbers looks spectacular and the low sulphur - high sulphur spread is now based on a functioning market and therefore likely to stay for the foreseeable future.

What is less certain are the external factors - territories banning discharge of scrubber water and supply impacted by regulations limiting the carriage of HFO.  There are also legitimate questions surrounding the readiness and availability of alternative fuels such as ammonia and hydrogen along-side the credentials of LNG as a fossil fuel, particularly around methane slip issues.

 

This is further compounded by the trend to watch and wait, with early adopters facing the risk of being locked into a technology that gets surpassed as regulations change. Canny investors like to see others take the lead.

Fleetwide vessel performance optimisation (VPO) needs consistent focus, measurement and management. It is a continuing process with a strong element of human behaviour. Unusually, the barrier to entry is often almost zero - any investment in the service is quickly returned through fuel savings. Secondly, it addresses the root cause of GHG emissions – by cutting fuel consumption - every tonne of fuel saves 3mt of CO2. It is hard to argue against machine learning VPO as an element of every vessel fleet’s path towards IMO 2030 compliance.

 

Mystery around the topic often leads to healthy scepticism, Embracing just enough knowledge of machine learning to understand why it the only feasible future approach to VPO is vital for owners and operators. But in order to truly understand its significance, we need to invest some time in to understanding what it is, how it works and how it can be applied effectively to provide actionable insights.

Back to basics

While it may be perceived by some as a passing phase, shipping and data have a long-standing relationship. The first publication of Lloyd’s List in 1734 is one of the earliest examples of data collection for maritime, listing vessels and their cargoes as they arrived into port. This data provided the backbone for successors to aggregate and develop into insightful information, that went on to inform the basis of future economic studies, shaping shipping intelligence as we know it.

Machine learning starts with data. The oceans are awash with vessels operating under all manner conditions. As noted above, collecting raw data is nothing new for the industry and neither is the challenge of transforming data into simple, useful information.

 

Analysing, understanding and acting upon data is what transforms this process and creates simple, tangible business outcomes. Machine learning is essential for this to happen.

A suite of tools and algorithms enable pattern recognition by computer systems. The more data the machine has access to, the more accurate it becomes through constant revisions and refinement. When it comes to using this learning to inform operational optimisation, machine learning helps operators and owners establish how a fleet functions at its most efficient using historical and, in the case of new-builds real time data.

This information can then be used as an indicator for future outcomes. What this boils down to is fuel savings, regardless of compliance route, along-side greener operations by optimising static and dynamic trim, speed profiles and fouling. 

Real-world application

In terms of accepting data, machine learning has very flexible dietary requirements. At the simplest level, the platform reads data from vessel noon reports. Once the model is established it continues to learn with each day’s new data. Worthwhile providers can discuss cost effective upgrade paths to invest simple data acquisition equipment and transform the value of the machine learning model’s output.

We see the path to better data running parallel to behavioural acceptance of alerts and notifications on board – they must progress together.

Fitting a vessel with dataloggers requires minimal downtime. Once fitted, data can be accessed at very low additional cost to operations. Prices of sensors and communication costs have fallen significantly over the last few years, and machine learning solutions are now seen as very low OPEX and CAPEX with a strong return on investment.

Once a ship's data has been provided and a machine learning model built, a dynamic performance baseline is established and can be used to target and measure potential saving.

If you were to consider a future voyage speed optimisation (which is currently in beta test), for example, the machine learning platform performs hundreds of thousands of simulations along a particular route which would take into account how the vessel behaves in the forecast sea state, wind, currents, depth etc.. The lowest cost speed profile is then used by the vessel.

Of course, this is all well and good, however machine learning solutions are only as effective as the human designers, developers and analysts that build them.

In general, operators and owners should be on the lookout for highly effective solutions based on the latest technology from trusted advisers, experienced in their field. Future-proofing investments has never been more important, so proven success and scalability are vital in order to ensure long-term profitability from a solution.

Reader Comments (0)

There are currently no comments on this article. Why not be the first and leave your thoughts below.

Leave Your Comment

Please keep your comment on topic, any inappropriate comments may be removed.

Return to index

Web Analytics