GreenSteam scientists have spent the last 12 months improving the clarity and accuracy of AI analytics for ship performance insights. Two new solutions have been launched to improve access to high quality, high-frequency AI-driven data.
One of two major challenges shipping companies face today when looking to optimise vessel performance is acquiring clear, accurate and trustworthy data. A second challenge is knowing how to gain knowledge from that data to improve vessel performance.
According to GreenSteam CEO Simon Whitford, while the data for improving ship performance exists, understanding what this data means for vessel performance is a “recipe for confusion,” he noted during VPO Global’s webinar ‘Analytics and advice using low frequency vessel data & data collection’ held on March 23. He explained that there are 17 critical parameters (see figure 1) that significantly affect vessel performance and in order to make noticeable improvements on the performance of the vessel, these parameters must be fully understood.
For owners and operators, it can be an overwhelming task trying to make sense of high-frequency, complex data sets. Moreover, identifying the impact of natural variables such as weather and sea temperature versus manmade decisions like course and speed, versus biological influences such as hull and propeller fouling, is an additional complexity that needs navigating to better understand vessel performance.
Figure 2 demonstrates the complexity of trying to understand these 17 variables their impact on vessel performance. The graph only plots six of the 17 parameters over a 12-hour time period but is already complex to look at. Adding more variables into the graph only exacerbates this. Mr Whitford likened this situation to 17 people from different countries speaking different languages at the same time, trying to communicate but without much success.
Greensteam has spent 13 years trying to master this complex subject and believes that AI is essential to managing this kind of data and in finding patterns that can makes sense of the data.
Non-artificial intelligence (AI) methods can be used, but the GreenSteam’s CEO says these solutions “cannot handle anything quite like these 17 dimensions and have to filter out 90 per cent of the data in an attempt to make the data manageable. This changes the problem you are trying to solve and doesn’t give you the clarity or accuracy you need.”
Mr Whitford noted that until now, the world of vessel performance management has been split into two parts. One is manual noon reports and the other is automated high-frequency data collection, of which currently only makes up a small proportion of the fleet.
According to GreenSteam, a major problem is that a lot of manual are misreported, which in turn holds back the accuracy of AI models. Mr Whitford noted that some of the 17 variables are more susceptible to misreporting than others. “Trim is really difficult, whilst fuel consumption is obviously subject to some biases here” – these biases are shown in figure 3.
“The overall problem is that artificial intelligence and high-quality vessel data are the two necessities. That’s clear. The problem is that both of these come with their own baggage. Nothing is simple. With all the problems in inaccuracies of noon reports, it looks like for the moment, we’re tied to relying on autolog data, which expensive and difficult to install. And then, since AI models are so data sensitive, it takes them quite a while to see enough data to get to know and understand vessel performance. This means the waiting time for good results is about 90 days.”
Mr Whitford went on to say that while this is possible for an owner operator that has the time to wait to ensure they have the highest accuracy of vessel specific models, for most vessels that are on a shorter term time charter, this 90 day wait time is too long.
He therefore believes that “artificial intelligence is essential,” and that “data hungry AI is the only workable approach to understanding vessel performance and emissions.”
In 2019, GreenSteam began a project to bring clarity and accuracy of AI analytics to every vessel across the world fleet. The first challenge they looked at was how to cut the time to get a decent AI model to one third of typical 90 day time period, making it suitable for a much wider group of ships. The second challenge looked at how to “evolutionise” in Mr Whitford’s own words, “noon reporting of the data needed for vessel performance to both increase frequency accuracy and make reporting transparent.”
The result was the launch of two new applications – a hybrid AI model and an app called GreenSteam Capture.
The hybrid AI model is based on 13 years of GreenSteam’s research and experience in various types of vessels and segments. Data scientists across the UK, Denmark and Poland have brought the hybrid model concept to life in the last 12 months. The hybrid AI model learns not only from the vessel specific data, but also takes more broader lessons and guidance from vessels that have previously used GreenSteam’s platform.
The hybrid model can deliver AI clarity after just 30 days of good data. The model makes very few assumptions, rather it continuously searches for patterns in the data to predict the power required from various inputs on the vessel.
The company’s second new technology, which it started working on in September 2019 is called GreenSteam Capture. The initial question that started off this project was how to get accurate vessel data, especially fuel consumption data, without manual entry from any gauge or metre on every vessel. The solution also aims to increase the frequency of data collection.
GreenSteam Capture is a downloadable app, which a seafarer can use by simply pointing it at a metre which captures the image and converts it to data. It also provides task lists to remind the operator when readings are required. Every reading is date and time stamped for accuracy. The app can read any metre on any vessel and can be used worldwide.
The app also triggers smart alerts so that onboard or onshore personnel can make timely decisions. Operational teams are also alerted to investigate when thresholds are breached, so that remedial action can be taken. For example, if the hull is underperforming, intervention can be arranged to precent additional fuel consumption.
“The key factor about GreenSteam Capture is it validates and captures audit images of every metre. With our new hybrid model, which gives clear, accurate AI analytics after 30 days of good data, and GreenSteam Capture, which provides three to four times as much data as noon reports, we’ve taken a good step towards democratising data collection for all, and making AI analytics available to every vessel in the world fleet,” Mr Whitford explained. “We don’t need equipment onboard, and we don’t need to test your patience waiting for good results.”
Learn more about GreenSteam’s vessel performance optimisation tools by clicking here.
To watch GreenSteam’s presentation on YouTube, click here.